Rolling bearing fault diagnosis method based on GAF-DRSN

Abstract

The invention provides a rolling bearing fault diagnosis method based on GAF-DRSN. The rolling bearing fault diagnosis method comprises the steps of performing fragmentation processing on a large-scale one-dimensional time sequence; converting the fragmented one-dimensional time sequence signal into a two-dimensional image with a reserved time feature based on a Gramb angle field principle; building a DRSN neural network model; a two-dimensional image obtained through one-dimensional time sequence conversion is used as input to train a DRSN network, and a fault diagnosis model for the rolling bearing is obtained; rolling bearing data with different fault features are mixed to build a two-dimensional data set, and the two-dimensional data set is used for testing and verifying the generalization ability of a diagnosis model. According to the method, the limitation that the ResNet network cannot adaptively extract time features in fault features is overcome, the capability of removing redundant information of the model is enhanced, and the reliability of the model diagnosis effect and the diagnosis accuracy are improved.

Description

Technical field [0001] The present invention relates to the field of vibration signal analysis and processing and fault diagnosis technology, specifically to a rolling bearing fault diagnosis method based on GAF- DRSN (Grame angle field - deep residual shrinkage network). Background [0002] With the rapid development of the current industrial technology, mechanical equipment is becoming more and more complex, and the harsh working environment such as tunnels and mines that are difficult for humans to set foot in is gradually replaced by various mechanical equipment, and the problem that follows is to achieve more and more complex functions, mechanical equipment becomes more and more sophisticated and intelligent, the equipment structure becomes more complex, and the parts are more numerous, so in the changeable and harsh working environment, the failure rate of mechanical parts is greatly increased. Therefore, it is particularly important to study the rapid, efficient and accura...

AI Technical Summary

Problems solved by technology

However, none of them can adaptively extract the time features in the fault features, so the model will ignore the time sequence in the fault information, and the obtained diagnosis results cannot be used as reliable output, and the more traditional one-dimensional convolution training parameters are too large, The model is not lightweight enough and is limited by hardware performance. In the absence of mobile workstations or poor performance of mobile workstations, the diagnostic performance will be greatly affected, and rapid diagnosis cannot be achieved.

Images