Acetic anhydride production plant by methyl acetate carbonylation.
Authorship
R.T.R.
Bachelor's Degree in Chemical Engeneering
R.T.R.
Bachelor's Degree in Chemical Engeneering
Defense date
07.16.2024 13:15
07.16.2024 13:15
Summary
The purpose of this project is to design a production plant for acetic anhydride through the carbonylation of methyl acetate. The production capacity of the plant to be designed will be 100,000 tons per year of acetic anhydride with a purity of 99.7% by weight, and it will operate continuously for 330 days a year, 24 hours a day. The key equipment of the plant includes the bubble reactor R-101, which will be designed by Rodrigo Tasende Reboredo, and the distillation column T-202, which will be designed by Jacobo Torres Pérez.
The purpose of this project is to design a production plant for acetic anhydride through the carbonylation of methyl acetate. The production capacity of the plant to be designed will be 100,000 tons per year of acetic anhydride with a purity of 99.7% by weight, and it will operate continuously for 330 days a year, 24 hours a day. The key equipment of the plant includes the bubble reactor R-101, which will be designed by Rodrigo Tasende Reboredo, and the distillation column T-202, which will be designed by Jacobo Torres Pérez.
Direction
FREIRE LEIRA, MARIA SONIA (Tutorships)
GONZALEZ ALVAREZ, JULIA (Co-tutorships)
FREIRE LEIRA, MARIA SONIA (Tutorships)
GONZALEZ ALVAREZ, JULIA (Co-tutorships)
Court
RODIL RODRIGUEZ, EVA (Chairman)
SOUTO GONZALEZ, JOSE ANTONIO (Secretary)
VAL DEL RIO, MARIA ANGELES (Member)
RODIL RODRIGUEZ, EVA (Chairman)
SOUTO GONZALEZ, JOSE ANTONIO (Secretary)
VAL DEL RIO, MARIA ANGELES (Member)
Dimethyl ether production plant from CO2 and green hydrogen
Authorship
M.D.V.
Bachelor's Degree in Chemical Engeneering
M.D.V.
Bachelor's Degree in Chemical Engeneering
Defense date
07.16.2024 10:40
07.16.2024 10:40
Summary
This project aims to design a dimethyl ether production plant using CO2 recovered through CCUS and green hydrogen produced via electrolysis, with the goal of decarbonizing a chemical process that nowadays relies on fossil-based raw materials. The proposed plant will have a production capacity of 6,000 tons per year of dimethyl ether with a molar purity of 98.7% in accordance with the required commercial purity grade, and the operation will be continuous for 330 days a year. The equipment subject to rigorous design will be the synthesis reactor R-201 and the distillation column T-302.
This project aims to design a dimethyl ether production plant using CO2 recovered through CCUS and green hydrogen produced via electrolysis, with the goal of decarbonizing a chemical process that nowadays relies on fossil-based raw materials. The proposed plant will have a production capacity of 6,000 tons per year of dimethyl ether with a molar purity of 98.7% in accordance with the required commercial purity grade, and the operation will be continuous for 330 days a year. The equipment subject to rigorous design will be the synthesis reactor R-201 and the distillation column T-302.
Direction
Rodríguez Figueiras, Óscar (Tutorships)
Rodríguez Figueiras, Óscar (Tutorships)
Court
Omil Prieto, Francisco (Chairman)
REGUEIRA LOPEZ, ALBERTE (Secretary)
MOREIRA MARTINEZ, RAMON FELIPE (Member)
Omil Prieto, Francisco (Chairman)
REGUEIRA LOPEZ, ALBERTE (Secretary)
MOREIRA MARTINEZ, RAMON FELIPE (Member)
Methylcyclohexane production plant
Authorship
A.R.N.
Bachelor's Degree in Chemical Engeneering
A.R.N.
Bachelor's Degree in Chemical Engeneering
Defense date
02.21.2024 11:30
02.21.2024 11:30
Summary
There is an undeniable need of achieveing sustainability by carrying out energy transition. Green hydrogen can be a sustainable energy vector that helps reducing polluting emissions. Spain, a country with vast renewable resources, is a possible exporter of this chemical product. This final project proposes the design of a methylcyclohexane production plant as a hydrogen carrier in Spanish territory, with a capacity of 1,485 tons of hydrogen per year, operating continuously 330 days a year. In the main document, the alternatives for hydrogen transport have been discussed, and a study of the technical and economic feasibility of the plant and the chosen LOHC cycle has been carried out. Other studies and the design of the main unit are also included. Current regulations and legislation have been taken into account for the realization of this project.
There is an undeniable need of achieveing sustainability by carrying out energy transition. Green hydrogen can be a sustainable energy vector that helps reducing polluting emissions. Spain, a country with vast renewable resources, is a possible exporter of this chemical product. This final project proposes the design of a methylcyclohexane production plant as a hydrogen carrier in Spanish territory, with a capacity of 1,485 tons of hydrogen per year, operating continuously 330 days a year. In the main document, the alternatives for hydrogen transport have been discussed, and a study of the technical and economic feasibility of the plant and the chosen LOHC cycle has been carried out. Other studies and the design of the main unit are also included. Current regulations and legislation have been taken into account for the realization of this project.
Direction
VIDAL TATO, MARIA ISABEL (Tutorships)
VIDAL TATO, MARIA ISABEL (Tutorships)
Court
HOSPIDO QUINTANA, ALMUDENA (Chairman)
MAURICIO IGLESIAS, MIGUEL (Secretary)
GONZALEZ GARCIA, SARA (Member)
HOSPIDO QUINTANA, ALMUDENA (Chairman)
MAURICIO IGLESIAS, MIGUEL (Secretary)
GONZALEZ GARCIA, SARA (Member)
Isoprene production plant from formaldehyde and 2-methylpropene
Authorship
L.C.S.
Bachelor's Degree in Chemical Engeneering
L.C.S.
Bachelor's Degree in Chemical Engeneering
Defense date
07.16.2024 11:20
07.16.2024 11:20
Summary
The objective of this TFG is the design of a plant for the production of isoprene from formaldehyde and 2-methylpropene, also known as isobutylene. The process is carried out in two reaction stages in a production plant with a capacity of 20,000 tons per year, operating 24 hours per day for 330 days per year, reserving the remaining days for cleaning and maintenance. The project includes a summary of the main characteristics and properties of the raw materials and product and a study of the market for them. It also indicates the location of the chemical plant to carry out the process. The alternative selected consists of two reactive phases: in the first, the intermediate product, 4,4-dimethyl-1,3-dioxane, is formed, and in the second, the product of interest, isoprene. The final product is 99% pure isoprene. In addition, an analysis of the technical and economic feasibility is carried out to verify that the necessary technology exists to carry out the project as well as its profitability. The calculation of the material and energy balances of the process and the design of the different units and equipment is carried out, focusing on the design of the first reaction section in which two stirred tank reactors configured in series are located. In reactors R-101-1 and R-101-2 the reaction of 37% formaldehyde with isobutylene in the presence of phosphoric acid takes place to give the reaction intermediate 4,4-dimethyl-1,3-dioxane. An analysis of the control and safety system is carried out, as well as an environmental impact study, and site plans, process diagram, R-101 reactor and control plan are drawn up.
The objective of this TFG is the design of a plant for the production of isoprene from formaldehyde and 2-methylpropene, also known as isobutylene. The process is carried out in two reaction stages in a production plant with a capacity of 20,000 tons per year, operating 24 hours per day for 330 days per year, reserving the remaining days for cleaning and maintenance. The project includes a summary of the main characteristics and properties of the raw materials and product and a study of the market for them. It also indicates the location of the chemical plant to carry out the process. The alternative selected consists of two reactive phases: in the first, the intermediate product, 4,4-dimethyl-1,3-dioxane, is formed, and in the second, the product of interest, isoprene. The final product is 99% pure isoprene. In addition, an analysis of the technical and economic feasibility is carried out to verify that the necessary technology exists to carry out the project as well as its profitability. The calculation of the material and energy balances of the process and the design of the different units and equipment is carried out, focusing on the design of the first reaction section in which two stirred tank reactors configured in series are located. In reactors R-101-1 and R-101-2 the reaction of 37% formaldehyde with isobutylene in the presence of phosphoric acid takes place to give the reaction intermediate 4,4-dimethyl-1,3-dioxane. An analysis of the control and safety system is carried out, as well as an environmental impact study, and site plans, process diagram, R-101 reactor and control plan are drawn up.
Direction
RODRIGUEZ MARTINEZ, HECTOR (Tutorships)
RODRIGUEZ MARTINEZ, HECTOR (Tutorships)
Court
Omil Prieto, Francisco (Chairman)
REGUEIRA LOPEZ, ALBERTE (Secretary)
MOREIRA MARTINEZ, RAMON FELIPE (Member)
Omil Prieto, Francisco (Chairman)
REGUEIRA LOPEZ, ALBERTE (Secretary)
MOREIRA MARTINEZ, RAMON FELIPE (Member)
Basic design of a plant for the production of butyl acetate and methanol from butan-1-ol and methyl acetate
Authorship
J.F.S.
Bachelor's Degree in Chemical Engeneering
J.F.S.
Bachelor's Degree in Chemical Engeneering
Defense date
02.21.2024 10:00
02.21.2024 10:00
Summary
The purpose of this project is to design a plant for the production of 50,000 tons per year of butyl acetate and 13,855 tons per year of methanol. The products are obtained through a transesterification reaction in a conventional distillation process from methyl acetate and butanol. The plant has a stirred tank reactor and three distillation columns to carry out the separation of the different compounds and obtain the desired products. The T-203 distillation column and reactor will be rigorously designed.
The purpose of this project is to design a plant for the production of 50,000 tons per year of butyl acetate and 13,855 tons per year of methanol. The products are obtained through a transesterification reaction in a conventional distillation process from methyl acetate and butanol. The plant has a stirred tank reactor and three distillation columns to carry out the separation of the different compounds and obtain the desired products. The T-203 distillation column and reactor will be rigorously designed.
Direction
VAL DEL RIO, MARIA ANGELES (Tutorships)
VAL DEL RIO, MARIA ANGELES (Tutorships)
Court
SOTO CAMPOS, ANA MARIA (Chairman)
Rodríguez Figueiras, Óscar (Secretary)
RODRIGUEZ MARTINEZ, HECTOR (Member)
SOTO CAMPOS, ANA MARIA (Chairman)
Rodríguez Figueiras, Óscar (Secretary)
RODRIGUEZ MARTINEZ, HECTOR (Member)
Dimethyl ether production plant from CO2 and green hydrogen
Authorship
A.N.C.P.
Bachelor's Degree in Chemical Engeneering
A.N.C.P.
Bachelor's Degree in Chemical Engeneering
Defense date
07.16.2024 10:40
07.16.2024 10:40
Summary
This project aims to design a dimethyl ether production plant using CO2 recovered through CCUS and green hydrogen produced via electrolysis, with the goal of decarbonizing a chemical process that nowadays relies on fossil-based raw materials. The proposed plant will have a production capacity of 6,000 tons per year of dimethyl ether with a molar purity of 98.7% in accordance with the required commercial purity grade, and the operation will be continuous for 330 days a year. The equipment subject to rigorous design will be the synthesis reactor R-201 and the distillation column T-302.
This project aims to design a dimethyl ether production plant using CO2 recovered through CCUS and green hydrogen produced via electrolysis, with the goal of decarbonizing a chemical process that nowadays relies on fossil-based raw materials. The proposed plant will have a production capacity of 6,000 tons per year of dimethyl ether with a molar purity of 98.7% in accordance with the required commercial purity grade, and the operation will be continuous for 330 days a year. The equipment subject to rigorous design will be the synthesis reactor R-201 and the distillation column T-302.
Direction
Rodríguez Figueiras, Óscar (Tutorships)
Rodríguez Figueiras, Óscar (Tutorships)
Court
Omil Prieto, Francisco (Chairman)
REGUEIRA LOPEZ, ALBERTE (Secretary)
MOREIRA MARTINEZ, RAMON FELIPE (Member)
Omil Prieto, Francisco (Chairman)
REGUEIRA LOPEZ, ALBERTE (Secretary)
MOREIRA MARTINEZ, RAMON FELIPE (Member)
Production plant of dimethyl ether from the catalytic dehydration of methanol
Authorship
M.D.A.
Bachelor's Degree in Chemical Engeneering
M.D.A.
Bachelor's Degree in Chemical Engeneering
Defense date
07.16.2024 09:15
07.16.2024 09:15
Summary
The objective of this project is to design a production plant for dimethyl ether (DME) from the catalytic dehydration of methanol. Its production capacity covers 50,000 tons annually at 99.95% purity by weight, operating continuously for 24 hours a day, 330 days a year, with maintenance shutdowns accounted for. Dimethyl ether is currently a product of interest as it is a competitive substitute for existing fuels. It is a clean gas, easily liquefiable, and easy to transport. It has potential in various markets, notably in power generation, as a substitute for liquefied petroleum gas, and as an automotive fuel due to its high cetane number and lack of sulfur oxide emissions. Its production is carried out in the R-201 multitubular catalytic reactor, which operates isothermally at 320ºC, producing dimethyl ether and water. This equipment is designed by Miguel Díaz Abalo. Next, in a train consisting of two distillation columns arranged in a direct sequence, the separation and purification of the product of interest take place, as well as the recirculation of the unreacted raw material. The first distillation column, designated T-301, is the equipment through which dimethyl ether is obtained at the required purity and flow conditions, and it is designed by Saleta Martínez Tella.
The objective of this project is to design a production plant for dimethyl ether (DME) from the catalytic dehydration of methanol. Its production capacity covers 50,000 tons annually at 99.95% purity by weight, operating continuously for 24 hours a day, 330 days a year, with maintenance shutdowns accounted for. Dimethyl ether is currently a product of interest as it is a competitive substitute for existing fuels. It is a clean gas, easily liquefiable, and easy to transport. It has potential in various markets, notably in power generation, as a substitute for liquefied petroleum gas, and as an automotive fuel due to its high cetane number and lack of sulfur oxide emissions. Its production is carried out in the R-201 multitubular catalytic reactor, which operates isothermally at 320ºC, producing dimethyl ether and water. This equipment is designed by Miguel Díaz Abalo. Next, in a train consisting of two distillation columns arranged in a direct sequence, the separation and purification of the product of interest take place, as well as the recirculation of the unreacted raw material. The first distillation column, designated T-301, is the equipment through which dimethyl ether is obtained at the required purity and flow conditions, and it is designed by Saleta Martínez Tella.
Direction
SOTO CAMPOS, ANA MARIA (Tutorships)
SOTO CAMPOS, ANA MARIA (Tutorships)
Court
ROCA BORDELLO, ENRIQUE (Chairman)
SANCHEZ FERNANDEZ, ADRIAN (Secretary)
MOREIRA VILAR, MARIA TERESA (Member)
ROCA BORDELLO, ENRIQUE (Chairman)
SANCHEZ FERNANDEZ, ADRIAN (Secretary)
MOREIRA VILAR, MARIA TERESA (Member)
Acrylic acid from glycerol production plant
Authorship
P.C.V.
Bachelor's Degree in Chemical Engeneering
P.C.V.
Bachelor's Degree in Chemical Engeneering
Defense date
07.16.2024 10:00
07.16.2024 10:00
Summary
Demonstration level facility of production of acrylic acid from glycerol as a first approach to the industrial scale development, as the maximum level of maturity achieved was the pilot plant scale. The traditional process for the obtention of acrylic acid is basen on the propylene oxidation, obtained directly from the combustion of fossil fuels. Due to the needs of the industry to support the energetic transition, it is necessary to search for alternatives which lessen the dependence on these type of non-renewable resources. Because of this, this work is based on the selective conversion of glycerol, a coproduct generated in the manufacture of biodiesel, which is obtained from vegetal oils and animal fats. Glycerol will react to produce acrolein in a first reactor, which will be converted into acrylic acid in a second step. This product will be purified by the usage of a flash drum, an absorption column, liquid-liquid extraction and a sequence of distillation columns. In terms of design, the student Daniel Fins Simón developes an hetereogeneous catalytic reactor in gas phase, while the student Pablo Carrera developes a liquid-liquid extraction unit.
Demonstration level facility of production of acrylic acid from glycerol as a first approach to the industrial scale development, as the maximum level of maturity achieved was the pilot plant scale. The traditional process for the obtention of acrylic acid is basen on the propylene oxidation, obtained directly from the combustion of fossil fuels. Due to the needs of the industry to support the energetic transition, it is necessary to search for alternatives which lessen the dependence on these type of non-renewable resources. Because of this, this work is based on the selective conversion of glycerol, a coproduct generated in the manufacture of biodiesel, which is obtained from vegetal oils and animal fats. Glycerol will react to produce acrolein in a first reactor, which will be converted into acrylic acid in a second step. This product will be purified by the usage of a flash drum, an absorption column, liquid-liquid extraction and a sequence of distillation columns. In terms of design, the student Daniel Fins Simón developes an hetereogeneous catalytic reactor in gas phase, while the student Pablo Carrera developes a liquid-liquid extraction unit.
Direction
VIDAL TATO, MARIA ISABEL (Tutorships)
VIDAL TATO, MARIA ISABEL (Tutorships)
Court
Omil Prieto, Francisco (Chairman)
REGUEIRA LOPEZ, ALBERTE (Secretary)
MOREIRA MARTINEZ, RAMON FELIPE (Member)
Omil Prieto, Francisco (Chairman)
REGUEIRA LOPEZ, ALBERTE (Secretary)
MOREIRA MARTINEZ, RAMON FELIPE (Member)
Ethyl Acetate production plant by esterification of acetic acid with ethanol
Authorship
J.F.G.
Bachelor's Degree in Chemical Engeneering
J.F.G.
Bachelor's Degree in Chemical Engeneering
Defense date
07.16.2024 09:55
07.16.2024 09:55
Summary
Ethyl acetate is a chemical compound used as a solvent in a large number of industrial processes, as well as an additive in both the food and cosmetic sectors. The plant's capacity is 25,000 tons per year since production in Spain is not of high tonnage. Ethyl acetate is produced through the esterification reaction between acetic acid and ethanol, with water obtained as a byproduct of this reaction. Additionally, as a sustainable alternative, the option of using bioethanol as a raw material is proposed. Joel Fernández García is responsible for designing the heterogeneous catalytic reactor, while Carlos Fuentes Varela designs the distillation column where the product is obtained. The separation train consists of three distillation columns and one liquid-liquid extraction unit.
Ethyl acetate is a chemical compound used as a solvent in a large number of industrial processes, as well as an additive in both the food and cosmetic sectors. The plant's capacity is 25,000 tons per year since production in Spain is not of high tonnage. Ethyl acetate is produced through the esterification reaction between acetic acid and ethanol, with water obtained as a byproduct of this reaction. Additionally, as a sustainable alternative, the option of using bioethanol as a raw material is proposed. Joel Fernández García is responsible for designing the heterogeneous catalytic reactor, while Carlos Fuentes Varela designs the distillation column where the product is obtained. The separation train consists of three distillation columns and one liquid-liquid extraction unit.
Direction
SOTO CAMPOS, ANA MARIA (Tutorships)
SOTO CAMPOS, ANA MARIA (Tutorships)
Court
ROCA BORDELLO, ENRIQUE (Chairman)
SANCHEZ FERNANDEZ, ADRIAN (Secretary)
MOREIRA VILAR, MARIA TERESA (Member)
ROCA BORDELLO, ENRIQUE (Chairman)
SANCHEZ FERNANDEZ, ADRIAN (Secretary)
MOREIRA VILAR, MARIA TERESA (Member)
Acrylic acid production plant from glycerol
Authorship
D.F.S.M.
Bachelor's Degree in Chemical Engeneering
D.F.S.M.
Bachelor's Degree in Chemical Engeneering
Defense date
07.16.2024 10:00
07.16.2024 10:00
Summary
Demonstration level facility of production of acrylic acid from glycerol as a first approach to the industrial scale development, as the maximum level of maturity achieved was the pilot plant scale. The traditional process for the obtention of acrylic acid is basen on the propylene oxidation, obtained directly from the combustion of fossil fuels. Due to the needs of the industry to support the energetic transition, it is necessary to search for alternatives which lessen the dependence on these type of non-renewable resources. Because of this, this work is based on the selective conversion of glycerol, a coproduct generated in the manufacture of biodiesel, which is obtained from vegetal oils and animal fats. Glycerol will react to produce acrolein in a first reactor, which will be converted into acrylic acid in a second step. This product will be purified by the usage of a flash drum, an absorption column, liquid-liquid extraction and a sequence of distillation columns. In terms of design, the student Daniel Fins Simón developes an hetereogeneous catalytic reactor in gas phase, while the student Pablo Carrera developes a liquid-liquid extraction unit.
Demonstration level facility of production of acrylic acid from glycerol as a first approach to the industrial scale development, as the maximum level of maturity achieved was the pilot plant scale. The traditional process for the obtention of acrylic acid is basen on the propylene oxidation, obtained directly from the combustion of fossil fuels. Due to the needs of the industry to support the energetic transition, it is necessary to search for alternatives which lessen the dependence on these type of non-renewable resources. Because of this, this work is based on the selective conversion of glycerol, a coproduct generated in the manufacture of biodiesel, which is obtained from vegetal oils and animal fats. Glycerol will react to produce acrolein in a first reactor, which will be converted into acrylic acid in a second step. This product will be purified by the usage of a flash drum, an absorption column, liquid-liquid extraction and a sequence of distillation columns. In terms of design, the student Daniel Fins Simón developes an hetereogeneous catalytic reactor in gas phase, while the student Pablo Carrera developes a liquid-liquid extraction unit.
Direction
VIDAL TATO, MARIA ISABEL (Tutorships)
VIDAL TATO, MARIA ISABEL (Tutorships)
Court
Omil Prieto, Francisco (Chairman)
REGUEIRA LOPEZ, ALBERTE (Secretary)
MOREIRA MARTINEZ, RAMON FELIPE (Member)
Omil Prieto, Francisco (Chairman)
REGUEIRA LOPEZ, ALBERTE (Secretary)
MOREIRA MARTINEZ, RAMON FELIPE (Member)
State of the art of Bluetooth pentesting tools and security analysis of an IoT device with Bluetooth communications
Authorship
A.R.C.
Computer Engineering Degree
A.R.C.
Computer Engineering Degree
Defense date
07.19.2024 18:00
07.19.2024 18:00
Summary
This work explores the state of the art of Bluetooth pentesting tools, evaluating their ability to identify vulnerabilities in IoT devices using this technology. It has been demonstrated that while no single tool can cover all possible threats, a combination of them allows for complete and effective security audits. The tools studied include Kismet, Ubertooth, Bluehydra, BtleJuice, GATTacker, Wireshark, BlueScanner, BlueDiving, Blooover, BlueSnarfer, BlueSpam, and BluePrinting, each with its specific capabilities and limitations. Additionally, the importance of a structured methodology, such as BSAM, has been highlighted to improve the coherence and depth of the audits. The implementation of these tools in real audits has allowed for the identification of critical vulnerabilities and the provision of recommendations to mitigate associated risks. This work concludes that the combined use of pentesting tools, along with a robust methodology, is essential to ensure the security of Bluetooth devices.
This work explores the state of the art of Bluetooth pentesting tools, evaluating their ability to identify vulnerabilities in IoT devices using this technology. It has been demonstrated that while no single tool can cover all possible threats, a combination of them allows for complete and effective security audits. The tools studied include Kismet, Ubertooth, Bluehydra, BtleJuice, GATTacker, Wireshark, BlueScanner, BlueDiving, Blooover, BlueSnarfer, BlueSpam, and BluePrinting, each with its specific capabilities and limitations. Additionally, the importance of a structured methodology, such as BSAM, has been highlighted to improve the coherence and depth of the audits. The implementation of these tools in real audits has allowed for the identification of critical vulnerabilities and the provision of recommendations to mitigate associated risks. This work concludes that the combined use of pentesting tools, along with a robust methodology, is essential to ensure the security of Bluetooth devices.
Direction
CARIÑENA AMIGO, MARIA PURIFICACION (Tutorships)
TARASCO ACUÑA, MIGUEL (Co-tutorships)
Vázquez Blanco, Antonio (Co-tutorships)
CARIÑENA AMIGO, MARIA PURIFICACION (Tutorships)
TARASCO ACUÑA, MIGUEL (Co-tutorships)
Vázquez Blanco, Antonio (Co-tutorships)
Court
Barro Ameneiro, Senén (Chairman)
TOBAR QUINTANAR, ALEJANDRO JOSE (Secretary)
LADRA GONZALEZ, MANUEL EULOGIO (Member)
Barro Ameneiro, Senén (Chairman)
TOBAR QUINTANAR, ALEJANDRO JOSE (Secretary)
LADRA GONZALEZ, MANUEL EULOGIO (Member)
Blockchain and smart contracts to guarantee the security and integrity of data
Authorship
E.P.R.
Computer Engineering Degree
E.P.R.
Computer Engineering Degree
Defense date
09.12.2024 12:30
09.12.2024 12:30
Summary
This document provides a systematic and practical approach to the development and testing of decentralized applications on a blockchain, demonstrating among other things the viability, efficiency and great usefulness of these platforms for projects, such as decentralized financial systems (DeFi), chains of provisioning, digital identity management and online voting applications among others. First, it condenses a large amount of relevant information about of five of the most popular blockchain platforms currently. This detailed information about its features, advantages and disadvantages along with a comparative analysis that includes aspects such as scalability, speed of transaction and costs among others can be of great value for any developer or researcher interested in the field of blockchain technologies. Additionally, the document offers a step-by-step guide for the installation, development and deployment of smart contracts on a blockchain platform and applications. decentralized applications (dApps) that will be in charge of interacting with said contracts to offer a more visual approach to users. With all this, it provides a solid foundation for future developments, contributing to knowledge and advancement in the field of blockchain technologies, the which is in constant development and which increasingly covers more types of projects in the world of computing.
This document provides a systematic and practical approach to the development and testing of decentralized applications on a blockchain, demonstrating among other things the viability, efficiency and great usefulness of these platforms for projects, such as decentralized financial systems (DeFi), chains of provisioning, digital identity management and online voting applications among others. First, it condenses a large amount of relevant information about of five of the most popular blockchain platforms currently. This detailed information about its features, advantages and disadvantages along with a comparative analysis that includes aspects such as scalability, speed of transaction and costs among others can be of great value for any developer or researcher interested in the field of blockchain technologies. Additionally, the document offers a step-by-step guide for the installation, development and deployment of smart contracts on a blockchain platform and applications. decentralized applications (dApps) that will be in charge of interacting with said contracts to offer a more visual approach to users. With all this, it provides a solid foundation for future developments, contributing to knowledge and advancement in the field of blockchain technologies, the which is in constant development and which increasingly covers more types of projects in the world of computing.
Direction
CARIÑENA AMIGO, MARIA PURIFICACION (Tutorships)
Fábregas Fernández, Jaime (Co-tutorships)
TARASCO ACUÑA, MIGUEL (Co-tutorships)
CARIÑENA AMIGO, MARIA PURIFICACION (Tutorships)
Fábregas Fernández, Jaime (Co-tutorships)
TARASCO ACUÑA, MIGUEL (Co-tutorships)
Court
Argüello Pedreira, Francisco Santiago (Chairman)
ALONSO MORAL, JOSE MARIA (Secretary)
MUCIENTES MOLINA, MANUEL FELIPE (Member)
Argüello Pedreira, Francisco Santiago (Chairman)
ALONSO MORAL, JOSE MARIA (Secretary)
MUCIENTES MOLINA, MANUEL FELIPE (Member)
API development for scheduling in heterogeneous ARM systems
Authorship
D.C.C.
Computer Engineering Degree
D.C.C.
Computer Engineering Degree
Defense date
07.19.2024 11:30
07.19.2024 11:30
Summary
Nowadays, society uses their technological devices for increasingly complex tasks that requires a huge use of the CPU, so the designers of the processor models and architectures need to adapt to these changes. Based on this need, it exists an architecture called big.LITTLE and designed by ARM that has cores with different efficiencies that allows the execution of the most computationally expensive tasks in the most efficient cores and the simple tasks in the cores with a lower performance and a lower energy consume. This work is developed in this context, creating an API that allows users to control the different cores of the processor. It also gives the users the ability to assign a core to a specific task. First, the API allows users to perform a speed test on a specific core and later, the assignment of one or more cores to a task.
Nowadays, society uses their technological devices for increasingly complex tasks that requires a huge use of the CPU, so the designers of the processor models and architectures need to adapt to these changes. Based on this need, it exists an architecture called big.LITTLE and designed by ARM that has cores with different efficiencies that allows the execution of the most computationally expensive tasks in the most efficient cores and the simple tasks in the cores with a lower performance and a lower energy consume. This work is developed in this context, creating an API that allows users to control the different cores of the processor. It also gives the users the ability to assign a core to a specific task. First, the API allows users to perform a speed test on a specific core and later, the assignment of one or more cores to a task.
Direction
PICHEL CAMPOS, JUAN CARLOS (Tutorships)
PIÑEIRO POMAR, CESAR ALFREDO (Co-tutorships)
PICHEL CAMPOS, JUAN CARLOS (Tutorships)
PIÑEIRO POMAR, CESAR ALFREDO (Co-tutorships)
Court
PENA BRAGE, FRANCISCO JOSE (Chairman)
MERA PEREZ, DAVID (Secretary)
CARIÑENA AMIGO, MARIA PURIFICACION (Member)
PENA BRAGE, FRANCISCO JOSE (Chairman)
MERA PEREZ, DAVID (Secretary)
CARIÑENA AMIGO, MARIA PURIFICACION (Member)
Population and correlation study in cardiac rehabilitation patients.
Authorship
C.P.I.
Computer Engineering Degree
C.P.I.
Computer Engineering Degree
Defense date
07.18.2024 12:00
07.18.2024 12:00
Summary
This final degree project is part of a research project in collaboration with the Department of Cardiology of the “Hospital Clínico Universitario de Santiago de Compostela” on the prediction of survival in patients involved in cardiac rehabilitation programs. This topic is of utmost interest since heart disease is responsible for a large number of deaths each year and there is a lack of information about the aetiology of many of these diseases. The medical team designed and proposed a three-year rehabilitation program to several patients with a history of cardiovascular disease. The medical follow-up resulted in a database of more than 3,000 patients and more than 300 variables. However, as usual in a medical context, mainly due to the lack of automation and standardisation of the data collection processes, the dataset was incomplete, unbalanced and with errors. We proposed a detailed framework to carry out a rigorous pre-processing of the data to obtain a complete and error-free database. This included the investigation of the meaning of variables (their typology, medical significance and fundamental dependency relationships), missing data management (including data imputation), automatic error and outlier detection and variable engineering (calculation of new variables from other variables). After this process we obtained a complete and error-free database of 445 patients and 83 variables. We then proposed the use of different statistical techniques to carry out a correlation study between the variables, and finally we addressed the prediction of adverse events such as hospital readmission or myocardial infarction. The correlation study showed significant relationships between different variables, including obvious ones (such as weight and abdominal circumference), others that would be expected from a cardiovascular point of view (such as severity of myocardial infarction and worse heart function), others that were unexpected (such as younger age with depression and worse diet), and others that require medical knowledge to interpret (such as atrial fibrillation and its usual medication). We found that it was not possible to attempt to predict any of the major cardiovascular events included in the dataset with the minimum of confidence, since after the necessary data preparation process, there were insufficient representative samples of these cases. This report presents a set of techniques for conducting a population-based, variable correlation, and event prediction study that can be used as a guide in similar medical studies.
This final degree project is part of a research project in collaboration with the Department of Cardiology of the “Hospital Clínico Universitario de Santiago de Compostela” on the prediction of survival in patients involved in cardiac rehabilitation programs. This topic is of utmost interest since heart disease is responsible for a large number of deaths each year and there is a lack of information about the aetiology of many of these diseases. The medical team designed and proposed a three-year rehabilitation program to several patients with a history of cardiovascular disease. The medical follow-up resulted in a database of more than 3,000 patients and more than 300 variables. However, as usual in a medical context, mainly due to the lack of automation and standardisation of the data collection processes, the dataset was incomplete, unbalanced and with errors. We proposed a detailed framework to carry out a rigorous pre-processing of the data to obtain a complete and error-free database. This included the investigation of the meaning of variables (their typology, medical significance and fundamental dependency relationships), missing data management (including data imputation), automatic error and outlier detection and variable engineering (calculation of new variables from other variables). After this process we obtained a complete and error-free database of 445 patients and 83 variables. We then proposed the use of different statistical techniques to carry out a correlation study between the variables, and finally we addressed the prediction of adverse events such as hospital readmission or myocardial infarction. The correlation study showed significant relationships between different variables, including obvious ones (such as weight and abdominal circumference), others that would be expected from a cardiovascular point of view (such as severity of myocardial infarction and worse heart function), others that were unexpected (such as younger age with depression and worse diet), and others that require medical knowledge to interpret (such as atrial fibrillation and its usual medication). We found that it was not possible to attempt to predict any of the major cardiovascular events included in the dataset with the minimum of confidence, since after the necessary data preparation process, there were insufficient representative samples of these cases. This report presents a set of techniques for conducting a population-based, variable correlation, and event prediction study that can be used as a guide in similar medical studies.
Direction
VILA BLANCO, NICOLAS (Tutorships)
CARREIRA NOUCHE, MARIA JOSE (Co-tutorships)
NUÑEZ GARCIA, MARTA (Co-tutorships)
VILA BLANCO, NICOLAS (Tutorships)
CARREIRA NOUCHE, MARIA JOSE (Co-tutorships)
NUÑEZ GARCIA, MARTA (Co-tutorships)
Court
VIDAL AGUIAR, JUAN CARLOS (Chairman)
DOSIL LAGO, RAQUEL (Secretary)
SAAVEDRA NIEVES, ALEJANDRO (Member)
VIDAL AGUIAR, JUAN CARLOS (Chairman)
DOSIL LAGO, RAQUEL (Secretary)
SAAVEDRA NIEVES, ALEJANDRO (Member)
Development of a tactile application to study the possibility of cognitive impairment in the elderly.
Authorship
S.S.C.
Computer Engineering Degree
S.S.C.
Computer Engineering Degree
Defense date
07.18.2024 10:30
07.18.2024 10:30
Summary
This final degree project focuses on the development of a tactile application for the assessment of cognitive impairment in the elderly. The application is based on the integration of recognised cognitive tests, adapted to an interactive and accessible digital format, with the aim of facilitating the early detection of cognitive disorders such as dementia and Alzheimer's disease. The project addresses the creation of a technological tool that not only allows for cognitive assessments, but also offers an attractive and user-friendly user experience. The tests included assess various cognitive functions, such as semantic memory and executive functions, through interactive games designed specifically for this purpose. The usability and technical tests conducted have shown that the application meets the established requirements, providing accurate and reliable results. In addition, the application allows users to consult and export their results for further analysis, which is crucial for health professionals.
This final degree project focuses on the development of a tactile application for the assessment of cognitive impairment in the elderly. The application is based on the integration of recognised cognitive tests, adapted to an interactive and accessible digital format, with the aim of facilitating the early detection of cognitive disorders such as dementia and Alzheimer's disease. The project addresses the creation of a technological tool that not only allows for cognitive assessments, but also offers an attractive and user-friendly user experience. The tests included assess various cognitive functions, such as semantic memory and executive functions, through interactive games designed specifically for this purpose. The usability and technical tests conducted have shown that the application meets the established requirements, providing accurate and reliable results. In addition, the application allows users to consult and export their results for further analysis, which is crucial for health professionals.
Direction
VALLADARES RODRIGUEZ, SONIA MARIA (Tutorships)
CATALA BOLOS, ALEJANDRO (Co-tutorships)
VALLADARES RODRIGUEZ, SONIA MARIA (Tutorships)
CATALA BOLOS, ALEJANDRO (Co-tutorships)
Court
Cotos Yáñez, José Manuel (Chairman)
QUESADA BARRIUSO, PABLO (Secretary)
GAGO COUSO, FELIPE (Member)
Cotos Yáñez, José Manuel (Chairman)
QUESADA BARRIUSO, PABLO (Secretary)
GAGO COUSO, FELIPE (Member)
Self-learning for re-identification of people in an open world environment without known starting identities
Authorship
P.L.P.
Computer Engineering Degree
P.L.P.
Computer Engineering Degree
Defense date
07.18.2024 12:00
07.18.2024 12:00
Summary
The re-identification of people is one of the big problems to be addressed in the field of computer vision. Variations in appearance and occlusions are some of the reasons that make this a complex task. Current models are not yet fully refined, so exploring various alternatives may be beneficial. This work is focused on studying several efficient methods for the initialization of an already existing model, without a prior extraction phase, and mechanisms for detecting duplicate identities. The purpose of this work is to study different alternatives and try to improve an already created person re-identification model, which can achieve comparable or superior performance to current methods, providing a solid basis for future research.
The re-identification of people is one of the big problems to be addressed in the field of computer vision. Variations in appearance and occlusions are some of the reasons that make this a complex task. Current models are not yet fully refined, so exploring various alternatives may be beneficial. This work is focused on studying several efficient methods for the initialization of an already existing model, without a prior extraction phase, and mechanisms for detecting duplicate identities. The purpose of this work is to study different alternatives and try to improve an already created person re-identification model, which can achieve comparable or superior performance to current methods, providing a solid basis for future research.
Direction
Pardo López, Xosé Manuel (Tutorships)
Díaz Parga, Cesar (Co-tutorships)
Pardo López, Xosé Manuel (Tutorships)
Díaz Parga, Cesar (Co-tutorships)
Court
Díaz Parga, Cesar (Student’s tutor)
Pardo López, Xosé Manuel (Student’s tutor)
Díaz Parga, Cesar (Student’s tutor)
Pardo López, Xosé Manuel (Student’s tutor)
NOBI: Dashboard Application
Authorship
P.U.C.
Computer Engineering Degree
P.U.C.
Computer Engineering Degree
Defense date
09.12.2024 13:00
09.12.2024 13:00
Summary
This Final Degree Project addresses the development of a cross-platform application designed for the dynamic visualization of business data. The main objective of the application is to enable users to visualize and manage data from different sources through customizable dashboards, thus facilitating strategic decision-making with a clear and consolidated view of key performance indicators (KPIs) and other relevant metrics. The application has been developed using a microservices architecture to ensure scalability and flexibility, integrating with various data sources. The back-end is implemented with microservices in Spring Boot, orchestrated using Docker, and utilizes an API Gateway to manage communications and user authentication via Firebase Auth. The database used is PostgreSQL, chosen for its ability to handle concurrent transactions and its scalability in distributed environments. The front-end is developed in Flutter, allowing the application to be compiled for mobile devices as well as desktop and web applications, ensuring a consistent and accessible user experience from any platform.
This Final Degree Project addresses the development of a cross-platform application designed for the dynamic visualization of business data. The main objective of the application is to enable users to visualize and manage data from different sources through customizable dashboards, thus facilitating strategic decision-making with a clear and consolidated view of key performance indicators (KPIs) and other relevant metrics. The application has been developed using a microservices architecture to ensure scalability and flexibility, integrating with various data sources. The back-end is implemented with microservices in Spring Boot, orchestrated using Docker, and utilizes an API Gateway to manage communications and user authentication via Firebase Auth. The database used is PostgreSQL, chosen for its ability to handle concurrent transactions and its scalability in distributed environments. The front-end is developed in Flutter, allowing the application to be compiled for mobile devices as well as desktop and web applications, ensuring a consistent and accessible user experience from any platform.
Direction
VARELA PET, JOSE (Tutorships)
Piñeiro Fernández, Silvia (Co-tutorships)
VARELA PET, JOSE (Tutorships)
Piñeiro Fernández, Silvia (Co-tutorships)
Court
Argüello Pedreira, Francisco Santiago (Chairman)
ALONSO MORAL, JOSE MARIA (Secretary)
MUCIENTES MOLINA, MANUEL FELIPE (Member)
Argüello Pedreira, Francisco Santiago (Chairman)
ALONSO MORAL, JOSE MARIA (Secretary)
MUCIENTES MOLINA, MANUEL FELIPE (Member)
Environment conditions monitoring of a building using Zigbee.
Authorship
P.G.L.
Computer Engineering Degree
P.G.L.
