A New Variable Conditions Intelligent Fault Diagnosis Method for Rotor-bearing Based on Vibration Image Dataset

Abstract

Modern industrial equipment is developing in the direction of automation and intelligence, and intelligent fault diagnosis based on deep learning (DL) has become a hot topic. Traditional fault diagnosis of rotating machinery is mostly based on the fault data obtained by the accelerometer, which has the problems of sparse vibration information and insignificant vibration characteristics. At the same time, the diagnosis algorithm is mostly based on the assumption that a large amount of labeled samples is available, the training and testing dataset are independent and identically distributed. When the mechanical equipment operates under complex and variable working conditions, the performance of traditional fault diagnosis algorithms will be degenerated. Visual vibration measurement has been gradually applied to the field of mechanical fault diagnosis because it can obtain the full-field vibration information with rich texture characteristics and does not produce mass load effect on the measured object. On this basis, this research proposes a new variable-condition fault diagnosis method based on image dataset, which encodes the full-field time-domain vibration information collected by vision into a gray-scale image sequence to enrich the texture to characterize the fault characteristics, instead of traditional accelerometer data for transfer fault diagnosis. The experimental results show that this method can achieve higher classification and recognition results in the task of fault diagnosis of rotor bearing variable working conditions.