A Bearing Fault Classification Method Based on Neural Network Attention Mechanism

Abstract

The present disclosure discloses a bearing fault classification method based on a neural network attention mechanism, including: using an acceleration sensor to collect the vibration acceleration time domain signal S of the bearing T , for the vibration acceleration time domain signal S T Perform envelope spectrum transformation to obtain the frequency domain signal S F , the frequency domain signal S F As the input sample of the neural network with attention mechanism; the input sample S F Divide into N different segments and input them into the neural network, process each segment and extract feature values, perform scoring operation and normalization on the feature values, and obtain the same value as the input sample S F Attention weights corresponding to different segments; establish input sample S F The connection between different segments and attention weights; use the attention weights to weight and sum the feature values ​​of each segment to obtain the attention mechanism output value y att ; output value y of the attention mechanism att After the full connection layer is normalized, the probability distribution with a sum of 1 is obtained, and the category corresponding to the highest probability is the fault type of the bearing.

Description

technical field [0001] The present disclosure belongs to the field of bearing fault detection, in particular to a bearing fault classification method based on a neural network attention mechanism. Background technique [0002] Today, in the context of industrial big data, the rapid progress of artificial intelligence and machine learning has made fault diagnosis gradually become intelligent. The use of data-driven intelligent algorithms for fault diagnosis has attracted more and more attention and has become a new research hotspot in the field of fault diagnosis. Bearing fault diagnosis is a popular research direction of mechanical condition monitoring, and intelligent diagnosis methods represented by deep learning are a development trend of bearing fault diagnosis in recent years. At present, the commonly used deep learning methods include convolutional neural networks, deep belief networks, recurrent neural networks, and adversarial neural networks, etc. Such "end-to-end" ...

AI Technical Summary

Problems solved by technology

The essence of deep neural network is a highly nonlinear and non-convex optimization problem. It is very difficult to analyze and explain purely mathematical theory. At present, a complete theory and method system has not been established. Therefore, most of them provide users with some visual information through visualization methods. Intuitive explanation on

Images