Computer Engineering Degree
Defense date
09.13.2024 12:00
09.13.2024 12:00
Summary
Human beings spend most of their time, especially today, inside a building, whether to live in it, for work reasons, for leisure... In short, human society focuses on the building throughout its life. Therefore, monitoring the aspects that may be of greatest importance is vital to improve quality of life and/or enjoyment. This fact makes building monitoring, within its different categories, a very useful way of providing information that is complex to obtain and process in another way. This information can significantly help improve the potential of the services that the building offers to its users. Among the various information that can be obtained through building monitoring, the most common is the monitoring of environmental conditions, since they help establish a suitable environment so that both people and other living beings, if applicable, can enjoy high environmental quality. . Furthermore, environmental comfort in a given space is one of the most important aspects when it comes to establishing a suitable place both to live and to visit at the right time. For these reasons, in this report it has been decided to implement a system that allows monitoring the environmental conditions of a building, so that information can be obtained in real time on different environmental factors such as temperature and humidity. Likewise, more measures will be incorporated that can be useful in the day-to-day life of a building, such as presence detection and door opening detection. All this information will be obtained through different types of sensors, which will send the measurements to an IoT node through the Zigbee protocol. Inside the node, all the logic necessary to interpret the measurements will be carried out, in addition to including a graphical interface for greater ease in configuring and monitoring the sensors. Finally, the information will be sent to a platform, where the end user will be able to see both the configured sensors and their respective measurements.
Human beings spend most of their time, especially today, inside a building, whether to live in it, for work reasons, for leisure... In short, human society focuses on the building throughout its life. Therefore, monitoring the aspects that may be of greatest importance is vital to improve quality of life and/or enjoyment. This fact makes building monitoring, within its different categories, a very useful way of providing information that is complex to obtain and process in another way. This information can significantly help improve the potential of the services that the building offers to its users. Among the various information that can be obtained through building monitoring, the most common is the monitoring of environmental conditions, since they help establish a suitable environment so that both people and other living beings, if applicable, can enjoy high environmental quality. . Furthermore, environmental comfort in a given space is one of the most important aspects when it comes to establishing a suitable place both to live and to visit at the right time. For these reasons, in this report it has been decided to implement a system that allows monitoring the environmental conditions of a building, so that information can be obtained in real time on different environmental factors such as temperature and humidity. Likewise, more measures will be incorporated that can be useful in the day-to-day life of a building, such as presence detection and door opening detection. All this information will be obtained through different types of sensors, which will send the measurements to an IoT node through the Zigbee protocol. Inside the node, all the logic necessary to interpret the measurements will be carried out, in addition to including a graphical interface for greater ease in configuring and monitoring the sensors. Finally, the information will be sent to a platform, where the end user will be able to see both the configured sensors and their respective measurements.
Direction
Taboada González, José Ángel (Tutorships)
Ruíz García, Miguel (Co-tutorships)
Taboada González, José Ángel (Tutorships)
Ruíz García, Miguel (Co-tutorships)
Court
Ruíz García, Miguel (Student’s tutor)
Taboada González, José Ángel (Student’s tutor)
Ruíz García, Miguel (Student’s tutor)
Taboada González, José Ángel (Student’s tutor)
Extension and improvement of a Multilanguage Framework for BigData processing
Authorship
M.R.D.
Bachelor's Degree in Computer Engineering
M.R.D.
Bachelor's Degree in Computer Engineering
Defense date
02.16.2024 09:30
02.16.2024 09:30
Summary
BigData is one of the fastest growing fields in computer science. Because of the big amounts of data and its variety, it started to be difficult or even unfeasible to process data with tradicional techniques. Because of this, new programming models such as MapReduce and new frameworks supporting this new models appeared. Current reference framework is Spark, which enables efficient massive data processing in a reasonable time. IgnisHPC is a data processing framework partially inspired on Spark and designed to improve some of its limitations. IgnisHPC has objectives such as the unification of the BigData and HPC workloads, multiple language compatibility, MPI integration, and a container virtualization design. In this work we develop the Java implementation for some of the IgnisHPC framework modules. We indicate the limitations between Spark and IgnisHPC, and compare performance with a test program.
BigData is one of the fastest growing fields in computer science. Because of the big amounts of data and its variety, it started to be difficult or even unfeasible to process data with tradicional techniques. Because of this, new programming models such as MapReduce and new frameworks supporting this new models appeared. Current reference framework is Spark, which enables efficient massive data processing in a reasonable time. IgnisHPC is a data processing framework partially inspired on Spark and designed to improve some of its limitations. IgnisHPC has objectives such as the unification of the BigData and HPC workloads, multiple language compatibility, MPI integration, and a container virtualization design. In this work we develop the Java implementation for some of the IgnisHPC framework modules. We indicate the limitations between Spark and IgnisHPC, and compare performance with a test program.
Direction
PICHEL CAMPOS, JUAN CARLOS (Tutorships)
PIÑEIRO POMAR, CESAR ALFREDO (Co-tutorships)
PICHEL CAMPOS, JUAN CARLOS (Tutorships)
PIÑEIRO POMAR, CESAR ALFREDO (Co-tutorships)
Court
SUAREZ GAREA, JORGE ALBERTO (Coordinator)
BUGARIN DIZ, ALBERTO JOSE (Chairman)
VILLARROYA FERNANDEZ, SEBASTIAN (Secretary)
BORRAJO GARCIA, MARIA ISABEL (Member)
SUAREZ GAREA, JORGE ALBERTO (Coordinator)
BUGARIN DIZ, ALBERTO JOSE (Chairman)
VILLARROYA FERNANDEZ, SEBASTIAN (Secretary)
BORRAJO GARCIA, MARIA ISABEL (Member)
Design and implementation of a networked system for cybersecurity practices
Authorship
J.E.C.
Bachelor's Degree in Computer Engineering
J.E.C.
Bachelor's Degree in Computer Engineering
Defense date
07.18.2024 11:00
07.18.2024 11:00
Summary
The subject of Cybersecurity taught in the Degree in Computer Engineering at the University of Santiago de Compostela's main objectives are to know and understand the main attacks that can receive a computer system. For the explanation and understanding of the different existing attacks, practices are carried out using a large number of virtual machines, many of them with numerous vulnerabilities. These practices have to be developed in an isolated and secure environment. Therefore, in this work we have implemented a networked system with a virtualization server that allows to configure and deploy all the virtual machines, a network has been configured to provide the required connectivity between the different systems, and several tests and pentesting exercises have been executed that can be performed in this safe and controlled environment.
The subject of Cybersecurity taught in the Degree in Computer Engineering at the University of Santiago de Compostela's main objectives are to know and understand the main attacks that can receive a computer system. For the explanation and understanding of the different existing attacks, practices are carried out using a large number of virtual machines, many of them with numerous vulnerabilities. These practices have to be developed in an isolated and secure environment. Therefore, in this work we have implemented a networked system with a virtualization server that allows to configure and deploy all the virtual machines, a network has been configured to provide the required connectivity between the different systems, and several tests and pentesting exercises have been executed that can be performed in this safe and controlled environment.
Direction
CARIÑENA AMIGO, MARIA PURIFICACION (Tutorships)
CARIÑENA AMIGO, MARIA PURIFICACION (Tutorships)
Court
CARIÑENA AMIGO, MARIA PURIFICACION (Student’s tutor)
CARIÑENA AMIGO, MARIA PURIFICACION (Student’s tutor)
Development of a scalable architecture for the integration of systems in intelligent buildings
Authorship
H.B.M.
Bachelor's Degree in Computer Engineering
H.B.M.
Bachelor's Degree in Computer Engineering
Defense date
07.19.2024 10:30
07.19.2024 10:30
Summary
The world of home automation has gone through a major growth phase in recent years and this has led to the emergence of many different manufacturers and systems, and even different communication protocols. This is very good in terms of competition and variety of options, but it can be a big problem to integrate several different systems in the same infrastructure. In this work we develop a simple and scalable architecture that allows us to integrate the information from these systems into a single system centered on an MQTT broker for real-time communication, a database to persist the information that flows through the system, an API to consume that information, a web portal to make information (both API and broker) available to users and a simulator that allows us to create an environment to test this architecture without the need for investment in hardware. And after finishing this work, we concluded that this integration is possible without the need to create interdependencies between the original systems. We achieved this thanks to the fact that the database is isolated behind the API and the latter does not need to interact with the MQTT broker if we use an adapter. In addition, MQTT is an open and lightweight protocol that can be implemented in any other system, or failing that, create new adapters that convert messages from those systems into MQTT messages.
The world of home automation has gone through a major growth phase in recent years and this has led to the emergence of many different manufacturers and systems, and even different communication protocols. This is very good in terms of competition and variety of options, but it can be a big problem to integrate several different systems in the same infrastructure. In this work we develop a simple and scalable architecture that allows us to integrate the information from these systems into a single system centered on an MQTT broker for real-time communication, a database to persist the information that flows through the system, an API to consume that information, a web portal to make information (both API and broker) available to users and a simulator that allows us to create an environment to test this architecture without the need for investment in hardware. And after finishing this work, we concluded that this integration is possible without the need to create interdependencies between the original systems. We achieved this thanks to the fact that the database is isolated behind the API and the latter does not need to interact with the MQTT broker if we use an adapter. In addition, MQTT is an open and lightweight protocol that can be implemented in any other system, or failing that, create new adapters that convert messages from those systems into MQTT messages.
Direction
Taboada González, José Ángel (Tutorships)
Martínez Casas, David (Co-tutorships)
Taboada González, José Ángel (Tutorships)
Martínez Casas, David (Co-tutorships)
Court
PENA BRAGE, FRANCISCO JOSE (Chairman)
MERA PEREZ, DAVID (Secretary)
CARIÑENA AMIGO, MARIA PURIFICACION (Member)
PENA BRAGE, FRANCISCO JOSE (Chairman)
MERA PEREZ, DAVID (Secretary)
CARIÑENA AMIGO, MARIA PURIFICACION (Member)
Configuration, software and infrastructure deployment through desired state configuration
Authorship
M.P.L.
Bachelor's Degree in Computer Engineering
M.P.L.
Bachelor's Degree in Computer Engineering
Defense date
02.16.2024 11:00
02.16.2024 11:00
Summary
This project consists of the analysis, design and implementation of a software solution that will help a customer to begin the migration of their infrastructure and corresponding configuration to IaC (Infrastructure as Code) paradigm. In the long term it also aims to encourage DevOps philosophy. Its adoption will involve both sysadmins and developers, bringing great benefits to the whole team by reducing the time required to configure and document the infrastructure's current state. Switching to this paradigm also results in a higher quality work and improves response times in multiple scenarios. Attempts to compromise the platform, security audits, deployment and hardening of new servers or publishing new application versions are just a few examples that will be approached more effectively and using less effort. Taking into consideration customer's constraints and existing infrastructure's characteristics, the best alternative will be chosen, the needed modules will be implemented to enable integration with current infrastructure and testing scenarios will be defined in order to validate that the solution works as expected.
This project consists of the analysis, design and implementation of a software solution that will help a customer to begin the migration of their infrastructure and corresponding configuration to IaC (Infrastructure as Code) paradigm. In the long term it also aims to encourage DevOps philosophy. Its adoption will involve both sysadmins and developers, bringing great benefits to the whole team by reducing the time required to configure and document the infrastructure's current state. Switching to this paradigm also results in a higher quality work and improves response times in multiple scenarios. Attempts to compromise the platform, security audits, deployment and hardening of new servers or publishing new application versions are just a few examples that will be approached more effectively and using less effort. Taking into consideration customer's constraints and existing infrastructure's characteristics, the best alternative will be chosen, the needed modules will be implemented to enable integration with current infrastructure and testing scenarios will be defined in order to validate that the solution works as expected.
Direction
CABALEIRO DOMINGUEZ, JOSE CARLOS (Tutorships)
Pena Rodríguez, Santiago (Co-tutorships)
CABALEIRO DOMINGUEZ, JOSE CARLOS (Tutorships)
Pena Rodríguez, Santiago (Co-tutorships)
Court
CABALEIRO DOMINGUEZ, JOSE CARLOS (Student’s tutor)
Pena Rodríguez, Santiago (Student’s tutor)
CABALEIRO DOMINGUEZ, JOSE CARLOS (Student’s tutor)
Pena Rodríguez, Santiago (Student’s tutor)
Implementation of data access structures on textual, spatial and temporal dimensions
Authorship
J.A.A.
Bachelor's Degree in Computer Engineering
J.A.A.
Bachelor's Degree in Computer Engineering
Defense date
09.13.2024 11:00
09.13.2024 11:00
Summary
This project consists of implementing an approach to the STILT model (Spatio-temporal Textual InterLeaved Trie), a data structure based on a binary Trie, with the aim of obtaining a performance improvement in data search compared to current technologies for data management. A description of those current technologies is made, explaining their differences with respect to the implemented model, also a description of the developed implementation and, in order to verify the efficiency of the searches by using the implemented data structure, a comparative analysis of the search times is carried out against a representative example of those current technologies, PostgreSQL.
This project consists of implementing an approach to the STILT model (Spatio-temporal Textual InterLeaved Trie), a data structure based on a binary Trie, with the aim of obtaining a performance improvement in data search compared to current technologies for data management. A description of those current technologies is made, explaining their differences with respect to the implemented model, also a description of the developed implementation and, in order to verify the efficiency of the searches by using the implemented data structure, a comparative analysis of the search times is carried out against a representative example of those current technologies, PostgreSQL.
Direction
RIOS VIQUEIRA, JOSE RAMON (Tutorships)
REY PRESAS, ANDREA (Co-tutorships)
RIOS VIQUEIRA, JOSE RAMON (Tutorships)
REY PRESAS, ANDREA (Co-tutorships)
Court
REY PRESAS, ANDREA (Student’s tutor)
RIOS VIQUEIRA, JOSE RAMON (Student’s tutor)
REY PRESAS, ANDREA (Student’s tutor)
RIOS VIQUEIRA, JOSE RAMON (Student’s tutor)
Analysis of Kubernetes Autoscaling Methods
Authorship
J.D.L.
Bachelor's Degree in Computer Engineering
J.D.L.
Bachelor's Degree in Computer Engineering
Defense date
09.12.2024 12:00
09.12.2024 12:00
Summary
Cloud Native technologies have been gaining traction within the business world, thanks largely to the advantages they offer over their direct competition, traditional cloud computing. These, being specifically designed to operate efficiently and scalably in cloud environments, allow you to make the most of the resources available within your environment. Even so, this technological model also brings with it problems that must be faced. One of them is their ability to adapt to work demands that are constantly changing, and which can lead to application availability problems if the system is not capable of supporting said load. This work aims to study the various automatic scaling tools available to Kubernetes, giving special emphasis to the Horizontal Pod Autoscaler (HPA) and traditional metrics, in order to verify their performance against various practical cases.
Cloud Native technologies have been gaining traction within the business world, thanks largely to the advantages they offer over their direct competition, traditional cloud computing. These, being specifically designed to operate efficiently and scalably in cloud environments, allow you to make the most of the resources available within your environment. Even so, this technological model also brings with it problems that must be faced. One of them is their ability to adapt to work demands that are constantly changing, and which can lead to application availability problems if the system is not capable of supporting said load. This work aims to study the various automatic scaling tools available to Kubernetes, giving special emphasis to the Horizontal Pod Autoscaler (HPA) and traditional metrics, in order to verify their performance against various practical cases.
Direction
PICHEL CAMPOS, JUAN CARLOS (Tutorships)
Maseda Muiño, Francisco (Co-tutorships)
PICHEL CAMPOS, JUAN CARLOS (Tutorships)
Maseda Muiño, Francisco (Co-tutorships)
Court
ARIAS RODRIGUEZ, JUAN ENRIQUE (Chairman)
VILA BLANCO, NICOLAS (Secretary)
GONZALEZ DIAZ, JULIO (Member)
ARIAS RODRIGUEZ, JUAN ENRIQUE (Chairman)
VILA BLANCO, NICOLAS (Secretary)
GONZALEZ DIAZ, JULIO (Member)
Mobile application for the implementation of actions in different areas
Authorship
J.A.I.
Bachelor's Degree in Computer Engineering
J.A.I.
Bachelor's Degree in Computer Engineering
Defense date
09.13.2024 10:00
09.13.2024 10:00
Summary
This project develops an application for Android mobile devices that allows actions to be carried out in any field (health, transport, industry...), facilitating the process for users and replacing other solutions such as spreadsheets or paper. It also has a server application that exposes a REST API with which the mobile application communicates in order to carry out the actions.
This project develops an application for Android mobile devices that allows actions to be carried out in any field (health, transport, industry...), facilitating the process for users and replacing other solutions such as spreadsheets or paper. It also has a server application that exposes a REST API with which the mobile application communicates in order to carry out the actions.
Direction
VIDAL AGUIAR, JUAN CARLOS (Tutorships)
Yañez Candales, Eva (Co-tutorships)
VIDAL AGUIAR, JUAN CARLOS (Tutorships)
Yañez Candales, Eva (Co-tutorships)
Court
VIDAL AGUIAR, JUAN CARLOS (Student’s tutor)
Yañez Candales, Eva (Student’s tutor)
VIDAL AGUIAR, JUAN CARLOS (Student’s tutor)
Yañez Candales, Eva (Student’s tutor)
Platform for the Integration of Similarity Rankings of Image Retrieval Algorithms
Authorship
C.M.M.B.
Bachelor's Degree in Computer Engineering
C.M.M.B.
Bachelor's Degree in Computer Engineering
Defense date
02.16.2024 10:00
02.16.2024 10:00
Summary
This project is focused on the development of an app that allows the user to obtain a ranking of images based on their similarity to a given question image, according to their shape, color and/or texture descriptors. It will also allow the user to choose the weight that each one of these types of descriptors will have at the time of generating the ranking. The extraction methods for these descriptors have been rewritten from three previous works, one for each type of them.
This project is focused on the development of an app that allows the user to obtain a ranking of images based on their similarity to a given question image, according to their shape, color and/or texture descriptors. It will also allow the user to choose the weight that each one of these types of descriptors will have at the time of generating the ranking. The extraction methods for these descriptors have been rewritten from three previous works, one for each type of them.
Direction
MOSQUERA GONZALEZ, ANTONIO (Tutorships)
MOSQUERA GONZALEZ, ANTONIO (Tutorships)
Court
MOSQUERA GONZALEZ, ANTONIO (Student’s tutor)
MOSQUERA GONZALEZ, ANTONIO (Student’s tutor)
CyberDefence and AI: Practical Cases
Authorship
P.S.S.
Computer Engineering Degree
P.S.S.
Computer Engineering Degree
Defense date
02.15.2024 11:30
02.15.2024 11:30
Summary
Due to the inexorable advance of the Internet, efforts to make it a safe place for everyone are increasingly necessary. One of these efforts is the abuse email boxes of companies dedicated to provide Internet services (ISPs). These abuse mailboxes aim to offer an effective way to communicate with the company about misuse perpetrated from their network. The high influx of reports received in these mailboxes forces such companies to seek efficient ways to manage them. It is from this need that this work arises, as its main objective is the development of a system for the efficient and proper management of the abuse mailbox of the company MásMóvil. To achieve this, on the one hand, we must automate the classification of received emails, as well as their response if necessary, and on the other hand, we must streamline the analysis of the reports themselves, especially the analysis of reported IP addresses. Both parts were successfully developed, as the designed method significantly optimise the task, saving the company significant effort and resources. The main tool for the work has been Microsoft Copilot, an artificial intelligence assistant recently released to the market, whose contributions and performance were to be tested. Additionally, for the completion of the process, tools such as Power Automate and PrivateGPT, an open-source project that integrates the advantages of chatbots with a locally private environment, were used. The Microsoft assistant ultimately played a less relevant role than expected, being surpassed in certain important aspects by PrivateGPT when it came to analyzing IPs. Power Automate, an application from the Microsoft 365 suite, took the lead role, as it handled practically all of the automation tasks.
Due to the inexorable advance of the Internet, efforts to make it a safe place for everyone are increasingly necessary. One of these efforts is the abuse email boxes of companies dedicated to provide Internet services (ISPs). These abuse mailboxes aim to offer an effective way to communicate with the company about misuse perpetrated from their network. The high influx of reports received in these mailboxes forces such companies to seek efficient ways to manage them. It is from this need that this work arises, as its main objective is the development of a system for the efficient and proper management of the abuse mailbox of the company MásMóvil. To achieve this, on the one hand, we must automate the classification of received emails, as well as their response if necessary, and on the other hand, we must streamline the analysis of the reports themselves, especially the analysis of reported IP addresses. Both parts were successfully developed, as the designed method significantly optimise the task, saving the company significant effort and resources. The main tool for the work has been Microsoft Copilot, an artificial intelligence assistant recently released to the market, whose contributions and performance were to be tested. Additionally, for the completion of the process, tools such as Power Automate and PrivateGPT, an open-source project that integrates the advantages of chatbots with a locally private environment, were used. The Microsoft assistant ultimately played a less relevant role than expected, being surpassed in certain important aspects by PrivateGPT when it came to analyzing IPs. Power Automate, an application from the Microsoft 365 suite, took the lead role, as it handled practically all of the automation tasks.
Direction
CARIÑENA AMIGO, MARIA PURIFICACION (Tutorships)
Saucedo Calderón, Artemio (Co-tutorships)
CARIÑENA AMIGO, MARIA PURIFICACION (Tutorships)
Saucedo Calderón, Artemio (Co-tutorships)
Court
PICHEL CAMPOS, JUAN CARLOS (Chairman)
CORES COSTA, DANIEL (Secretary)
Rodeiro Pazos, David (Member)
PICHEL CAMPOS, JUAN CARLOS (Chairman)
CORES COSTA, DANIEL (Secretary)
Rodeiro Pazos, David (Member)
Parallelization of point cloud processing algorithms in distributed memory
Authorship
T.G.B.
Computer Engineering Degree
T.G.B.
Computer Engineering Degree
Defense date
07.18.2024 17:00
07.18.2024 17:00
Summary
LiDAR technology allows us to obtain three-dimensional scans of a surface or object at a very high resolution, but its processing, especially the calculation of the neighbors of a point, is very expensive both in terms of time and memory the more points a cloud has. One way to improve the performance of point cloud processing algorithms is to parallelize them, thus distributing the workload between different processing units. The objective of this work is to implement a solution to this problem, developing a point cloud partitioning algorithm and subsequently processing the clouds in parallel using MPI in distributed memory. In the strategy presented in this work, first the point clouds are equally partitioned between the different MPI processes. Then, each process reads its corresponding part of the point cloud and creates an octree with its points. Finally, the processes calculate the points' neighborhood and their geometric descriptors and save the results in a plain text file. An analysis of the performance and efficiency of the new implementation will also be carried out based on the number of processes, using different point clouds and comparing the results with those obtained sequentially.
LiDAR technology allows us to obtain three-dimensional scans of a surface or object at a very high resolution, but its processing, especially the calculation of the neighbors of a point, is very expensive both in terms of time and memory the more points a cloud has. One way to improve the performance of point cloud processing algorithms is to parallelize them, thus distributing the workload between different processing units. The objective of this work is to implement a solution to this problem, developing a point cloud partitioning algorithm and subsequently processing the clouds in parallel using MPI in distributed memory. In the strategy presented in this work, first the point clouds are equally partitioned between the different MPI processes. Then, each process reads its corresponding part of the point cloud and creates an octree with its points. Finally, the processes calculate the points' neighborhood and their geometric descriptors and save the results in a plain text file. An analysis of the performance and efficiency of the new implementation will also be carried out based on the number of processes, using different point clouds and comparing the results with those obtained sequentially.
Direction
CABALEIRO DOMINGUEZ, JOSE CARLOS (Tutorships)
Yermo García, Miguel (Co-tutorships)
CABALEIRO DOMINGUEZ, JOSE CARLOS (Tutorships)
Yermo García, Miguel (Co-tutorships)
Court
Pardo López, Xosé Manuel (Chairman)
Puentes Cociña, Beltrán (Secretary)
LOPEZ FANDIÑO, JAVIER (Member)
Pardo López, Xosé Manuel (Chairman)
Puentes Cociña, Beltrán (Secretary)
LOPEZ FANDIÑO, JAVIER (Member)
Let's Be Healthy: Mobile training app
Authorship
A.V.M.
Computer Engineering Degree
A.V.M.
Computer Engineering Degree
Defense date
09.12.2024 11:00
09.12.2024 11:00
Summary
The objective of this Final Degree Project is to develop a mobile application for the Android operating system whose purpose is to guide and advise different user profiles in the world of training. The application will provide different routines, both for the gym and for home. The user will be able to keep track of the routines they have completed, their weight, calories burned, etc.
The objective of this Final Degree Project is to develop a mobile application for the Android operating system whose purpose is to guide and advise different user profiles in the world of training. The application will provide different routines, both for the gym and for home. The user will be able to keep track of the routines they have completed, their weight, calories burned, etc.
Direction
VARELA PET, JOSE (Tutorships)
VARELA PET, JOSE (Tutorships)
Court
Fernández Rivera, Francisco (Chairman)
SANTOS MATEOS, ROI (Secretary)
López Vilariño, David (Member)
Fernández Rivera, Francisco (Chairman)
SANTOS MATEOS, ROI (Secretary)
López Vilariño, David (Member)
Emobook. Evolution of a Life Stories Application for Music Therapy Sessions
Authorship
A.V.R.
Computer Engineering Degree
A.V.R.
Computer Engineering Degree
Defense date
07.18.2024 12:30
07.18.2024 12:30
Summary
Life Story Books are widely used in reminiscence therapy for treating older adults with dementia, as they help stimulate memory, improve communication, and strengthen personal identity. Emobook, an Android application designed for tablets, assists caregivers in preparing these stories and allows the addition of elements such as videos or audio recordings. It also facilitates the capture of emotional responses associated with the memories, enabling the tracking of therapy sessions. During this project, a significant evolution of this application was undertaken, introducing new functionalities and improvements to expand its capabilities and enhance the user experience. Among the main improvements was the transition from local storage to cloud storage, making it easier to access and edit memories from any tablet with an internet connection, ensuring the necessary accessibility and flexibility. Additionally, the application was enhanced with the ability to add audio from Spotify, allowing caregivers to easily include meaningful music in the life stories. Features were also introduced to allow the transfer of memories between groups and to log attendance and disconnections, as well as the ability to modify configuration options and edit stories during therapy sessions directly from the viewing interface. These enhancements have optimized the Emobook application, facilitating caregivers' personalization and effective management of life stories, thereby improving the care and quality of life for older adults with dementia.
Life Story Books are widely used in reminiscence therapy for treating older adults with dementia, as they help stimulate memory, improve communication, and strengthen personal identity. Emobook, an Android application designed for tablets, assists caregivers in preparing these stories and allows the addition of elements such as videos or audio recordings. It also facilitates the capture of emotional responses associated with the memories, enabling the tracking of therapy sessions. During this project, a significant evolution of this application was undertaken, introducing new functionalities and improvements to expand its capabilities and enhance the user experience. Among the main improvements was the transition from local storage to cloud storage, making it easier to access and edit memories from any tablet with an internet connection, ensuring the necessary accessibility and flexibility. Additionally, the application was enhanced with the ability to add audio from Spotify, allowing caregivers to easily include meaningful music in the life stories. Features were also introduced to allow the transfer of memories between groups and to log attendance and disconnections, as well as the ability to modify configuration options and edit stories during therapy sessions directly from the viewing interface. These enhancements have optimized the Emobook application, facilitating caregivers' personalization and effective management of life stories, thereby improving the care and quality of life for older adults with dementia.
Direction
CATALA BOLOS, ALEJANDRO (Tutorships)
CONDORI FERNANDEZ, NELLY (Co-tutorships)
CATALA BOLOS, ALEJANDRO (Tutorships)
CONDORI FERNANDEZ, NELLY (Co-tutorships)
Court
Cotos Yáñez, José Manuel (Chairman)
QUESADA BARRIUSO, PABLO (Secretary)
GAGO COUSO, FELIPE (Member)
Cotos Yáñez, José Manuel (Chairman)
QUESADA BARRIUSO, PABLO (Secretary)
GAGO COUSO, FELIPE (Member)
Evolution of the business intelligence application for sports management of the Obradoiro CAB basketball club
Authorship
S.L.Q.
Computer Engineering Degree
S.L.Q.
Computer Engineering Degree
Defense date
07.18.2024 17:00
07.18.2024 17:00
Summary
This work is about the improvement and evolution of the application used by the Monbus Obradoiro basketball team to make sports desicions. This club, participating in the Liga Endesa ACB in the 2023/2024 season, makes use of this resource to study the performance of the team, its players and its rivals. The development has been focused on the improvement of the quality of the application's data, given that its reliability at the beginning of the project was quite doubtful; and on the beggining of a new importation module, so that a new modern version of the application can be created to migrate before this one becomes obsolete. To conclude, the impact of the work carried out and the possible directions that can be taken in future extensions will be reviewed.
This work is about the improvement and evolution of the application used by the Monbus Obradoiro basketball team to make sports desicions. This club, participating in the Liga Endesa ACB in the 2023/2024 season, makes use of this resource to study the performance of the team, its players and its rivals. The development has been focused on the improvement of the quality of the application's data, given that its reliability at the beginning of the project was quite doubtful; and on the beggining of a new importation module, so that a new modern version of the application can be created to migrate before this one becomes obsolete. To conclude, the impact of the work carried out and the possible directions that can be taken in future extensions will be reviewed.
Direction
Losada Carril, David Enrique (Tutorships)
Sanz Anchelergues, Adolfo (Co-tutorships)
Losada Carril, David Enrique (Tutorships)
Sanz Anchelergues, Adolfo (Co-tutorships)
Court
VAZQUEZ CENDON, MARIA ELENA (Chairman)
CHAVES FRAGA, DAVID (Secretary)
SUAREZ GAREA, JORGE ALBERTO (Member)
VAZQUEZ CENDON, MARIA ELENA (Chairman)
CHAVES FRAGA, DAVID (Secretary)
SUAREZ GAREA, JORGE ALBERTO (Member)
New variational architecture for anomaly detection in process mining
Authorship
A.C.L.
Computer Engineering Degree
A.C.L.
Computer Engineering Degree
Defense date
07.19.2024 11:00
07.19.2024 11:00
Summary
Process mining is a discipline focused on the analysis of the execution of business processes. Its objective is to obtain a detailed, accurate and actionable vision of these processes with the aim of improving their efficiency, conformity, performance and predictive capacity. It is, therefore, a critical tool today in many industrial and financial fields, but also in other areas less related to this technology such as health and education. However, the data extracted from the execution of such processes are not usually perfect and generally include problems derived from data capture, either by humans or by automated systems. This Final Degree Project (TFG) will focus on one of these problems, keys for mining techniques to maintain a high degree of effectiveness, integrity and efficiency: the detection of anomalies. Recently, due to the limitations of the classical methods of anomaly detection, deep-learning techniques have been employed. The latter offer significantly higher capabilities, allowing more accurate and efficient detection of anomalies compared to traditional approaches. These techniques include autoencoder-based architectures, a type of neural network that has proven to achieve very competent results in the identification of anomalies, according to the existing literature on the problem. These proposals are based on taking advantage of the great capacity of the autoencoder for the reconstruction of the input data: as normal data are more frequent, the trained model tends to reconstruct them with a lower error than the anomalous data, which allows us to detect the latter. This work proposes a new architecture based on one of the variants of the autoencoders: the CVAE (Conditional Variational Autoencoder). This proposal seeks to achieve a higher performance in the detection of anomalies compared to the approximations of the state of the art. In addition to the design process of the architecture, efforts have also focused on carrying out an appropriate preparation of the data and defining an appropriate strategy for the identification of anomalies. This proposal has been validated using 24 records of events associated with 6 synthetic processes of different complexity, achieving results higher than the state of art in deep learning approaches in the field.
Process mining is a discipline focused on the analysis of the execution of business processes. Its objective is to obtain a detailed, accurate and actionable vision of these processes with the aim of improving their efficiency, conformity, performance and predictive capacity. It is, therefore, a critical tool today in many industrial and financial fields, but also in other areas less related to this technology such as health and education. However, the data extracted from the execution of such processes are not usually perfect and generally include problems derived from data capture, either by humans or by automated systems. This Final Degree Project (TFG) will focus on one of these problems, keys for mining techniques to maintain a high degree of effectiveness, integrity and efficiency: the detection of anomalies. Recently, due to the limitations of the classical methods of anomaly detection, deep-learning techniques have been employed. The latter offer significantly higher capabilities, allowing more accurate and efficient detection of anomalies compared to traditional approaches. These techniques include autoencoder-based architectures, a type of neural network that has proven to achieve very competent results in the identification of anomalies, according to the existing literature on the problem. These proposals are based on taking advantage of the great capacity of the autoencoder for the reconstruction of the input data: as normal data are more frequent, the trained model tends to reconstruct them with a lower error than the anomalous data, which allows us to detect the latter. This work proposes a new architecture based on one of the variants of the autoencoders: the CVAE (Conditional Variational Autoencoder). This proposal seeks to achieve a higher performance in the detection of anomalies compared to the approximations of the state of the art. In addition to the design process of the architecture, efforts have also focused on carrying out an appropriate preparation of the data and defining an appropriate strategy for the identification of anomalies. This proposal has been validated using 24 records of events associated with 6 synthetic processes of different complexity, achieving results higher than the state of art in deep learning approaches in the field.
Direction
VIDAL AGUIAR, JUAN CARLOS (Tutorships)
LAMA PENIN, MANUEL (Co-tutorships)
GAMALLO FERNANDEZ, PEDRO (Co-tutorships)
VIDAL AGUIAR, JUAN CARLOS (Tutorships)
LAMA PENIN, MANUEL (Co-tutorships)
GAMALLO FERNANDEZ, PEDRO (Co-tutorships)
Court
PENA BRAGE, FRANCISCO JOSE (Chairman)
MERA PEREZ, DAVID (Secretary)
CARIÑENA AMIGO, MARIA PURIFICACION (Member)
PENA BRAGE, FRANCISCO JOSE (Chairman)
MERA PEREZ, DAVID (Secretary)
CARIÑENA AMIGO, MARIA PURIFICACION (Member)
Estimation of the Breathing Signal from ECG Using Machine Learning.
Authorship
P.M.G.
Computer Engineering Degree
P.M.G.
Computer Engineering Degree
Defense date
02.15.2024 12:30
02.15.2024 12:30
Summary
This Bachelor's Thesis addresses the challenge of predicting the respiratory signal from electrocardiogram (ECG) signals using advanced machine learning techniques, specifically Recurrent Neural Networks (RNN) and Neural Ordinary Differential Equations (ODE-RNN). Given the irregular nature of biomedical time series, a methodology is proposed that not only focuses on the ECG signal directly but also incorporates a derived signal known as the warping signal, which captures significant temporal variations between heartbeats. This approach aims to improve the accuracy of predictions by providing the model with a richer representation of the underlying information in biomedical signals. Two differently trained models are developed and evaluated to see the influence of the warping signal on training for the prediction of the respiratory signal. Through this study, the ability of ODE-RNNs to effectively handle irregular temporal data, a critical aspect in the biomedical field, is explored. This work not only contributes to the field of time series processing and artificial intelligence but also lays the groundwork for future research in the prediction and analysis of complex biomedical signals, opening new possibilities for health monitoring and medical diagnosis.
This Bachelor's Thesis addresses the challenge of predicting the respiratory signal from electrocardiogram (ECG) signals using advanced machine learning techniques, specifically Recurrent Neural Networks (RNN) and Neural Ordinary Differential Equations (ODE-RNN). Given the irregular nature of biomedical time series, a methodology is proposed that not only focuses on the ECG signal directly but also incorporates a derived signal known as the warping signal, which captures significant temporal variations between heartbeats. This approach aims to improve the accuracy of predictions by providing the model with a richer representation of the underlying information in biomedical signals. Two differently trained models are developed and evaluated to see the influence of the warping signal on training for the prediction of the respiratory signal. Through this study, the ability of ODE-RNNs to effectively handle irregular temporal data, a critical aspect in the biomedical field, is explored. This work not only contributes to the field of time series processing and artificial intelligence but also lays the groundwork for future research in the prediction and analysis of complex biomedical signals, opening new possibilities for health monitoring and medical diagnosis.
Direction
FELIX LAMAS, PAULO (Tutorships)
PEREZ HERRERO, ADRIAN (Co-tutorships)
FELIX LAMAS, PAULO (Tutorships)
PEREZ HERRERO, ADRIAN (Co-tutorships)
Court
PICHEL CAMPOS, JUAN CARLOS (Chairman)
CORES COSTA, DANIEL (Secretary)
Rodeiro Pazos, David (Member)
PICHEL CAMPOS, JUAN CARLOS (Chairman)
CORES COSTA, DANIEL (Secretary)
Rodeiro Pazos, David (Member)
Search and classification technologies for the automatic analysis of gambling disorder texts
Authorship
P.F.F.
Computer Engineering Degree
P.F.F.
Computer Engineering Degree
Defense date
07.19.2024 11:30
07.19.2024 11:30
Summary
Data mining for psychological disorders has been a prolific area of study in recent times. With the exponential growth of social media activity, the number of messages posted by users can provide valuable insights into their mental health status. On the other hand, pathological gambling is being targeted by many Western governments due to its rise with the advent of new technologies, especially among adolescents. The main objective of this work is to explore and analyse messages from an online gambling forum in order to detect signs, behavioural patterns or symptoms of pathological gambling. For this purpose, different data mining techniques and search models will be tested in order to extract statistical data and phrases reflecting symptoms of gambling disorder. Finally, the results obtained will be discussed, evaluating the effectiveness of the techniques applied and the data they reflect.
Data mining for psychological disorders has been a prolific area of study in recent times. With the exponential growth of social media activity, the number of messages posted by users can provide valuable insights into their mental health status. On the other hand, pathological gambling is being targeted by many Western governments due to its rise with the advent of new technologies, especially among adolescents. The main objective of this work is to explore and analyse messages from an online gambling forum in order to detect signs, behavioural patterns or symptoms of pathological gambling. For this purpose, different data mining techniques and search models will be tested in order to extract statistical data and phrases reflecting symptoms of gambling disorder. Finally, the results obtained will be discussed, evaluating the effectiveness of the techniques applied and the data they reflect.
Direction
Losada Carril, David Enrique (Tutorships)
FERNANDEZ PICHEL, MARCOS (Co-tutorships)
Losada Carril, David Enrique (Tutorships)
FERNANDEZ PICHEL, MARCOS (Co-tutorships)
Court
BARJA PEREZ, JAVIER (Chairman)
ORDOÑEZ IGLESIAS, ALVARO (Secretary)
MOSQUERA GONZALEZ, ANTONIO (Member)
BARJA PEREZ, JAVIER (Chairman)
ORDOÑEZ IGLESIAS, ALVARO (Secretary)
MOSQUERA GONZALEZ, ANTONIO (Member)
Methyl acetate production plant
Authorship
S.G.G.
Bachelor's Degree in Chemical Engeneering
S.G.G.
Bachelor's Degree in Chemical Engeneering
Defense date
09.13.2024 11:40
09.13.2024 11:40
Summary
The objetive of this proyect is the design of a methyl acetate production plant using methanol and acetic acid as raw materials. Furthermore, the idea of implementing the ionic liquid in the chemical industry as a new way to separate azeotropic mixtures is developed in order to open a new vision on separation methods.
The objetive of this proyect is the design of a methyl acetate production plant using methanol and acetic acid as raw materials. Furthermore, the idea of implementing the ionic liquid in the chemical industry as a new way to separate azeotropic mixtures is developed in order to open a new vision on separation methods.
Direction
SOTO CAMPOS, ANA MARIA (Tutorships)
SOTO CAMPOS, ANA MARIA (Tutorships)
Court
GIL GONZALEZ, ALVARO (Chairman)
FRANCO URIA, MARIA AMAYA (Secretary)
GOMEZ DIAZ, DIEGO (Member)
GIL GONZALEZ, ALVARO (Chairman)
FRANCO URIA, MARIA AMAYA (Secretary)
GOMEZ DIAZ, DIEGO (Member)
Methyl mercaptan production plant from hydrogen sulfide and methanol.
Authorship
S.C.P.
Bachelor's Degree in Chemical Engeneering
S.C.P.
Bachelor's Degree in Chemical Engeneering
Defense date
07.16.2024 13:05
07.16.2024 13:05
Summary
This final degree project is based on the design of a methyl mercaptan production plant from hydrogen sulfide and methanol, with the aim of obtaining a capacity of 20.000 t/year of the product at 99.5% purity. The document addresses in detail various aspects of a chemical plant, including its mode of operation and production capacity. Different alternatives are analyzed to select the most viable, both technically and economically. The site chosen for the plant is also specified and complete process material and energy balances are provided. In addition, the process and equipment used are described, along with the instrumentation and control system. A safety study is included, the mechanical design of the main reactor (R-201), and various plans, which are: site plan, location plan, layout plan, process flow diagram (PFD), and piping and instrumentation diagram. Finally, the budget and economic viability of the facility are evaluated, the equipment is sized and environmental impact studies are carried out, as well as health and safety, and detailed specification is attached.
