Detecção de falhas em rolamentos de motores de indução trifásicos baseada em máquina de vetores de suporte e redes de Bragg em fibra óptica

• Ano de defesa: 2022
• Autor(a) principal: Brusamarello, Beatriz
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Tecnológica Federal do Paraná; Pato Branco; Brasil; Programa de Pós-Graduação em Engenharia Elétrica; UTFPR
• Palavras-chave em Português: Análise de componentes principais; Localização de falhas (Engenharia); Rolamentos; Motores elétricos de indução; Redes de Bragg; Principal components analysis; Fault location (Engineering); Bearings (Machinery); Electric motors, Induction; Bragg gratings; CNPQ::ENGENHARIAS::ENGENHARIA ELETRICA; Engenharia/Tecnologia/Gestão
• Link de acesso: http://repositorio.utfpr.edu.br/jspui/handle/1/28144

Due to its robustness and great cost-benefit ratio, the induction motor has become the most widespread electric machine today. However, it is vulnerable to a fault like any other equipment, with bearing faults being the most common in induction motors. This work presents a supervised system for detecting and diagnosing faults in the outer bearing’s raceway based on support vector machine (SVM) and motor dynamic strain signals, collected by a sensor based on fiber Bragg grating (FBG) installed inside the motor end shield. Three different degrees of severity were considered for faults in the outer bearing’s raceway: early-stage fault, intermediate fault, and severe fault. The tests were carried out on the motor operating at no load, with 47 different power frequencies. Tests were also carried out on the motor under load with the bearing showing fault at an early stage, with a supply frequency of 60 Hz. The measured signals were treated using Fast Fourier Transform and power spectral density in the frequency domain. The SVM classifier was trained with two different datasets, resulting from two-dimensionality reduction techniques via feature extraction: the selection of the four highest peaks of the frequency spectra and the principal component analysis (PCA). The optimization and definition of the SVM parameters were performed using the grid Search and cross Validation techniques. For the no-load motor tests, the results of the SVM classifiers show that the set of characteristics formed from the PCA presented a higher hit rate than the set of characteristics constituted by the four highest peaks of the frequency spectra, with values of 99.82% and 99.73%, respectively. This fact was repeated for the SVM classifiers trained with the datasets that contained the tests performed on the motor with load; the dimensionality reduction via PCA presented a greater precision than the dataset formed by the four highest peaks of the frequency spectra, 99.31%, and 92.27%, respectively. From the methodology presented in this work, it was possible to validate the use of FBG to detect faults in bearings since, regardless of the degree of severity of the fault tested, the FBG has enough sensitivity to detect all fault conditions in the outer bearing’s raceway.