Diagnóstico de fallas por rotura de barras en motores de inducción trifásicos mediante el uso de modelos de deep learning

Three-phase induction motors are used in various industrial applications due to their efficiency and reliability. However, rotor failures such as a broken rotor bar can affect their performance, causing electromagnetic instability, vibrations, and loss of energy efficiency, leading to additional mai...

Full description

Bibliographic Details
Author: Garcia Tucci, José Luis
Format: master thesis
Publication Date:2025
Country:España
Institution:Consejo Superior de Investigaciones Científicas (CSIC)
Repository:O2, repositorio institucional de la UOC
OAI Identifier:oai:openaccess.uoc.edu:10609/153873
Online Access:https://hdl.handle.net/10609/153873
Access Level:Open access
Keyword:aprendizaje automático
aprendizaje profundo
motores de inducción
detección de fallas
mantenimiento predictivo
barras rotas
machine learning
deep learning
induction motors
fault detection
predictive maintenance
broken bars
Deep learning -- TFM
Aprenentatge profund -- TFM
Description
Summary:Three-phase induction motors are used in various industrial applications due to their efficiency and reliability. However, rotor failures such as a broken rotor bar can affect their performance, causing electromagnetic instability, vibrations, and loss of energy efficiency, leading to additional maintenance costs and losses in industrial production. Early detection of these failures allows predictive maintenance strategies to be implemented to reduce costs and avoid unplanned interruptions in industrial systems. This work proposes the development of a deep learning-based classification model for the automatic detection of broken rotor bar faults in three-phase induction motors. The aim is to identify the number of broken bars from the analysis of current signals based on the angular position of the rotor, using a non-invasive method that combines current monitoring with automatic detection techniques. The methodology includes experimental data acquisition, current signal preprocessing, feature extraction, and training of deep neural network models. The model architecture is designed to improve its generalization capacity and performance in detecting faults in motors with different rotor bar configurations. Experimental data obtained in collaboration with the Universitat Politècnica de València will be used, which will allow the evaluation of the model and its capacity to recognize characteristic patterns of these faults. The system starts from signal acquisition → codification → classification → results visualization in the GUI.