This final degree project is based on the design of a methyl mercaptan production plant from hydrogen sulfide and methanol, with the aim of obtaining a capacity of 20.000 t/year of the product at 99.5% purity. The document addresses in detail various aspects of a chemical plant, including its mode of operation and production capacity. Different alternatives are analyzed to select the most viable, both technically and economically. The site chosen for the plant is also specified and complete process material and energy balances are provided. In addition, the process and equipment used are described, along with the instrumentation and control system. A safety study is included, the mechanical design of the main reactor (R-201), and various plans, which are: site plan, location plan, layout plan, process flow diagram (PFD), and piping and instrumentation diagram. Finally, the budget and economic viability of the facility are evaluated, the equipment is sized and environmental impact studies are carried out, as well as health and safety, and detailed specification is attached.
Direction
GONZALEZ ALVAREZ, JULIA (Tutorships)
FREIRE LEIRA, MARIA SONIA (Co-tutorships)
GONZALEZ ALVAREZ, JULIA (Tutorships)
FREIRE LEIRA, MARIA SONIA (Co-tutorships)
Court
SOTO CAMPOS, ANA MARIA (Chairman)
Rodríguez Figueiras, Óscar (Secretary)
RODRIGUEZ MARTINEZ, HECTOR (Member)
SOTO CAMPOS, ANA MARIA (Chairman)
Rodríguez Figueiras, Óscar (Secretary)
RODRIGUEZ MARTINEZ, HECTOR (Member)
Protein production plant for pharmaceutical purposes
Authorship
P.A.F.
Bachelor's Degree in Chemical Engeneering
P.A.F.
Bachelor's Degree in Chemical Engeneering
Defense date
09.13.2024 13:00
09.13.2024 13:00
Summary
Design of a production plant for a therapeutic protein, specifically a monoclonal antibody, by cultivating mammalian cells. The capacity is one batch of product every 47 days, each batch containing 205.94 kg of protein. The process consists of a first stage of cell culture, followed by purification (separation and concentration of product) and finally filling. The detailed design of ultrafiltration-diafiltration equipment is carried out for product concentration and final buffer addition.
Design of a production plant for a therapeutic protein, specifically a monoclonal antibody, by cultivating mammalian cells. The capacity is one batch of product every 47 days, each batch containing 205.94 kg of protein. The process consists of a first stage of cell culture, followed by purification (separation and concentration of product) and finally filling. The detailed design of ultrafiltration-diafiltration equipment is carried out for product concentration and final buffer addition.
Direction
VIDAL TATO, MARIA ISABEL (Tutorships)
Gato Tasende, Lucía (Co-tutorships)
VIDAL TATO, MARIA ISABEL (Tutorships)
Gato Tasende, Lucía (Co-tutorships)
Court
GARRIDO FERNANDEZ, JUAN MANUEL (Chairman)
FRANCO RUIZ, DANIEL JOSE (Secretary)
FREIRE LEIRA, MARIA SONIA (Member)
GARRIDO FERNANDEZ, JUAN MANUEL (Chairman)
FRANCO RUIZ, DANIEL JOSE (Secretary)
FREIRE LEIRA, MARIA SONIA (Member)
Hydrazine hydrate production plant
Authorship
X.L.V.
Bachelor's Degree in Chemical Engeneering
X.L.V.
Bachelor's Degree in Chemical Engeneering
Defense date
07.16.2024 11:10
07.16.2024 11:10
Summary
The objective of this project is to develop a plant for the production of hydrazine hydrate with a mass concentration of 64 percent from ammonia and sodium hypochlorite, both in solution, using the Raschig method. The design is approached with the purpose of producing 12000T/year of product operating continuously, 330 days a year, 24 hours a day. The rigorous design of the most relevant equipment of the process is described, being the hydrazine reactor and the distillation column to obtain the product. It has been decided to address the design of a hydrazine hydrate production plant because of its current research in the energy field for its use as chemical storage of hydrogen in fuel cells due to its high hydrogen content, ease of transport and storage, in addition to the fact that the only by-product of its decomposition would be nitrogen. This is a study of current importance, since it offers an alternative to fossil fuels. On the other hand, hydrazine hydrate is currently used in various industrial fields; due to its ability to eliminate oxygen from water, it is used in high-pressure steam generators and industrial boilers, but it also has other uses, such as foaming agent for plastic foams, polymerization initiators, pesticides, pharmaceutical products, photographic chemical products or dyes. Derivatives of this component also have a great place in today's industry, such as, for example, hydrazine anhydrous, which is synthesized by the Olin Raschig method and is notable for its use as a rocket propellant. The facility is located in the north of Spain, more specifically in the industrial park of Lloreda, belonging to the Asturian city of Gijón, with the aim of facilitating the commercialization of hydrazine hydrate by sea to northern Europe and other commercial areas of interest. This document represents the Final Degree Project by means of which the students Mara Meitín Fraga and Xeila Liñayo Vicente opt for the university degree in Chemical Engineering awarded by the Escola Técnica Superior de Enxeñaría of the Universidade de Santiago de Compostela.
The objective of this project is to develop a plant for the production of hydrazine hydrate with a mass concentration of 64 percent from ammonia and sodium hypochlorite, both in solution, using the Raschig method. The design is approached with the purpose of producing 12000T/year of product operating continuously, 330 days a year, 24 hours a day. The rigorous design of the most relevant equipment of the process is described, being the hydrazine reactor and the distillation column to obtain the product. It has been decided to address the design of a hydrazine hydrate production plant because of its current research in the energy field for its use as chemical storage of hydrogen in fuel cells due to its high hydrogen content, ease of transport and storage, in addition to the fact that the only by-product of its decomposition would be nitrogen. This is a study of current importance, since it offers an alternative to fossil fuels. On the other hand, hydrazine hydrate is currently used in various industrial fields; due to its ability to eliminate oxygen from water, it is used in high-pressure steam generators and industrial boilers, but it also has other uses, such as foaming agent for plastic foams, polymerization initiators, pesticides, pharmaceutical products, photographic chemical products or dyes. Derivatives of this component also have a great place in today's industry, such as, for example, hydrazine anhydrous, which is synthesized by the Olin Raschig method and is notable for its use as a rocket propellant. The facility is located in the north of Spain, more specifically in the industrial park of Lloreda, belonging to the Asturian city of Gijón, with the aim of facilitating the commercialization of hydrazine hydrate by sea to northern Europe and other commercial areas of interest. This document represents the Final Degree Project by means of which the students Mara Meitín Fraga and Xeila Liñayo Vicente opt for the university degree in Chemical Engineering awarded by the Escola Técnica Superior de Enxeñaría of the Universidade de Santiago de Compostela.
Direction
MOREIRA MARTINEZ, RAMON FELIPE (Tutorships)
GOMEZ DIAZ, DIEGO (Co-tutorships)
MOREIRA MARTINEZ, RAMON FELIPE (Tutorships)
GOMEZ DIAZ, DIEGO (Co-tutorships)
Court
MOSQUERA CORRAL, ANUSKA (Chairman)
SINEIRO TORRES, JORGE (Secretary)
VIDAL TATO, MARIA ISABEL (Member)
MOSQUERA CORRAL, ANUSKA (Chairman)
SINEIRO TORRES, JORGE (Secretary)
VIDAL TATO, MARIA ISABEL (Member)
2-methoxy-2-methyl-heptane production plant
Authorship
P.R.S.
Bachelor's Degree in Chemical Engeneering
P.R.S.
Bachelor's Degree in Chemical Engeneering
Defense date
07.16.2024 12:25
07.16.2024 12:25
Summary
2-Methoxy-2-methylheptane (MMH) is used in various fields, primarily as a gasoline additive or pharmaceutical intermediate. The plant's capacity is 30,000 tons per year since both traditional applications and new chemical-grade alternatives are not high-tonnage applications. MMH is produced from methanol and 2-methyl-1-heptene. As a sustainable alternative, it is proposed to use green methanol from a nearby plant. Pablo Rodríguez-Villamil is responsible for the design of a packed distillation column (T-203). Raúl Veiga Moo is in charge of the design of a catalytic heterogeneous fixed-bed reactor (R-101). Finally, Hugo Vidal Cabanelas is responsible for the design of a plate distillation column (T-202).
2-Methoxy-2-methylheptane (MMH) is used in various fields, primarily as a gasoline additive or pharmaceutical intermediate. The plant's capacity is 30,000 tons per year since both traditional applications and new chemical-grade alternatives are not high-tonnage applications. MMH is produced from methanol and 2-methyl-1-heptene. As a sustainable alternative, it is proposed to use green methanol from a nearby plant. Pablo Rodríguez-Villamil is responsible for the design of a packed distillation column (T-203). Raúl Veiga Moo is in charge of the design of a catalytic heterogeneous fixed-bed reactor (R-101). Finally, Hugo Vidal Cabanelas is responsible for the design of a plate distillation column (T-202).
Direction
VIDAL TATO, MARIA ISABEL (Tutorships)
VIDAL TATO, MARIA ISABEL (Tutorships)
Court
RODIL RODRIGUEZ, EVA (Chairman)
SOUTO GONZALEZ, JOSE ANTONIO (Secretary)
VAL DEL RIO, MARIA ANGELES (Member)
RODIL RODRIGUEZ, EVA (Chairman)
SOUTO GONZALEZ, JOSE ANTONIO (Secretary)
VAL DEL RIO, MARIA ANGELES (Member)
2-methoxy-2-methyl-heptane production plant
Authorship
H.V.C.
Bachelor's Degree in Chemical Engeneering
H.V.C.
Bachelor's Degree in Chemical Engeneering
Defense date
07.16.2024 12:25
07.16.2024 12:25
Summary
2-methoxy-2-methyl-heptane (MMH) is used in various fields, mostly as a gasoline additive or pharmaceutical intermediate. The plant capacity is 30,000 tons per year as both traditional applications and new chemical grade alternatives are not large tonnage applications. MMH is produced from methanol and 2-methyl-1-heptene. As a sustainable alternative, green methanol from a nearby plant is being considered. Pablo Rodríguez-Villamil Seoane is in charge of the design of a packed distillation column (T-203). Raul Veiga Moo is in charge of the design of a catalytic heterogeneous fixed bed reactor (R-101). Finally, Hugo Vidal Cabanelas is in charge of the design of a distillation plate column (T-202).
2-methoxy-2-methyl-heptane (MMH) is used in various fields, mostly as a gasoline additive or pharmaceutical intermediate. The plant capacity is 30,000 tons per year as both traditional applications and new chemical grade alternatives are not large tonnage applications. MMH is produced from methanol and 2-methyl-1-heptene. As a sustainable alternative, green methanol from a nearby plant is being considered. Pablo Rodríguez-Villamil Seoane is in charge of the design of a packed distillation column (T-203). Raul Veiga Moo is in charge of the design of a catalytic heterogeneous fixed bed reactor (R-101). Finally, Hugo Vidal Cabanelas is in charge of the design of a distillation plate column (T-202).
Direction
VIDAL TATO, MARIA ISABEL (Tutorships)
VIDAL TATO, MARIA ISABEL (Tutorships)
Court
RODIL RODRIGUEZ, EVA (Chairman)
SOUTO GONZALEZ, JOSE ANTONIO (Secretary)
VAL DEL RIO, MARIA ANGELES (Member)
RODIL RODRIGUEZ, EVA (Chairman)
SOUTO GONZALEZ, JOSE ANTONIO (Secretary)
VAL DEL RIO, MARIA ANGELES (Member)
Hydrogen carrier dehydrogenation plant (perhydro-dibenzyltoluene)
Authorship
F.A.S.B.
Bachelor's Degree in Chemical Engeneering
F.A.S.B.
Bachelor's Degree in Chemical Engeneering
Defense date
09.13.2024 12:20
09.13.2024 12:20
Summary
In this bachelor´s thesis, the design and development of a hydrogen carrier (perhydro-dibenzyltoluene) dehydrogenation plant are proposed and developed for the effective production and separation of 50,000 t/year of hydrogen. Additionally, a feasibility analysis of the plant is conducted, along with the design of the process reactor.
In this bachelor´s thesis, the design and development of a hydrogen carrier (perhydro-dibenzyltoluene) dehydrogenation plant are proposed and developed for the effective production and separation of 50,000 t/year of hydrogen. Additionally, a feasibility analysis of the plant is conducted, along with the design of the process reactor.
Direction
VIDAL TATO, MARIA ISABEL (Tutorships)
VIDAL TATO, MARIA ISABEL (Tutorships)
Court
GIL GONZALEZ, ALVARO (Chairman)
FRANCO URIA, MARIA AMAYA (Secretary)
GOMEZ DIAZ, DIEGO (Member)
GIL GONZALEZ, ALVARO (Chairman)
FRANCO URIA, MARIA AMAYA (Secretary)
GOMEZ DIAZ, DIEGO (Member)
Production plant for copper sulphate pentahydrate
Authorship
B.G.A.
Bachelor's Degree in Chemical Engeneering
B.G.A.
Bachelor's Degree in Chemical Engeneering
Defense date
07.16.2024 10:35
07.16.2024 10:35
Summary
This essay focuses on the design of a production plant for 378.79 kg/h of copper sulfate pentahydrate from copper scrap. The process has an irreversible chemical reaction, which will be carried out in a continuous stirred tank reactor with an aeration system. The product resulting from the reaction is copper sulfate, which will be transferred to a crystallizer to obtain the desired product. The crystals formed will be separated from the rest of the mother solution through a filtration process that will be carried out in tandem with a continuous washing process to remove the acidity of the product. This entire process culminates in a drying process through which the excess water will be removed to give rise to a copper sulfate pentahydrate product with 5% humidity.
This essay focuses on the design of a production plant for 378.79 kg/h of copper sulfate pentahydrate from copper scrap. The process has an irreversible chemical reaction, which will be carried out in a continuous stirred tank reactor with an aeration system. The product resulting from the reaction is copper sulfate, which will be transferred to a crystallizer to obtain the desired product. The crystals formed will be separated from the rest of the mother solution through a filtration process that will be carried out in tandem with a continuous washing process to remove the acidity of the product. This entire process culminates in a drying process through which the excess water will be removed to give rise to a copper sulfate pentahydrate product with 5% humidity.
Direction
FRANCO URIA, MARIA AMAYA (Tutorships)
SINEIRO TORRES, JORGE (Co-tutorships)
FRANCO URIA, MARIA AMAYA (Tutorships)
SINEIRO TORRES, JORGE (Co-tutorships)
Court
ROCA BORDELLO, ENRIQUE (Chairman)
SANCHEZ FERNANDEZ, ADRIAN (Secretary)
MOREIRA VILAR, MARIA TERESA (Member)
ROCA BORDELLO, ENRIQUE (Chairman)
SANCHEZ FERNANDEZ, ADRIAN (Secretary)
MOREIRA VILAR, MARIA TERESA (Member)
Evolution of an interactive questionnaires tool for children
Authorship
O.T.P.
Computer Engineering Degree
O.T.P.
Computer Engineering Degree
Defense date
07.18.2024 12:00
07.18.2024 12:00
Summary
This Final Degree Project focuses on the evolution of the interactive questionnaire tool for children “FeedbackScales4Children”, initially developed by Jaime Iglesias Navas and tutored by Julián Flores González and Alejandro Catalá Bolós. The project addresses several significant improvements, including a complete redesign of the user interface to optimize its usability and aesthetics by clearly differentiating between the roles of evaluators and evaluated, the incorporation of new functionalities such as music integration and greater flexibility in the types of questionnaires available, as well as the implementation of a cloud storage subsystem for centralized and efficient management of data regardless of the device. As a result, a more intuitive, functional and attractive application has been achieved, which better responds to the needs of end users and ensures a improved accessibility and data handling.
This Final Degree Project focuses on the evolution of the interactive questionnaire tool for children “FeedbackScales4Children”, initially developed by Jaime Iglesias Navas and tutored by Julián Flores González and Alejandro Catalá Bolós. The project addresses several significant improvements, including a complete redesign of the user interface to optimize its usability and aesthetics by clearly differentiating between the roles of evaluators and evaluated, the incorporation of new functionalities such as music integration and greater flexibility in the types of questionnaires available, as well as the implementation of a cloud storage subsystem for centralized and efficient management of data regardless of the device. As a result, a more intuitive, functional and attractive application has been achieved, which better responds to the needs of end users and ensures a improved accessibility and data handling.
Direction
CATALA BOLOS, ALEJANDRO (Tutorships)
CATALA BOLOS, ALEJANDRO (Tutorships)
Court
Cotos Yáñez, José Manuel (Chairman)
QUESADA BARRIUSO, PABLO (Secretary)
GAGO COUSO, FELIPE (Member)
Cotos Yáñez, José Manuel (Chairman)
QUESADA BARRIUSO, PABLO (Secretary)
GAGO COUSO, FELIPE (Member)
Spatial prediction of traffic flows in a smart city
Authorship
P.L.P.
Computer Engineering Degree
P.L.P.
Computer Engineering Degree
Defense date
07.18.2024 17:30
07.18.2024 17:30
Summary
The motivation for this work comes from the lack of knowledge, on the part of the city councils of the large Galician cities, of the state of traffic in most of their streets. They only have information about those in which sensors have been installed. Therefore, it is proposed to create an artificial intelligence model capable of determining what the traffic is like on these streets. It is proposed to create a Machine Learning model based on LSTM (Long Short Term Memory) neural networks that is capable of determining the traffic status at any point in the city of Santiago de Compostela. We categorize the traffic state into 4 classes: free, fluid, congested and traffic jam. This traffic categorization is made from the two measurements obtained from the sensors already installed in the city. These measures are the flow or flow of cars and the occupancy or percentage of occupancy. An algorithm is also implemented that allows knowing the distance from a point to the other sensors along the shortest path, based on Dijkstra's algorithm. In this way, we seek to determine how strong the correlation will be between the selected point and the sensors. This solution will be an efficient, reliable and economical alternative for the municipalities of Galicia, compared to the installation of more sensors that monitor traffic.
The motivation for this work comes from the lack of knowledge, on the part of the city councils of the large Galician cities, of the state of traffic in most of their streets. They only have information about those in which sensors have been installed. Therefore, it is proposed to create an artificial intelligence model capable of determining what the traffic is like on these streets. It is proposed to create a Machine Learning model based on LSTM (Long Short Term Memory) neural networks that is capable of determining the traffic status at any point in the city of Santiago de Compostela. We categorize the traffic state into 4 classes: free, fluid, congested and traffic jam. This traffic categorization is made from the two measurements obtained from the sensors already installed in the city. These measures are the flow or flow of cars and the occupancy or percentage of occupancy. An algorithm is also implemented that allows knowing the distance from a point to the other sensors along the shortest path, based on Dijkstra's algorithm. In this way, we seek to determine how strong the correlation will be between the selected point and the sensors. This solution will be an efficient, reliable and economical alternative for the municipalities of Galicia, compared to the installation of more sensors that monitor traffic.
Direction
RIOS VIQUEIRA, JOSE RAMON (Tutorships)
MERA PEREZ, DAVID (Co-tutorships)
RIOS VIQUEIRA, JOSE RAMON (Tutorships)
MERA PEREZ, DAVID (Co-tutorships)
Court
VAZQUEZ CENDON, MARIA ELENA (Chairman)
CHAVES FRAGA, DAVID (Secretary)
SUAREZ GAREA, JORGE ALBERTO (Member)
VAZQUEZ CENDON, MARIA ELENA (Chairman)
CHAVES FRAGA, DAVID (Secretary)
SUAREZ GAREA, JORGE ALBERTO (Member)
Performance Analysis of the Speculative Computing Tool SpecLib
Authorship
O.V.D.
Computer Engineering Degree
O.V.D.
Computer Engineering Degree
Defense date
09.12.2024 13:00
09.12.2024 13:00
Summary
Since the introduction of multicore processors, shared memory parallelization has been essential for improving the performance of computer systems by dividing tasks into smaller tasks that can be executed simultaneously on multiple processors or cores. Speculative parallelization allows this technique to be applied to problems with complex or unpredictable data dependencies, enhancing efficiency and reducing waiting times. In this work, we analyze the performance of SpecLib, an automatic speculative parallelization library, evaluating its effectiveness in various scenarios, varying factors such as the number of threads and the size of code blocks. The work aims to identify the factors that can affect the library's efficiency in different scenarios.
Since the introduction of multicore processors, shared memory parallelization has been essential for improving the performance of computer systems by dividing tasks into smaller tasks that can be executed simultaneously on multiple processors or cores. Speculative parallelization allows this technique to be applied to problems with complex or unpredictable data dependencies, enhancing efficiency and reducing waiting times. In this work, we analyze the performance of SpecLib, an automatic speculative parallelization library, evaluating its effectiveness in various scenarios, varying factors such as the number of threads and the size of code blocks. The work aims to identify the factors that can affect the library's efficiency in different scenarios.
Direction
Fernández Rivera, Francisco (Tutorships)
CABALEIRO DOMINGUEZ, JOSE CARLOS (Co-tutorships)
Fernández Rivera, Francisco (Tutorships)
CABALEIRO DOMINGUEZ, JOSE CARLOS (Co-tutorships)
Court
ARIAS RODRIGUEZ, JUAN ENRIQUE (Chairman)
VILA BLANCO, NICOLAS (Secretary)
GONZALEZ DIAZ, JULIO (Member)
ARIAS RODRIGUEZ, JUAN ENRIQUE (Chairman)
VILA BLANCO, NICOLAS (Secretary)
GONZALEZ DIAZ, JULIO (Member)
UniLink: An application to connect university students in Santiago
Authorship
M.D.U.O.
Computer Engineering Degree
M.D.U.O.
Computer Engineering Degree
Defense date
09.12.2024 10:00
09.12.2024 10:00
Summary
This Final Degree Project presents the development of UniLink, a mobile application designed to facilitate and strengthen the social interaction and connection between university students in the city of Santiago de Compostela. Its main objective is to function as a social network focused on the university community and allow users to discover people, events, places of interest and other elements in their environment. The application will try to address an important and current social need within the university context, where some students, especially newcomers, may feel disoriented or may have more difficulties when it comes to finding classmates, accommodation, places of leisure... UniLink will provide a digital space where users can connect, collaborate, share resources and participate in relevant events, improving their overall university experience.
This Final Degree Project presents the development of UniLink, a mobile application designed to facilitate and strengthen the social interaction and connection between university students in the city of Santiago de Compostela. Its main objective is to function as a social network focused on the university community and allow users to discover people, events, places of interest and other elements in their environment. The application will try to address an important and current social need within the university context, where some students, especially newcomers, may feel disoriented or may have more difficulties when it comes to finding classmates, accommodation, places of leisure... UniLink will provide a digital space where users can connect, collaborate, share resources and participate in relevant events, improving their overall university experience.
Direction
VARELA PET, JOSE (Tutorships)
VARELA PET, JOSE (Tutorships)
Court
Fernández Rivera, Francisco (Chairman)
SANTOS MATEOS, ROI (Secretary)
López Vilariño, David (Member)
Fernández Rivera, Francisco (Chairman)
SANTOS MATEOS, ROI (Secretary)
López Vilariño, David (Member)
Semantic Emofinder: Construction and Exploitation of a Knowledge Graph for the F.A.I.R. Management of Emotional Word Bases
Authorship
O.H.G.
Computer Engineering Degree
O.H.G.
Computer Engineering Degree
Defense date
09.12.2024 11:30
09.12.2024 11:30
Summary
This end-of-degree project focuses on the development of a system that allows the construction and exploitation of a knowledge graph from emotional word bases. For this purpose, a vocabulary will be created to manage these word bases, and a set of procedures to popularise the knowledge base in a (semi)automatic way. Finally, a web application will be developed that consumes such structured data and allows its management and downloading by other users/scientists based on a set of defined filters. The main contribution, therefore, will be the updating of the previous system (https://www.usc.gal/pcc/app/emofinder/) through the use and exploitation of semantic technologies to ensure the F.A.I.R. (findable, accessible, interoperable and reusable) principles in the integrated word bases.
This end-of-degree project focuses on the development of a system that allows the construction and exploitation of a knowledge graph from emotional word bases. For this purpose, a vocabulary will be created to manage these word bases, and a set of procedures to popularise the knowledge base in a (semi)automatic way. Finally, a web application will be developed that consumes such structured data and allows its management and downloading by other users/scientists based on a set of defined filters. The main contribution, therefore, will be the updating of the previous system (https://www.usc.gal/pcc/app/emofinder/) through the use and exploitation of semantic technologies to ensure the F.A.I.R. (findable, accessible, interoperable and reusable) principles in the integrated word bases.
Direction
CHAVES FRAGA, DAVID (Tutorships)
LAMA PENIN, MANUEL (Co-tutorships)
CHAVES FRAGA, DAVID (Tutorships)
LAMA PENIN, MANUEL (Co-tutorships)
Court
Argüello Pedreira, Francisco Santiago (Chairman)
ALONSO MORAL, JOSE MARIA (Secretary)
MUCIENTES MOLINA, MANUEL FELIPE (Member)
Argüello Pedreira, Francisco Santiago (Chairman)
ALONSO MORAL, JOSE MARIA (Secretary)
MUCIENTES MOLINA, MANUEL FELIPE (Member)
Multitemporal image alignment of fluvial ecosystems using feature-based algorithms
Authorship
D.F.R.
Computer Engineering Degree
D.F.R.
Computer Engineering Degree
Defense date
07.18.2024 16:30
07.18.2024 16:30
Summary
In recent years, technological advances have enabled the use of multidimensional images in a wide range of contexts and applications. In many of these cases, Unmanned Aerial Vehicles (UAVs) with sensors capable of capturing images containing reflectance information from different bands of the electromagnetic spectrum are used. Problems associated with processing this type of image, such as registration or alignment, have become necessary for a multitude of applications of these images, such as change detection or the analysis of the evolution of foreign vegetation. The registration of two multidimensional images involves finding a geometric transformation (scale, angle, and rotation) that allows aligning images captured in the same area by using all the available spectral information to improve the quality of the registration. These images may be obtained at different time instances, with different perspectives or lighting conditions. Difficulties that, combined with the conditions of capture from UAVs, result in high variability between the images, despite having the same number and properties of spectral bands. These conditions are the main challenges of this work, which seeks to solve the problem of efficient registration of high spatial resolution multispectral multitemporal images of river ecosystems in Galicia. The aim is to align images captured in 2023 and 2024 with respect to images obtained in 2018 for the same areas. Given the large size of the images, portions of the complete images will be used. Additionally, due to the difficulties mentioned earlier in the registration process, a two-level process will be proposed: initial registration and refinement. First, an analysis and selection of the different registration algorithms available in the state of the art, capable of registering multi-band images, particularly hiperspectral ones, has been carried out. The selected methods have been implemented and evaluated on high spatial resolution multispectral multitemporal images of land cover. In the proposed experimentation, different tests have been conducted: evaluation of accuracy for different image sizes, with different numbers and types of natural and artificial elements, and with different levels of overlap or visual coincidence. The main contribution of this work is to address the registration of images that were captured as part of a project carried out by the research group in which this Final Degree Project was developed and which, therefore, suffer from all the problems typical of real images. For example, the pairs of images being registered correspond to different years and were captured by sensors on board drones, so the capture conditions between images from different dates are very different, both in terms of climate and drone inclination or altitude. Furthermore, it addresses the application of methods adapted to images with hundreds of bands and medium spatial resolution to images that have only five spectral bands but very high spatial resolution. A process was designed to systematically address image registration, and it was determined which algorithms are most suitable for both initial registration and final refinement. The evaluation is done both in terms of accuracies obtained and resilience to image overlaps and variations, as well as in terms of execution time.
In recent years, technological advances have enabled the use of multidimensional images in a wide range of contexts and applications. In many of these cases, Unmanned Aerial Vehicles (UAVs) with sensors capable of capturing images containing reflectance information from different bands of the electromagnetic spectrum are used. Problems associated with processing this type of image, such as registration or alignment, have become necessary for a multitude of applications of these images, such as change detection or the analysis of the evolution of foreign vegetation. The registration of two multidimensional images involves finding a geometric transformation (scale, angle, and rotation) that allows aligning images captured in the same area by using all the available spectral information to improve the quality of the registration. These images may be obtained at different time instances, with different perspectives or lighting conditions. Difficulties that, combined with the conditions of capture from UAVs, result in high variability between the images, despite having the same number and properties of spectral bands. These conditions are the main challenges of this work, which seeks to solve the problem of efficient registration of high spatial resolution multispectral multitemporal images of river ecosystems in Galicia. The aim is to align images captured in 2023 and 2024 with respect to images obtained in 2018 for the same areas. Given the large size of the images, portions of the complete images will be used. Additionally, due to the difficulties mentioned earlier in the registration process, a two-level process will be proposed: initial registration and refinement. First, an analysis and selection of the different registration algorithms available in the state of the art, capable of registering multi-band images, particularly hiperspectral ones, has been carried out. The selected methods have been implemented and evaluated on high spatial resolution multispectral multitemporal images of land cover. In the proposed experimentation, different tests have been conducted: evaluation of accuracy for different image sizes, with different numbers and types of natural and artificial elements, and with different levels of overlap or visual coincidence. The main contribution of this work is to address the registration of images that were captured as part of a project carried out by the research group in which this Final Degree Project was developed and which, therefore, suffer from all the problems typical of real images. For example, the pairs of images being registered correspond to different years and were captured by sensors on board drones, so the capture conditions between images from different dates are very different, both in terms of climate and drone inclination or altitude. Furthermore, it addresses the application of methods adapted to images with hundreds of bands and medium spatial resolution to images that have only five spectral bands but very high spatial resolution. A process was designed to systematically address image registration, and it was determined which algorithms are most suitable for both initial registration and final refinement. The evaluation is done both in terms of accuracies obtained and resilience to image overlaps and variations, as well as in terms of execution time.
Direction
ORDOÑEZ IGLESIAS, ALVARO (Tutorships)
Blanco Heras, Dora (Co-tutorships)
ORDOÑEZ IGLESIAS, ALVARO (Tutorships)
Blanco Heras, Dora (Co-tutorships)
Court
Pardo López, Xosé Manuel (Chairman)
Puentes Cociña, Beltrán (Secretary)
LOPEZ FANDIÑO, JAVIER (Member)
Pardo López, Xosé Manuel (Chairman)
Puentes Cociña, Beltrán (Secretary)
LOPEZ FANDIÑO, JAVIER (Member)
Detecting malicious or noisy clients in federated learning
Authorship
R.M.E.
Computer Engineering Degree
R.M.E.
Computer Engineering Degree
Defense date
07.18.2024 18:30
07.18.2024 18:30
Summary
Federated learning (FL) emerges as a promising solution for collaborative development of artificial intelligence models without compromising the privacy of distributed data. However, data heterogeneity, diverse local learning methodologies, and the presence of malicious attacks pose significant challenges that hinder the effectiveness and security of the global model. This work explores solutions to enhance the robustness of federated learning, not only in scenarios with high data heterogeneity but also in environments affected by malicious attacks aiming to significantly degrade the performance of the global model. Two main lines of work are examined to mitigate the impact of divergent clients: robust aggregation and active defense techniques. Moreover, a new methodology within the field of adapted personalized federated learning is introduced, employing local personalized models alongside a global model, allowing for more precise adaptation to the specific conditions of each client. Empirical studies, conducted in both synthetic and real environments, validate the effectiveness of the proposed strategies against attacks and demonstrate how proper management of heterogeneity and anomaly detection can significantly improve the performance of the federated model. This work, in addition to contributing to the existing literature on federated learning, also offers a practical perspective for its application in robotics, where data diversity and task specificity present unique challenges.
Federated learning (FL) emerges as a promising solution for collaborative development of artificial intelligence models without compromising the privacy of distributed data. However, data heterogeneity, diverse local learning methodologies, and the presence of malicious attacks pose significant challenges that hinder the effectiveness and security of the global model. This work explores solutions to enhance the robustness of federated learning, not only in scenarios with high data heterogeneity but also in environments affected by malicious attacks aiming to significantly degrade the performance of the global model. Two main lines of work are examined to mitigate the impact of divergent clients: robust aggregation and active defense techniques. Moreover, a new methodology within the field of adapted personalized federated learning is introduced, employing local personalized models alongside a global model, allowing for more precise adaptation to the specific conditions of each client. Empirical studies, conducted in both synthetic and real environments, validate the effectiveness of the proposed strategies against attacks and demonstrate how proper management of heterogeneity and anomaly detection can significantly improve the performance of the federated model. This work, in addition to contributing to the existing literature on federated learning, also offers a practical perspective for its application in robotics, where data diversity and task specificity present unique challenges.
Direction
IGLESIAS RODRIGUEZ, ROBERTO (Tutorships)
Barro Ameneiro, Senén (Co-tutorships)
IGLESIAS RODRIGUEZ, ROBERTO (Tutorships)
Barro Ameneiro, Senén (Co-tutorships)
Court
VAZQUEZ CENDON, MARIA ELENA (Chairman)
CHAVES FRAGA, DAVID (Secretary)
SUAREZ GAREA, JORGE ALBERTO (Member)
VAZQUEZ CENDON, MARIA ELENA (Chairman)
CHAVES FRAGA, DAVID (Secretary)
SUAREZ GAREA, JORGE ALBERTO (Member)
Development of a prototype for an autonomous intelligent conversational agent to study depression using artificial intelligence techniques
Authorship
A.G.C.
Computer Engineering Degree
A.G.C.
Computer Engineering Degree
Defense date
07.19.2024 16:00
07.19.2024 16:00
Summary
Depression has become one of the major chronic health problems globally, and its diagnosis often presents challenges. In Spain, it is estimated that 50% of people with depression are not diagnosed. Although society is advancing in understanding this condition, the stigma of visiting a psychiatrist persists, and it continues to be confused with sadness. Simultaneously, in recent years, the use of artificial intelligence techniques has significantly advanced in the field of mental health. For example, natural language processing (NLP) techniques have been employed to analyze sentiments on social media. However, the number of research studies or projects applying these methodologies for the early detection of depression remains limited. Many people who require psychological intervention are not identified due to the aforementioned obstacles or the global shortage of mental health professionals, among other factors. Therefore, it is important to have a mass screening approach that allows for the detection of depression in its early stages and facilitates the intervention of mental health professionals. The solution proposed in this project is a prototype virtual assistant designed to manage the Geriatric Depression Scale (GDS-VE). This virtual assistant interacts with the user through a fluid conversation, analyzes their responses based on the questionnaire options, and determines the presence or absence of depressive symptoms. The capability of Dialogflow, by Google Cloud, to carry out this task has been experimentally tested. Additionally, the efficacy of pre-trained natural language inference (NLI) models in analyzing and understanding the context of responses and user conversation has been evaluated. The observations and results obtained place these tools in a favorable position for their purpose: to serve as support in facilitating a quick diagnosis of this type of illness. Under no circumstances are they proposed as a substitute for mental health professionals.
Depression has become one of the major chronic health problems globally, and its diagnosis often presents challenges. In Spain, it is estimated that 50% of people with depression are not diagnosed. Although society is advancing in understanding this condition, the stigma of visiting a psychiatrist persists, and it continues to be confused with sadness. Simultaneously, in recent years, the use of artificial intelligence techniques has significantly advanced in the field of mental health. For example, natural language processing (NLP) techniques have been employed to analyze sentiments on social media. However, the number of research studies or projects applying these methodologies for the early detection of depression remains limited. Many people who require psychological intervention are not identified due to the aforementioned obstacles or the global shortage of mental health professionals, among other factors. Therefore, it is important to have a mass screening approach that allows for the detection of depression in its early stages and facilitates the intervention of mental health professionals. The solution proposed in this project is a prototype virtual assistant designed to manage the Geriatric Depression Scale (GDS-VE). This virtual assistant interacts with the user through a fluid conversation, analyzes their responses based on the questionnaire options, and determines the presence or absence of depressive symptoms. The capability of Dialogflow, by Google Cloud, to carry out this task has been experimentally tested. Additionally, the efficacy of pre-trained natural language inference (NLI) models in analyzing and understanding the context of responses and user conversation has been evaluated. The observations and results obtained place these tools in a favorable position for their purpose: to serve as support in facilitating a quick diagnosis of this type of illness. Under no circumstances are they proposed as a substitute for mental health professionals.
Direction
VALLADARES RODRIGUEZ, SONIA MARIA (Tutorships)
Losada Carril, David Enrique (Co-tutorships)
FERNANDEZ PICHEL, MARCOS (Co-tutorships)
VALLADARES RODRIGUEZ, SONIA MARIA (Tutorships)
Losada Carril, David Enrique (Co-tutorships)
FERNANDEZ PICHEL, MARCOS (Co-tutorships)
Court
Fernández Pena, Anselmo Tomás (Chairman)
SACO LOPEZ, PEDRO JOSE (Secretary)
RODRIGUEZ PRESEDO, JESUS MARIA (Member)
Fernández Pena, Anselmo Tomás (Chairman)
SACO LOPEZ, PEDRO JOSE (Secretary)
RODRIGUEZ PRESEDO, JESUS MARIA (Member)
2-methoxy-2-methylheptane production plant
Authorship
R.V.M.
Bachelor's Degree in Chemical Engeneering
R.V.M.
Bachelor's Degree in Chemical Engeneering
Defense date
07.16.2024 12:25
07.16.2024 12:25
Summary
2-methoxy-2-methyl-heptane (MMH) is used in various fields, mostly as a gasoline additive or pharmaceutical intermediate. The capacity of the plant is 30 000 tonnes per year as both traditional applications and new chemical grade alternatives are not large tonnage applications. MMH is produced from methanol and 2-methyl-1-heptene. As a sustainable alternative, green methanol from a nearby plant is proposed. Pablo Rodríguez-Villamil Seoane is in charge of the design of a packed distillation column (T-203). Raúl Veiga Moo is in charge of the design of a catalytic heterogeneous fixed bed reactor (R-101). Finally, Hugo Vidal Cabanelas is in charge of the design of a distillation plate column (T-202).
2-methoxy-2-methyl-heptane (MMH) is used in various fields, mostly as a gasoline additive or pharmaceutical intermediate. The capacity of the plant is 30 000 tonnes per year as both traditional applications and new chemical grade alternatives are not large tonnage applications. MMH is produced from methanol and 2-methyl-1-heptene. As a sustainable alternative, green methanol from a nearby plant is proposed. Pablo Rodríguez-Villamil Seoane is in charge of the design of a packed distillation column (T-203). Raúl Veiga Moo is in charge of the design of a catalytic heterogeneous fixed bed reactor (R-101). Finally, Hugo Vidal Cabanelas is in charge of the design of a distillation plate column (T-202).
Direction
VIDAL TATO, MARIA ISABEL (Tutorships)
VIDAL TATO, MARIA ISABEL (Tutorships)
Court
RODIL RODRIGUEZ, EVA (Chairman)
SOUTO GONZALEZ, JOSE ANTONIO (Secretary)
VAL DEL RIO, MARIA ANGELES (Member)
RODIL RODRIGUEZ, EVA (Chairman)
SOUTO GONZALEZ, JOSE ANTONIO (Secretary)
VAL DEL RIO, MARIA ANGELES (Member)
Plant for the production of pentanoic acid from y-valerolactone
Authorship
N.M.G.
Bachelor's Degree in Chemical Engeneering
N.M.G.
Bachelor's Degree in Chemical Engeneering
Defense date
07.16.2024 11:50
07.16.2024 11:50
Summary
The aim of this proyect is to desing a plant for the production of pentanoic acid from y-valerolactone. The proposed plant will produce 50,000t/year of pentanoic acid with a purity of 99% and will operate continously for 24 hours, 330 days a year. The process will start with the conditioning of the raw materials, green hydrogen and y-valerolactone, together with a reaction stage. The latter will aim at obtaining pentanoic acid and 2-methyltetrahydrofuran as a by-product. The reaction stage will take place in a catalytic reactor R-101 by hydroxigenation of y-valerolactone. This equipment will be designed by Manuel Martínez García. The process continues with a first separation stage where the unreacted hydrogen is recovered. Finally, the rest of the compounds are taken to a separation train consisting of three distillation towers. Of these three columns, Noemí Mallón González will carry out the rigorous design of the last, T-303, where the pentanoic acid is separated from the y-valerolactone.
The aim of this proyect is to desing a plant for the production of pentanoic acid from y-valerolactone. The proposed plant will produce 50,000t/year of pentanoic acid with a purity of 99% and will operate continously for 24 hours, 330 days a year. The process will start with the conditioning of the raw materials, green hydrogen and y-valerolactone, together with a reaction stage. The latter will aim at obtaining pentanoic acid and 2-methyltetrahydrofuran as a by-product. The reaction stage will take place in a catalytic reactor R-101 by hydroxigenation of y-valerolactone. This equipment will be designed by Manuel Martínez García. The process continues with a first separation stage where the unreacted hydrogen is recovered. Finally, the rest of the compounds are taken to a separation train consisting of three distillation towers. Of these three columns, Noemí Mallón González will carry out the rigorous design of the last, T-303, where the pentanoic acid is separated from the y-valerolactone.
