Intelligent Methods for Feature Extraction and Diagnosis of Rolling Bearing Composite Faults Based on Machine Learning: A Review

Abstract

As a core component of rotating machinery, rolling bearing composite fault diagnosis is of great significance to ensure the stable operation of equipment. The rise of machine learning technology provides a new way for fault feature extraction and diagnosis. This paper reviews the intelligent method of rolling bearing composite fault feature extraction and diagnosis based on machine learning. First of all, it analyzes the background of the field, the importance and the practical significance of the research topic, and describes the intelligent analysis of feature extraction and fault diagnosis. Then the rolling bearing composite fault feature extraction methods are introduced in detail, including wavelet transform, Hilbert-Yellow transform, spectral crag, etc. and their improvement methods, and the corresponding research directions are discussed. Then the intelligent methods of rolling bearing composite fault diagnosis based on machine learning are classified and elaborated, covering neural network, support vector machine, random forest, convolutional neural network, recurrent neural network and its variants, deep confidence network, etc., and the research direction of each method is analyzed. Finally, the current problems and future development trends are summarized, aiming to provide a comprehensive reference for related research.