DEFECT RECOGNITION AND CLASSIFICATION IN ROLLING ELEMENT BEARINGS USING A NOVEL MACHINE LEARNING TECHNIQUE

Abstract

The rising advancements in Industry 4.0 technologies have made more usual to acquire significant volumes of machine operating data in real time. In response to inconsistent data distribution and label scarcity in target domains, this work suggests a machine learning (ML) approach for rolling element bearing failure identification under a variety of circumstances. This study presents, a new method called Composite coyote optimized resilient linear regression (CCO-RLR) for defect recognition and classification in rolling element bearings. Early rolling bearing failure diagnosis is a crucial and time-sensitive operation that guarantees the dependability and security of mechanical fault systems. Initially, the rolling element bearings dataset is collected and preprocessed using Min-max normalization. For extracting the feature, Fourier transform (FT) is employed. The result shows that the CCO-RLR accuracy is 97.8% when compared with those existing methods. Our suggested method offers an effective means of quantifying flaws and significantly improving classification effectiveness.