Direction
SOTO CAMPOS, ANA MARIA (Tutorships)
SOTO CAMPOS, ANA MARIA (Tutorships)
Court
MOSQUERA CORRAL, ANUSKA (Chairman)
SINEIRO TORRES, JORGE (Secretary)
VIDAL TATO, MARIA ISABEL (Member)
MOSQUERA CORRAL, ANUSKA (Chairman)
SINEIRO TORRES, JORGE (Secretary)
VIDAL TATO, MARIA ISABEL (Member)
Production plant of polyvinyl alcohol
Authorship
D.P.L.
Bachelor's Degree in Chemical Engeneering
D.P.L.
Bachelor's Degree in Chemical Engeneering
Defense date
07.16.2024 11:55
07.16.2024 11:55
Summary
This final degree project covers the design of a plant for the polyvinyl alcohol polymer, produced from polyvinyl acetate in a hydrolysis reaction with methanol in the presence of a sodium hydroxide catalyst. After the reaction and separation of the solid phase from the liquid phase, the main byproduct is separated by distillation and the product is washed with water. The project has a report with its respective annexes; material and energy balances, supporting calculations, economic viability and substance safety sheets. It also has blueprints, studies with its own entity, specifications and budget.
This final degree project covers the design of a plant for the polyvinyl alcohol polymer, produced from polyvinyl acetate in a hydrolysis reaction with methanol in the presence of a sodium hydroxide catalyst. After the reaction and separation of the solid phase from the liquid phase, the main byproduct is separated by distillation and the product is washed with water. The project has a report with its respective annexes; material and energy balances, supporting calculations, economic viability and substance safety sheets. It also has blueprints, studies with its own entity, specifications and budget.
Direction
SINEIRO TORRES, JORGE (Tutorships)
FRANCO URIA, MARIA AMAYA (Co-tutorships)
SINEIRO TORRES, JORGE (Tutorships)
FRANCO URIA, MARIA AMAYA (Co-tutorships)
Court
RODIL RODRIGUEZ, EVA (Chairman)
SOUTO GONZALEZ, JOSE ANTONIO (Secretary)
VAL DEL RIO, MARIA ANGELES (Member)
RODIL RODRIGUEZ, EVA (Chairman)
SOUTO GONZALEZ, JOSE ANTONIO (Secretary)
VAL DEL RIO, MARIA ANGELES (Member)
Peracetic acid production plant
Authorship
M.A.C.P.
Bachelor's Degree in Chemical Engeneering
M.A.C.P.
Bachelor's Degree in Chemical Engeneering
Defense date
09.13.2024 09:30
09.13.2024 09:30
Summary
The purpose of this project is to design a basic production plant for peracetic acid from hydrogen peroxide and acetic acid. The annual production of the industrial plant will be 20,000 tons per year of a solution containing more than 32% peracetic acid. The process includes a pre-conditioning stage, where sulfuric acid in the liquid phase will be used to protonate the acetic acid that will be used in the reactor, with sulfuric acid playing a crucial role in the synthesis of the reaction as a catalyst. This reaction will take place in the tubular reactor R-101, designed by Brais Queiro Lamas. At the reactor outlet, two separation columns, T-201 and T-202, will be used respectively. The first is intended for the recovery of sulfuric acid, and the second to concentrate the peracetic acid to obtain the desired product. The T-202 column will be designed by Manuel Antonio Castro Pereiras. This project is presented as a Final Degree Project through which the students Manuel Antonio Castro Pereiras and Brais Queiro Lamas aim to obtain the Degree in Chemical Engineering awarded by the Higher Technical School of Engineering belonging to the University of Santiago de Compostela.
The purpose of this project is to design a basic production plant for peracetic acid from hydrogen peroxide and acetic acid. The annual production of the industrial plant will be 20,000 tons per year of a solution containing more than 32% peracetic acid. The process includes a pre-conditioning stage, where sulfuric acid in the liquid phase will be used to protonate the acetic acid that will be used in the reactor, with sulfuric acid playing a crucial role in the synthesis of the reaction as a catalyst. This reaction will take place in the tubular reactor R-101, designed by Brais Queiro Lamas. At the reactor outlet, two separation columns, T-201 and T-202, will be used respectively. The first is intended for the recovery of sulfuric acid, and the second to concentrate the peracetic acid to obtain the desired product. The T-202 column will be designed by Manuel Antonio Castro Pereiras. This project is presented as a Final Degree Project through which the students Manuel Antonio Castro Pereiras and Brais Queiro Lamas aim to obtain the Degree in Chemical Engineering awarded by the Higher Technical School of Engineering belonging to the University of Santiago de Compostela.
Direction
MOREIRA MARTINEZ, RAMON FELIPE (Tutorships)
GOMEZ DIAZ, DIEGO (Co-tutorships)
MOREIRA MARTINEZ, RAMON FELIPE (Tutorships)
GOMEZ DIAZ, DIEGO (Co-tutorships)
Court
GARRIDO FERNANDEZ, JUAN MANUEL (Chairman)
FRANCO RUIZ, DANIEL JOSE (Secretary)
FREIRE LEIRA, MARIA SONIA (Member)
GARRIDO FERNANDEZ, JUAN MANUEL (Chairman)
FRANCO RUIZ, DANIEL JOSE (Secretary)
FREIRE LEIRA, MARIA SONIA (Member)
Acetic anhydride production plant by methyl acetate carbonylation.
Authorship
J.T.P.
Bachelor's Degree in Chemical Engeneering
J.T.P.
Bachelor's Degree in Chemical Engeneering
Defense date
07.16.2024 13:15
07.16.2024 13:15
Summary
The purpose of this project is to design a production plant for acetic anhydride through the carbonylation of methyl acetate. The production capacity of the plant to be designed will be 100,000 tons per year of acetic anhydride with a purity of 99.7% by weight, and it will operate continuously for 330 days a year, 24 hours a day. The key equipment of the plant includes the bubble reactor R-101, which will be designed by Rodrigo Tasende Reboredo, and the distillation column T-202, which will be designed by Jacobo Torres Pérez.
The purpose of this project is to design a production plant for acetic anhydride through the carbonylation of methyl acetate. The production capacity of the plant to be designed will be 100,000 tons per year of acetic anhydride with a purity of 99.7% by weight, and it will operate continuously for 330 days a year, 24 hours a day. The key equipment of the plant includes the bubble reactor R-101, which will be designed by Rodrigo Tasende Reboredo, and the distillation column T-202, which will be designed by Jacobo Torres Pérez.
Direction
FREIRE LEIRA, MARIA SONIA (Tutorships)
GONZALEZ ALVAREZ, JULIA (Co-tutorships)
FREIRE LEIRA, MARIA SONIA (Tutorships)
GONZALEZ ALVAREZ, JULIA (Co-tutorships)
Court
RODIL RODRIGUEZ, EVA (Chairman)
SOUTO GONZALEZ, JOSE ANTONIO (Secretary)
VAL DEL RIO, MARIA ANGELES (Member)
RODIL RODRIGUEZ, EVA (Chairman)
SOUTO GONZALEZ, JOSE ANTONIO (Secretary)
VAL DEL RIO, MARIA ANGELES (Member)
Methyl acetate production plant
Authorship
M.G.R.
Bachelor's Degree in Chemical Engeneering
M.G.R.
Bachelor's Degree in Chemical Engeneering
Defense date
09.13.2024 11:40
09.13.2024 11:40
Summary
The objective of this project is the design of a methyl acetate production plant using methanol and acetic acid as raw materials. Furthermore, the idea of implementing the ionic liquid in the chemical industry as a new way to separate azeotropic mixtures is developed in order to open a new vision on separation methods.
The objective of this project is the design of a methyl acetate production plant using methanol and acetic acid as raw materials. Furthermore, the idea of implementing the ionic liquid in the chemical industry as a new way to separate azeotropic mixtures is developed in order to open a new vision on separation methods.
Direction
SOTO CAMPOS, ANA MARIA (Tutorships)
SOTO CAMPOS, ANA MARIA (Tutorships)
Court
GIL GONZALEZ, ALVARO (Chairman)
FRANCO URIA, MARIA AMAYA (Secretary)
GOMEZ DIAZ, DIEGO (Member)
GIL GONZALEZ, ALVARO (Chairman)
FRANCO URIA, MARIA AMAYA (Secretary)
GOMEZ DIAZ, DIEGO (Member)
Monoethanolamine production plant from Ammonia and Ethylene Oxide
Authorship
C.R.G.
Bachelor's Degree in Chemical Engeneering
C.R.G.
Bachelor's Degree in Chemical Engeneering
Defense date
09.13.2024 10:50
09.13.2024 10:50
Summary
For the design of a plant for the production of monoethanolamine from ammonia and ethylene oxide, the production process is divided into three sections: Conditioning: contains an absorption column in which ammonia is absorbed in water, the objective of which is to achieve a given ammonia composition at the reactor inlet. This absorption column will be designed by Irene Vidal. Reaction: this section can be composed of one or two reactors. In both cases, the equipment is isothermal. Everything will depend on the catalyst and the selected separation conditions. The exothermic reactions that occur inside the reactor are those produced by ammonia and ethylene oxide, obtaining ethanolamines (ETAS). They are carried out in liquid phase with second order kinetics. Adrián Ramos will design a single reactor). Separation: the last section is composed of a set of distillation columns, whose function is to separate the monoethanolamine from the different by-products. The distillation column will be designed by Carlos Rodriguez.
For the design of a plant for the production of monoethanolamine from ammonia and ethylene oxide, the production process is divided into three sections: Conditioning: contains an absorption column in which ammonia is absorbed in water, the objective of which is to achieve a given ammonia composition at the reactor inlet. This absorption column will be designed by Irene Vidal. Reaction: this section can be composed of one or two reactors. In both cases, the equipment is isothermal. Everything will depend on the catalyst and the selected separation conditions. The exothermic reactions that occur inside the reactor are those produced by ammonia and ethylene oxide, obtaining ethanolamines (ETAS). They are carried out in liquid phase with second order kinetics. Adrián Ramos will design a single reactor). Separation: the last section is composed of a set of distillation columns, whose function is to separate the monoethanolamine from the different by-products. The distillation column will be designed by Carlos Rodriguez.
Direction
RODIL RODRIGUEZ, EVA (Tutorships)
RODIL RODRIGUEZ, EVA (Tutorships)
Court
CARBALLA ARCOS, MARTA (Chairman)
EIBES GONZALEZ, GEMMA MARIA (Secretary)
BELLO BUGALLO, PASTORA MARIA (Member)
CARBALLA ARCOS, MARTA (Chairman)
EIBES GONZALEZ, GEMMA MARIA (Secretary)
BELLO BUGALLO, PASTORA MARIA (Member)
Plant of production of pentanoic acid from y-valerolactone
Authorship
M.M.G.
Bachelor's Degree in Chemical Engeneering
M.M.G.
Bachelor's Degree in Chemical Engeneering
Defense date
07.16.2024 11:50
07.16.2024 11:50
Summary
The aim of this project is to design a plant for the production of pentanoic acid from y-valerolactone. The proposed plant will produce 50,000 t/year of pentanoic acid with a purity of 99% and will operate continously for 24 hours, 330 days a year. The process will start with the conditioning of the raw materials, green hydrogen and y-valerolactone, together with a reaction stage. The latter willl aim at obtaning pentaoic acid and 2-methyltetrahydrofuran as a by-product. The reaction stage takes place in a catalytic reactor R-101 by hidrodeoxygenation of y-valerolactone. This equipment will be designed by Manuel Martínez García. The process continues with a first separaction stage where the unreacted hydrogen is recovered. Finally, the rest of the compounds are taken to a separation train consisting of three destillation towers. Of these three columns Noemí Mallón González will carry out with the rigorous desing of the last, T-303, where the pentanoic acid will be separated from the y-valerolactone.
The aim of this project is to design a plant for the production of pentanoic acid from y-valerolactone. The proposed plant will produce 50,000 t/year of pentanoic acid with a purity of 99% and will operate continously for 24 hours, 330 days a year. The process will start with the conditioning of the raw materials, green hydrogen and y-valerolactone, together with a reaction stage. The latter willl aim at obtaning pentaoic acid and 2-methyltetrahydrofuran as a by-product. The reaction stage takes place in a catalytic reactor R-101 by hidrodeoxygenation of y-valerolactone. This equipment will be designed by Manuel Martínez García. The process continues with a first separaction stage where the unreacted hydrogen is recovered. Finally, the rest of the compounds are taken to a separation train consisting of three destillation towers. Of these three columns Noemí Mallón González will carry out with the rigorous desing of the last, T-303, where the pentanoic acid will be separated from the y-valerolactone.
Direction
SOTO CAMPOS, ANA MARIA (Tutorships)
SOTO CAMPOS, ANA MARIA (Tutorships)
Court
MOSQUERA CORRAL, ANUSKA (Chairman)
SINEIRO TORRES, JORGE (Secretary)
VIDAL TATO, MARIA ISABEL (Member)
MOSQUERA CORRAL, ANUSKA (Chairman)
SINEIRO TORRES, JORGE (Secretary)
VIDAL TATO, MARIA ISABEL (Member)
Plant to produce monoethanolamine from ammonia and ethylene oxide
Authorship
A.R.R.
Bachelor's Degree in Chemical Engeneering
A.R.R.
Bachelor's Degree in Chemical Engeneering
Defense date
09.13.2024 10:50
09.13.2024 10:50
Summary
For the design of a monoethanolamine production plant from ammonia and ethylene oxide, the production process is divided into three sections: -Conditioning: This section contains an absorption column where ammonia is absorbed in water, with the goal of achieving a specific ammonia composition at the reactor inlet. This absorption column will be designed by Irene Vidal. -Reaction: This section can consist of one or two reactors. In both cases, the equipment is of the isothermal type. Everything will depend on the catalyst and the selected separation conditions. The exothermic reactions that occur inside the reactor are produced by ammonia and ethylene oxide, yielding ethanolamines (ETAS). These reactions take place in the liquid phase with second-order kinetics. Adrián Ramos will design a single reactor. -Separation: The final section consists of a set of distillation columns, whose function is to separate monoethanolamine from the various by-products. The distillation column will be designed by Carlos Rodríguez.
For the design of a monoethanolamine production plant from ammonia and ethylene oxide, the production process is divided into three sections: -Conditioning: This section contains an absorption column where ammonia is absorbed in water, with the goal of achieving a specific ammonia composition at the reactor inlet. This absorption column will be designed by Irene Vidal. -Reaction: This section can consist of one or two reactors. In both cases, the equipment is of the isothermal type. Everything will depend on the catalyst and the selected separation conditions. The exothermic reactions that occur inside the reactor are produced by ammonia and ethylene oxide, yielding ethanolamines (ETAS). These reactions take place in the liquid phase with second-order kinetics. Adrián Ramos will design a single reactor. -Separation: The final section consists of a set of distillation columns, whose function is to separate monoethanolamine from the various by-products. The distillation column will be designed by Carlos Rodríguez.
Direction
RODIL RODRIGUEZ, EVA (Tutorships)
RODIL RODRIGUEZ, EVA (Tutorships)
Court
CARBALLA ARCOS, MARTA (Chairman)
EIBES GONZALEZ, GEMMA MARIA (Secretary)
BELLO BUGALLO, PASTORA MARIA (Member)
CARBALLA ARCOS, MARTA (Chairman)
EIBES GONZALEZ, GEMMA MARIA (Secretary)
BELLO BUGALLO, PASTORA MARIA (Member)
Detection of placement and collection of acoustic buoys with low-resolution GPS positions.
Authorship
C.R.F.
Computer Engineering Degree
C.R.F.
Computer Engineering Degree
Defense date
02.15.2024 13:00
02.15.2024 13:00
Summary
Tuna fishing boats use satellite buoys with acoustic probes and GPS to locate tuna schools. These buoys are deployed in the ocean by tuna vessels and support ships. In their usual operation, tuna boats receive acoustic information from the buoys onboard via satellite, which they must interpret to determine the presence of tuna. This allows for fine-tuning the search for tuna schools, avoiding unnecessary travel and consequently reducing fuel consumption. The placement and collection locations of these buoys are key to optimizing this operation. Marine Instruments is a leading company in the manufacturing of buoys for tuna fishing. In the context of this company, and due to the importance of these maneuvers, the objective of this project is the development, implementation, and deployment of an algorithm that, using the GPS positions of the boats and buoys, can automatically locate the places where these actions occur. The first part of the project involves preprocessing the GPS position information transmitted by the acoustic buoys and the boats themselves to primarily filter and complement the information that may be useful for the algorithms to be developed. The next stage involves the development of algorithms to detect the target maneuvers. Utilizing knowledge about the typical operation of tuna boats and the maneuvers they perform, the necessary conditions or rules will be obtained to apply to the information from the previous stage and detect when and where the maneuvers occur. Finally, the output information from our algorithm will need to be stored in a database. The objective is to maintain a record of all detected activity, both present and past, being able to use visualization tools to monitor the information and conduct studies on fishing activity. Managing the flow of information, from the moment it is generated by the algorithm, to its database loading and subsequent visualization, will also be a task covered within this project.
Tuna fishing boats use satellite buoys with acoustic probes and GPS to locate tuna schools. These buoys are deployed in the ocean by tuna vessels and support ships. In their usual operation, tuna boats receive acoustic information from the buoys onboard via satellite, which they must interpret to determine the presence of tuna. This allows for fine-tuning the search for tuna schools, avoiding unnecessary travel and consequently reducing fuel consumption. The placement and collection locations of these buoys are key to optimizing this operation. Marine Instruments is a leading company in the manufacturing of buoys for tuna fishing. In the context of this company, and due to the importance of these maneuvers, the objective of this project is the development, implementation, and deployment of an algorithm that, using the GPS positions of the boats and buoys, can automatically locate the places where these actions occur. The first part of the project involves preprocessing the GPS position information transmitted by the acoustic buoys and the boats themselves to primarily filter and complement the information that may be useful for the algorithms to be developed. The next stage involves the development of algorithms to detect the target maneuvers. Utilizing knowledge about the typical operation of tuna boats and the maneuvers they perform, the necessary conditions or rules will be obtained to apply to the information from the previous stage and detect when and where the maneuvers occur. Finally, the output information from our algorithm will need to be stored in a database. The objective is to maintain a record of all detected activity, both present and past, being able to use visualization tools to monitor the information and conduct studies on fishing activity. Managing the flow of information, from the moment it is generated by the algorithm, to its database loading and subsequent visualization, will also be a task covered within this project.
Direction
RIOS VIQUEIRA, JOSE RAMON (Tutorships)
Garaboa Paz, Ángel Daniel (Co-tutorships)
RIOS VIQUEIRA, JOSE RAMON (Tutorships)
Garaboa Paz, Ángel Daniel (Co-tutorships)
Court
PICHEL CAMPOS, JUAN CARLOS (Chairman)
CORES COSTA, DANIEL (Secretary)
Rodeiro Pazos, David (Member)
PICHEL CAMPOS, JUAN CARLOS (Chairman)
CORES COSTA, DANIEL (Secretary)
Rodeiro Pazos, David (Member)
SportPal: an app for creating and finding amateur sports events
Authorship
D.P.S.
Computer Engineering Degree
D.P.S.
Computer Engineering Degree
Defense date
09.12.2024 12:00
09.12.2024 12:00
Summary
This Final Degree Project aims to develop a web application that facilitates the organization of amateur sports matches, removing obstacles such as the lack of participants or the lack of knowledge of locations for practicing a specific sport, thus promoting exercise in an entertaining way.
This Final Degree Project aims to develop a web application that facilitates the organization of amateur sports matches, removing obstacles such as the lack of participants or the lack of knowledge of locations for practicing a specific sport, thus promoting exercise in an entertaining way.
Direction
VARELA PET, JOSE (Tutorships)
VARELA PET, JOSE (Tutorships)
Court
Argüello Pedreira, Francisco Santiago (Chairman)
ALONSO MORAL, JOSE MARIA (Secretary)
MUCIENTES MOLINA, MANUEL FELIPE (Member)
Argüello Pedreira, Francisco Santiago (Chairman)
ALONSO MORAL, JOSE MARIA (Secretary)
MUCIENTES MOLINA, MANUEL FELIPE (Member)
Zero-shot object segmentation in images
Authorship
A.B.R.
Computer Engineering Degree
A.B.R.
Computer Engineering Degree
Defense date
07.19.2024 10:00
07.19.2024 10:00
Summary
Training object detection and segmentation models represents a considerable challenge in the field of computer vision. This process not only requires extensive datasets of annotated images, which may not be available for all classes of objects of interest, but also demands significant computational resources. The need for image annotation, high-performance GPUs, and powerful computer equipment implies a substantial investment in terms of time, economic resources, and human effort that not everyone can address. In this context, the development of models that, after a single initial training, are capable of detecting, segmenting, and classifying any type of object, can be a great advantage. This approach, known as zero-shot panoptic segmentation, seeks to address some of the current limitations of traditional models, which can only classify objects of classes with which they have been trained. To date, this area of research has been relatively unexplored due to its difficulty, both due to the lack of annotated data and the challenge of generalizing to unknown classes. The recently published CONCAT model is currently the only work in this field, and its results suggest that there is still considerable room for improvement. In order to improve the current state of the art in this task, this work proposes a new zero-shot panoptic segmentation algorithm, achieving significantly better results than CONCAT.
Training object detection and segmentation models represents a considerable challenge in the field of computer vision. This process not only requires extensive datasets of annotated images, which may not be available for all classes of objects of interest, but also demands significant computational resources. The need for image annotation, high-performance GPUs, and powerful computer equipment implies a substantial investment in terms of time, economic resources, and human effort that not everyone can address. In this context, the development of models that, after a single initial training, are capable of detecting, segmenting, and classifying any type of object, can be a great advantage. This approach, known as zero-shot panoptic segmentation, seeks to address some of the current limitations of traditional models, which can only classify objects of classes with which they have been trained. To date, this area of research has been relatively unexplored due to its difficulty, both due to the lack of annotated data and the challenge of generalizing to unknown classes. The recently published CONCAT model is currently the only work in this field, and its results suggest that there is still considerable room for improvement. In order to improve the current state of the art in this task, this work proposes a new zero-shot panoptic segmentation algorithm, achieving significantly better results than CONCAT.
Direction
MUCIENTES MOLINA, MANUEL FELIPE (Tutorships)
CORES COSTA, DANIEL (Co-tutorships)
MUCIENTES MOLINA, MANUEL FELIPE (Tutorships)
CORES COSTA, DANIEL (Co-tutorships)
Court
PENA BRAGE, FRANCISCO JOSE (Chairman)
MERA PEREZ, DAVID (Secretary)
CARIÑENA AMIGO, MARIA PURIFICACION (Member)
PENA BRAGE, FRANCISCO JOSE (Chairman)
MERA PEREZ, DAVID (Secretary)
CARIÑENA AMIGO, MARIA PURIFICACION (Member)
Predicting business process times based on machine learning techniques
Authorship
J.A.F.A.
Computer Engineering Degree
J.A.F.A.
Computer Engineering Degree
Defense date
09.12.2024 12:30
09.12.2024 12:30
Summary
This final degree project delves into predictive process monitoring, specifically, the prediction of remaining time. This final degree project proposes a model based on the structure of transformers and applies various modifications to it to enhance its learning capacity of the temporal dependencies between different events, using self-attention and cross-attention mechanisms together. In addition, experiments are carried out with the representation of time to observe which measure is more useful and facilitates learning for the model. All of this is tested with various state-of-the-art models that use techniques such as Long Short-Term Memory (LSTM). After analyzing the results, it is concluded that transformers offer a greater learning capacity of these dependencies than other models such as recurrent neural networks (RNN), generating better models. In addition, the combined use of self-attention with cross-attention, and a temporal representation that indicates the duration of the events obtains the best results for most event logs. This work not only provides a new approach to time prediction in business processes, but also establishes a solid foundation for future research and development in this field.
This final degree project delves into predictive process monitoring, specifically, the prediction of remaining time. This final degree project proposes a model based on the structure of transformers and applies various modifications to it to enhance its learning capacity of the temporal dependencies between different events, using self-attention and cross-attention mechanisms together. In addition, experiments are carried out with the representation of time to observe which measure is more useful and facilitates learning for the model. All of this is tested with various state-of-the-art models that use techniques such as Long Short-Term Memory (LSTM). After analyzing the results, it is concluded that transformers offer a greater learning capacity of these dependencies than other models such as recurrent neural networks (RNN), generating better models. In addition, the combined use of self-attention with cross-attention, and a temporal representation that indicates the duration of the events obtains the best results for most event logs. This work not only provides a new approach to time prediction in business processes, but also establishes a solid foundation for future research and development in this field.
Direction
LAMA PENIN, MANUEL (Tutorships)
VIDAL AGUIAR, JUAN CARLOS (Co-tutorships)
Suárez Barro, Noel (Co-tutorships)
LAMA PENIN, MANUEL (Tutorships)
VIDAL AGUIAR, JUAN CARLOS (Co-tutorships)
Suárez Barro, Noel (Co-tutorships)
Court
ARIAS RODRIGUEZ, JUAN ENRIQUE (Chairman)
VILA BLANCO, NICOLAS (Secretary)
GONZALEZ DIAZ, JULIO (Member)
ARIAS RODRIGUEZ, JUAN ENRIQUE (Chairman)
VILA BLANCO, NICOLAS (Secretary)
GONZALEZ DIAZ, JULIO (Member)
Medical image analysis for disease detection
Authorship
S.S.C.
Computer Engineering Degree
S.S.C.
Computer Engineering Degree
Defense date
07.18.2024 11:00
07.18.2024 11:00
Summary
The new coronavirus (COVID-19), which appeared in the Chinese city of Wuhan in 2019 and rapidly spread worldwide and became a pandemic, caused extensive damage, both human and material. Early detection of cases is therefore essential to prevent the spread of the virus. In this context, machine learning techniques, especially computer vision and convolutional neural networks, are becoming popular for their high accuracy in disease detection and classification. This work aims to analyse medical images through the use of deep learning techniques, specifically convolutional neural networks (CNN). The study focuses on three types of clinical cases: healthy lungs, lungs with pneumonia and lungs with COVID-19. Different network architectures are used, such as VGG16, VGG2 and VGG3, applying various enhancement techniques to optimise the results. The methodology includes the collection and pre-processing of an X-ray image dataset, followed by the implementation and training of CNN models. Techniques used to improve the performance of the networks include data augmentation, dropout, transfer learning and fine tuning. In addition, a web application was created that allows the user to upload an image and receive a classification, indicating whether the image corresponds to a normal lung, COVID-19 or pneumonia. This application facilitates the practical visualisation of the results obtained and their clinical utility. The results obtained show that the binary classification model with VGG16 and transfer learning and fine tuning techniques provides the best results, especially in the context of unbalanced data. The experiments performed demonstrate the effectiveness of the techniques applied to mitigate overfitting and improve model accuracy. The implementation of improvement techniques, such as data augmentation and transfer learning, is crucial to obtain more accurate models.
The new coronavirus (COVID-19), which appeared in the Chinese city of Wuhan in 2019 and rapidly spread worldwide and became a pandemic, caused extensive damage, both human and material. Early detection of cases is therefore essential to prevent the spread of the virus. In this context, machine learning techniques, especially computer vision and convolutional neural networks, are becoming popular for their high accuracy in disease detection and classification. This work aims to analyse medical images through the use of deep learning techniques, specifically convolutional neural networks (CNN). The study focuses on three types of clinical cases: healthy lungs, lungs with pneumonia and lungs with COVID-19. Different network architectures are used, such as VGG16, VGG2 and VGG3, applying various enhancement techniques to optimise the results. The methodology includes the collection and pre-processing of an X-ray image dataset, followed by the implementation and training of CNN models. Techniques used to improve the performance of the networks include data augmentation, dropout, transfer learning and fine tuning. In addition, a web application was created that allows the user to upload an image and receive a classification, indicating whether the image corresponds to a normal lung, COVID-19 or pneumonia. This application facilitates the practical visualisation of the results obtained and their clinical utility. The results obtained show that the binary classification model with VGG16 and transfer learning and fine tuning techniques provides the best results, especially in the context of unbalanced data. The experiments performed demonstrate the effectiveness of the techniques applied to mitigate overfitting and improve model accuracy. The implementation of improvement techniques, such as data augmentation and transfer learning, is crucial to obtain more accurate models.
Direction
GARCIA TAHOCES, PABLO (Tutorships)
GARCIA TAHOCES, PABLO (Tutorships)
Court
Cotos Yáñez, José Manuel (Chairman)
QUESADA BARRIUSO, PABLO (Secretary)
GAGO COUSO, FELIPE (Member)
Cotos Yáñez, José Manuel (Chairman)
QUESADA BARRIUSO, PABLO (Secretary)
GAGO COUSO, FELIPE (Member)
PERIOSTAGE: Application for the annotation of the Periodontogram and calculation of the Stage of Periodontal Disease
Authorship
U.C.T.
Computer Engineering Degree
U.C.T.
Computer Engineering Degree
Defense date
07.19.2024 10:00
07.19.2024 10:00
Summary
Periodontal diseases are a group of chronic conditions that affect the tissues that hold teeth in place. The main way of assessing the state of these oral tissues to diagnose patients and analyse their evolution is by performing periodontograms. The data obtained when performing a periodontogram can be recorded manually or digitally using different tools. However, most of the existing tools for periodontograms have some limitations, such as the need for an Internet connection. This report presents the work carried out for the implementation of a new desktop tool for periodontograms, PERIOSTAGE, which provides a number of significant advantages over existing solutions. It not only allows offline data recording, eliminating the dependence on a network connection, but also provides graphs and visual summaries of the patient's periodontal status. In addition, the application analyses the periodontogram to determine a periodontal diagnosis according to the most widely accepted criteria in the dental community, improving the accuracy of this assessment and facilitating treatment planning accordingly. Finally, all of this is integrated into an intuitive and easy-to-use interface.
Periodontal diseases are a group of chronic conditions that affect the tissues that hold teeth in place. The main way of assessing the state of these oral tissues to diagnose patients and analyse their evolution is by performing periodontograms. The data obtained when performing a periodontogram can be recorded manually or digitally using different tools. However, most of the existing tools for periodontograms have some limitations, such as the need for an Internet connection. This report presents the work carried out for the implementation of a new desktop tool for periodontograms, PERIOSTAGE, which provides a number of significant advantages over existing solutions. It not only allows offline data recording, eliminating the dependence on a network connection, but also provides graphs and visual summaries of the patient's periodontal status. In addition, the application analyses the periodontogram to determine a periodontal diagnosis according to the most widely accepted criteria in the dental community, improving the accuracy of this assessment and facilitating treatment planning accordingly. Finally, all of this is integrated into an intuitive and easy-to-use interface.
Direction
VILA BLANCO, NICOLAS (Tutorships)
Regueira Iglesias, Alba (Co-tutorships)
VILA BLANCO, NICOLAS (Tutorships)
Regueira Iglesias, Alba (Co-tutorships)
Court
BARJA PEREZ, JAVIER (Chairman)
ORDOÑEZ IGLESIAS, ALVARO (Secretary)
MOSQUERA GONZALEZ, ANTONIO (Member)
BARJA PEREZ, JAVIER (Chairman)
ORDOÑEZ IGLESIAS, ALVARO (Secretary)
MOSQUERA GONZALEZ, ANTONIO (Member)
NoTimePrinting: Locator for available and well-functioning printers in a corporate environment with a fleet of multiple printers
Authorship
N.B.Q.
Computer Engineering Degree
N.B.Q.
Computer Engineering Degree
Defense date
07.18.2024 16:00
07.18.2024 16:00
Summary
In corporate environments with multiple printing devices, efficient printer management is essential to reduce wait times, minimize unnecessary movements, and ensure users choose the most suitable printer for their needs. The variety of features among printers can make selecting the right device challenging, leading to ineffective management and reduced productivity. This problem intensifies in large organizations where the lack of clear information about the status and location of available printers can result in suboptimal utilization of printing resources. To address this challenge, this Final Degree Project focuses on the development of NoTimePrinting, a web application designed to optimize printer usage in corporate environments. This solution presents an interactive map that displays the distribution of printers, facilitating visual and accessible management of printing resources. NoTimePrinting also includes an advanced recommendation system that suggests the optimal printer based on parameters such as print job characteristics, user location, or the status and features of printers registered in the system.
In corporate environments with multiple printing devices, efficient printer management is essential to reduce wait times, minimize unnecessary movements, and ensure users choose the most suitable printer for their needs. The variety of features among printers can make selecting the right device challenging, leading to ineffective management and reduced productivity. This problem intensifies in large organizations where the lack of clear information about the status and location of available printers can result in suboptimal utilization of printing resources. To address this challenge, this Final Degree Project focuses on the development of NoTimePrinting, a web application designed to optimize printer usage in corporate environments. This solution presents an interactive map that displays the distribution of printers, facilitating visual and accessible management of printing resources. NoTimePrinting also includes an advanced recommendation system that suggests the optimal printer based on parameters such as print job characteristics, user location, or the status and features of printers registered in the system.
Direction
RIOS VIQUEIRA, JOSE RAMON (Tutorships)
Ménguez Álvarez, Guillermo (Co-tutorships)
Gutierrez Rodríguez, Jorge (Co-tutorships)
RIOS VIQUEIRA, JOSE RAMON (Tutorships)
Ménguez Álvarez, Guillermo (Co-tutorships)
Gutierrez Rodríguez, Jorge (Co-tutorships)
Court
VAZQUEZ CENDON, MARIA ELENA (Chairman)
CHAVES FRAGA, DAVID (Secretary)
SUAREZ GAREA, JORGE ALBERTO (Member)
VAZQUEZ CENDON, MARIA ELENA (Chairman)
CHAVES FRAGA, DAVID (Secretary)
SUAREZ GAREA, JORGE ALBERTO (Member)
PRISMATIC: Machine learning for multimodal cardiac medical image segmentation
Authorship
R.J.S.G.
Computer Engineering Degree
R.J.S.G.
Computer Engineering Degree
Defense date
07.19.2024 18:30
07.19.2024 18:30
Summary
Cardiovascular diseases are currently the main cause of death globally. In research, different topics are being addressed both from the Cardiology Service of the Hospital Clínico Universitario de Santiago (SC-CHUS) and from the Centro Singular de Investigación en Tecnologías Inteligentes (CiTIUS) of the Universidade de Santiago de Compostela. In this context, a common goal of collaboration between both entities arose in the search for a method for the automatic segmentation of the regions of interest of the heart that was fast and reliable to be able to process large amount of data efficiently. In this project, a basic model of convolutional neural network based on the U-Net architecture has been implemented, which has evolved with successive improvements until the final proposal, which is a convolutional neural network based on the U-Net architecture that performs a cascade segmentation in two phases. In the first phase, it works with low-resolution images to cut them to the cube of interest (which encloses the heart) and in the second phase in works with the cut images at a higher resolution to segment the heart into 7 substructures: the 4 chambers, the myocardium of the left ventricle and the great vessels. To train the model, since SC-CHUS does not have expert annotated data, images from the Multi-Modality Whole Heart Segmentation (MM-WHS) competition were used. The proposal of this project has also been evaluated with the criteria of the competition, obtaining better results than those of the methods that make up the state of the art but with the advantage of being less computationally complex, which means that it requires less inference time and has the capacity to be used even if powerful computing resources are not available. The final proposal has been validated with images from clinical practice provided by the SC-CHUS, which are very different from those of the MM-WHS. Even so, very satisfactory results have been achieved with correct segmentations in most cases and with a computational complexity that can be assumed in portable computers.
Cardiovascular diseases are currently the main cause of death globally. In research, different topics are being addressed both from the Cardiology Service of the Hospital Clínico Universitario de Santiago (SC-CHUS) and from the Centro Singular de Investigación en Tecnologías Inteligentes (CiTIUS) of the Universidade de Santiago de Compostela. In this context, a common goal of collaboration between both entities arose in the search for a method for the automatic segmentation of the regions of interest of the heart that was fast and reliable to be able to process large amount of data efficiently. In this project, a basic model of convolutional neural network based on the U-Net architecture has been implemented, which has evolved with successive improvements until the final proposal, which is a convolutional neural network based on the U-Net architecture that performs a cascade segmentation in two phases. In the first phase, it works with low-resolution images to cut them to the cube of interest (which encloses the heart) and in the second phase in works with the cut images at a higher resolution to segment the heart into 7 substructures: the 4 chambers, the myocardium of the left ventricle and the great vessels. To train the model, since SC-CHUS does not have expert annotated data, images from the Multi-Modality Whole Heart Segmentation (MM-WHS) competition were used. The proposal of this project has also been evaluated with the criteria of the competition, obtaining better results than those of the methods that make up the state of the art but with the advantage of being less computationally complex, which means that it requires less inference time and has the capacity to be used even if powerful computing resources are not available. The final proposal has been validated with images from clinical practice provided by the SC-CHUS, which are very different from those of the MM-WHS. Even so, very satisfactory results have been achieved with correct segmentations in most cases and with a computational complexity that can be assumed in portable computers.
Direction
VILA BLANCO, NICOLAS (Tutorships)
CARREIRA NOUCHE, MARIA JOSE (Co-tutorships)
NUÑEZ GARCIA, MARTA (Co-tutorships)
VILA BLANCO, NICOLAS (Tutorships)
CARREIRA NOUCHE, MARIA JOSE (Co-tutorships)
NUÑEZ GARCIA, MARTA (Co-tutorships)
Court
Barro Ameneiro, Senén (Chairman)
TOBAR QUINTANAR, ALEJANDRO JOSE (Secretary)
LADRA GONZALEZ, MANUEL EULOGIO (Member)
Barro Ameneiro, Senén (Chairman)
TOBAR QUINTANAR, ALEJANDRO JOSE (Secretary)
LADRA GONZALEZ, MANUEL EULOGIO (Member)
Isopropyl alcohol production plant from acetone
Authorship
A.E.G.R.
Bachelor's Degree in Chemical Engeneering
A.E.G.R.
Bachelor's Degree in Chemical Engeneering
Defense date
07.16.2024 09:15
07.16.2024 09:15
Summary
Isopropyl alcohol (IPA) is a product in progressive growth in the world market, both in the medium and long term. Among the many uses of this compound, its use in the pharmaceutical, cosmetics, cleaning, automotive and industrial sectors stands out. The process carried out consists of the production of IPA from acetone. A hydrogenation reaction is carried out, in which the reactants are acetone (liquid phase) and hydrogen and the products are isopropyl alcohol, propene and water. The IPA is then separated and is obtained in a very pure concentration. To take advantage of the by-products, hydrogen is recirculated, and a fuel is also obtained from the by-products obtained. The plant would be designed to obtain 97,000 tons of IPA at 97% purity and 29,000 tons of azeotropic IPA. It should be noted that this year (2023) the construction of an IPA production plant from renewable materials has been confirmed in Huelva, Spain. The company carrying out this project is CEPSA and the plant will be operational from 2025. Finally, the distribution of the equipment design will be the reactor with Javier Bermudez and the distillation column for Alicia Gomez.
Isopropyl alcohol (IPA) is a product in progressive growth in the world market, both in the medium and long term. Among the many uses of this compound, its use in the pharmaceutical, cosmetics, cleaning, automotive and industrial sectors stands out. The process carried out consists of the production of IPA from acetone. A hydrogenation reaction is carried out, in which the reactants are acetone (liquid phase) and hydrogen and the products are isopropyl alcohol, propene and water. The IPA is then separated and is obtained in a very pure concentration. To take advantage of the by-products, hydrogen is recirculated, and a fuel is also obtained from the by-products obtained. The plant would be designed to obtain 97,000 tons of IPA at 97% purity and 29,000 tons of azeotropic IPA. It should be noted that this year (2023) the construction of an IPA production plant from renewable materials has been confirmed in Huelva, Spain. The company carrying out this project is CEPSA and the plant will be operational from 2025. Finally, the distribution of the equipment design will be the reactor with Javier Bermudez and the distillation column for Alicia Gomez.
Direction
Rodríguez Figueiras, Óscar (Tutorships)
Rodríguez Figueiras, Óscar (Tutorships)
Court
HOSPIDO QUINTANA, ALMUDENA (Chairman)
MAURICIO IGLESIAS, MIGUEL (Secretary)
GONZALEZ GARCIA, SARA (Member)
HOSPIDO QUINTANA, ALMUDENA (Chairman)
MAURICIO IGLESIAS, MIGUEL (Secretary)
GONZALEZ GARCIA, SARA (Member)
Ethyl acetate production plant by esterification of acetic acid with ethanol
Authorship
C.F.V.
