Identificação de falhas em rolamentos aplicados em motores de indução por meio de sensores de corrente de efeito hall, técnicas de processamento de sinais e sistemas inteligentes
Carregando...
Arquivos
Data
2022-08-03
Autores
Souza, Wallace Gabriel de
Título da Revista
ISSN da Revista
Título de Volume
Editor
Universidade Estadual Paulista (Unesp)
Resumo
Os motores de indução trifásicos pertencem ao conjunto das máquinas elétricas mais encontradas nas aplicações industriais. Devido às vantagens de baixo custo, robustez, confiabilidade e disponibilidade, muitas vezes são utilizados para operar em condições com elevadas taxas de umidade, altas temperaturas e dimensionamentos inadequados. A interrupção de um motor de indução levará à altos custos de manutenção e graves perdas financeiras devido aos tempos de paradas não programados das plantas industriais. Nesse caso, os rolamentos representam cerca de 40 a 50 % dos tipos de danos em máquinas elétricas, que afetam as condições operacionais do motor de indução, reduzindo a sua vida útil por meio de falhas. Como componente principal dos sistemas rotativos, os rolamentos ajudam a reduzir as perdas rotacionais e garantem o alinhamento do rotor em relação ao estator. Dessa maneira, a detecção e o diagnóstico precoce dos danos nos rolamentos garantem a segurança e o suporte para a manutenção preditiva das máquinas elétricas. À vista disso, inúmeras técnicas foram desenvolvidas para identificar anomalias na operação da máquina. Dentre elas, o monitoramento da corrente se destaca devido ao baixo custo e fácil acesso às plantas industriais. Com esse propósito, o presente trabalho propõe uma nova abordagem que utiliza métricas de agrupamentos de dados combinada com sistemas inteligentes na identificação de falhas nos rolamentos de um motor de indução trifásico pela assinatura da corrente do estator. Nesse caso, foram abordadas falhas que ocorrem na indústria, como os desgastes das pistas e das esferas dos rolamentos, a oxidação e os danos originados pelas das correntes de eixo. Nessa situação, foram utilizadas a transformada wavelet discreta, a largura de banda eficaz e a frequência média no processamento dos sinais amostrados. Essas técnicas possibilitam o treinamento de um sistema inteligente para identificar os danos nos rolamentos. Assim sendo, foi comparado o desempenho de uma rede neural artificial. Além disso, o índice silhouette foi utilizado como validação estatística da qualidade dos agrupamentos de dados obtidos a partir do conjunto de falhas nos rolamentos abordadas neste trabalho. A base de dados foi obtida por meio de experimentos laboratoriais realizados em uma máquina de indução de 4 CV sob condições de variação do conjugado para demonstrar que o método proposto não depende do torque do motor e pode ser aplicado em qualquer carregamento. Os resultados experimentais revelaram que a técnica proposta foi eficaz na identificação das falhas nos rolamentos neste trabalho.
Threephase induction motors are part of the electrical machines most commonly found in industrial applications. Due to their vantages of low cost, robustness, reliability, and availability, they are often used to operate in conditions with high humidity rates, high temperatures, and inadequate sizing. Stopping an induction motor will lead to high maintenance costs and severe financial losses due to unscheduled d owntime of industrial plants. In this case, bearings represent about 40 to 50 % of the types of damage in electrical machines, which affect the operating conditions of the induction motor, reducing its lifetime through fault of r s. As a principal component otating systems, bearings help reduce rotational losses and ensure alignment of the rotor relative to the stator. Thus, early detection and diagnosis of bearing damage provide safety and support for predictive maintenance of electrical machines. Given this , numerous techniques have been developed to identify anomalies in machine operation. Among them, current monitoring stands out due to its low cost and easy access to industrial plants. With this purpose, the present work proposes a new approach that uses data grouping metrics combined with intelligent systems to identify faults in the bearings of a three current signature. In this case, the faults phase induction motor by the stator present in the industry, such as the wear of bearing races and balls, oxid ation, and damage from shaft currents, were addressed. the discrete wavelet transform, RMSIn this situation, Bandwidth, and average frequency were used in the digital signal processing of the sampled signals. These techniques make it possible the training an intelligent system to identify bearing damage. Therefore, the performance of an artificial neural network was compared. In addition, the silhouette index was used as a statistical validation of the data clusters obtained from the bearing faults group a ddressed in this paper. was obtained through laboratory experiments performed on a 4 HP The database induction machine under conditions of varying torque to demonstrate that the proposed method does not depend on motor torque and can be applied at any loading. The experimental results revealed that the proposed technique was efficacious in the bearing fa ult identification in this work.
Threephase induction motors are part of the electrical machines most commonly found in industrial applications. Due to their vantages of low cost, robustness, reliability, and availability, they are often used to operate in conditions with high humidity rates, high temperatures, and inadequate sizing. Stopping an induction motor will lead to high maintenance costs and severe financial losses due to unscheduled d owntime of industrial plants. In this case, bearings represent about 40 to 50 % of the types of damage in electrical machines, which affect the operating conditions of the induction motor, reducing its lifetime through fault of r s. As a principal component otating systems, bearings help reduce rotational losses and ensure alignment of the rotor relative to the stator. Thus, early detection and diagnosis of bearing damage provide safety and support for predictive maintenance of electrical machines. Given this , numerous techniques have been developed to identify anomalies in machine operation. Among them, current monitoring stands out due to its low cost and easy access to industrial plants. With this purpose, the present work proposes a new approach that uses data grouping metrics combined with intelligent systems to identify faults in the bearings of a three current signature. In this case, the faults phase induction motor by the stator present in the industry, such as the wear of bearing races and balls, oxid ation, and damage from shaft currents, were addressed. the discrete wavelet transform, RMSIn this situation, Bandwidth, and average frequency were used in the digital signal processing of the sampled signals. These techniques make it possible the training an intelligent system to identify bearing damage. Therefore, the performance of an artificial neural network was compared. In addition, the silhouette index was used as a statistical validation of the data clusters obtained from the bearing faults group a ddressed in this paper. was obtained through laboratory experiments performed on a 4 HP The database induction machine under conditions of varying torque to demonstrate that the proposed method does not depend on motor torque and can be applied at any loading. The experimental results revealed that the proposed technique was efficacious in the bearing fa ult identification in this work.
Descrição
Palavras-chave
Motores de Indução Trifásicos, Falhas em Rolamentos, Processamento Digital de Sinais, Sistemas Inteligentes, Classificadores de Padrões, Threephase induction motors, Bearing faults, Digital signal processing, Intelligent systems, Pattern classifiers