Título: Diagnóstico de Falhas em Rolamentos de Motores Elétricos com Base na Análise da Assinatura da Corrente do Motor
Autores: Marcos Romulo de Oliveira, Luiz Alberto Pinto, Cassius Zanetti Resende
Resumo: This article presents a comparative analysis of the performance of classification algorithms applied to fault diagnosis in electric motor bearings. To represent the faults, 13 statistical descriptors of the electric current signals available in the Paderborn dataset were extracted. The classification problem consisted of nine failure classes in addition to the normal operating condition class. The models were built both in the time domain and in the time scale domain, using wavelet transform filters, Coiflet 5, Daubechies 4 (Daub 4) and Symlet 8, at predefined resolution levels. The k-NN, SVM and Decision Tree algorithms were used to build the models. The performance of the models was evaluated based on the metrics of accuracy, sensitivity and F1 score. The best accuracy results were obtained using the SVM algorithm, with Symlet and Coiflet 5 filters, reaching a value of 0.9967%. The articles main contribution, in addition to the use of Wavelets and SVM combined, was that it demonstrated results that were comparable when using electric current signals as entry to vibration signals that are already widely used in the industry and that the use of the Wavelet Transform, along which Machine Learning algorithms, when using good descriptors, can be a viable solution to implement diagnostic systems automatic detection of more accurate faults in electric motor bearings
Palavras-chave: bearing failure classification, eletric motors, support vector machine, electric current signal, pattern recognition
Páginas: 8
Código DOI: 10.21528/CBIC2023-048
Artigo em pdf: CBIC_2023_paper048.pdf
Arquivo BibTeX: CBIC_2023_048.bib