Bachelor's Degree in Chemical Engeneering
C.F.V.
Bachelor's Degree in Chemical Engeneering
Defense date
07.16.2024 09:55
07.16.2024 09:55
Summary
Ethyl acetate is a chemical compound used as a solvent in a large number of industrial processes, as well as an additive in both the food and cosmetic sectors. The plant's capacity is 25,000 tons per year since production in Spain is not of high tonnage. Ethyl acetate is produced through the esterification reaction between acetic acid and ethanol, with water obtained as a byproduct of this reaction. Additionally, as a sustainable alternative, the option of using bioethanol as a raw material is proposed. Joel Fernández García is responsible for designing the heterogeneous catalytic reactor, while Carlos Fuentes Varela designs the distillation column where the product is obtained. The separation train consists of three distillation columns and one liquid-liquid extraction unit.
Ethyl acetate is a chemical compound used as a solvent in a large number of industrial processes, as well as an additive in both the food and cosmetic sectors. The plant's capacity is 25,000 tons per year since production in Spain is not of high tonnage. Ethyl acetate is produced through the esterification reaction between acetic acid and ethanol, with water obtained as a byproduct of this reaction. Additionally, as a sustainable alternative, the option of using bioethanol as a raw material is proposed. Joel Fernández García is responsible for designing the heterogeneous catalytic reactor, while Carlos Fuentes Varela designs the distillation column where the product is obtained. The separation train consists of three distillation columns and one liquid-liquid extraction unit.
Direction
SOTO CAMPOS, ANA MARIA (Tutorships)
SOTO CAMPOS, ANA MARIA (Tutorships)
Court
ROCA BORDELLO, ENRIQUE (Chairman)
SANCHEZ FERNANDEZ, ADRIAN (Secretary)
MOREIRA VILAR, MARIA TERESA (Member)
ROCA BORDELLO, ENRIQUE (Chairman)
SANCHEZ FERNANDEZ, ADRIAN (Secretary)
MOREIRA VILAR, MARIA TERESA (Member)
Acetaldehyde production plant through ethanol dehydrogenation
Authorship
M.V.R.
Bachelor's Degree in Chemical Engeneering
M.V.R.
Bachelor's Degree in Chemical Engeneering
Defense date
09.13.2024 11:00
09.13.2024 11:00
Summary
In this work, the design of an acetaldehyde production plant was carried out, using the ethanol dehydrogenation method. Starting from ethanol, the endothermic dehydrogenation reaction occurs, giving rise to acetaldehyde and hydrogen.The process consists of a section where the raw material is conditioned to the reaction conditions, then the section where the products are separated and purified, and the compounds that are not desired to be eliminated are also recirculated.
In this work, the design of an acetaldehyde production plant was carried out, using the ethanol dehydrogenation method. Starting from ethanol, the endothermic dehydrogenation reaction occurs, giving rise to acetaldehyde and hydrogen.The process consists of a section where the raw material is conditioned to the reaction conditions, then the section where the products are separated and purified, and the compounds that are not desired to be eliminated are also recirculated.
Direction
RODIL RODRIGUEZ, EVA (Tutorships)
RODIL RODRIGUEZ, EVA (Tutorships)
Court
GIL GONZALEZ, ALVARO (Chairman)
FRANCO URIA, MARIA AMAYA (Secretary)
GOMEZ DIAZ, DIEGO (Member)
GIL GONZALEZ, ALVARO (Chairman)
FRANCO URIA, MARIA AMAYA (Secretary)
GOMEZ DIAZ, DIEGO (Member)
Isopropyl alcohol production plant from acetone
Authorship
J.B.B.
Bachelor's Degree in Chemical Engeneering
J.B.B.
Bachelor's Degree in Chemical Engeneering
Defense date
07.16.2024 09:15
07.16.2024 09:15
Summary
Isopropyl alcohol (IPA) is a product in progressive growth in the world market, both in the medium and long term. Among the many uses of this compound, its use in the pharmaceutical, cosmetics, cleaning, automotive and industrial sectors stands out. pharmaceutical, cosmetic, cleaning, automotive and industrial sectors. The process carried out consists of the production of IPA from acetone. A hydrogenation reaction is carried out, in which the reagents are acetone (liquid phase) and hydrogen and the products are isopropyl alcohol, isopropyl alcohol and hydrogen. products are isopropyl alcohol, propene and water. The IPA is then separated and is obtained in a very pure concentration. To take advantage of the by-products, hydrogen is recirculated and a fuel is obtained from the by-products. also a fuel is obtained from the by-products obtained. The plant would be designed to obtain 97,000 tons of IPA at 97% purity and 29,000 tons of azeotropic IPA. It should be noted that this year (2023) This year (2023) has confirmed the construction of a plant to produce IPA from renewable materials in Huelva, Spain. The company carrying out this project is CEPSA and the plant will be operational from 2025. Finally, the distribution of the equipment design will be the reactor with Javier Bermudez and the distillation column for Alicia Gomez.
Isopropyl alcohol (IPA) is a product in progressive growth in the world market, both in the medium and long term. Among the many uses of this compound, its use in the pharmaceutical, cosmetics, cleaning, automotive and industrial sectors stands out. pharmaceutical, cosmetic, cleaning, automotive and industrial sectors. The process carried out consists of the production of IPA from acetone. A hydrogenation reaction is carried out, in which the reagents are acetone (liquid phase) and hydrogen and the products are isopropyl alcohol, isopropyl alcohol and hydrogen. products are isopropyl alcohol, propene and water. The IPA is then separated and is obtained in a very pure concentration. To take advantage of the by-products, hydrogen is recirculated and a fuel is obtained from the by-products. also a fuel is obtained from the by-products obtained. The plant would be designed to obtain 97,000 tons of IPA at 97% purity and 29,000 tons of azeotropic IPA. It should be noted that this year (2023) This year (2023) has confirmed the construction of a plant to produce IPA from renewable materials in Huelva, Spain. The company carrying out this project is CEPSA and the plant will be operational from 2025. Finally, the distribution of the equipment design will be the reactor with Javier Bermudez and the distillation column for Alicia Gomez.
Direction
Rodríguez Figueiras, Óscar (Tutorships)
Rodríguez Figueiras, Óscar (Tutorships)
Court
HOSPIDO QUINTANA, ALMUDENA (Chairman)
MAURICIO IGLESIAS, MIGUEL (Secretary)
GONZALEZ GARCIA, SARA (Member)
HOSPIDO QUINTANA, ALMUDENA (Chairman)
MAURICIO IGLESIAS, MIGUEL (Secretary)
GONZALEZ GARCIA, SARA (Member)
Methylamines production process by methanol and ammonia.
Authorship
A.D.G.
Bachelor's Degree in Chemical Engeneering
A.D.G.
Bachelor's Degree in Chemical Engeneering
Defense date
09.13.2024 10:10
09.13.2024 10:10
Summary
This project has as a goal the design of methylamine production process by methanol and ammonia. The capacity will be 60.000 T/year of dimethylamine and methylamina, considering continuous operating regime during 330 days and 24 h/day. It is split in two sections. Section 100 includes the conditioning of raw materials and the reaction of reactor R-101 (designed by Alex Dacosta García). Section 200 separates and purifies the product, obtaining 98% pure dimethylamine and 95% pure methylamine. The destillation column will be designed by Leire Gallego Portabales. There will also be taken into account security, environmental and economic viability studies.
This project has as a goal the design of methylamine production process by methanol and ammonia. The capacity will be 60.000 T/year of dimethylamine and methylamina, considering continuous operating regime during 330 days and 24 h/day. It is split in two sections. Section 100 includes the conditioning of raw materials and the reaction of reactor R-101 (designed by Alex Dacosta García). Section 200 separates and purifies the product, obtaining 98% pure dimethylamine and 95% pure methylamine. The destillation column will be designed by Leire Gallego Portabales. There will also be taken into account security, environmental and economic viability studies.
Direction
FRANCO URIA, MARIA AMAYA (Tutorships)
SINEIRO TORRES, JORGE (Co-tutorships)
FRANCO URIA, MARIA AMAYA (Tutorships)
SINEIRO TORRES, JORGE (Co-tutorships)
Court
GARRIDO FERNANDEZ, JUAN MANUEL (Chairman)
FRANCO RUIZ, DANIEL JOSE (Secretary)
FREIRE LEIRA, MARIA SONIA (Member)
GARRIDO FERNANDEZ, JUAN MANUEL (Chairman)
FRANCO RUIZ, DANIEL JOSE (Secretary)
FREIRE LEIRA, MARIA SONIA (Member)
Hydrazine hydrate production plant
Authorship
M.M.F.
Bachelor's Degree in Chemical Engeneering
M.M.F.
Bachelor's Degree in Chemical Engeneering
Defense date
07.16.2024 11:10
07.16.2024 11:10
Summary
The objective of this project is to develop a plant for the production of hydrazine hydrate with a mass concentration of 64 percent from ammonia and sodium hypochlorite, both in solution, using the Raschig method. The design is approached with the purpose of producing 12000T/year of product operating continuously, 330 days a year, 24 hours a day. The rigorous design of the most relevant equipment of the process is described, being the hydrazine reactor and the distillation column to obtain the product. It has been decided to address the design of a hydrazine hydrate production plant because of its current research in the energy field for its use as chemical storage of hydrogen in fuel cells due to its high hydrogen content, ease of transport and storage, in addition to the fact that the only by-product of its decomposition would be nitrogen. This is a study of current importance, since it offers an alternative to fossil fuels. On the other hand, hydrazine hydrate is currently used in various industrial fields; due to its ability to eliminate oxygen from water, it is used in high-pressure steam generators and industrial boilers, but it also has other uses, such as foaming agent for plastic foams, polymerization initiators, pesticides, pharmaceutical products, photographic chemical products or dyes. Derivatives of this component also have a great place in today's industry, such as, for example, hydrazine anhydrous, which is synthesized by the Olin Raschig method and is notable for its use as a rocket propellant. The facility is located in the north of Spain, more specifically in the industrial park of Lloreda, belonging to the Asturian city of Gijón, with the aim of facilitating the commercialization of hydrazine hydrate by sea to northern Europe and other commercial areas of interest. This document represents the Final Degree Project by means of which the students Mara Meitín Fraga and Xeila Liñayo Vicente opt for the university degree in Chemical Engineering awarded by the Escola Técnica Superior de Enxeñaría of the Universidade de Santiago de Compostela.
The objective of this project is to develop a plant for the production of hydrazine hydrate with a mass concentration of 64 percent from ammonia and sodium hypochlorite, both in solution, using the Raschig method. The design is approached with the purpose of producing 12000T/year of product operating continuously, 330 days a year, 24 hours a day. The rigorous design of the most relevant equipment of the process is described, being the hydrazine reactor and the distillation column to obtain the product. It has been decided to address the design of a hydrazine hydrate production plant because of its current research in the energy field for its use as chemical storage of hydrogen in fuel cells due to its high hydrogen content, ease of transport and storage, in addition to the fact that the only by-product of its decomposition would be nitrogen. This is a study of current importance, since it offers an alternative to fossil fuels. On the other hand, hydrazine hydrate is currently used in various industrial fields; due to its ability to eliminate oxygen from water, it is used in high-pressure steam generators and industrial boilers, but it also has other uses, such as foaming agent for plastic foams, polymerization initiators, pesticides, pharmaceutical products, photographic chemical products or dyes. Derivatives of this component also have a great place in today's industry, such as, for example, hydrazine anhydrous, which is synthesized by the Olin Raschig method and is notable for its use as a rocket propellant. The facility is located in the north of Spain, more specifically in the industrial park of Lloreda, belonging to the Asturian city of Gijón, with the aim of facilitating the commercialization of hydrazine hydrate by sea to northern Europe and other commercial areas of interest. This document represents the Final Degree Project by means of which the students Mara Meitín Fraga and Xeila Liñayo Vicente opt for the university degree in Chemical Engineering awarded by the Escola Técnica Superior de Enxeñaría of the Universidade de Santiago de Compostela.
Direction
MOREIRA MARTINEZ, RAMON FELIPE (Tutorships)
GOMEZ DIAZ, DIEGO (Co-tutorships)
MOREIRA MARTINEZ, RAMON FELIPE (Tutorships)
GOMEZ DIAZ, DIEGO (Co-tutorships)
Court
MOSQUERA CORRAL, ANUSKA (Chairman)
SINEIRO TORRES, JORGE (Secretary)
VIDAL TATO, MARIA ISABEL (Member)
MOSQUERA CORRAL, ANUSKA (Chairman)
SINEIRO TORRES, JORGE (Secretary)
VIDAL TATO, MARIA ISABEL (Member)
Production Plant for Drying Oil from Waste Cooking Oils
Authorship
J.M.R.
Bachelor's Degree in Chemical Engeneering
J.M.R.
Bachelor's Degree in Chemical Engeneering
Defense date
07.16.2024 12:30
07.16.2024 12:30
Summary
Drying oil is used as a finish for paints and varnishes. In this process, waste cooking oils are used in a process that involves the use of a reactor (which will be designed in detail) and several separation equipment such as distillation columns. Waste cooking oils are introduced into the process, from which palmitic acid is extracted and fed into the process battery to ultimately form a product with good drying properties. This process involves typical equipment in the field of Chemical Engineering at high pressures and temperatures to meet the proposed plant capacity of 20,000 tons/year.
Drying oil is used as a finish for paints and varnishes. In this process, waste cooking oils are used in a process that involves the use of a reactor (which will be designed in detail) and several separation equipment such as distillation columns. Waste cooking oils are introduced into the process, from which palmitic acid is extracted and fed into the process battery to ultimately form a product with good drying properties. This process involves typical equipment in the field of Chemical Engineering at high pressures and temperatures to meet the proposed plant capacity of 20,000 tons/year.
Direction
MOREIRA MARTINEZ, RAMON FELIPE (Tutorships)
FRANCO RUIZ, DANIEL JOSE (Co-tutorships)
MOREIRA MARTINEZ, RAMON FELIPE (Tutorships)
FRANCO RUIZ, DANIEL JOSE (Co-tutorships)
Court
MOSQUERA CORRAL, ANUSKA (Chairman)
SINEIRO TORRES, JORGE (Secretary)
VIDAL TATO, MARIA ISABEL (Member)
MOSQUERA CORRAL, ANUSKA (Chairman)
SINEIRO TORRES, JORGE (Secretary)
VIDAL TATO, MARIA ISABEL (Member)
Production plant of dimethyl ether from the catalytic dehydration of methanol
Authorship
S.M.T.
Bachelor's Degree in Chemical Engeneering
S.M.T.
Bachelor's Degree in Chemical Engeneering
Defense date
07.16.2024 09:15
07.16.2024 09:15
Summary
The objective of this project is to design a production plant for dimethyl ether (DME) from the catalytic dehydration of methanol. Its production capacity covers 50,000 tons annually at 99.95% purity by weight, operating continuously for 24 hours a day, 330 days a year, with maintenance shutdowns accounted for. Dimethyl ether is currently a product of interest as it is a competitive substitute for existing fuels. It is a clean gas, easily liquefiable, and easy to transport. It has potential in various markets, notably in power generation, as a substitute for liquefied petroleum gas, and as an automotive fuel due to its high cetane number and lack of sulfur oxide emissions. Its production is carried out in the R-201 multitubular catalytic reactor, which operates isothermally at 320ºC, producing dimethyl ether and water. This equipment is designed by Miguel Díaz Abalo. Next, in a train consisting of two distillation columns arranged in a direct sequence, the separation and purification of the product of interest take place, as well as the recirculation of the unreacted raw material. The first distillation column, designated T-301, is the equipment through which dimethyl ether is obtained at the required purity and flow conditions, and it is designed by Saleta Martínez Tella.
The objective of this project is to design a production plant for dimethyl ether (DME) from the catalytic dehydration of methanol. Its production capacity covers 50,000 tons annually at 99.95% purity by weight, operating continuously for 24 hours a day, 330 days a year, with maintenance shutdowns accounted for. Dimethyl ether is currently a product of interest as it is a competitive substitute for existing fuels. It is a clean gas, easily liquefiable, and easy to transport. It has potential in various markets, notably in power generation, as a substitute for liquefied petroleum gas, and as an automotive fuel due to its high cetane number and lack of sulfur oxide emissions. Its production is carried out in the R-201 multitubular catalytic reactor, which operates isothermally at 320ºC, producing dimethyl ether and water. This equipment is designed by Miguel Díaz Abalo. Next, in a train consisting of two distillation columns arranged in a direct sequence, the separation and purification of the product of interest take place, as well as the recirculation of the unreacted raw material. The first distillation column, designated T-301, is the equipment through which dimethyl ether is obtained at the required purity and flow conditions, and it is designed by Saleta Martínez Tella.
Direction
SOTO CAMPOS, ANA MARIA (Tutorships)
SOTO CAMPOS, ANA MARIA (Tutorships)
Court
ROCA BORDELLO, ENRIQUE (Chairman)
SANCHEZ FERNANDEZ, ADRIAN (Secretary)
MOREIRA VILAR, MARIA TERESA (Member)
ROCA BORDELLO, ENRIQUE (Chairman)
SANCHEZ FERNANDEZ, ADRIAN (Secretary)
MOREIRA VILAR, MARIA TERESA (Member)
Automatic stock control in warehouses using cameras
Authorship
A.G.F.
Computer Engineering Degree
A.G.F.
Computer Engineering Degree
Defense date
07.19.2024 18:00
07.19.2024 18:00
Summary
Currently, artificial intelligence is making it possible to solve problems and automate processes that, until now, required human labor. In this work, the problem of stock detection in warehouses will be addressed from the field of artificial intelligence and automation. Manual or semi-automatic methods are the most used for inventory control, but in this work a system based on image processing will be developed that allows situations of stock breakage or replacement to be detected in an automated manner, providing a more efficient and precise solution. To achieve this, this problem will be divided into different smaller challenges that will be addressed sequentially, finally building a complete system that will be able to detect movement, crop areas of interest in the images using ArUco markers, transmit the data efficiently using Kafka and classify the images using the CLIP model to monitor these inventory changes. Finally, the performance of this system will be evaluated to determine its viability as a real alternative to traditional inventory management methods used in warehouses.
Currently, artificial intelligence is making it possible to solve problems and automate processes that, until now, required human labor. In this work, the problem of stock detection in warehouses will be addressed from the field of artificial intelligence and automation. Manual or semi-automatic methods are the most used for inventory control, but in this work a system based on image processing will be developed that allows situations of stock breakage or replacement to be detected in an automated manner, providing a more efficient and precise solution. To achieve this, this problem will be divided into different smaller challenges that will be addressed sequentially, finally building a complete system that will be able to detect movement, crop areas of interest in the images using ArUco markers, transmit the data efficiently using Kafka and classify the images using the CLIP model to monitor these inventory changes. Finally, the performance of this system will be evaluated to determine its viability as a real alternative to traditional inventory management methods used in warehouses.
Direction
CORES COSTA, DANIEL (Tutorships)
MUCIENTES MOLINA, MANUEL FELIPE (Co-tutorships)
CORES COSTA, DANIEL (Tutorships)
MUCIENTES MOLINA, MANUEL FELIPE (Co-tutorships)
Court
Fernández Pena, Anselmo Tomás (Chairman)
SACO LOPEZ, PEDRO JOSE (Secretary)
RODRIGUEZ PRESEDO, JESUS MARIA (Member)
Fernández Pena, Anselmo Tomás (Chairman)
SACO LOPEZ, PEDRO JOSE (Secretary)
RODRIGUEZ PRESEDO, JESUS MARIA (Member)
Implementation of Classifiers for Stress Recognition Based on Biosignals
Authorship
M.C.L.
Computer Engineering Degree
M.C.L.
Computer Engineering Degree
Defense date
07.19.2024 09:30
07.19.2024 09:30
Summary
Stress has become an increasingly prevalent illness, and along with the growing concern of contemporary society about mental health issues, it has become of great interest both to detect it and to study its causes. In this context, the following work has focused on the detection of stress through machine learning and deep learning techniques applied to physiological signals, specifically electrodermal activity (EDA) signals. To approach this research, an analysis of the current state of the art was conducted, replicating previous studies to understand and then improve them. Two public datasets were used: WESAD and AMIGOS, which provide records of physiological signals in various contexts and through different experiments. The methodology included data preprocessing, extraction of relevant features, and implementation of various classification models, both machine learning (SVM, Random Forest) and deep learning (neural networks). An advanced model combining Convolutional Neural Networks (CNN) with Long Short-Term Memory (LSTM) was developed, exploring the impact of modifying their configuration parameters, resulting in high accuracy. Additionally, validation techniques were enhanced by combining cross-validation and the leave-one-subject-out technique. The experimental results demonstrated that deep learning models significantly outperform machine learning models in terms of accuracy, achieving up to 98% precision with LSTMs. The generalization capability of the trained models was also evaluated using both datasets, observing superior consistency in deep learning models compared to notable disparities in machine learning models. These findings underscore the effectiveness of deep learning techniques in capturing complex relationships in physiological signals and improving stress detection.
Stress has become an increasingly prevalent illness, and along with the growing concern of contemporary society about mental health issues, it has become of great interest both to detect it and to study its causes. In this context, the following work has focused on the detection of stress through machine learning and deep learning techniques applied to physiological signals, specifically electrodermal activity (EDA) signals. To approach this research, an analysis of the current state of the art was conducted, replicating previous studies to understand and then improve them. Two public datasets were used: WESAD and AMIGOS, which provide records of physiological signals in various contexts and through different experiments. The methodology included data preprocessing, extraction of relevant features, and implementation of various classification models, both machine learning (SVM, Random Forest) and deep learning (neural networks). An advanced model combining Convolutional Neural Networks (CNN) with Long Short-Term Memory (LSTM) was developed, exploring the impact of modifying their configuration parameters, resulting in high accuracy. Additionally, validation techniques were enhanced by combining cross-validation and the leave-one-subject-out technique. The experimental results demonstrated that deep learning models significantly outperform machine learning models in terms of accuracy, achieving up to 98% precision with LSTMs. The generalization capability of the trained models was also evaluated using both datasets, observing superior consistency in deep learning models compared to notable disparities in machine learning models. These findings underscore the effectiveness of deep learning techniques in capturing complex relationships in physiological signals and improving stress detection.
Direction
CONDORI FERNANDEZ, NELLY (Tutorships)
CATALA BOLOS, ALEJANDRO (Co-tutorships)
CONDORI FERNANDEZ, NELLY (Tutorships)
CATALA BOLOS, ALEJANDRO (Co-tutorships)
Court
BARJA PEREZ, JAVIER (Chairman)
ORDOÑEZ IGLESIAS, ALVARO (Secretary)
MOSQUERA GONZALEZ, ANTONIO (Member)
BARJA PEREZ, JAVIER (Chairman)
ORDOÑEZ IGLESIAS, ALVARO (Secretary)
MOSQUERA GONZALEZ, ANTONIO (Member)
galerIA: Web application for multimedia gallery with machine learning functions
Authorship
L.M.A.S.
Computer Engineering Degree
L.M.A.S.
Computer Engineering Degree
Defense date
07.19.2024 09:30
07.19.2024 09:30
Summary
galerIA is a web application that implements a multimedia gallery where users can store and organize their images and videos. The main difference from other similar applications is that galerIA implements innovative functionalities based on machine learning models on the client side. Users, in a completely transparent way and without the need for technical knowledge, can benefit from applying super-resolution to their images, filtering them by the objects that appear in them, or generating new images, without the need for a dedicated server to do so, minimizing the cost of serving the application and the required bandwidth at the expense of higher resource consumption on the client device.
galerIA is a web application that implements a multimedia gallery where users can store and organize their images and videos. The main difference from other similar applications is that galerIA implements innovative functionalities based on machine learning models on the client side. Users, in a completely transparent way and without the need for technical knowledge, can benefit from applying super-resolution to their images, filtering them by the objects that appear in them, or generating new images, without the need for a dedicated server to do so, minimizing the cost of serving the application and the required bandwidth at the expense of higher resource consumption on the client device.
Direction
MUCIENTES MOLINA, MANUEL FELIPE (Tutorships)
CORES COSTA, DANIEL (Co-tutorships)
MUCIENTES MOLINA, MANUEL FELIPE (Tutorships)
CORES COSTA, DANIEL (Co-tutorships)
Court
PENA BRAGE, FRANCISCO JOSE (Chairman)
MERA PEREZ, DAVID (Secretary)
CARIÑENA AMIGO, MARIA PURIFICACION (Member)
PENA BRAGE, FRANCISCO JOSE (Chairman)
MERA PEREZ, DAVID (Secretary)
CARIÑENA AMIGO, MARIA PURIFICACION (Member)
Identification of hypertrophic cardiomyopathy from electrocardiographic recording using machine learning
Authorship
I.G.P.
Computer Engineering Degree
I.G.P.
Computer Engineering Degree
Defense date
07.18.2024 18:00
07.18.2024 18:00
Summary
Hypertrophic cardiomyopathy is a relatively common genetic disease with a very heterogeneous expression. Although most patients can lead a normal or nearly normal life, a small number of patients are at very high risk of ventricular arrhythmia and sudden death. Some of the reference tests for evaluating the risk of sudden death in these patients are cardiac magnetic resonance or echocardiography. This work raises the possibility of identifying new risk predictors from the electrocardiographic test, whose low cost and wide availability in the service portfolio of public health systems would ensure greater accessibility for the general population. To do this, machine learning techniques will be put into play for the development of a prediction model, as well as variable selection techniques for better interpretability of the results. A database compiled by the Cardiology Service of the Hospital Princesa de Madrid and annotated by hand by expert cardiologists will be used.
Hypertrophic cardiomyopathy is a relatively common genetic disease with a very heterogeneous expression. Although most patients can lead a normal or nearly normal life, a small number of patients are at very high risk of ventricular arrhythmia and sudden death. Some of the reference tests for evaluating the risk of sudden death in these patients are cardiac magnetic resonance or echocardiography. This work raises the possibility of identifying new risk predictors from the electrocardiographic test, whose low cost and wide availability in the service portfolio of public health systems would ensure greater accessibility for the general population. To do this, machine learning techniques will be put into play for the development of a prediction model, as well as variable selection techniques for better interpretability of the results. A database compiled by the Cardiology Service of the Hospital Princesa de Madrid and annotated by hand by expert cardiologists will be used.
Direction
FELIX LAMAS, PAULO (Tutorships)
RODRIGUEZ PRESEDO, JESUS MARIA (Co-tutorships)
FELIX LAMAS, PAULO (Tutorships)
RODRIGUEZ PRESEDO, JESUS MARIA (Co-tutorships)
Court
Pardo López, Xosé Manuel (Chairman)
Puentes Cociña, Beltrán (Secretary)
LOPEZ FANDIÑO, JAVIER (Member)
Pardo López, Xosé Manuel (Chairman)
Puentes Cociña, Beltrán (Secretary)
LOPEZ FANDIÑO, JAVIER (Member)
Inventory Management Module in Chemical Research Laboratories
Authorship
C.T.C.
Computer Engineering Degree
C.T.C.
Computer Engineering Degree
Defense date
02.16.2024 10:00
02.16.2024 10:00
Summary
Currently, in a chemical research laboratory, there is a need for an alternative to the traditional paper notebook where chemists record their experiments. This is because there is a desire to improve aspects such as collaboration, organization, and information retrieval in the laboratory. Therefore, this Bachelor's Thesis involves the design and implementation of a functional backend module that allows for the management of a chemical laboratory inventory, including the handling of consumables and the chemical compounds that constitute it. Regarding the compounds, the management of their safety information and the procedures for controlling hazardous substances to health (COSHH in English) is also required. The testing plan to evaluate the module, including unit tests and tests on the API, indicates that it meets the stated requirements and demonstrates its proper functioning. Finally, it is important to note that the completion of this work represents a contribution to the development of applications used in the fields of chemical, pharmaceutical, and academic research. Its use brings improvements in speed, organization, and record-keeping within a chemical laboratory.
Currently, in a chemical research laboratory, there is a need for an alternative to the traditional paper notebook where chemists record their experiments. This is because there is a desire to improve aspects such as collaboration, organization, and information retrieval in the laboratory. Therefore, this Bachelor's Thesis involves the design and implementation of a functional backend module that allows for the management of a chemical laboratory inventory, including the handling of consumables and the chemical compounds that constitute it. Regarding the compounds, the management of their safety information and the procedures for controlling hazardous substances to health (COSHH in English) is also required. The testing plan to evaluate the module, including unit tests and tests on the API, indicates that it meets the stated requirements and demonstrates its proper functioning. Finally, it is important to note that the completion of this work represents a contribution to the development of applications used in the fields of chemical, pharmaceutical, and academic research. Its use brings improvements in speed, organization, and record-keeping within a chemical laboratory.
Direction
TABOADA IGLESIAS, MARÍA JESÚS (Tutorships)
FRAGA CASTRO, SANTIAGO (Co-tutorships)
Lalín Rodríguez, María del Rosario (Co-tutorships)
TABOADA IGLESIAS, MARÍA JESÚS (Tutorships)
FRAGA CASTRO, SANTIAGO (Co-tutorships)
Lalín Rodríguez, María del Rosario (Co-tutorships)
Court
SUAREZ GAREA, JORGE ALBERTO (Coordinator)
BUGARIN DIZ, ALBERTO JOSE (Chairman)
VILLARROYA FERNANDEZ, SEBASTIAN (Secretary)
BORRAJO GARCIA, MARIA ISABEL (Member)
SUAREZ GAREA, JORGE ALBERTO (Coordinator)
BUGARIN DIZ, ALBERTO JOSE (Chairman)
VILLARROYA FERNANDEZ, SEBASTIAN (Secretary)
BORRAJO GARCIA, MARIA ISABEL (Member)
Information Retrieval v3: Local Processing
Authorship
R.D.L.I.R.
Computer Engineering Degree
R.D.L.I.R.
Computer Engineering Degree
Defense date
07.18.2024 17:00
07.18.2024 17:00
Summary
In the current technological era, the demands for processing and transmitting multimedia content are increasing. The vast amount of existing data necessitates efficient systems for information extraction, digitization, storage, and ultimately retrieval. The latter point is crucial as new technologies increasingly demand speed and effectiveness in utilizing all this information. This project aims to emphasize the idea of retrieving images from databases, specifically from storage systems where all this information is kept. Specifically, the development of a tool is proposed that allows retrieving a similarity ranking between images based on their descriptors. These descriptors are sets of numerical values that reflect characteristics of an image such as color or texture. Furthermore, the functionality is proposed to compute these descriptors locally, i.e., on a specific part of the image. This approach gives users more options when retrieving information. This comparison will be carried out between images in a database and a query image chosen by the user. Thus, the tool also integrates all corresponding database management actions, such as adding or deleting a database, or adding or removing an image within a database. Throughout this project, the knowledge acquired up to this point, combined with new knowledge gained during its development, will be applied to carry out valuable work in the field of computer engineering. This could potentially provide an opportunity to contribute value to the current technological world.
In the current technological era, the demands for processing and transmitting multimedia content are increasing. The vast amount of existing data necessitates efficient systems for information extraction, digitization, storage, and ultimately retrieval. The latter point is crucial as new technologies increasingly demand speed and effectiveness in utilizing all this information. This project aims to emphasize the idea of retrieving images from databases, specifically from storage systems where all this information is kept. Specifically, the development of a tool is proposed that allows retrieving a similarity ranking between images based on their descriptors. These descriptors are sets of numerical values that reflect characteristics of an image such as color or texture. Furthermore, the functionality is proposed to compute these descriptors locally, i.e., on a specific part of the image. This approach gives users more options when retrieving information. This comparison will be carried out between images in a database and a query image chosen by the user. Thus, the tool also integrates all corresponding database management actions, such as adding or deleting a database, or adding or removing an image within a database. Throughout this project, the knowledge acquired up to this point, combined with new knowledge gained during its development, will be applied to carry out valuable work in the field of computer engineering. This could potentially provide an opportunity to contribute value to the current technological world.
Direction
MOSQUERA GONZALEZ, ANTONIO (Tutorships)
MOSQUERA GONZALEZ, ANTONIO (Tutorships)
Court
MOSQUERA GONZALEZ, ANTONIO (Student’s tutor)
MOSQUERA GONZALEZ, ANTONIO (Student’s tutor)
Development of a Web Application for the Sports Management of a Rowing Club
Authorship
R.M.T.
Computer Engineering Degree
R.M.T.
Computer Engineering Degree
Defense date
07.18.2024 18:00
07.18.2024 18:00
Summary
This Final Degree Project presents the development of a web application for the sports management of a rowing club, designed to be used by athletes, coaches, managers, and administrators. The application allows the consultation, publication, and updating of the club's sports information, addressing the need to improve the current management of the “Club do Mar de Noia”. Therefore, the work can be categorized under B modality.
This Final Degree Project presents the development of a web application for the sports management of a rowing club, designed to be used by athletes, coaches, managers, and administrators. The application allows the consultation, publication, and updating of the club's sports information, addressing the need to improve the current management of the “Club do Mar de Noia”. Therefore, the work can be categorized under B modality.
Direction
TOBAR QUINTANAR, ALEJANDRO JOSE (Tutorships)
TOBAR QUINTANAR, ALEJANDRO JOSE (Tutorships)
Court
VAZQUEZ CENDON, MARIA ELENA (Chairman)
CHAVES FRAGA, DAVID (Secretary)
SUAREZ GAREA, JORGE ALBERTO (Member)
VAZQUEZ CENDON, MARIA ELENA (Chairman)
CHAVES FRAGA, DAVID (Secretary)
SUAREZ GAREA, JORGE ALBERTO (Member)
Prediction of Paroxysmal Atrial Fibrillation using Machine Learning in Healthy Individuals
Authorship
A.G.S.
Computer Engineering Degree
A.G.S.
Computer Engineering Degree
Defense date
07.18.2024 18:30
07.18.2024 18:30
Summary
The objective of this work is to develop and validate a model for predicting paroxysmal atrial fibrillation (AF) in healthy individuals using machine learning techniques. Unlike other approaches that focus on detecting AF in already sick individuals, this study aims to anticipate the onset of the disease in still healthy individuals. To this end, a dataset of 58,582 electrocardiograms (ECGs) from 11,636 patients was used, which includes an identifier and a total of 493 electrocardiographic features. Some patients remain healthy across all ECGs, while others develop persistent or paroxysmal atrial fibrillation at some point. An important factor in this problem is the irregularity in the timing of ECG samples, as they are performed on patients in an unplanned manner, resulting in data that are not evenly spaced in time. This complicates the resolution of the problem since traditional time series analysis methods cannot adequately handle these irregularities. Therefore, a Latent-ODE (Latent Ordinary Differential Equations) model was chosen, which is particularly suitable for handling these temporal irregularities. Latent-ODEs allow modeling continuous dynamics in a latent space using ordinary differential equations in combination with recurrent neural networks (RNNs). In this latent space, it is possible to capture the underlying temporal evolution of the ECG features, thus providing a more robust and accurate representation of the temporal data. In this way, a Latent-ODE model was developed, along with feature selection and hyperparameter optimization techniques, which allowed us to achieve a best result of an AUC of 0.8812 and an F1-score of 0.7038.
The objective of this work is to develop and validate a model for predicting paroxysmal atrial fibrillation (AF) in healthy individuals using machine learning techniques. Unlike other approaches that focus on detecting AF in already sick individuals, this study aims to anticipate the onset of the disease in still healthy individuals. To this end, a dataset of 58,582 electrocardiograms (ECGs) from 11,636 patients was used, which includes an identifier and a total of 493 electrocardiographic features. Some patients remain healthy across all ECGs, while others develop persistent or paroxysmal atrial fibrillation at some point. An important factor in this problem is the irregularity in the timing of ECG samples, as they are performed on patients in an unplanned manner, resulting in data that are not evenly spaced in time. This complicates the resolution of the problem since traditional time series analysis methods cannot adequately handle these irregularities. Therefore, a Latent-ODE (Latent Ordinary Differential Equations) model was chosen, which is particularly suitable for handling these temporal irregularities. Latent-ODEs allow modeling continuous dynamics in a latent space using ordinary differential equations in combination with recurrent neural networks (RNNs). In this latent space, it is possible to capture the underlying temporal evolution of the ECG features, thus providing a more robust and accurate representation of the temporal data. In this way, a Latent-ODE model was developed, along with feature selection and hyperparameter optimization techniques, which allowed us to achieve a best result of an AUC of 0.8812 and an F1-score of 0.7038.
Direction
FELIX LAMAS, PAULO (Tutorships)
RODRIGUEZ PRESEDO, JESUS MARIA (Co-tutorships)
FELIX LAMAS, PAULO (Tutorships)
RODRIGUEZ PRESEDO, JESUS MARIA (Co-tutorships)
Court
Pardo López, Xosé Manuel (Chairman)
Puentes Cociña, Beltrán (Secretary)
LOPEZ FANDIÑO, JAVIER (Member)
Pardo López, Xosé Manuel (Chairman)
Puentes Cociña, Beltrán (Secretary)
LOPEZ FANDIÑO, JAVIER (Member)
Anomaly Detection on Multispectral Images based on the Isolation Forest Algorithm and Deep Learning networks
Authorship
D.G.F.
Computer Engineering Degree
D.G.F.
Computer Engineering Degree
Defense date
07.19.2024 18:30
07.19.2024 18:30
Summary
In the scientific field of remote sensing, multispectral and hyperspectral images capture detailed information in different bands of the electromagnetic spectrum, allowing high-precision analysis of the Earth's surface. This ability to capture multiple wavelengths is ideal for observing and studying the Earth's surface through satellites, drones, or airplanes. In recent years, advances in machine learning have significantly enhanced research in anomaly detection in these images, successfully identifying distinct elements within a specific environment. Recently, the technique known as Isolation Forest has demonstrated superior performance for anomaly detection compared to traditional distance-based methods. Moreover, its approach, based on the construction of isolation trees, allows for more precise anomaly detection, making it a powerful tool in the field of anomaly detection and pushing the state-of-the-art forward. The use of deep learning techniques in anomaly detection offers several advantages compared to traditional algorithms based on pixel distance calculations. Among these advantages is greater adaptability, as neural networks can be retrained with new data to adjust to changes in the environment or the characteristics of anomalies. Additionally, deep learning methods are less sensitive to noise and small variations in data, allowing for more robust performance in complex data scenarios, such as in MSI (Multispectral Images). Among the available algorithms of this type in the literature, Deep SAD offers a deep learning-based approach for anomaly detection that has demonstrated good performance on RGB image datasets. In this work, the Isolation Forest and Deep SAD algorithms have been adapted for use in multispectral images. Additionally, for both algorithms, a hyperparameter exploration was conducted to obtain a parametrization that fits the datasets used, belonging to the hydrographic basins of the Oitavén River. The results obtained by Isolation Forest and Deep SAD on these datasets have been compared with those obtained by the RX method, a consolidated method in the state-of-the-art for anomaly detection in this type of images. The results of the conducted experimentation showed that both proposed methods, Isolation Forest and Deep SAD, achieved good results in anomaly detection in MSI (Multispectral Images). Isolation Forest proved to be particularly effective, detecting a high percentage of the total existing anomalies in the images and significantly surpassing the detection percentages achieved by the distance-based RX algorithm. On the other hand, Deep SAD also managed to improve the number of detected anomalous pixels compared to the RX method, also showing a high capacity to distinguish between classes, achieving a good balance between the number of true positives and false positives.
In the scientific field of remote sensing, multispectral and hyperspectral images capture detailed information in different bands of the electromagnetic spectrum, allowing high-precision analysis of the Earth's surface. This ability to capture multiple wavelengths is ideal for observing and studying the Earth's surface through satellites, drones, or airplanes. In recent years, advances in machine learning have significantly enhanced research in anomaly detection in these images, successfully identifying distinct elements within a specific environment. Recently, the technique known as Isolation Forest has demonstrated superior performance for anomaly detection compared to traditional distance-based methods. Moreover, its approach, based on the construction of isolation trees, allows for more precise anomaly detection, making it a powerful tool in the field of anomaly detection and pushing the state-of-the-art forward. The use of deep learning techniques in anomaly detection offers several advantages compared to traditional algorithms based on pixel distance calculations. Among these advantages is greater adaptability, as neural networks can be retrained with new data to adjust to changes in the environment or the characteristics of anomalies. Additionally, deep learning methods are less sensitive to noise and small variations in data, allowing for more robust performance in complex data scenarios, such as in MSI (Multispectral Images). Among the available algorithms of this type in the literature, Deep SAD offers a deep learning-based approach for anomaly detection that has demonstrated good performance on RGB image datasets. In this work, the Isolation Forest and Deep SAD algorithms have been adapted for use in multispectral images. Additionally, for both algorithms, a hyperparameter exploration was conducted to obtain a parametrization that fits the datasets used, belonging to the hydrographic basins of the Oitavén River. The results obtained by Isolation Forest and Deep SAD on these datasets have been compared with those obtained by the RX method, a consolidated method in the state-of-the-art for anomaly detection in this type of images. The results of the conducted experimentation showed that both proposed methods, Isolation Forest and Deep SAD, achieved good results in anomaly detection in MSI (Multispectral Images). Isolation Forest proved to be particularly effective, detecting a high percentage of the total existing anomalies in the images and significantly surpassing the detection percentages achieved by the distance-based RX algorithm. On the other hand, Deep SAD also managed to improve the number of detected anomalous pixels compared to the RX method, also showing a high capacity to distinguish between classes, achieving a good balance between the number of true positives and false positives.
Direction
LOPEZ FANDIÑO, JAVIER (Tutorships)
Argüello Pedreira, Francisco Santiago (Co-tutorships)
LOPEZ FANDIÑO, JAVIER (Tutorships)
Argüello Pedreira, Francisco Santiago (Co-tutorships)
Court
Fernández Pena, Anselmo Tomás (Chairman)
SACO LOPEZ, PEDRO JOSE (Secretary)
RODRIGUEZ PRESEDO, JESUS MARIA (Member)
Fernández Pena, Anselmo Tomás (Chairman)
SACO LOPEZ, PEDRO JOSE (Secretary)
RODRIGUEZ PRESEDO, JESUS MARIA (Member)
Pollution monitoring of Spanish cities using TROPOMI data
Authorship
H.G.S.
Computer Engineering Degree
H.G.S.
Computer Engineering Degree
Defense date
02.16.2024 16:00
02.16.2024 16:00
Summary
The presence of different pollutants in the atmosphere is the cause not only of climate change, but also of many health problems for people. Nowadays, we have multiple satellites that allow us to monitor the levels of these parameters remotely. However, analyzing a multitude of files and data, for various parameters and days, can be a tedious task. That is why this thesis is proposed, which aims to design a system to monitor the variability and trend of different parameters obtained through satellite data. By downloading this data and its transformation and aggregation, we will be able to significantly reduce the amount of files that will be necessary to analyze, which eases the monitoring of pollution by end users. The development of this system will be done in such a way that it can be easily accessed, and also taking into account the interoperability of the system.
The presence of different pollutants in the atmosphere is the cause not only of climate change, but also of many health problems for people. Nowadays, we have multiple satellites that allow us to monitor the levels of these parameters remotely. However, analyzing a multitude of files and data, for various parameters and days, can be a tedious task. That is why this thesis is proposed, which aims to design a system to monitor the variability and trend of different parameters obtained through satellite data. By downloading this data and its transformation and aggregation, we will be able to significantly reduce the amount of files that will be necessary to analyze, which eases the monitoring of pollution by end users. The development of this system will be done in such a way that it can be easily accessed, and also taking into account the interoperability of the system.
Direction
Triñanes Fernández, Joaquín Ángel (Tutorships)
Triñanes Fernández, Joaquín Ángel (Tutorships)
Court
SUAREZ GAREA, JORGE ALBERTO (Coordinator)
BUGARIN DIZ, ALBERTO JOSE (Chairman)
VILLARROYA FERNANDEZ, SEBASTIAN (Secretary)
BORRAJO GARCIA, MARIA ISABEL (Member)
SUAREZ GAREA, JORGE ALBERTO (Coordinator)
BUGARIN DIZ, ALBERTO JOSE (Chairman)
VILLARROYA FERNANDEZ, SEBASTIAN (Secretary)
BORRAJO GARCIA, MARIA ISABEL (Member)
Development of a web application to visualise the result of applying artificial intelligence and explainability techniques in the prediction of cognitive impairment in the elderly
Authorship
J.R.C.
Computer Engineering Degree
J.R.C.
Computer Engineering Degree
Defense date
07.19.2024 09:30
07.19.2024 09:30
Summary
The aim of this Final Degree Project is to design and develop a dashboard-type web application that allows the representation of predictions and data in the neuropsychological field. It is aimed at users who are not experts in datascience, algorithm programming or the use of technologies related to Artificial Intelligence. Once developed, the application has been validated at a technical and usability level, thus allowing us to evaluate the features implemented to address its usability and functionality.
The aim of this Final Degree Project is to design and develop a dashboard-type web application that allows the representation of predictions and data in the neuropsychological field. It is aimed at users who are not experts in datascience, algorithm programming or the use of technologies related to Artificial Intelligence. Once developed, the application has been validated at a technical and usability level, thus allowing us to evaluate the features implemented to address its usability and functionality.
Direction
VALLADARES RODRIGUEZ, SONIA MARIA (Tutorships)
CONDORI FERNANDEZ, NELLY (Co-tutorships)
VALLADARES RODRIGUEZ, SONIA MARIA (Tutorships)
CONDORI FERNANDEZ, NELLY (Co-tutorships)
Court
Barro Ameneiro, Senén (Chairman)
TOBAR QUINTANAR, ALEJANDRO JOSE (Secretary)
LADRA GONZALEZ, MANUEL EULOGIO (Member)
Barro Ameneiro, Senén (Chairman)
TOBAR QUINTANAR, ALEJANDRO JOSE (Secretary)
LADRA GONZALEZ, MANUEL EULOGIO (Member)
Evolution of Formula 1 data analysis system: Migration to the Cloud
Authorship
L.P.M.
Computer Engineering Degree
L.P.M.
Computer Engineering Degree
Defense date
07.19.2024 09:00
07.19.2024 09:00
Summary
This TFG is devoted to the optimisation and migration of a Formula 1 data analysis system to the cloud. Its main objective is to improve the structure and functionality of the project to manage large volumes of data in real time and provide interactive visualisations, taking advantage of the abundance of data generated by the sport. First of all, reliable and cloud-accessible data sources have been identified and selected to collect the relevant information. The entire ETL (Extraction, Transformation and Loading) process is performed in the cloud, where the data is transformed and integrated into a model designed for optimal structuring, with the added capability of automatic weekly updating. Once the database is updated, a visualisation tool is employed that connects directly to the cloud database, allowing users to access a vast amount of information in an interactive and visually appealing way. With this tool, users can select different datasets and generate dynamic visualisations that facilitate the understanding of the data, enabling comparison and exploration of historical trends, as well as the assessment of performance over time. The ultimate goal is to ensure effective and useful interpretation of Formula 1 related data, facilitating informed decision making and improving the overall performance of the analytics system. This integration of advanced data visualisation and analysis tools ensures that users can take full advantage of the wealth of information available to derive valuable insights.
This TFG is devoted to the optimisation and migration of a Formula 1 data analysis system to the cloud. Its main objective is to improve the structure and functionality of the project to manage large volumes of data in real time and provide interactive visualisations, taking advantage of the abundance of data generated by the sport. First of all, reliable and cloud-accessible data sources have been identified and selected to collect the relevant information. The entire ETL (Extraction, Transformation and Loading) process is performed in the cloud, where the data is transformed and integrated into a model designed for optimal structuring, with the added capability of automatic weekly updating. Once the database is updated, a visualisation tool is employed that connects directly to the cloud database, allowing users to access a vast amount of information in an interactive and visually appealing way. With this tool, users can select different datasets and generate dynamic visualisations that facilitate the understanding of the data, enabling comparison and exploration of historical trends, as well as the assessment of performance over time. The ultimate goal is to ensure effective and useful interpretation of Formula 1 related data, facilitating informed decision making and improving the overall performance of the analytics system. This integration of advanced data visualisation and analysis tools ensures that users can take full advantage of the wealth of information available to derive valuable insights.
Direction
VIDAL AGUIAR, JUAN CARLOS (Tutorships)
Curra Durán, Alberto (Co-tutorships)
VIDAL AGUIAR, JUAN CARLOS (Tutorships)
Curra Durán, Alberto (Co-tutorships)
Court
Barro Ameneiro, Senén (Chairman)
TOBAR QUINTANAR, ALEJANDRO JOSE (Secretary)
LADRA GONZALEZ, MANUEL EULOGIO (Member)
Barro Ameneiro, Senén (Chairman)
TOBAR QUINTANAR, ALEJANDRO JOSE (Secretary)
LADRA GONZALEZ, MANUEL EULOGIO (Member)
Development of a web application for monitoring telecommunication infrastructures
Authorship
M.M.M.
Computer Engineering Degree
M.M.M.
Computer Engineering Degree
Defense date
02.15.2024 12:00
02.15.2024 12:00
Summary
This TFG (Final Degree Project) is a collaboration between a company called Itelsis and the aim is to develop a software that enables telecommunications management for the company’s customers. The main focus of the development is to create a network map that can be customized and updated in real-time. The application includes various modules that complement the map and aid in network management. The main objective of the application is to display telecommunications network data in a simplified and easy-to-understand manner for users. The network map should show the status of different hosts in the network, allowing users to modify host positions and customize map icons according to their preference. Additionally, complementary modules are being developed to provide users with detailed and real-time information on hosts and services. Users will also have the option to configure the main webpage of the software and customize certain user interactions with the application. To access the application you must have a valid user given by Itelsis, the user can be of different types, changing the way they interact with the application and their permissions on the application objects. This Final Degree Project corresponds to the type B established in the regulations (Specification, analysis, design and implementation of different aspects related to a module of a Computer System or a complete Computer System).
This TFG (Final Degree Project) is a collaboration between a company called Itelsis and the aim is to develop a software that enables telecommunications management for the company’s customers. The main focus of the development is to create a network map that can be customized and updated in real-time. The application includes various modules that complement the map and aid in network management. The main objective of the application is to display telecommunications network data in a simplified and easy-to-understand manner for users. The network map should show the status of different hosts in the network, allowing users to modify host positions and customize map icons according to their preference. Additionally, complementary modules are being developed to provide users with detailed and real-time information on hosts and services. Users will also have the option to configure the main webpage of the software and customize certain user interactions with the application. To access the application you must have a valid user given by Itelsis, the user can be of different types, changing the way they interact with the application and their permissions on the application objects. This Final Degree Project corresponds to the type B established in the regulations (Specification, analysis, design and implementation of different aspects related to a module of a Computer System or a complete Computer System).
Direction
TOBAR QUINTANAR, ALEJANDRO JOSE (Tutorships)
Conde Balado, Miguel Ángel (Co-tutorships)
Franco Romero, Carlos (Co-tutorships)
TOBAR QUINTANAR, ALEJANDRO JOSE (Tutorships)
Conde Balado, Miguel Ángel (Co-tutorships)
Franco Romero, Carlos (Co-tutorships)
Court
PICHEL CAMPOS, JUAN CARLOS (Chairman)
CORES COSTA, DANIEL (Secretary)
Rodeiro Pazos, David (Member)
PICHEL CAMPOS, JUAN CARLOS (Chairman)
CORES COSTA, DANIEL (Secretary)
Rodeiro Pazos, David (Member)
Exploration of LLMs for fine-grained classification of texts annotated with MeSH descriptors
Authorship
A.R.P.
Computer Engineering Degree
A.R.P.
Computer Engineering Degree
Defense date
07.19.2024 10:30
07.19.2024 10:30
Summary
This Bachelor's Thesis explores the semantic classification of biomedical texts using the generative language model Llama3, aiming to improve access to information in biomedicine through MeSH descriptors. The proposed method consists of different stages, the first being the preprocessing of the MeSH ontology obtained through BioPortal. Then, biomedical abstracts annotated with generic categories are selected and prepared, ensuring a representative sample for subsequent classification. The methodology involves the implementation of an algorithm based on a zero-shot approach of the Llama-3 model, developing specific prompts and normalizing the generated responses to ensure consistency. Finally, exhaustive tests are conducted, applying standardized metrics such as precision, recall, and F-measure to evaluate the performance of the classification technique. The results show that Llama-3 surpasses weak supervision methods in terms of precision, although it presents challenges in terms of recall. The discussion includes a detailed analysis of specific descriptors, identifying patterns and proposing improvement strategies. In conclusion, the effectiveness of generative language models with closed queries is highlighted, as well as the importance of future optimizations in terms of computing capacity and prompt engineering techniques. This work lays the foundation for future research in the semantic classification of biomedical texts, suggesting expansions to optimize the obtained results.
This Bachelor's Thesis explores the semantic classification of biomedical texts using the generative language model Llama3, aiming to improve access to information in biomedicine through MeSH descriptors. The proposed method consists of different stages, the first being the preprocessing of the MeSH ontology obtained through BioPortal. Then, biomedical abstracts annotated with generic categories are selected and prepared, ensuring a representative sample for subsequent classification. The methodology involves the implementation of an algorithm based on a zero-shot approach of the Llama-3 model, developing specific prompts and normalizing the generated responses to ensure consistency. Finally, exhaustive tests are conducted, applying standardized metrics such as precision, recall, and F-measure to evaluate the performance of the classification technique. The results show that Llama-3 surpasses weak supervision methods in terms of precision, although it presents challenges in terms of recall. The discussion includes a detailed analysis of specific descriptors, identifying patterns and proposing improvement strategies. In conclusion, the effectiveness of generative language models with closed queries is highlighted, as well as the importance of future optimizations in terms of computing capacity and prompt engineering techniques. This work lays the foundation for future research in the semantic classification of biomedical texts, suggesting expansions to optimize the obtained results.
Direction
TABOADA IGLESIAS, MARÍA JESÚS (Tutorships)
TABOADA IGLESIAS, MARÍA JESÚS (Tutorships)
Court
Barro Ameneiro, Senén (Chairman)
TOBAR QUINTANAR, ALEJANDRO JOSE (Secretary)
LADRA GONZALEZ, MANUEL EULOGIO (Member)
Barro Ameneiro, Senén (Chairman)
TOBAR QUINTANAR, ALEJANDRO JOSE (Secretary)
LADRA GONZALEZ, MANUEL EULOGIO (Member)
Segmentation of remote sensing images using deep learning algorithms in Pytorch
Authorship
J.C.M.
Computer Engineering Degree
J.C.M.
Computer Engineering Degree
Defense date
07.19.2024 12:00
07.19.2024 12:00
Summary
Semantic segmentation of multispectral remote sensing images has seen a significant increase in the use of deep learning algorithms in the last decade, due to the high accuracy these methods can achieve. Multispectral images are those captured through multiple bands of the electromagnetic spectrum, beyond the visible bands of red, green, and blue, allowing detailed information about the properties and characteristics of observed objects to be captured. This work focuses on adapting deep neural network models, originally designed for RGB images, to the domain of multispectral images using the PyTorch environment. The main objective is to develop and evaluate efficient and robust models for semantic segmentation, validating their performance with multispectral image datasets. One of the main contributions is the experimental demonstration that deep learning models designed for the semantic segmentation of RGB images can be effectively adapted to multispectral images, utilizing all the spectral and spatial information available in this type of images and achieving high-quality results. This validates the initial hypothesis that these algorithms can be successfully applied in the context of multispectral images, surpassing state-of-the-art models in terms of precision and effectiveness. In the development of this work, typical architectures of deep neural networks are studied and different models are tested to determine which offers better performance in the semantic segmentation of multispectral images. It is identified that, for the dataset used, after approximately 60 training epochs, the increases in precision and mean IoU become minimal, indicating a saturation point. This finding suggests that training models beyond this point does not provide significant improvements in performance but does considerably increase computation time. Additionally, signs of overfitting are observed in some models that continue training beyond this point, which does not improve performance and, in some cases, even harms it. Another important contribution is the verification that deeper architectures do not necessarily offer the best results. Lighter models, such as the ResNet18dilated + C1_deepsup, achieve high precision and good performance with less training time compared to more complex and deep models like the ResNet101dilated + PPM_deepsup. The final tests, conducted with the model that presented the best results in the preliminary tests (ResNet18dilated + C1_deepsup), reveal notable variability in the model's performance among different test images, suggesting that the size of the training dataset might be insufficient to capture all the necessary variability for consistent and generalized performance. This opens the possibility of future improvements, such as increasing the size and diversity of the training dataset to improve the robustness and consistency of the models, optimizing hyperparameters to discover more efficient and effective configurations for the semantic segmentation of multispectral images.
Semantic segmentation of multispectral remote sensing images has seen a significant increase in the use of deep learning algorithms in the last decade, due to the high accuracy these methods can achieve. Multispectral images are those captured through multiple bands of the electromagnetic spectrum, beyond the visible bands of red, green, and blue, allowing detailed information about the properties and characteristics of observed objects to be captured. This work focuses on adapting deep neural network models, originally designed for RGB images, to the domain of multispectral images using the PyTorch environment. The main objective is to develop and evaluate efficient and robust models for semantic segmentation, validating their performance with multispectral image datasets. One of the main contributions is the experimental demonstration that deep learning models designed for the semantic segmentation of RGB images can be effectively adapted to multispectral images, utilizing all the spectral and spatial information available in this type of images and achieving high-quality results. This validates the initial hypothesis that these algorithms can be successfully applied in the context of multispectral images, surpassing state-of-the-art models in terms of precision and effectiveness. In the development of this work, typical architectures of deep neural networks are studied and different models are tested to determine which offers better performance in the semantic segmentation of multispectral images. It is identified that, for the dataset used, after approximately 60 training epochs, the increases in precision and mean IoU become minimal, indicating a saturation point. This finding suggests that training models beyond this point does not provide significant improvements in performance but does considerably increase computation time. Additionally, signs of overfitting are observed in some models that continue training beyond this point, which does not improve performance and, in some cases, even harms it. Another important contribution is the verification that deeper architectures do not necessarily offer the best results. Lighter models, such as the ResNet18dilated + C1_deepsup, achieve high precision and good performance with less training time compared to more complex and deep models like the ResNet101dilated + PPM_deepsup. The final tests, conducted with the model that presented the best results in the preliminary tests (ResNet18dilated + C1_deepsup), reveal notable variability in the model's performance among different test images, suggesting that the size of the training dataset might be insufficient to capture all the necessary variability for consistent and generalized performance. This opens the possibility of future improvements, such as increasing the size and diversity of the training dataset to improve the robustness and consistency of the models, optimizing hyperparameters to discover more efficient and effective configurations for the semantic segmentation of multispectral images.
Direction
Argüello Pedreira, Francisco Santiago (Tutorships)
Blanco Heras, Dora (Co-tutorships)
Argüello Pedreira, Francisco Santiago (Tutorships)
Blanco Heras, Dora (Co-tutorships)
Court
PENA BRAGE, FRANCISCO JOSE (Chairman)
MERA PEREZ, DAVID (Secretary)
CARIÑENA AMIGO, MARIA PURIFICACION (Member)
PENA BRAGE, FRANCISCO JOSE (Chairman)
MERA PEREZ, DAVID (Secretary)
CARIÑENA AMIGO, MARIA PURIFICACION (Member)
Improving the access to the knowledge graph agrovoc
Authorship
M.F.M.
Computer Engineering Degree
M.F.M.
Computer Engineering Degree
Defense date
02.16.2024 10:30
02.16.2024 10:30
Summary
Currently, the expansion of quality synonyms is a necessary task to evaluate the sensitivity of linguistic models to small alterations introduced through synonym substitution. Using knowledge graphs to query and extract quality synonyms makes this task easier. However, knowledge graphs contain a lot of implicit knowledge that cannot be directly recovered with common query languages. Along these lines, this final degree project explores a new method to successfully expand the query to the AGROVOC knowledge graph, in the domain of agriculture. Taking advantage of the hierarchical and logical context of the concepts and the lexical richness of the terms in AGROVOC, the method allows inferring the most precise category for entities not included in the KG that verify two restrictions, that they are multiword and nest other entities affirmed in AGROVOC.
Currently, the expansion of quality synonyms is a necessary task to evaluate the sensitivity of linguistic models to small alterations introduced through synonym substitution. Using knowledge graphs to query and extract quality synonyms makes this task easier. However, knowledge graphs contain a lot of implicit knowledge that cannot be directly recovered with common query languages. Along these lines, this final degree project explores a new method to successfully expand the query to the AGROVOC knowledge graph, in the domain of agriculture. Taking advantage of the hierarchical and logical context of the concepts and the lexical richness of the terms in AGROVOC, the method allows inferring the most precise category for entities not included in the KG that verify two restrictions, that they are multiword and nest other entities affirmed in AGROVOC.
Direction
TABOADA IGLESIAS, MARÍA JESÚS (Tutorships)
TABOADA IGLESIAS, MARÍA JESÚS (Tutorships)
Court
TABOADA IGLESIAS, MARÍA JESÚS (Student’s tutor)
TABOADA IGLESIAS, MARÍA JESÚS (Student’s tutor)
Methylamine production process by methanol and ammonia
Authorship
L.G.P.
Bachelor's Degree in Chemical Engeneering
L.G.P.
Bachelor's Degree in Chemical Engeneering
Defense date
09.13.2024 10:10
09.13.2024 10:10
Summary
This project has as a goal the production of methylamine and dimethylamine by methanol and ammonia. The capacity will be 60000 t/year of methylamine and dimethylamine, considering continuous operating regime during 330 days a year, 24 hours a day. It is split in two section. Section 100 includes the conditioning of raw materials and the reaction on the reactor R-101 (designed by Alex Dacosta García). Section 200 separates and purifies the products, obtaining dimethylamine 98% pure and methylamine 95% pure. The distillation column that separates both products was designed by Leire Gallego Portabales. There will also be taken into account security, environmental and economy viability studies.
This project has as a goal the production of methylamine and dimethylamine by methanol and ammonia. The capacity will be 60000 t/year of methylamine and dimethylamine, considering continuous operating regime during 330 days a year, 24 hours a day. It is split in two section. Section 100 includes the conditioning of raw materials and the reaction on the reactor R-101 (designed by Alex Dacosta García). Section 200 separates and purifies the products, obtaining dimethylamine 98% pure and methylamine 95% pure. The distillation column that separates both products was designed by Leire Gallego Portabales. There will also be taken into account security, environmental and economy viability studies.
Direction
FRANCO URIA, MARIA AMAYA (Tutorships)
SINEIRO TORRES, JORGE (Co-tutorships)
FRANCO URIA, MARIA AMAYA (Tutorships)
SINEIRO TORRES, JORGE (Co-tutorships)
Court
GARRIDO FERNANDEZ, JUAN MANUEL (Chairman)
FRANCO RUIZ, DANIEL JOSE (Secretary)
FREIRE LEIRA, MARIA SONIA (Member)
GARRIDO FERNANDEZ, JUAN MANUEL (Chairman)
FRANCO RUIZ, DANIEL JOSE (Secretary)
FREIRE LEIRA, MARIA SONIA (Member)
Peracetic acid production plant
Authorship
B.Q.L.
Bachelor's Degree in Chemical Engeneering
B.Q.L.
Bachelor's Degree in Chemical Engeneering
Defense date
09.13.2024 09:30
09.13.2024 09:30
Summary
The purpose of this project is to design a basic production plant for peracetic acid from hydrogen peroxide and acetic acid. The annual production of the industrial plant will be 20,000 tons per year of a solution containing more than 32% peracetic acid. The process includes a pre-conditioning stage, where sulfuric acid in the liquid phase will be used to protonate the acetic acid that will be used in the reactor, with sulfuric acid playing a crucial role in the synthesis of the reaction as a catalyst. This reaction will take place in the tubular reactor R-101, designed by Brais Queiro Lamas. At the reactor outlet, two separation columns, T-201 and T-202, will be used respectively. The first is intended for the recovery of sulfuric acid, and the second to concentrate the peracetic acid to obtain the desired product. The T-202 column will be designed by Manuel Antonio Castro Pereiras. This project is presented as a Final Degree Project through which the students Manuel Antonio Castro Pereiras and Brais Queiro Lamas aim to obtain the Degree in Chemical Engineering awarded by the Higher Technical School of Engineering belonging to the University of Santiago de Compostela.
The purpose of this project is to design a basic production plant for peracetic acid from hydrogen peroxide and acetic acid. The annual production of the industrial plant will be 20,000 tons per year of a solution containing more than 32% peracetic acid. The process includes a pre-conditioning stage, where sulfuric acid in the liquid phase will be used to protonate the acetic acid that will be used in the reactor, with sulfuric acid playing a crucial role in the synthesis of the reaction as a catalyst. This reaction will take place in the tubular reactor R-101, designed by Brais Queiro Lamas. At the reactor outlet, two separation columns, T-201 and T-202, will be used respectively. The first is intended for the recovery of sulfuric acid, and the second to concentrate the peracetic acid to obtain the desired product. The T-202 column will be designed by Manuel Antonio Castro Pereiras. This project is presented as a Final Degree Project through which the students Manuel Antonio Castro Pereiras and Brais Queiro Lamas aim to obtain the Degree in Chemical Engineering awarded by the Higher Technical School of Engineering belonging to the University of Santiago de Compostela.
Direction
MOREIRA MARTINEZ, RAMON FELIPE (Tutorships)
GOMEZ DIAZ, DIEGO (Co-tutorships)
MOREIRA MARTINEZ, RAMON FELIPE (Tutorships)
GOMEZ DIAZ, DIEGO (Co-tutorships)
Court
GARRIDO FERNANDEZ, JUAN MANUEL (Chairman)
FRANCO RUIZ, DANIEL JOSE (Secretary)
FREIRE LEIRA, MARIA SONIA (Member)
GARRIDO FERNANDEZ, JUAN MANUEL (Chairman)
FRANCO RUIZ, DANIEL JOSE (Secretary)
FREIRE LEIRA, MARIA SONIA (Member)
Plant for the production of resorcinol from benzene.
Authorship
V.D.V.
Bachelor's Degree in Chemical Engeneering
V.D.V.
Bachelor's Degree in Chemical Engeneering
Defense date
09.13.2024 09:30
09.13.2024 09:30
Summary
In the present project, the design and feasibility study of a plant for the production of resorcinol from benzene is carried out. For this purpose, the hydroperoxidation of m-diisopropylbenzene will be followed, in which 1,3-diisopropylbenzene dihydroperoxide will be produced as an intermediate product. This process will be carried out in a plant with a production capacity of 10.794 tons of resorcinol per year, with a purity of 99.25% molar. The project will be carried out by Vanesa Domínguez Valladares, in charge of the design and sizing of the R-301 reactor, and Ángela González Bruzos, in charge of the design and sizing of the T-302 distillation column. As consequence, the Chemical Engineering Final Degree Project is completed, opting for the title of Graduates in Chemical Engineering from the University of Santiago de Compostela.
In the present project, the design and feasibility study of a plant for the production of resorcinol from benzene is carried out. For this purpose, the hydroperoxidation of m-diisopropylbenzene will be followed, in which 1,3-diisopropylbenzene dihydroperoxide will be produced as an intermediate product. This process will be carried out in a plant with a production capacity of 10.794 tons of resorcinol per year, with a purity of 99.25% molar. The project will be carried out by Vanesa Domínguez Valladares, in charge of the design and sizing of the R-301 reactor, and Ángela González Bruzos, in charge of the design and sizing of the T-302 distillation column. As consequence, the Chemical Engineering Final Degree Project is completed, opting for the title of Graduates in Chemical Engineering from the University of Santiago de Compostela.
Direction
MOREIRA MARTINEZ, RAMON FELIPE (Tutorships)
GOMEZ DIAZ, DIEGO (Co-tutorships)
MOREIRA MARTINEZ, RAMON FELIPE (Tutorships)
GOMEZ DIAZ, DIEGO (Co-tutorships)
Court
HOSPIDO QUINTANA, ALMUDENA (Chairman)
MAURICIO IGLESIAS, MIGUEL (Secretary)
GONZALEZ GARCIA, SARA (Member)
HOSPIDO QUINTANA, ALMUDENA (Chairman)
MAURICIO IGLESIAS, MIGUEL (Secretary)
GONZALEZ GARCIA, SARA (Member)
Plant for the production of resorcinol from benzene
Authorship
A.G.B.
Bachelor's Degree in Chemical Engeneering
A.G.B.
Bachelor's Degree in Chemical Engeneering
Defense date
02.21.2024 10:45
02.21.2024 10:45
Summary
The objective of the present project is the design of a plant to produce 10794 tons of resorcinol and 10907 tons of acetone as a secondary product. These products are obtained by a three-step process: Friedel-Crafts alkylation of benzene with propylene to produce m-DIPB, oxidation of m-DIPB with air to produce DHP and, finally, cleavage of DHP to obtain resorcinol and acetone. The plant consists of four CSTR reactors, six distillation columns, three Karr columns, a flash tank, and a spray dryer, in addition to pressure and temperature adjustment equipment such as valves, compressors, heat exchangers or pumps. In this project, the rigorous design of the R-301 catalytic reactor and the T-302 distillation column is carried out.
The objective of the present project is the design of a plant to produce 10794 tons of resorcinol and 10907 tons of acetone as a secondary product. These products are obtained by a three-step process: Friedel-Crafts alkylation of benzene with propylene to produce m-DIPB, oxidation of m-DIPB with air to produce DHP and, finally, cleavage of DHP to obtain resorcinol and acetone. The plant consists of four CSTR reactors, six distillation columns, three Karr columns, a flash tank, and a spray dryer, in addition to pressure and temperature adjustment equipment such as valves, compressors, heat exchangers or pumps. In this project, the rigorous design of the R-301 catalytic reactor and the T-302 distillation column is carried out.
Direction
MOREIRA MARTINEZ, RAMON FELIPE (Tutorships)
GOMEZ DIAZ, DIEGO (Co-tutorships)
MOREIRA MARTINEZ, RAMON FELIPE (Tutorships)
GOMEZ DIAZ, DIEGO (Co-tutorships)
Court
HOSPIDO QUINTANA, ALMUDENA (Chairman)
MAURICIO IGLESIAS, MIGUEL (Secretary)
GONZALEZ GARCIA, SARA (Member)
HOSPIDO QUINTANA, ALMUDENA (Chairman)
MAURICIO IGLESIAS, MIGUEL (Secretary)
GONZALEZ GARCIA, SARA (Member)
Production plant of synthetic fuel from methanol
Authorship
I.M.F.
Bachelor's Degree in Chemical Engeneering
I.M.F.
Bachelor's Degree in Chemical Engeneering
Defense date
09.13.2024 10:10
09.13.2024 10:10
Summary
This project is based on the production of synthetic fuel from pure methanol from carbon dioxide capture in the atmosphere, wishing to process 44204 tonnes of raw material per year. This process was carried out by designing a distillation column which would remove all the water from the first equilibrium reactor; another multi-tubular fixed-bed reactor was designed with a zeolite catalyst, allowing us to obtain the purest possible hydrocarbon product for subsequent addition of additives. The main objective of this project is to study an alternative to conventional petrol in order to obtain a more environmentally sustainable fuel.
This project is based on the production of synthetic fuel from pure methanol from carbon dioxide capture in the atmosphere, wishing to process 44204 tonnes of raw material per year. This process was carried out by designing a distillation column which would remove all the water from the first equilibrium reactor; another multi-tubular fixed-bed reactor was designed with a zeolite catalyst, allowing us to obtain the purest possible hydrocarbon product for subsequent addition of additives. The main objective of this project is to study an alternative to conventional petrol in order to obtain a more environmentally sustainable fuel.
Direction
Rodríguez Figueiras, Óscar (Tutorships)
Rodríguez Figueiras, Óscar (Tutorships)
Court
CARBALLA ARCOS, MARTA (Chairman)
EIBES GONZALEZ, GEMMA MARIA (Secretary)
BELLO BUGALLO, PASTORA MARIA (Member)
CARBALLA ARCOS, MARTA (Chairman)
EIBES GONZALEZ, GEMMA MARIA (Secretary)
BELLO BUGALLO, PASTORA MARIA (Member)
Acetaldehyde production plant through ethanol dehydrogenation
Authorship
P.F.V.
Bachelor's Degree in Chemical Engeneering
P.F.V.
Bachelor's Degree in Chemical Engeneering
Defense date
09.13.2024 11:00
09.13.2024 11:00
Summary
In this work, the design of an acetaldehyde production plant was carried out, using the ethanol dehydrogenation method. Sating from ethanol, the endothermic dehydrogenation reaction occurs, giving rise to acetaldehyde and hydrogen. The process consists of a section where the raw material is conditioned to the reaction conditions, then the section where the products are separated and purified, and the compounds that are not desired to be eliminated are also recirculated.
In this work, the design of an acetaldehyde production plant was carried out, using the ethanol dehydrogenation method. Sating from ethanol, the endothermic dehydrogenation reaction occurs, giving rise to acetaldehyde and hydrogen. The process consists of a section where the raw material is conditioned to the reaction conditions, then the section where the products are separated and purified, and the compounds that are not desired to be eliminated are also recirculated.
Direction
RODIL RODRIGUEZ, EVA (Tutorships)
RODIL RODRIGUEZ, EVA (Tutorships)
Court
GIL GONZALEZ, ALVARO (Chairman)
FRANCO URIA, MARIA AMAYA (Secretary)
GOMEZ DIAZ, DIEGO (Member)
GIL GONZALEZ, ALVARO (Chairman)
FRANCO URIA, MARIA AMAYA (Secretary)
GOMEZ DIAZ, DIEGO (Member)
Production plant of monoisopropylamine from isopropanol and ammonia
Authorship
L.C.F.
Bachelor's Degree in Chemical Engeneering
L.C.F.
Bachelor's Degree in Chemical Engeneering
Defense date
07.16.2024 13:35
07.16.2024 13:35
Summary
In this project, a monoisopropylamine (MIPA) production plant will be designed fro, isopropanol and ammonia. An annual production of 20000 tonnes is estimated. In this project, the reactor and distillation column required to obtain the product will be rigurously designed. The production process takes place as follows. The raw materials are introduced into the process and preconditioned witb the purpose of facilitating the reaction. The raw materials must be introduced into the reactor under optimal conditions of 12 atmospheres of pressure and 130ºC. The product stream from the reactor is directed to the sequence of separation structures, where all compounds will be separated. On the one hand, monoisopropylamine will be obtained as a product. On the other hand, raw materials that do not react in the reactor will be recobvered, allowing the profitability of the process to increase.
In this project, a monoisopropylamine (MIPA) production plant will be designed fro, isopropanol and ammonia. An annual production of 20000 tonnes is estimated. In this project, the reactor and distillation column required to obtain the product will be rigurously designed. The production process takes place as follows. The raw materials are introduced into the process and preconditioned witb the purpose of facilitating the reaction. The raw materials must be introduced into the reactor under optimal conditions of 12 atmospheres of pressure and 130ºC. The product stream from the reactor is directed to the sequence of separation structures, where all compounds will be separated. On the one hand, monoisopropylamine will be obtained as a product. On the other hand, raw materials that do not react in the reactor will be recobvered, allowing the profitability of the process to increase.
Direction
FREIRE LEIRA, MARIA SONIA (Tutorships)
GONZALEZ ALVAREZ, JULIA (Co-tutorships)
FREIRE LEIRA, MARIA SONIA (Tutorships)
GONZALEZ ALVAREZ, JULIA (Co-tutorships)
Court
SOTO CAMPOS, ANA MARIA (Chairman)
Rodríguez Figueiras, Óscar (Secretary)
RODRIGUEZ MARTINEZ, HECTOR (Member)
SOTO CAMPOS, ANA MARIA (Chairman)
Rodríguez Figueiras, Óscar (Secretary)
RODRIGUEZ MARTINEZ, HECTOR (Member)
Monoethanolamine production plant from ethylene oxide and ammonia
Authorship
I.V.F.
Bachelor's Degree in Chemical Engeneering
I.V.F.
Bachelor's Degree in Chemical Engeneering
Defense date
09.13.2024 10:50
09.13.2024 10:50
Summary
For the design of a monoethanolamine production plant from ammonia and ethylene oxide, the production process is divided into three sections: -Conditioning: the main equipment is an absorption column in which ammonia is absorbed in water to achieve a certain ammonia composition at the reactor inlet. This equipment will be designed by Irene Vidal. -Reaction: it is mainly composed of two istodermal reactors in which the second-order reaction occurs in the liquid phase between the aqueous solution of ammonia in water and ethylene oxide. This is highly exothermic. It is obtained as a monoethanolamine product and diethanolamine and triethanolamine as a by-product. One of these reactors will be designed by Adrián Ramos. -Separation: to separate the desired product (MEA) it will be necessary to use a distillation column, which will be designed by Carlos Rodríguez.
For the design of a monoethanolamine production plant from ammonia and ethylene oxide, the production process is divided into three sections: -Conditioning: the main equipment is an absorption column in which ammonia is absorbed in water to achieve a certain ammonia composition at the reactor inlet. This equipment will be designed by Irene Vidal. -Reaction: it is mainly composed of two istodermal reactors in which the second-order reaction occurs in the liquid phase between the aqueous solution of ammonia in water and ethylene oxide. This is highly exothermic. It is obtained as a monoethanolamine product and diethanolamine and triethanolamine as a by-product. One of these reactors will be designed by Adrián Ramos. -Separation: to separate the desired product (MEA) it will be necessary to use a distillation column, which will be designed by Carlos Rodríguez.
Direction
RODIL RODRIGUEZ, EVA (Tutorships)
RODIL RODRIGUEZ, EVA (Tutorships)
Court
CARBALLA ARCOS, MARTA (Chairman)
EIBES GONZALEZ, GEMMA MARIA (Secretary)
BELLO BUGALLO, PASTORA MARIA (Member)
CARBALLA ARCOS, MARTA (Chairman)
EIBES GONZALEZ, GEMMA MARIA (Secretary)
BELLO BUGALLO, PASTORA MARIA (Member)
Production plant of monoisopropylamine from isopropanol and ammonia
Authorship
R.R.S.
Bachelor's Degree in Chemical Engeneering
R.R.S.
Bachelor's Degree in Chemical Engeneering
Defense date
07.16.2024 13:35
07.16.2024 13:35
Summary
En el presente proyecto se diseñará una planta de producción de monoisopropilamina (MIPA) a partir de isopropanol y amoníaco. Se estima una producción anu al de 20000 toneladas. En el presente proyecto se diseñarán de forma rigurosa el reactor y la columna de destilación requerida para la obtención de producto. El proceso de producción tiene lugar de la siguiente manera. Las materias primas son introducidas en el proceso y preacondicionadas con el propósito de facilitar la reacción. Las materias primas han de ser introducidas al reactor en unas condiciones óptimas de 12 atmósferas de presión y 130ºC. La corriente de producto del reactor es dirigida a la secuencia de estructuras de separación, donde se separarán todos los compuestos. Por un lado se obtendrá monoisopropilamina como producto. Por otro, se recuperarán las materias primas que no reaccionan en el reactor, permitiendo aumentar la rentabilidad del proceso.
En el presente proyecto se diseñará una planta de producción de monoisopropilamina (MIPA) a partir de isopropanol y amoníaco. Se estima una producción anu al de 20000 toneladas. En el presente proyecto se diseñarán de forma rigurosa el reactor y la columna de destilación requerida para la obtención de producto. El proceso de producción tiene lugar de la siguiente manera. Las materias primas son introducidas en el proceso y preacondicionadas con el propósito de facilitar la reacción. Las materias primas han de ser introducidas al reactor en unas condiciones óptimas de 12 atmósferas de presión y 130ºC. La corriente de producto del reactor es dirigida a la secuencia de estructuras de separación, donde se separarán todos los compuestos. Por un lado se obtendrá monoisopropilamina como producto. Por otro, se recuperarán las materias primas que no reaccionan en el reactor, permitiendo aumentar la rentabilidad del proceso.
Direction
FREIRE LEIRA, MARIA SONIA (Tutorships)
GONZALEZ ALVAREZ, JULIA (Co-tutorships)
FREIRE LEIRA, MARIA SONIA (Tutorships)
GONZALEZ ALVAREZ, JULIA (Co-tutorships)
Court
SOTO CAMPOS, ANA MARIA (Chairman)
Rodríguez Figueiras, Óscar (Secretary)
RODRIGUEZ MARTINEZ, HECTOR (Member)
SOTO CAMPOS, ANA MARIA (Chairman)
Rodríguez Figueiras, Óscar (Secretary)
RODRIGUEZ MARTINEZ, HECTOR (Member)
Production plant of synthetic gasoline from methanol
Authorship
T.S.R.
Bachelor's Degree in Chemical Engeneering
T.S.R.
Bachelor's Degree in Chemical Engeneering
Defense date
09.13.2024 10:10
09.13.2024 10:10
Summary
This project is based on the production of synthetic fuel from pure methanol from carbon dioxide capture in the atmosphere, wishing to process 44204 tonnes of raw material per year. This process was carried out by designing a distillation column which would remove all the water from the first equilibrium reactor; another multi-tubular fixed-bed reactor was designed with a zeolite catalyst, allowing us to obtain the purest possible hydrocarbon product for subsequent addition of additives. The main objective of this project is to study an alternative to conventional petrol in order to obtain a more environmentally sustainable fuel.
This project is based on the production of synthetic fuel from pure methanol from carbon dioxide capture in the atmosphere, wishing to process 44204 tonnes of raw material per year. This process was carried out by designing a distillation column which would remove all the water from the first equilibrium reactor; another multi-tubular fixed-bed reactor was designed with a zeolite catalyst, allowing us to obtain the purest possible hydrocarbon product for subsequent addition of additives. The main objective of this project is to study an alternative to conventional petrol in order to obtain a more environmentally sustainable fuel.
Direction
Rodríguez Figueiras, Óscar (Tutorships)
Rodríguez Figueiras, Óscar (Tutorships)
Court
CARBALLA ARCOS, MARTA (Chairman)
EIBES GONZALEZ, GEMMA MARIA (Secretary)
BELLO BUGALLO, PASTORA MARIA (Member)
CARBALLA ARCOS, MARTA (Chairman)
EIBES GONZALEZ, GEMMA MARIA (Secretary)
BELLO BUGALLO, PASTORA MARIA (Member)
Detection of ionic imbalances in the myocardium from electrocardiographic records using machine learning
Authorship
M.G.L.
Computer Engineering Degree
M.G.L.
Computer Engineering Degree
Defense date
07.18.2024 17:30
07.18.2024 17:30
Summary
In this work, various techniques from statistics and machine learning are studied with the objective of developing a model for detecting imbalances in blood potassium concentration using electrocardiographic records. Additionally, it also aims to provide explainability to the results through the identification of the most relevant electrocardiographic features for obtaining the solution. After a cleaning process on the original dataset, we have 2483 electrocardiograms (ECGs) associated with patients with normal, low (hypokalemia), and high (hyperkalemia) potassium concentrations. A feature selection process is carried out on these to obtain a reduced set of variables that allow the correct detection of imbalances in the ECGs and the subsequent explanation of the results. These variables were used to train different models that approached the problem from both a classification perspective (logistic regression, regression trees, Random Forest Classifier, and XGBoost) and a regression perspective (linear regression, decision trees, Random Forest, and XGBoost). Generally, the best results were obtained using the Random Forest model. In the classification of hypokalemia and hyperkalemia, specificities of 95.66% and 86.26%, sensitivities of 13.76% and 48.26%, and AUCs (area under the curve) of 0.73 and 0.76 were achieved. For regression, the R2 obtained was 0.17 and an RMSE (root mean square error) of 0.73. Therefore, none of the models allow the identification of ionic imbalances with a quality comparable to that obtained in previous works, although an advance was made in terms of the explainability of the results.
In this work, various techniques from statistics and machine learning are studied with the objective of developing a model for detecting imbalances in blood potassium concentration using electrocardiographic records. Additionally, it also aims to provide explainability to the results through the identification of the most relevant electrocardiographic features for obtaining the solution. After a cleaning process on the original dataset, we have 2483 electrocardiograms (ECGs) associated with patients with normal, low (hypokalemia), and high (hyperkalemia) potassium concentrations. A feature selection process is carried out on these to obtain a reduced set of variables that allow the correct detection of imbalances in the ECGs and the subsequent explanation of the results. These variables were used to train different models that approached the problem from both a classification perspective (logistic regression, regression trees, Random Forest Classifier, and XGBoost) and a regression perspective (linear regression, decision trees, Random Forest, and XGBoost). Generally, the best results were obtained using the Random Forest model. In the classification of hypokalemia and hyperkalemia, specificities of 95.66% and 86.26%, sensitivities of 13.76% and 48.26%, and AUCs (area under the curve) of 0.73 and 0.76 were achieved. For regression, the R2 obtained was 0.17 and an RMSE (root mean square error) of 0.73. Therefore, none of the models allow the identification of ionic imbalances with a quality comparable to that obtained in previous works, although an advance was made in terms of the explainability of the results.
Direction
FELIX LAMAS, PAULO (Tutorships)
RODRIGUEZ PRESEDO, JESUS MARIA (Co-tutorships)
FELIX LAMAS, PAULO (Tutorships)
RODRIGUEZ PRESEDO, JESUS MARIA (Co-tutorships)
Court
Pardo López, Xosé Manuel (Chairman)
Puentes Cociña, Beltrán (Secretary)
LOPEZ FANDIÑO, JAVIER (Member)
Pardo López, Xosé Manuel (Chairman)
Puentes Cociña, Beltrán (Secretary)
LOPEZ FANDIÑO, JAVIER (Member)
Extension of a Shuffling Robot for Tangible Card Games
Authorship
D.B.P.
Computer Engineering Degree
D.B.P.
Computer Engineering Degree
Defense date
02.16.2024 16:30
02.16.2024 16:30
Summary
In our society, Spanish deck card games stand out for their social function and popularity, especially among older generations. These games, which encourage social interaction and cognitive exercise, such as counting cards and developing strategies, face the challenge of adapting to the digital age to remain relevant and accessible. This thesis focuses on the improvement of a previous prototype of an intelligent card shuffling system for the Spanish deck. The goal is to advance the automation and robustness of the system, allowing its use without manual intervention and improving its accuracy. The objective is to make the system more accessible and attractive to a wide audience, helping to maintain the cultural and social importance of card games in a modern context.
In our society, Spanish deck card games stand out for their social function and popularity, especially among older generations. These games, which encourage social interaction and cognitive exercise, such as counting cards and developing strategies, face the challenge of adapting to the digital age to remain relevant and accessible. This thesis focuses on the improvement of a previous prototype of an intelligent card shuffling system for the Spanish deck. The goal is to advance the automation and robustness of the system, allowing its use without manual intervention and improving its accuracy. The objective is to make the system more accessible and attractive to a wide audience, helping to maintain the cultural and social importance of card games in a modern context.
Direction
CATALA BOLOS, ALEJANDRO (Tutorships)
CORRALES RAMON, JUAN ANTONIO (Co-tutorships)
CATALA BOLOS, ALEJANDRO (Tutorships)
CORRALES RAMON, JUAN ANTONIO (Co-tutorships)
Court
SUAREZ GAREA, JORGE ALBERTO (Coordinator)
BUGARIN DIZ, ALBERTO JOSE (Chairman)
VILLARROYA FERNANDEZ, SEBASTIAN (Secretary)
BORRAJO GARCIA, MARIA ISABEL (Member)
SUAREZ GAREA, JORGE ALBERTO (Coordinator)
BUGARIN DIZ, ALBERTO JOSE (Chairman)
VILLARROYA FERNANDEZ, SEBASTIAN (Secretary)
BORRAJO GARCIA, MARIA ISABEL (Member)
FreshStock: Grocery stock management app for refrigerator and pantry
Authorship
B.G.L.
Computer Engineering Degree
B.G.L.
Computer Engineering Degree
Defense date
07.19.2024 16:30
07.19.2024 16:30
Summary
Food waste in Spanish households exceeds 1200 million kg or liters, according to 2022 data (Ministry of Agriculture). Of this, more than 70% refers to unused food, as purchased, due to expiration or spoilage. This indicates a widespread excess in purchasing products at the supermarket and a lack of control over these in the refrigerator or pantry. As a contribution to reducing the impact of this problem, this thesis develops FreshStock, an Android mobile application that helps individuals and households to know and manage their food inventory and expiration dates, assisting in making more conscious consumption decisions. To achieve this, it will focus on greater usability and simplicity, providing useful features such as generating complete shopping lists with a button, organizing products by labels, and a barcode scanner with automatic Internet search for product information.
Food waste in Spanish households exceeds 1200 million kg or liters, according to 2022 data (Ministry of Agriculture). Of this, more than 70% refers to unused food, as purchased, due to expiration or spoilage. This indicates a widespread excess in purchasing products at the supermarket and a lack of control over these in the refrigerator or pantry. As a contribution to reducing the impact of this problem, this thesis develops FreshStock, an Android mobile application that helps individuals and households to know and manage their food inventory and expiration dates, assisting in making more conscious consumption decisions. To achieve this, it will focus on greater usability and simplicity, providing useful features such as generating complete shopping lists with a button, organizing products by labels, and a barcode scanner with automatic Internet search for product information.
Direction
Triñanes Fernández, Joaquín Ángel (Tutorships)
Álvarez de Celis, Francisco Javier (Co-tutorships)
Gutiérrez García, Francisco (Co-tutorships)
Triñanes Fernández, Joaquín Ángel (Tutorships)
Álvarez de Celis, Francisco Javier (Co-tutorships)
Gutiérrez García, Francisco (Co-tutorships)
Court
Fernández Pena, Anselmo Tomás (Chairman)
SACO LOPEZ, PEDRO JOSE (Secretary)
RODRIGUEZ PRESEDO, JESUS MARIA (Member)
Fernández Pena, Anselmo Tomás (Chairman)
SACO LOPEZ, PEDRO JOSE (Secretary)
RODRIGUEZ PRESEDO, JESUS MARIA (Member)
MIS PICTOCUENTOS: animated and interactive stories for everyone
Authorship
M.O.N.
Computer Engineering Degree
M.O.N.
Computer Engineering Degree
Defense date
07.18.2024 10:30
07.18.2024 10:30
Summary
Pictocuentos is an application born from the desire to provide an interactive form of entertainment based on the use of classic children's literature for girls and boys with communication difficulties, such as those with Autism Spectrum Disorder (ASD). It serves as a tool that can aid their cognitive development by adapting to their needs through the use of pictograms, thus helping them better understand their environment and stimulate language-related aspects. The goal of this Final Year Project is to develop a web app platform that allows users to create their own scenes using preloaded characters and settings taken from popular children's literature. Users can contextualize the scene through dialogues translated into pictograms and can enhance the immersive experience with background music or narration through voice bots. The design, implementation, and testing processes were approached differently for the frontend and backend. Notably, the user testing was conducted by graduates of the Speech Therapy Degree at the University of A Coruña and graduates of the Double Degree in Primary and Early Childhood Education at the University of Santiago de Compostela. This feedback came from individuals experienced in working with children with ASD. The results obtained suggest that Pictocuentos can become a leisure application that caregivers of such individuals can use to teach vocabulary in a fun and familiar way, fostering creativity while working on communication skills through pictograms.
Pictocuentos is an application born from the desire to provide an interactive form of entertainment based on the use of classic children's literature for girls and boys with communication difficulties, such as those with Autism Spectrum Disorder (ASD). It serves as a tool that can aid their cognitive development by adapting to their needs through the use of pictograms, thus helping them better understand their environment and stimulate language-related aspects. The goal of this Final Year Project is to develop a web app platform that allows users to create their own scenes using preloaded characters and settings taken from popular children's literature. Users can contextualize the scene through dialogues translated into pictograms and can enhance the immersive experience with background music or narration through voice bots. The design, implementation, and testing processes were approached differently for the frontend and backend. Notably, the user testing was conducted by graduates of the Speech Therapy Degree at the University of A Coruña and graduates of the Double Degree in Primary and Early Childhood Education at the University of Santiago de Compostela. This feedback came from individuals experienced in working with children with ASD. The results obtained suggest that Pictocuentos can become a leisure application that caregivers of such individuals can use to teach vocabulary in a fun and familiar way, fostering creativity while working on communication skills through pictograms.
Direction
TABOADA IGLESIAS, MARÍA JESÚS (Tutorships)
García García, Miguel (Co-tutorships)
TABOADA IGLESIAS, MARÍA JESÚS (Tutorships)
García García, Miguel (Co-tutorships)
Court
VIDAL AGUIAR, JUAN CARLOS (Chairman)
DOSIL LAGO, RAQUEL (Secretary)
SAAVEDRA NIEVES, ALEJANDRO (Member)
VIDAL AGUIAR, JUAN CARLOS (Chairman)
DOSIL LAGO, RAQUEL (Secretary)
SAAVEDRA NIEVES, ALEJANDRO (Member)
BUHO - Best Universal Hardening Operatives
Authorship
B.M.V.
Computer Engineering Degree
B.M.V.
Computer Engineering Degree
Defense date
07.18.2024 10:00
07.18.2024 10:00
Summary
Nowadays, as computers have become an essential part of people's lives, it is necessary to implement mechanisms to protect all the information that is being handled. These mechanisms can be overwhelming for the average user, who may not have extensive knowledge about computers. To solve this problem and help users protect their systems, BUHO (Best Universal Hardening Operatives) has been developed. In this application, users select one of the available technologies and complete a survey about it. In this survey, they answer questions related to the settings, current state, and future usage of the system. Based on their responses, they receive a series of recommendations to improve the security of their systems according to their specific situation.
Nowadays, as computers have become an essential part of people's lives, it is necessary to implement mechanisms to protect all the information that is being handled. These mechanisms can be overwhelming for the average user, who may not have extensive knowledge about computers. To solve this problem and help users protect their systems, BUHO (Best Universal Hardening Operatives) has been developed. In this application, users select one of the available technologies and complete a survey about it. In this survey, they answer questions related to the settings, current state, and future usage of the system. Based on their responses, they receive a series of recommendations to improve the security of their systems according to their specific situation.
Direction
CARIÑENA AMIGO, MARIA PURIFICACION (Tutorships)
Viqueira Hierro, Bernardo (Co-tutorships)
CARIÑENA AMIGO, MARIA PURIFICACION (Tutorships)
Viqueira Hierro, Bernardo (Co-tutorships)
Court
VIDAL AGUIAR, JUAN CARLOS (Chairman)
DOSIL LAGO, RAQUEL (Secretary)
SAAVEDRA NIEVES, ALEJANDRO (Member)
VIDAL AGUIAR, JUAN CARLOS (Chairman)
DOSIL LAGO, RAQUEL (Secretary)
SAAVEDRA NIEVES, ALEJANDRO (Member)
Design and Development of a Mobile Application for the Stock Management and Shopping Lists
Authorship
J.R.G.
Computer Engineering Degree
J.R.G.
Computer Engineering Degree
Defense date
07.19.2024 10:00
07.19.2024 10:00
Summary
Within the framework of the development of mobile applications for product management, this project has addressed the efficient management of stock and shopping lists, resulting in a mobile application designed to optimize the organization and tracking of stored products and necessary purchases. The application focuses on two main functionalities: stock management and shopping list management. It allows users to add products to inventory via barcode scanning or manual entry, categorize them, and assign expiration dates, helping reduce food waste through proactive notifications. It also offers complete details of each product, including nutritional information. Regarding shopping list management, the application allows you to create and organize multiple lists, search for products, add external products and create custom categories. Additionally, it automatically suggests adding out-of-stock products from stock to purchase lists, maintaining efficient synchronization between inventory and purchasing needs. Usability has been emphasized, ensuring an intuitive and efficient user experience. The conclusions of the project indicate that the proposed objectives have been achieved, providing a useful and effective tool for managing inventories and shopping lists, and contributing to the reduction of food waste.
Within the framework of the development of mobile applications for product management, this project has addressed the efficient management of stock and shopping lists, resulting in a mobile application designed to optimize the organization and tracking of stored products and necessary purchases. The application focuses on two main functionalities: stock management and shopping list management. It allows users to add products to inventory via barcode scanning or manual entry, categorize them, and assign expiration dates, helping reduce food waste through proactive notifications. It also offers complete details of each product, including nutritional information. Regarding shopping list management, the application allows you to create and organize multiple lists, search for products, add external products and create custom categories. Additionally, it automatically suggests adding out-of-stock products from stock to purchase lists, maintaining efficient synchronization between inventory and purchasing needs. Usability has been emphasized, ensuring an intuitive and efficient user experience. The conclusions of the project indicate that the proposed objectives have been achieved, providing a useful and effective tool for managing inventories and shopping lists, and contributing to the reduction of food waste.
Direction
Taboada González, José Ángel (Tutorships)
AL-RAHAMNEH , ANAS JAMAL SAAD (Co-tutorships)
Taboada González, José Ángel (Tutorships)
AL-RAHAMNEH , ANAS JAMAL SAAD (Co-tutorships)
Court
Barro Ameneiro, Senén (Chairman)
TOBAR QUINTANAR, ALEJANDRO JOSE (Secretary)
LADRA GONZALEZ, MANUEL EULOGIO (Member)
Barro Ameneiro, Senén (Chairman)
TOBAR QUINTANAR, ALEJANDRO JOSE (Secretary)
LADRA GONZALEZ, MANUEL EULOGIO (Member)
Training and Sports Tracking Platform
Authorship
M.C.D.
Computer Engineering Degree
M.C.D.
Computer Engineering Degree
Defense date
02.16.2024 17:00
02.16.2024 17:00
Summary
The aim of this Final Degree Project is to develop a multiplatform application (web, iOS, Android) to facilitate communication between sports coaches and athletes. The application will allow monitoring the athlete's progress and creating personalised exercise routines adapted to their individual circumstances. Users will be able to access the application from their preferred devices to record exercise performance, describe feelings throughout the sessions and visualise their personal progress. The implementation using a cross-platform framework stands out in this system, as it provides similar performance to native development on each device with fewer lines of code, giving the user the flexibility to work on any platform. Throughout the development of the project, the decisions made to ensure the efficiency of the system will be addressed, with a particular focus on the user experience.
The aim of this Final Degree Project is to develop a multiplatform application (web, iOS, Android) to facilitate communication between sports coaches and athletes. The application will allow monitoring the athlete's progress and creating personalised exercise routines adapted to their individual circumstances. Users will be able to access the application from their preferred devices to record exercise performance, describe feelings throughout the sessions and visualise their personal progress. The implementation using a cross-platform framework stands out in this system, as it provides similar performance to native development on each device with fewer lines of code, giving the user the flexibility to work on any platform. Throughout the development of the project, the decisions made to ensure the efficiency of the system will be addressed, with a particular focus on the user experience.
Direction
VARELA PET, JOSE (Tutorships)
VARELA PET, JOSE (Tutorships)
Court
SUAREZ GAREA, JORGE ALBERTO (Coordinator)
BUGARIN DIZ, ALBERTO JOSE (Chairman)
VILLARROYA FERNANDEZ, SEBASTIAN (Secretary)
BORRAJO GARCIA, MARIA ISABEL (Member)
SUAREZ GAREA, JORGE ALBERTO (Coordinator)
BUGARIN DIZ, ALBERTO JOSE (Chairman)
VILLARROYA FERNANDEZ, SEBASTIAN (Secretary)
BORRAJO GARCIA, MARIA ISABEL (Member)
AIWantTheJob
Authorship
J.A.V.
Computer Engineering Degree
J.A.V.
Computer Engineering Degree
Defense date
07.18.2024 09:30
07.18.2024 09:30
Summary
In the last ten years, the popularity, capability, and investment in Artificial Intelligence techniques based on deep learning has increased considerably. Large Language Models (LLMs) are currently the most widely spread tools for text generation in a wide range of tasks. The size of these models and the corpus with which they are trained have turned them into large knowledge bases. One of the most promising lines of work focuses on using these models as study aid tools. An example of this trend could be Google Workspaces for Education. Job interviews are an integral part of our lives, and surprisingly, there aren't many interactive tools to help in their preparation. This bachelor's project titled AIWantTheJob is created to explore the use of LLMs as learning aid tools in order to cover this detected gap. AIWantTheJob aims to develop a job interview simulator to prepare candidates for a specific position. Using Deep Learning and NLP techniques, the system generates realistic and personalized interview scenarios that provide real-time assessments to improve interviewees' skills. To enhance the results of the responsible LLM, this work proposes a RAG (Retrieval-Augmented Generation) that generates and provides access to the specific knowledge base of the job position for which the user is preparing to interview. Additionally, a parallel tool called Nemo Guardrails is used, which serves, among many other things, to define conversation flows between the person and the LLM. The system has been implemented as an app that can be deployed in virtualized environments quickly and requires low-cost computing resources (considering the standard requirements of AI-based applications). Furthermore, the application is equipped with realistic deployment techniques to simulate a real production environment. Beyond the business case of the interviewer, the created system allows any developer to implement chatbots on other topics with specific actions, requiring minimal time investment and great ease. The proyect provides a chatbot that adequately performs the three proposed tasks (asking, evaluating, and simulating responses), adhering well to the context and hallucinating relatively little. Moreover, we can conclude that the results can be improved by using a more intelligent model, which would require more computational resources than those available during the project.
In the last ten years, the popularity, capability, and investment in Artificial Intelligence techniques based on deep learning has increased considerably. Large Language Models (LLMs) are currently the most widely spread tools for text generation in a wide range of tasks. The size of these models and the corpus with which they are trained have turned them into large knowledge bases. One of the most promising lines of work focuses on using these models as study aid tools. An example of this trend could be Google Workspaces for Education. Job interviews are an integral part of our lives, and surprisingly, there aren't many interactive tools to help in their preparation. This bachelor's project titled AIWantTheJob is created to explore the use of LLMs as learning aid tools in order to cover this detected gap. AIWantTheJob aims to develop a job interview simulator to prepare candidates for a specific position. Using Deep Learning and NLP techniques, the system generates realistic and personalized interview scenarios that provide real-time assessments to improve interviewees' skills. To enhance the results of the responsible LLM, this work proposes a RAG (Retrieval-Augmented Generation) that generates and provides access to the specific knowledge base of the job position for which the user is preparing to interview. Additionally, a parallel tool called Nemo Guardrails is used, which serves, among many other things, to define conversation flows between the person and the LLM. The system has been implemented as an app that can be deployed in virtualized environments quickly and requires low-cost computing resources (considering the standard requirements of AI-based applications). Furthermore, the application is equipped with realistic deployment techniques to simulate a real production environment. Beyond the business case of the interviewer, the created system allows any developer to implement chatbots on other topics with specific actions, requiring minimal time investment and great ease. The proyect provides a chatbot that adequately performs the three proposed tasks (asking, evaluating, and simulating responses), adhering well to the context and hallucinating relatively little. Moreover, we can conclude that the results can be improved by using a more intelligent model, which would require more computational resources than those available during the project.
Direction
TABOADA IGLESIAS, MARÍA JESÚS (Tutorships)
Vinci , Walter (Co-tutorships)
TABOADA IGLESIAS, MARÍA JESÚS (Tutorships)
Vinci , Walter (Co-tutorships)
Court
Cotos Yáñez, José Manuel (Chairman)
QUESADA BARRIUSO, PABLO (Secretary)
GAGO COUSO, FELIPE (Member)
Cotos Yáñez, José Manuel (Chairman)
QUESADA BARRIUSO, PABLO (Secretary)
GAGO COUSO, FELIPE (Member)
De Rham cohomology
Authorship
C.L.A.
Double bachelor degree of Engeneering in Information Technology and Mathematics
C.L.A.
Double bachelor degree of Engeneering in Information Technology and Mathematics
Defense date
07.17.2024 10:10
07.17.2024 10:10
Summary
Differential forms constitute a highly relevant mathematical object in Differential Topology and Differential Geometry with important applications to Physics. The aim of this work is to study different uses of differential forms and to show how they allow for effectively addressing numerous problems. First, basic notions about fiber bundle theory will be given, which are fundamental to address the rest of the work. Following this, the concept of orientation of a manifold will be defined, and a brief introduction to semi-Riemannian manifolds will be presented. This will allow us to generalize basic concepts of vector calculus to differentiable manifolds. Next, integration on manifolds will be studied, and the generalized Stokes' Theorem will be proven. These concepts will also be addressed for the case when the manifold is non-orientable. Finally, de Rham cohomology and its main properties will be studied. In addition to all this, applications to electromagnetism of the studied concepts will be included throughout the work. Specifically, Maxwell's Equations will be formulated using differential forms, Stokes' Theorem will be used to give an integral version of the equations, and de Rham cohomology will be used to illustrate wormholes and magnetic monopoles, physical phenomena never detected experimentally.
Differential forms constitute a highly relevant mathematical object in Differential Topology and Differential Geometry with important applications to Physics. The aim of this work is to study different uses of differential forms and to show how they allow for effectively addressing numerous problems. First, basic notions about fiber bundle theory will be given, which are fundamental to address the rest of the work. Following this, the concept of orientation of a manifold will be defined, and a brief introduction to semi-Riemannian manifolds will be presented. This will allow us to generalize basic concepts of vector calculus to differentiable manifolds. Next, integration on manifolds will be studied, and the generalized Stokes' Theorem will be proven. These concepts will also be addressed for the case when the manifold is non-orientable. Finally, de Rham cohomology and its main properties will be studied. In addition to all this, applications to electromagnetism of the studied concepts will be included throughout the work. Specifically, Maxwell's Equations will be formulated using differential forms, Stokes' Theorem will be used to give an integral version of the equations, and de Rham cohomology will be used to illustrate wormholes and magnetic monopoles, physical phenomena never detected experimentally.
Direction
Álvarez López, Jesús Antonio (Tutorships)
Álvarez López, Jesús Antonio (Tutorships)
Court
GARCIA RIO, EDUARDO (Chairman)
RIVERO SALGADO, OSCAR (Secretary)
CASAS MENDEZ, BALBINA VIRGINIA (Member)
GARCIA RIO, EDUARDO (Chairman)
RIVERO SALGADO, OSCAR (Secretary)
CASAS MENDEZ, BALBINA VIRGINIA (Member)
Semi-supervised learning for object detection
Authorship
C.L.A.
Double bachelor degree of Engeneering in Information Technology and Mathematics
C.L.A.
Double bachelor degree of Engeneering in Information Technology and Mathematics
Defense date
07.18.2024 16:30
07.18.2024 16:30
Summary
Object detection is one of the main challenges addressed in the field of Computer Vision. Traditional detectors require a large amount of labeled images for training. This presents a significant limitation, as image annotation is costly and data availability is often limited. In this context, semi-supervised learning for object detection emerges, addressing the scenario where there are few labeled images but a large number of unlabeled images. This work falls within this area and focuses on the study of Teacher-Student architectures. Specifically, the Unbiased Teacher v2 detector is studied. The aim of this work is to study different techniques that are beneficial in other learning contexts and analyze their applicability to this detector. Specifically, the following proposals will be analyzed: incorporation of modules typical of few-shot architectures (specifically the Gradient Decoupled Layer (GDL) and Prototypical Calibration Block (PCB) proposed in the DeFRCN detector); replacement of the fixed threshold strategy for filtering pseudo-labels with a flexible threshold strategy; and the use of a label assignment strategy based on Optimal Transport Assignment (OTA). The experiments conducted show that the use of GDL is beneficial for the detector's performance. Additionally, although the other proposals did not improve the detector, the experimentation revealed several challenges that arise when adapting these strategies to a detector with the characteristics of Unbiased Teacher v2. This information may be relevant for future analyses of the problem.
Object detection is one of the main challenges addressed in the field of Computer Vision. Traditional detectors require a large amount of labeled images for training. This presents a significant limitation, as image annotation is costly and data availability is often limited. In this context, semi-supervised learning for object detection emerges, addressing the scenario where there are few labeled images but a large number of unlabeled images. This work falls within this area and focuses on the study of Teacher-Student architectures. Specifically, the Unbiased Teacher v2 detector is studied. The aim of this work is to study different techniques that are beneficial in other learning contexts and analyze their applicability to this detector. Specifically, the following proposals will be analyzed: incorporation of modules typical of few-shot architectures (specifically the Gradient Decoupled Layer (GDL) and Prototypical Calibration Block (PCB) proposed in the DeFRCN detector); replacement of the fixed threshold strategy for filtering pseudo-labels with a flexible threshold strategy; and the use of a label assignment strategy based on Optimal Transport Assignment (OTA). The experiments conducted show that the use of GDL is beneficial for the detector's performance. Additionally, although the other proposals did not improve the detector, the experimentation revealed several challenges that arise when adapting these strategies to a detector with the characteristics of Unbiased Teacher v2. This information may be relevant for future analyses of the problem.
Direction
MUCIENTES MOLINA, MANUEL FELIPE (Tutorships)
CORES COSTA, DANIEL (Co-tutorships)
MUCIENTES MOLINA, MANUEL FELIPE (Tutorships)
CORES COSTA, DANIEL (Co-tutorships)
Court
VAZQUEZ CENDON, MARIA ELENA (Chairman)
CHAVES FRAGA, DAVID (Secretary)
SUAREZ GAREA, JORGE ALBERTO (Member)
VAZQUEZ CENDON, MARIA ELENA (Chairman)
CHAVES FRAGA, DAVID (Secretary)
SUAREZ GAREA, JORGE ALBERTO (Member)
Dental Diagnosis: Automatic Classification of Oral Panoramic X-ray Images Using Deep Learning Techniques
Authorship
E.P.V.
Double bachelor degree of Engeneering in Information Technology and Mathematics
E.P.V.
Double bachelor degree of Engeneering in Information Technology and Mathematics
Defense date
07.18.2024 12:30
07.18.2024 12:30
Summary
n the field of medical image analysis, the classification of dental X-rays is crucial for detecting pathologies, as it allows each image to be labeled with the presence or absence of a specific condition. Recently, convolutional neural networks (CNNs) have proven effective in this area. This thesis develops a model based on ResNet architectures to classify panoramic dental X-rays. Using a dataset provided by the OSRG from the University of Santiago de Compostela, which includes 8 classes of pathologies, data augmentation and class balancing techniques were applied to address class imbalance. Various ResNet architectures were explored, and hyperparameters such as batch size and image resolution were adjusted. This report details the experiments conducted, the results obtained, and the conclusions and potential extensions of the work.
n the field of medical image analysis, the classification of dental X-rays is crucial for detecting pathologies, as it allows each image to be labeled with the presence or absence of a specific condition. Recently, convolutional neural networks (CNNs) have proven effective in this area. This thesis develops a model based on ResNet architectures to classify panoramic dental X-rays. Using a dataset provided by the OSRG from the University of Santiago de Compostela, which includes 8 classes of pathologies, data augmentation and class balancing techniques were applied to address class imbalance. Various ResNet architectures were explored, and hyperparameters such as batch size and image resolution were adjusted. This report details the experiments conducted, the results obtained, and the conclusions and potential extensions of the work.
Direction
VILA BLANCO, NICOLAS (Tutorships)
CARREIRA NOUCHE, MARIA JOSE (Co-tutorships)
TOMAS CARMONA, INMACULADA (Co-tutorships)
VILA BLANCO, NICOLAS (Tutorships)
CARREIRA NOUCHE, MARIA JOSE (Co-tutorships)
TOMAS CARMONA, INMACULADA (Co-tutorships)
Court
VIDAL AGUIAR, JUAN CARLOS (Chairman)
DOSIL LAGO, RAQUEL (Secretary)
SAAVEDRA NIEVES, ALEJANDRO (Member)
VIDAL AGUIAR, JUAN CARLOS (Chairman)
DOSIL LAGO, RAQUEL (Secretary)
SAAVEDRA NIEVES, ALEJANDRO (Member)
Existence of periodic solutions to the Mathieu equation
Authorship
E.P.V.
Double bachelor degree of Engeneering in Information Technology and Mathematics
E.P.V.
Double bachelor degree of Engeneering in Information Technology and Mathematics
Defense date
07.16.2024 17:45
07.16.2024 17:45
Summary
In this work, we will consider the dynamics of an electron beam guided by an axially symmetric periodic magnetic field. This type of mechanism is known as an electron gun or microwave valve and is a part of many scientific, industrial, and domestic electronic devices. Its dynamics can be modeled using the Mathieu equation, which is a second-order equation with singularities and can be treated as a particular case of the Hill equation. Thus, we will focus on modeling the phenomenon and describe some results that guarantee the existence of periodic solutions with constant sign.
In this work, we will consider the dynamics of an electron beam guided by an axially symmetric periodic magnetic field. This type of mechanism is known as an electron gun or microwave valve and is a part of many scientific, industrial, and domestic electronic devices. Its dynamics can be modeled using the Mathieu equation, which is a second-order equation with singularities and can be treated as a particular case of the Hill equation. Thus, we will focus on modeling the phenomenon and describe some results that guarantee the existence of periodic solutions with constant sign.
Direction
CABADA FERNANDEZ, ALBERTO (Tutorships)
CABADA FERNANDEZ, ALBERTO (Tutorships)
Court
FEBRERO BANDE, MANUEL (Chairman)
BUEDO FERNANDEZ, SEBASTIAN (Secretary)
RODRIGUEZ GARCIA, JERONIMO (Member)
FEBRERO BANDE, MANUEL (Chairman)
BUEDO FERNANDEZ, SEBASTIAN (Secretary)
RODRIGUEZ GARCIA, JERONIMO (Member)
Piecewise linear motion planning in robotics
Authorship
M.T.L.
Double bachelor degree of Engeneering in Information Technology and Mathematics
M.T.L.
Double bachelor degree of Engeneering in Information Technology and Mathematics
Defense date
07.17.2024 12:50
07.17.2024 12:50
Summary
The topological complexity is an invariant proposed by Michael Farber, arising in the context of motion planning in robotics. In this thesis, a discrete analogy proposed by Jesús González is developed: the simplicial complexity of a simplicial complex, which is equivalent to the topological complexity of its geometric realization but is computable using combinatorial methods.
The topological complexity is an invariant proposed by Michael Farber, arising in the context of motion planning in robotics. In this thesis, a discrete analogy proposed by Jesús González is developed: the simplicial complexity of a simplicial complex, which is equivalent to the topological complexity of its geometric realization but is computable using combinatorial methods.
Direction
Macías Virgós, Enrique (Tutorships)
MOSQUERA LOIS, DAVID (Co-tutorships)
Macías Virgós, Enrique (Tutorships)
MOSQUERA LOIS, DAVID (Co-tutorships)
Court
GARCIA RIO, EDUARDO (Chairman)
RIVERO SALGADO, OSCAR (Secretary)
CASAS MENDEZ, BALBINA VIRGINIA (Member)
GARCIA RIO, EDUARDO (Chairman)
RIVERO SALGADO, OSCAR (Secretary)
CASAS MENDEZ, BALBINA VIRGINIA (Member)
Review and analysis of quantum artificial intelligence models
Authorship
M.T.L.
Double bachelor degree of Engeneering in Information Technology and Mathematics
M.T.L.
Double bachelor degree of Engeneering in Information Technology and Mathematics
Defense date
07.18.2024 11:30
07.18.2024 11:30
Summary
Quantum computing is a promising field nowadays, aiming to a revolution in computer science by leveraging the advantages of quantum mechanics. Its combination with the field of artificial intelligence gives rise to the new field of Quantum Machine Learning. This thesis provides an introduction to its techniques and models. After a literature review, a specific model is selected: a classifier based on a novel technique known as data re-uploading. This model will be implemented using the standard Qiskit library and experimentation will be conducted to evaluate it.
Quantum computing is a promising field nowadays, aiming to a revolution in computer science by leveraging the advantages of quantum mechanics. Its combination with the field of artificial intelligence gives rise to the new field of Quantum Machine Learning. This thesis provides an introduction to its techniques and models. After a literature review, a specific model is selected: a classifier based on a novel technique known as data re-uploading. This model will be implemented using the standard Qiskit library and experimentation will be conducted to evaluate it.
Direction
BUGARIN DIZ, ALBERTO JOSE (Tutorships)
Fernández Pena, Anselmo Tomás (Co-tutorships)
BUGARIN DIZ, ALBERTO JOSE (Tutorships)
Fernández Pena, Anselmo Tomás (Co-tutorships)
Court
Cotos Yáñez, José Manuel (Chairman)
QUESADA BARRIUSO, PABLO (Secretary)
GAGO COUSO, FELIPE (Member)
Cotos Yáñez, José Manuel (Chairman)
QUESADA BARRIUSO, PABLO (Secretary)
GAGO COUSO, FELIPE (Member)
Methods of Classification and Ensemble of Classifiers in Supervised Learning
Authorship
A.G.L.
Double bachelor degree of Engeneering in Information Technology and Mathematics
A.G.L.
Double bachelor degree of Engeneering in Information Technology and Mathematics
Defense date
07.16.2024 12:30
07.16.2024 12:30
Summary
In this work, various ensemble techniques in supervised learning are analyzed, focusing on bagging, random forests, and AdaBoost. Initially, the fundamentals of statistical classification and supervised learning are explained. Subsequently, different strategies for combining classifier outputs when they consist of predictions and continuous values are examined. Finally, the ensemble methods are detailed, highlighting the characteristics that differentiate them.
In this work, various ensemble techniques in supervised learning are analyzed, focusing on bagging, random forests, and AdaBoost. Initially, the fundamentals of statistical classification and supervised learning are explained. Subsequently, different strategies for combining classifier outputs when they consist of predictions and continuous values are examined. Finally, the ensemble methods are detailed, highlighting the characteristics that differentiate them.
Direction
PATEIRO LOPEZ, BEATRIZ (Tutorships)
Rodríguez Acevedo, Iria (Co-tutorships)
PATEIRO LOPEZ, BEATRIZ (Tutorships)
Rodríguez Acevedo, Iria (Co-tutorships)
Court
GONZALEZ MANTEIGA, WENCESLAO (Chairman)
PAEZ GUILLAN, MARIA PILAR (Secretary)
ALVAREZ DIOS, JOSE ANTONIO (Member)
GONZALEZ MANTEIGA, WENCESLAO (Chairman)
PAEZ GUILLAN, MARIA PILAR (Secretary)
ALVAREZ DIOS, JOSE ANTONIO (Member)
Effective adaptation of generative adversarial networks for processing multidimensional remote sensing images
Authorship
A.G.L.
Double bachelor degree of Engeneering in Information Technology and Mathematics
A.G.L.
Double bachelor degree of Engeneering in Information Technology and Mathematics
Defense date
07.19.2024 17:00
07.19.2024 17:00
Summary
In recent years, various deep learning classification models have been proposed, adding to the existing methods for classifying multispectral remote sensing images. Within this framework, in this work, a conditional generative adversarial network (GAN), based on the StyleGAN2 model architecture, was adapted to the problem of classifying multispectral remote sensing images. Subsequently, a study was conducted on the network's capability to both generate and classify high spatial resolution multispectral images corresponding to Galician rivers. Finally, the results achieved with this classification model were compared to other models used in multispectral remote sensing problems. It was observed that the conditional StyleGAN2 achieves results close to other classification schemes that do not use generated datasets, such as convolutional neural networks, but falls short of methods specifically designed to generate samples for classification problems with imbalanced classes, such as ResBaGAN.
In recent years, various deep learning classification models have been proposed, adding to the existing methods for classifying multispectral remote sensing images. Within this framework, in this work, a conditional generative adversarial network (GAN), based on the StyleGAN2 model architecture, was adapted to the problem of classifying multispectral remote sensing images. Subsequently, a study was conducted on the network's capability to both generate and classify high spatial resolution multispectral images corresponding to Galician rivers. Finally, the results achieved with this classification model were compared to other models used in multispectral remote sensing problems. It was observed that the conditional StyleGAN2 achieves results close to other classification schemes that do not use generated datasets, such as convolutional neural networks, but falls short of methods specifically designed to generate samples for classification problems with imbalanced classes, such as ResBaGAN.
Direction
Argüello Pedreira, Francisco Santiago (Tutorships)
Blanco Heras, Dora (Co-tutorships)
Argüello Pedreira, Francisco Santiago (Tutorships)
Blanco Heras, Dora (Co-tutorships)
Court
Fernández Pena, Anselmo Tomás (Chairman)
SACO LOPEZ, PEDRO JOSE (Secretary)
RODRIGUEZ PRESEDO, JESUS MARIA (Member)
Fernández Pena, Anselmo Tomás (Chairman)
SACO LOPEZ, PEDRO JOSE (Secretary)
RODRIGUEZ PRESEDO, JESUS MARIA (Member)
Statistical learning for algorithm selection in optimisation problems
Authorship
A.F.E.
Double bachelor degree of Engeneering in Information Technology and Mathematics
A.F.E.
Double bachelor degree of Engeneering in Information Technology and Mathematics
Defense date
07.16.2024 18:00
07.16.2024 18:00
Summary
In this work, statistical learning techniques will be used to predict the best performing global optimiser for a non-linear mathematical programming problem. First of all, a method for solving integer linear programming problems and its adaptation to the nonlinear case, where new difficulties will arise, will be explained. Subsequently, the statistical learning problem and two techniques that allow to fit a model and create predictions will be presented: linear regression and single hidden layer neural networks. These techniques will allow learning to be performed on a set of problems and the results to be obtained, looking at the performance of the different optimisers.
In this work, statistical learning techniques will be used to predict the best performing global optimiser for a non-linear mathematical programming problem. First of all, a method for solving integer linear programming problems and its adaptation to the nonlinear case, where new difficulties will arise, will be explained. Subsequently, the statistical learning problem and two techniques that allow to fit a model and create predictions will be presented: linear regression and single hidden layer neural networks. These techniques will allow learning to be performed on a set of problems and the results to be obtained, looking at the performance of the different optimisers.
Direction
GONZALEZ DIAZ, JULIO (Tutorships)
GOMEZ CASARES, IGNACIO (Co-tutorships)
GONZALEZ DIAZ, JULIO (Tutorships)
GOMEZ CASARES, IGNACIO (Co-tutorships)
Court
GARCIA RIO, EDUARDO (Chairman)
RIVERO SALGADO, OSCAR (Secretary)
CASAS MENDEZ, BALBINA VIRGINIA (Member)
GARCIA RIO, EDUARDO (Chairman)
RIVERO SALGADO, OSCAR (Secretary)
CASAS MENDEZ, BALBINA VIRGINIA (Member)
YourTurn!
Authorship
A.F.E.
Double bachelor degree of Engeneering in Information Technology and Mathematics
A.F.E.
Double bachelor degree of Engeneering in Information Technology and Mathematics
Defense date
07.19.2024 11:00
07.19.2024 11:00
Summary
The aim of this work is to develop a web application to allow a Magic game to be played in real time and remotely. The developed system will allow sharing the players' game table by using a video camera and will identify the cards, displaying their information and translating it in case the players speak a different language.
The aim of this work is to develop a web application to allow a Magic game to be played in real time and remotely. The developed system will allow sharing the players' game table by using a video camera and will identify the cards, displaying their information and translating it in case the players speak a different language.
Direction
TOBAR QUINTANAR, ALEJANDRO JOSE (Tutorships)
GARCIA LLORENS, LUIS VICENTE (Co-tutorships)
Ibán Sánchez, Armando (Co-tutorships)
TOBAR QUINTANAR, ALEJANDRO JOSE (Tutorships)
GARCIA LLORENS, LUIS VICENTE (Co-tutorships)
Ibán Sánchez, Armando (Co-tutorships)
Court
BARJA PEREZ, JAVIER (Chairman)
ORDOÑEZ IGLESIAS, ALVARO (Secretary)
MOSQUERA GONZALEZ, ANTONIO (Member)
BARJA PEREZ, JAVIER (Chairman)
ORDOÑEZ IGLESIAS, ALVARO (Secretary)
MOSQUERA GONZALEZ, ANTONIO (Member)
Complex multiplication theory and Kronecker's Jugendtraum
Authorship
H.V.R.
Double bachelor degree of Engeneering in Information Technology and Mathematics
H.V.R.
Double bachelor degree of Engeneering in Information Technology and Mathematics
Defense date
09.12.2024 17:20
09.12.2024 17:20
Summary
Kronecker's Jugendtraum is related to the problem of finding all abelian extensions of a quadratic imaginary field, therefore generalising the Kronecker-Weber theorem. To study this issue, the main tool is a certain class of elliptic curves, with a specially rich arithmetic structure, called complex multiplication. In this work we will explore the connection between these two topics. We will start by introducing, on one side, the elliptic curves, defining over them a group structure, focusing on finite order points; and on the other, extensions of Q, emphasising those with an abelian Galois group, as well as those generated by roots of unity or elliptic curve points. Afterwards, we will present the concept of complex multiplication and, finally, with the help from group representations, all the previous theory will be used to study a particular case of Kronecker's Jugendtraum, looking at the abelian extensions of Q(i).
Kronecker's Jugendtraum is related to the problem of finding all abelian extensions of a quadratic imaginary field, therefore generalising the Kronecker-Weber theorem. To study this issue, the main tool is a certain class of elliptic curves, with a specially rich arithmetic structure, called complex multiplication. In this work we will explore the connection between these two topics. We will start by introducing, on one side, the elliptic curves, defining over them a group structure, focusing on finite order points; and on the other, extensions of Q, emphasising those with an abelian Galois group, as well as those generated by roots of unity or elliptic curve points. Afterwards, we will present the concept of complex multiplication and, finally, with the help from group representations, all the previous theory will be used to study a particular case of Kronecker's Jugendtraum, looking at the abelian extensions of Q(i).
Direction
GAGO COUSO, FELIPE (Tutorships)
RIVERO SALGADO, OSCAR (Co-tutorships)
GAGO COUSO, FELIPE (Tutorships)
RIVERO SALGADO, OSCAR (Co-tutorships)
Court
OTERO ESPINAR, MARIA VICTORIA (Chairman)
GONZALEZ DIAZ, JULIO (Secretary)
Jeremías López, Ana (Member)
OTERO ESPINAR, MARIA VICTORIA (Chairman)
GONZALEZ DIAZ, JULIO (Secretary)
Jeremías López, Ana (Member)
Recovery water treatment plant for agriculture
Authorship
S.T.B.
Bachelor's Degree in Chemical Engeneering
S.T.B.
Bachelor's Degree in Chemical Engeneering
Defense date
07.16.2024 14:00
07.16.2024 14:00
Summary
Design of a water recovery plant that uses reverse osmosis as a treatment to reduce electrical conductivity. It utilizes water from the Mazarrón wastewater treatment plant and adjusts the characteristics of this effluent so that it can be used by the Irrigation Community.
Design of a water recovery plant that uses reverse osmosis as a treatment to reduce electrical conductivity. It utilizes water from the Mazarrón wastewater treatment plant and adjusts the characteristics of this effluent so that it can be used by the Irrigation Community.
Direction
GARRIDO FERNANDEZ, JUAN MANUEL (Tutorships)
GARRIDO FERNANDEZ, JUAN MANUEL (Tutorships)
Court
CARBALLA ARCOS, MARTA (Chairman)
EIBES GONZALEZ, GEMMA MARIA (Secretary)
BELLO BUGALLO, PASTORA MARIA (Member)
CARBALLA ARCOS, MARTA (Chairman)
EIBES GONZALEZ, GEMMA MARIA (Secretary)
BELLO BUGALLO, PASTORA MARIA (Member)
Use of Google Earth Engine within a framework for estimating indicators in urban areas
Authorship
A.P.M.
Computer Engineering Degree
A.P.M.
Computer Engineering Degree
Defense date
07.18.2024 11:30
07.18.2024 11:30
Summary
The development of this project has included several significant contributions in the field of monitoring land use in urban environments through the use of advanced technologies. These contributions are: Integration of Google Earth Engine for Urban Monitoring. A system has been implemented that uses Google Earth Engine to obtain near real-time geospatial data. This integration allows access to up-to-date and accurate land use data, improving the ability for continuous monitoring and informed decision-making in the urban context. Development of Automated Script in Python. A Python script has been developed to automate the acquisition of land use data. This script enables the classification of data obtained from satellite images, facilitating the analysis and calculation of key indicators. Interactive Data Visualization with Power BI. An interactive dashboard has been created in Power BI that allows users to visualize and analyze land use data intuitively. Contribution to Sustainable Urban Development. By providing an accurate and efficient tool for land use monitoring, the project contributes to sustainable urban development. The ability to analyze changes in land use and calculate key indicators allows authorities and urban planners to make data-driven decisions, promoting responsible and sustainable environmental management. Foundation for Future Expansions and Improvements. The developed system establishes a solid foundation for future research and development. Additional improvements and new functionalities can be incorporated, such as the integration of other geospatial data sources or the application of artificial intelligence techniques for predictive analysis.
The development of this project has included several significant contributions in the field of monitoring land use in urban environments through the use of advanced technologies. These contributions are: Integration of Google Earth Engine for Urban Monitoring. A system has been implemented that uses Google Earth Engine to obtain near real-time geospatial data. This integration allows access to up-to-date and accurate land use data, improving the ability for continuous monitoring and informed decision-making in the urban context. Development of Automated Script in Python. A Python script has been developed to automate the acquisition of land use data. This script enables the classification of data obtained from satellite images, facilitating the analysis and calculation of key indicators. Interactive Data Visualization with Power BI. An interactive dashboard has been created in Power BI that allows users to visualize and analyze land use data intuitively. Contribution to Sustainable Urban Development. By providing an accurate and efficient tool for land use monitoring, the project contributes to sustainable urban development. The ability to analyze changes in land use and calculate key indicators allows authorities and urban planners to make data-driven decisions, promoting responsible and sustainable environmental management. Foundation for Future Expansions and Improvements. The developed system establishes a solid foundation for future research and development. Additional improvements and new functionalities can be incorporated, such as the integration of other geospatial data sources or the application of artificial intelligence techniques for predictive analysis.
Direction
Triñanes Fernández, Joaquín Ángel (Tutorships)
Triñanes Fernández, Joaquín Ángel (Tutorships)
Court
VIDAL AGUIAR, JUAN CARLOS (Chairman)
DOSIL LAGO, RAQUEL (Secretary)
SAAVEDRA NIEVES, ALEJANDRO (Member)
VIDAL AGUIAR, JUAN CARLOS (Chairman)
DOSIL LAGO, RAQUEL (Secretary)
SAAVEDRA NIEVES, ALEJANDRO (Member)
Visual Perception of Objects for Collaborative Manipulation Tasks
Authorship
D.P.V.
Computer Engineering Degree
D.P.V.
Computer Engineering Degree
Defense date
07.18.2024 11:00
07.18.2024 11:00
Summary
The industry drives the economic growth of a country, so its evolution is crucial. Recent innovations are based on the integration of new technologies that improve results and increase productivity. In all of this, collaborative arms or cobots are having a great impact due to their flexibility when working with human operators and their great adaptability to various tasks. For all this, and with the purpose of investigating the operation of new technologies, a system has been designed that integrates a Franka Panda collaborative arm, a GelSight Mini touch sensor, and Aruco markers recognized through an Orbbec Astra camera. The entire integration process is accompanied by challenges and difficulties such as the calibration of the Gelsight Mini or the transformation of coordinates between different reference systems. All designed and developed with the purpose of being able to perceive and manipulate objects of different sizes and materials through a robotic arm, which could later be incorporated into a production line or industrial environments. Furthermore, once the results are obtained, possible extensions and improvements are proposed that would complement the system.
The industry drives the economic growth of a country, so its evolution is crucial. Recent innovations are based on the integration of new technologies that improve results and increase productivity. In all of this, collaborative arms or cobots are having a great impact due to their flexibility when working with human operators and their great adaptability to various tasks. For all this, and with the purpose of investigating the operation of new technologies, a system has been designed that integrates a Franka Panda collaborative arm, a GelSight Mini touch sensor, and Aruco markers recognized through an Orbbec Astra camera. The entire integration process is accompanied by challenges and difficulties such as the calibration of the Gelsight Mini or the transformation of coordinates between different reference systems. All designed and developed with the purpose of being able to perceive and manipulate objects of different sizes and materials through a robotic arm, which could later be incorporated into a production line or industrial environments. Furthermore, once the results are obtained, possible extensions and improvements are proposed that would complement the system.
Direction
CATALA BOLOS, ALEJANDRO (Tutorships)
CORRALES RAMON, JUAN ANTONIO (Co-tutorships)
CATALA BOLOS, ALEJANDRO (Tutorships)
CORRALES RAMON, JUAN ANTONIO (Co-tutorships)
Court
VIDAL AGUIAR, JUAN CARLOS (Chairman)
DOSIL LAGO, RAQUEL (Secretary)
SAAVEDRA NIEVES, ALEJANDRO (Member)
VIDAL AGUIAR, JUAN CARLOS (Chairman)
DOSIL LAGO, RAQUEL (Secretary)
SAAVEDRA NIEVES, ALEJANDRO (Member)
Platform for Planning the Execution and Evaluation of Object Detection Algorithms
Authorship
R.C.J.
Computer Engineering Degree
R.C.J.
Computer Engineering Degree
Defense date
09.12.2024 09:30
09.12.2024 09:30
Summary
In the field of artificial intelligence research focused on object detection, the pursuit of improvements in the process of analyzing and evaluating datasets and the respective results of detection models is a crucial part of optimizing the entire process both in terms of time and associated resource costs. These evaluations and analyses must be carried out accurately and avoiding, as much as possible, any type of bias. For this, researchers need a wide variety of tools, and coordinating all of them can be complicated and confusing. Moreover, to perform all these training and analysis of detection models, it is necessary to use distributed clusters due to the high computational load their execution entails, causing each researcher to constantly monitor these clusters to stay up to date not only with the status of their own tasks but also with the general status of the job queue. In this work, a platform will be developed that will unify all the necessary tools to maximize the optimization of the research process both in the management and querying of the status of tasks and the job queue in these distributed clusters, as well as in the evaluation and analysis of image datasets in COCO format and the results and detections of detection models. This platform aims to provide a comprehensive and efficient solution, facilitating the researchers' work and improving the accuracy and coordination of their tasks.
In the field of artificial intelligence research focused on object detection, the pursuit of improvements in the process of analyzing and evaluating datasets and the respective results of detection models is a crucial part of optimizing the entire process both in terms of time and associated resource costs. These evaluations and analyses must be carried out accurately and avoiding, as much as possible, any type of bias. For this, researchers need a wide variety of tools, and coordinating all of them can be complicated and confusing. Moreover, to perform all these training and analysis of detection models, it is necessary to use distributed clusters due to the high computational load their execution entails, causing each researcher to constantly monitor these clusters to stay up to date not only with the status of their own tasks but also with the general status of the job queue. In this work, a platform will be developed that will unify all the necessary tools to maximize the optimization of the research process both in the management and querying of the status of tasks and the job queue in these distributed clusters, as well as in the evaluation and analysis of image datasets in COCO format and the results and detections of detection models. This platform aims to provide a comprehensive and efficient solution, facilitating the researchers' work and improving the accuracy and coordination of their tasks.
Direction
CORES COSTA, DANIEL (Tutorships)
García Fernández, Pablo (Co-tutorships)
CORES COSTA, DANIEL (Tutorships)
García Fernández, Pablo (Co-tutorships)
Court
Fernández Rivera, Francisco (Chairman)
SANTOS MATEOS, ROI (Secretary)
López Vilariño, David (Member)
Fernández Rivera, Francisco (Chairman)
SANTOS MATEOS, ROI (Secretary)
López Vilariño, David (Member)
Open Source Vulnerability Detection and Remediation Software
Authorship
A.B.L.
Computer Engineering Degree
A.B.L.
Computer Engineering Degree
Defense date
07.19.2024 09:00
07.19.2024 09:00
Summary
When it comes to protecting an organization's assets, the use of security software tools is one of the main security measures adopted by tech companies, as it allows them to monitor and analyze large amounts of information in search of suspicious activity. Among the functionalities provided by these tools is the detection of vulnerabilities in the applications installed on the devices. However, there are certain technical and economic limitations that mean that most small and medium-sized organizations do not consider an automatic detection and remediation system for these types of vulnerabilities. Currently, there are open-source solutions that allow for vulnerability detection, but the remediation of these vulnerabilities is usually not available in the free tools, or at least not in an automatic way, as seen in the study of the documentation provided by several tools specifically recommended by the organization in which this project is developed. Therefore, this work proposes a partial solution to this problem by extending the capabilities of the open-source tool Velociraptor. On one hand, methods are developed to enable the collection and analysis of the necessary information during vulnerability detection. Additionally, through an application repository and action protocols, an effective way to remediate these vulnerabilities with minimal intervention is created.
When it comes to protecting an organization's assets, the use of security software tools is one of the main security measures adopted by tech companies, as it allows them to monitor and analyze large amounts of information in search of suspicious activity. Among the functionalities provided by these tools is the detection of vulnerabilities in the applications installed on the devices. However, there are certain technical and economic limitations that mean that most small and medium-sized organizations do not consider an automatic detection and remediation system for these types of vulnerabilities. Currently, there are open-source solutions that allow for vulnerability detection, but the remediation of these vulnerabilities is usually not available in the free tools, or at least not in an automatic way, as seen in the study of the documentation provided by several tools specifically recommended by the organization in which this project is developed. Therefore, this work proposes a partial solution to this problem by extending the capabilities of the open-source tool Velociraptor. On one hand, methods are developed to enable the collection and analysis of the necessary information during vulnerability detection. Additionally, through an application repository and action protocols, an effective way to remediate these vulnerabilities with minimal intervention is created.
Direction
CARIÑENA AMIGO, MARIA PURIFICACION (Tutorships)
Carro Fernández, Marcos (Co-tutorships)
CARIÑENA AMIGO, MARIA PURIFICACION (Tutorships)
Carro Fernández, Marcos (Co-tutorships)
Court
BARJA PEREZ, JAVIER (Chairman)
ORDOÑEZ IGLESIAS, ALVARO (Secretary)
MOSQUERA GONZALEZ, ANTONIO (Member)
BARJA PEREZ, JAVIER (Chairman)
ORDOÑEZ IGLESIAS, ALVARO (Secretary)
MOSQUERA GONZALEZ, ANTONIO (Member)
Digital audio signal processing on programmable logic
Authorship
D.C.P.
Computer Engineering Degree
D.C.P.
Computer Engineering Degree
Defense date
07.19.2024 12:30
07.19.2024 12:30
Summary
In this Final Degree Project, an audio processing and musical transcription system has been developed using a heterogeneous CPU+FPGA platform, specifically the Pynq-Z2 board and the PYNQ development framework. The main objective has been to address and solve several key problems in the field of audio processing and musical transcription, achieving an effective and reconfigurable system that performs its functions efficiently. The system is designed to read audio files in Python and transmit them to the programmable logic for processing, either applying audio effects or performing spectral analysis. Once the data is processed, the system reconstructs the audio file with the new signal or performs the transcription to a MIDI file, including the identification of the corresponding musical scale.
In this Final Degree Project, an audio processing and musical transcription system has been developed using a heterogeneous CPU+FPGA platform, specifically the Pynq-Z2 board and the PYNQ development framework. The main objective has been to address and solve several key problems in the field of audio processing and musical transcription, achieving an effective and reconfigurable system that performs its functions efficiently. The system is designed to read audio files in Python and transmit them to the programmable logic for processing, either applying audio effects or performing spectral analysis. Once the data is processed, the system reconstructs the audio file with the new signal or performs the transcription to a MIDI file, including the identification of the corresponding musical scale.
Direction
López Vilariño, David (Tutorships)
RODRIGUEZ ALCARAZ, SILVIA (Co-tutorships)
López Vilariño, David (Tutorships)
RODRIGUEZ ALCARAZ, SILVIA (Co-tutorships)
Court
PENA BRAGE, FRANCISCO JOSE (Chairman)
MERA PEREZ, DAVID (Secretary)
CARIÑENA AMIGO, MARIA PURIFICACION (Member)
PENA BRAGE, FRANCISCO JOSE (Chairman)
MERA PEREZ, DAVID (Secretary)
CARIÑENA AMIGO, MARIA PURIFICACION (Member)
Development of a prototype for a wearable or smartphone to study physical frailty in elderly people using artificial intelligence techniques
Authorship
A.F.G.
Computer Engineering Degree
A.F.G.
Computer Engineering Degree
Defense date
07.18.2024 16:00
07.18.2024 16:00
Summary
This final degree project investigates the feasibility of developing a prototype for a wearable or smartphone device to assess physical frailty in elderly people. The prototype will calculate measures such as gait speed and other factors related to frailty, and will be able to make predictions about the frailty status using artificial intelligence techniques, offering an innovative tool for early identification and monitoring of frailty in the elderly population.
This final degree project investigates the feasibility of developing a prototype for a wearable or smartphone device to assess physical frailty in elderly people. The prototype will calculate measures such as gait speed and other factors related to frailty, and will be able to make predictions about the frailty status using artificial intelligence techniques, offering an innovative tool for early identification and monitoring of frailty in the elderly population.
Direction
VALLADARES RODRIGUEZ, SONIA MARIA (Tutorships)
RODRIGUEZ PRESEDO, JESUS MARIA (Co-tutorships)
VALLADARES RODRIGUEZ, SONIA MARIA (Tutorships)
RODRIGUEZ PRESEDO, JESUS MARIA (Co-tutorships)
Court
Pardo López, Xosé Manuel (Chairman)
Puentes Cociña, Beltrán (Secretary)
LOPEZ FANDIÑO, JAVIER (Member)
Pardo López, Xosé Manuel (Chairman)
Puentes Cociña, Beltrán (Secretary)
LOPEZ FANDIÑO, JAVIER (Member)
Evaluation of stress detectors during interaction with serious games that stimulate cognitive activities
Authorship
L.M.L.
Computer Engineering Degree
L.M.L.
Computer Engineering Degree
Defense date
07.18.2024 19:00
07.18.2024 19:00
Summary
Automatic stress detection is a relevant area of research due to the growing concerns about mental health in today's society. Thanks to advances in the field of affective computing, various methodologies and systems have been investigated in this area over the years. Although numerous monitoring systems exist, evidence on the congruence of their accuracy and applicability in a test with the same characteristics is still limited and varied. Therefore, this Bachelor's thesis aims to conduct a comparative evaluation to identify common patterns, strengths and areas for improvement between two methods of stress detection: one based on physiological signals using an Empatica EmbracePlus wristband, and the other on facial expressions using the Kopernica tool. The study was set in a controlled laboratory environment, with participants interacting with a cognitive reasoning game designed to induce stress. The process was divided into several phases: a relaxation period, interaction with the game in three different modes and a perceived stress questionnaire. Several key objectives were achieved. First, according to the EDA signal, it was established that the serious game used is a valid stress inducer, generating an increase in measured stress levels. Moderate congruence was found between the two detection methods, although with some discrepancies in their sensitivity to the complexity of the game tasks. The method based on physiological signals was shown to be more sensitive to slight variations in stress levels compared to facial expressions. Users were found to be unaware of their actual stress levels. Furthermore, it was concluded that reducing the resolution of the video impacts the accuracy of stress detection by Kopernica, with 640x360 being the minimum acceptable resolution. The results provide a clearer understanding of the strengths and limitations of each method, suggesting the need for specific adjustments in data processing to improve the accuracy and consistency of measurements. It also highlights the importance of extending the research to other emotions measured by Kopernica and to different populations to observe the manifestation of stress at different life stages.
Automatic stress detection is a relevant area of research due to the growing concerns about mental health in today's society. Thanks to advances in the field of affective computing, various methodologies and systems have been investigated in this area over the years. Although numerous monitoring systems exist, evidence on the congruence of their accuracy and applicability in a test with the same characteristics is still limited and varied. Therefore, this Bachelor's thesis aims to conduct a comparative evaluation to identify common patterns, strengths and areas for improvement between two methods of stress detection: one based on physiological signals using an Empatica EmbracePlus wristband, and the other on facial expressions using the Kopernica tool. The study was set in a controlled laboratory environment, with participants interacting with a cognitive reasoning game designed to induce stress. The process was divided into several phases: a relaxation period, interaction with the game in three different modes and a perceived stress questionnaire. Several key objectives were achieved. First, according to the EDA signal, it was established that the serious game used is a valid stress inducer, generating an increase in measured stress levels. Moderate congruence was found between the two detection methods, although with some discrepancies in their sensitivity to the complexity of the game tasks. The method based on physiological signals was shown to be more sensitive to slight variations in stress levels compared to facial expressions. Users were found to be unaware of their actual stress levels. Furthermore, it was concluded that reducing the resolution of the video impacts the accuracy of stress detection by Kopernica, with 640x360 being the minimum acceptable resolution. The results provide a clearer understanding of the strengths and limitations of each method, suggesting the need for specific adjustments in data processing to improve the accuracy and consistency of measurements. It also highlights the importance of extending the research to other emotions measured by Kopernica and to different populations to observe the manifestation of stress at different life stages.
Direction
CONDORI FERNANDEZ, NELLY (Tutorships)
CATALA BOLOS, ALEJANDRO (Co-tutorships)
CONDORI FERNANDEZ, NELLY (Tutorships)
CATALA BOLOS, ALEJANDRO (Co-tutorships)
Court
VAZQUEZ CENDON, MARIA ELENA (Chairman)
CHAVES FRAGA, DAVID (Secretary)
SUAREZ GAREA, JORGE ALBERTO (Member)
VAZQUEZ CENDON, MARIA ELENA (Chairman)
CHAVES FRAGA, DAVID (Secretary)
SUAREZ GAREA, JORGE ALBERTO (Member)
Machine learning for carbon flux estimation from model and satellite data
Authorship
P.A.G.P.
Computer Engineering Degree
P.A.G.P.
Computer Engineering Degree
Defense date
07.19.2024 17:30
07.19.2024 17:30
Summary
This work tackles the partial pressure of carbon dioxide in the ocean surface prediction problem for carbon flux estimation. The set of oceanic data used comes from different satellites and oceanographic investigators organizations. To this end, the results of training nine different regression models with this set of data and with three other subsets of data created from the global set delimiting the geographic regions of the Atlantic, Indic and Pacific oceans were compared amongst them. The compared regression models were Random Forest, ExtraTrees, the method of the nearest neighbours, the support vectors machine, Gradient Boosting, Adaptive Boosting, CatBoost, XGBoost and LightGBM. After the analysis of the model comparison for each dataset, it was determined that the one which offered better results in the four cases was the ExtraTrees model, so its hiperparameters of number of estimators and maximum depth were adjusted to optimize the model’s output for each case. Once these models were adjusted, predictions using each dataset were made and files in the netCDF format were generated both for the input datasets and the sets of predicted data.
This work tackles the partial pressure of carbon dioxide in the ocean surface prediction problem for carbon flux estimation. The set of oceanic data used comes from different satellites and oceanographic investigators organizations. To this end, the results of training nine different regression models with this set of data and with three other subsets of data created from the global set delimiting the geographic regions of the Atlantic, Indic and Pacific oceans were compared amongst them. The compared regression models were Random Forest, ExtraTrees, the method of the nearest neighbours, the support vectors machine, Gradient Boosting, Adaptive Boosting, CatBoost, XGBoost and LightGBM. After the analysis of the model comparison for each dataset, it was determined that the one which offered better results in the four cases was the ExtraTrees model, so its hiperparameters of number of estimators and maximum depth were adjusted to optimize the model’s output for each case. Once these models were adjusted, predictions using each dataset were made and files in the netCDF format were generated both for the input datasets and the sets of predicted data.
Direction
VARELA PET, JOSE (Tutorships)
Triñanes Fernández, Joaquín Ángel (Co-tutorships)
VARELA PET, JOSE (Tutorships)
Triñanes Fernández, Joaquín Ángel (Co-tutorships)
Court
Fernández Pena, Anselmo Tomás (Chairman)
SACO LOPEZ, PEDRO JOSE (Secretary)
RODRIGUEZ PRESEDO, JESUS MARIA (Member)
Fernández Pena, Anselmo Tomás (Chairman)
SACO LOPEZ, PEDRO JOSE (Secretary)
RODRIGUEZ PRESEDO, JESUS MARIA (Member)
Survey and evaluation of homomorphic encryption algorithms
Authorship
H.V.R.
Double bachelor degree of Engeneering in Information Technology and Mathematics
H.V.R.
Double bachelor degree of Engeneering in Information Technology and Mathematics
Defense date
09.12.2024 10:30
09.12.2024 10:30
Summary
Homomorphic cryptography allows operations to be performed on encrypted data without having to decrypt it. This feature is very attractive because, in particular, it prevents third parties to whom calculations with that data are delegated from accessing it, opening up a wide range of applications. The problem of building a completely homomorphic cryptosystem, which allows arbitrary calculations with encrypted data, was considered the holy grail of cryptography for 30 years. During that time, countless attempts have been made to solve it, obtaining only partial solutions. In 2009, the work of Craig Gentry represented a revolution by introducing the bootstrapping technique, through which some cryptosystems could be made completely homomorphic. From this moment on, different families of cryptosystems with this property have emerged, all of them making use of Gentry's invention. The aim of this paper is to carry out a small comparative study between some of the most widely used completely homomorphic cryptosystems today, specifically BGV, BFV and TFHE, comparing the algorithms that compose them and measuring the execution times. Previously, homomorphic cryptography will be introduced theoretically, defining bootstrapping and making certain considerations regarding the security of these cryptosystems. In addition, a brief historical context and a brief review of the state of the art will be given, presenting the main families of completely homomorphic cryptosystems.
Homomorphic cryptography allows operations to be performed on encrypted data without having to decrypt it. This feature is very attractive because, in particular, it prevents third parties to whom calculations with that data are delegated from accessing it, opening up a wide range of applications. The problem of building a completely homomorphic cryptosystem, which allows arbitrary calculations with encrypted data, was considered the holy grail of cryptography for 30 years. During that time, countless attempts have been made to solve it, obtaining only partial solutions. In 2009, the work of Craig Gentry represented a revolution by introducing the bootstrapping technique, through which some cryptosystems could be made completely homomorphic. From this moment on, different families of cryptosystems with this property have emerged, all of them making use of Gentry's invention. The aim of this paper is to carry out a small comparative study between some of the most widely used completely homomorphic cryptosystems today, specifically BGV, BFV and TFHE, comparing the algorithms that compose them and measuring the execution times. Previously, homomorphic cryptography will be introduced theoretically, defining bootstrapping and making certain considerations regarding the security of these cryptosystems. In addition, a brief historical context and a brief review of the state of the art will be given, presenting the main families of completely homomorphic cryptosystems.
Direction
CARIÑENA AMIGO, MARIA PURIFICACION (Tutorships)
GAGO COUSO, FELIPE (Co-tutorships)
CARIÑENA AMIGO, MARIA PURIFICACION (Tutorships)
GAGO COUSO, FELIPE (Co-tutorships)
Court
Fernández Rivera, Francisco (Chairman)
SANTOS MATEOS, ROI (Secretary)
López Vilariño, David (Member)
Fernández Rivera, Francisco (Chairman)
SANTOS MATEOS, ROI (Secretary)
López Vilariño, David (Member)
Introduction to nonlinear optimization
Authorship
D.P.V.
Double bachelor degree of Engeneering in Information Technology and Mathematics
D.P.V.
Double bachelor degree of Engeneering in Information Technology and Mathematics
Defense date
07.17.2024 12:30
07.17.2024 12:30
Summary
This project offers a theorical approach of the fundamentals of nonlinear optimization. It begins with a review of the optimization-related content covered throughout the degree, aiming to establish the starting point. After introducing the necessary preliminaries for a correct formalization, the main points of the work are presented: optimality conditions and Lagrangian duality. Firstly, the Fritz John and Karush-Kuhn-Tucker optimality conditions for nonlinear problems are developed; and finally, the Lagrangian dual problem and the duality theorems are discussed. The relationships between the concepts presented and their generalizations or particular cases are addresed as the work progresses.
This project offers a theorical approach of the fundamentals of nonlinear optimization. It begins with a review of the optimization-related content covered throughout the degree, aiming to establish the starting point. After introducing the necessary preliminaries for a correct formalization, the main points of the work are presented: optimality conditions and Lagrangian duality. Firstly, the Fritz John and Karush-Kuhn-Tucker optimality conditions for nonlinear problems are developed; and finally, the Lagrangian dual problem and the duality theorems are discussed. The relationships between the concepts presented and their generalizations or particular cases are addresed as the work progresses.
Direction
GONZALEZ DIAZ, JULIO (Tutorships)
GONZALEZ DIAZ, JULIO (Tutorships)
Court
GONZALEZ MANTEIGA, WENCESLAO (Chairman)
PAEZ GUILLAN, MARIA PILAR (Secretary)
ALVAREZ DIOS, JOSE ANTONIO (Member)
GONZALEZ MANTEIGA, WENCESLAO (Chairman)
PAEZ GUILLAN, MARIA PILAR (Secretary)
ALVAREZ DIOS, JOSE ANTONIO (Member)
Anomaly detection in power lines
Authorship
D.P.V.
Double bachelor degree of Engeneering in Information Technology and Mathematics
D.P.V.
Double bachelor degree of Engeneering in Information Technology and Mathematics
Defense date
07.18.2024 13:00
07.18.2024 13:00
Summary
Power line inspection is a field in constant progress due to the importance of the entire electrical infrastructure in today's society. Most articles that address this topic focus on the monitoring of insulators, leaving aside a large number of components that also have relevance in the power grid. This paper explores the capability of versions of the RetinaNet, Single Shot Multibox Detector (SSD), You Only Look Once (YOLO) and Faster R-CNN models for the detection of a wider number of power grid elements. Finally, the performance of ResNet and EfficientNet classification models for the categorization of insulator defects is also studied.
Power line inspection is a field in constant progress due to the importance of the entire electrical infrastructure in today's society. Most articles that address this topic focus on the monitoring of insulators, leaving aside a large number of components that also have relevance in the power grid. This paper explores the capability of versions of the RetinaNet, Single Shot Multibox Detector (SSD), You Only Look Once (YOLO) and Faster R-CNN models for the detection of a wider number of power grid elements. Finally, the performance of ResNet and EfficientNet classification models for the categorization of insulator defects is also studied.
Direction
MUCIENTES MOLINA, MANUEL FELIPE (Tutorships)
Abado Bóveda, Silvia (Co-tutorships)
MUCIENTES MOLINA, MANUEL FELIPE (Tutorships)
Abado Bóveda, Silvia (Co-tutorships)
Court
VIDAL AGUIAR, JUAN CARLOS (Chairman)
DOSIL LAGO, RAQUEL (Secretary)
SAAVEDRA NIEVES, ALEJANDRO (Member)
VIDAL AGUIAR, JUAN CARLOS (Chairman)
DOSIL LAGO, RAQUEL (Secretary)
SAAVEDRA NIEVES, ALEJANDRO (Member)
Reformulation in mixed integer nonlinear programming problems.
Authorship
S.A.P.
Double bachelor degree of Engeneering in Information Technology and Mathematics
S.A.P.
Double bachelor degree of Engeneering in Information Technology and Mathematics
Defense date
07.16.2024 16:00
07.16.2024 16:00
Summary
In this project we study a class of reformulations for polynomial programming problems. Initially, we present the theoretical foundation of these reformulations, which were originally developed for binary polynomial optimization problems. Then we extend these results to the broader context of polynomial optimization problems, including non-binary variables (integer and continuous). Finally, we evaluate the numeric results obtained after adapting and implementing these reformulations in a solver.
In this project we study a class of reformulations for polynomial programming problems. Initially, we present the theoretical foundation of these reformulations, which were originally developed for binary polynomial optimization problems. Then we extend these results to the broader context of polynomial optimization problems, including non-binary variables (integer and continuous). Finally, we evaluate the numeric results obtained after adapting and implementing these reformulations in a solver.
Direction
GONZALEZ DIAZ, JULIO (Tutorships)
Rodríguez Acevedo, Iria (Co-tutorships)
GONZALEZ DIAZ, JULIO (Tutorships)
Rodríguez Acevedo, Iria (Co-tutorships)
Court
GARCIA RIO, EDUARDO (Chairman)
RIVERO SALGADO, OSCAR (Secretary)
CASAS MENDEZ, BALBINA VIRGINIA (Member)
GARCIA RIO, EDUARDO (Chairman)
RIVERO SALGADO, OSCAR (Secretary)
CASAS MENDEZ, BALBINA VIRGINIA (Member)
Lifelong learning for edge computing applications in industry 4.0
Authorship
S.A.P.
Double bachelor degree of Engeneering in Information Technology and Mathematics
S.A.P.
Double bachelor degree of Engeneering in Information Technology and Mathematics
Defense date
07.18.2024 10:00
07.18.2024 10:00
Summary
In this work we develop an unsupervised time series encoder with continual learning, for its integration with a classification model. Specifically, we adapt an existing time series encoder to be trained with continual learning, allowing the generated encodings to be used as input for the classification model. In addition, in order to evaluate the quality of the encodings, we design a metrics and visualization methods. The experiments show that the performance when using the encoder with continual learning is comparable to the performance obtained when using offline learning, the former situation being more difficult for the model. The integration experiments show that when using the encodings as input for the classifier with continual learning, the performance obtained is satisfactory.
In this work we develop an unsupervised time series encoder with continual learning, for its integration with a classification model. Specifically, we adapt an existing time series encoder to be trained with continual learning, allowing the generated encodings to be used as input for the classification model. In addition, in order to evaluate the quality of the encodings, we design a metrics and visualization methods. The experiments show that the performance when using the encoder with continual learning is comparable to the performance obtained when using offline learning, the former situation being more difficult for the model. The integration experiments show that when using the encodings as input for the classifier with continual learning, the performance obtained is satisfactory.
Direction
MERA PEREZ, DAVID (Tutorships)
Fernández Castro, Bruno (Co-tutorships)
García Santaclara, Pablo (Co-tutorships)
MERA PEREZ, DAVID (Tutorships)
Fernández Castro, Bruno (Co-tutorships)
García Santaclara, Pablo (Co-tutorships)
Court
Cotos Yáñez, José Manuel (Chairman)
QUESADA BARRIUSO, PABLO (Secretary)
GAGO COUSO, FELIPE (Member)
Cotos Yáñez, José Manuel (Chairman)
QUESADA BARRIUSO, PABLO (Secretary)
GAGO COUSO, FELIPE (Member)
Biological dynamic systems optimization with algebraic differential equations
Authorship
C.C.P.
Double bachelor degree of Engeneering in Information Technology and Mathematics
C.C.P.
Double bachelor degree of Engeneering in Information Technology and Mathematics
Defense date
07.16.2024 16:40
07.16.2024 16:40
Summary
Mathematical modeling is essential for the understanding and handling of biological dynamic systems. Optimal control problems, which usually include algebraic differential equations as constrains, need to be used. For solving these kinds of problems, of infinite dimension, there are several different types of strategies. In this work, we will delve into direct methods, which perform a discretization of the optimal control problem, transforming it into a finite dimension nonlinear programming problem. In biology, most existing problems involve estimating system parameters from a series of experimental observations. For such a biological problem, a numerical study will be proposed for its solution, varying its configuration, and the results obtained will be commented in depth.
Mathematical modeling is essential for the understanding and handling of biological dynamic systems. Optimal control problems, which usually include algebraic differential equations as constrains, need to be used. For solving these kinds of problems, of infinite dimension, there are several different types of strategies. In this work, we will delve into direct methods, which perform a discretization of the optimal control problem, transforming it into a finite dimension nonlinear programming problem. In biology, most existing problems involve estimating system parameters from a series of experimental observations. For such a biological problem, a numerical study will be proposed for its solution, varying its configuration, and the results obtained will be commented in depth.
Direction
GONZALEZ RUEDA, ANGEL MANUEL (Tutorships)
GONZALEZ RUEDA, ANGEL MANUEL (Tutorships)
Court
GARCIA RIO, EDUARDO (Chairman)
RIVERO SALGADO, OSCAR (Secretary)
CASAS MENDEZ, BALBINA VIRGINIA (Member)
GARCIA RIO, EDUARDO (Chairman)
RIVERO SALGADO, OSCAR (Secretary)
CASAS MENDEZ, BALBINA VIRGINIA (Member)
Automatic lip syncing with the Furhat social robot
Authorship
C.C.P.
Double bachelor degree of Engeneering in Information Technology and Mathematics
C.C.P.
Double bachelor degree of Engeneering in Information Technology and Mathematics
Defense date
07.19.2024 10:30
07.19.2024 10:30
Summary
The rise of artificial intelligence in social robotics and the proliferation of TTS (Text To Speech) systems mean that there can be an integration of both fields, as can be seen in the Furhat robot. This technology is limited to a few languages, recently it was successfully managed to integrate the Proxecto Nos’s TTS into this robot, which allowed Galician speech. However, lip syncing was not adjusted to the Galician language, resulting in an incomplete user experience. This work proposes an alternative lip synchronization for the social robot Furhat during its Galician speech. To achieve this, forced alignment will be used, through the Montreal Forced Aligner (MFA) tool. To align with MFA, we have to create a pronunciation dictionary and train acoustic models with this tool, as none of these resources are currently available for the Galician language.
The rise of artificial intelligence in social robotics and the proliferation of TTS (Text To Speech) systems mean that there can be an integration of both fields, as can be seen in the Furhat robot. This technology is limited to a few languages, recently it was successfully managed to integrate the Proxecto Nos’s TTS into this robot, which allowed Galician speech. However, lip syncing was not adjusted to the Galician language, resulting in an incomplete user experience. This work proposes an alternative lip synchronization for the social robot Furhat during its Galician speech. To achieve this, forced alignment will be used, through the Montreal Forced Aligner (MFA) tool. To align with MFA, we have to create a pronunciation dictionary and train acoustic models with this tool, as none of these resources are currently available for the Galician language.
Direction
CATALA BOLOS, ALEJANDRO (Tutorships)
BUGARIN DIZ, ALBERTO JOSE (Co-tutorships)
MAGARIÑOS IGLESIAS, MARIA DEL CARMEN (Co-tutorships)
CATALA BOLOS, ALEJANDRO (Tutorships)
BUGARIN DIZ, ALBERTO JOSE (Co-tutorships)
MAGARIÑOS IGLESIAS, MARIA DEL CARMEN (Co-tutorships)
Court
BARJA PEREZ, JAVIER (Chairman)
ORDOÑEZ IGLESIAS, ALVARO (Secretary)
MOSQUERA GONZALEZ, ANTONIO (Member)
BARJA PEREZ, JAVIER (Chairman)
ORDOÑEZ IGLESIAS, ALVARO (Secretary)
MOSQUERA GONZALEZ, ANTONIO (Member)