Method for diagnosing rolling bearing fault based on stochastic resonance and autoencoder

Abstract

The invention discloses a method for diagnosing rolling bearing faults based on a stochastic resonance and an autoencoder. The method comprises the following steps: establishing a rolling bearing fault diagnosis model; collecting a rolling bearing vibration signal when diagnosing; inputting the rolling bearing vibration signal into the rolling bearing fault diagnosis model; and obtaining a rollingbearing fault diagnosis result. The method selects and optimizes a plurality of parameters of a stochastic resonance system through a genetic algorithm in parallel, adaptively screens out a stochastic resonance system optimal matched with the input signal, and overcomes defects of SR method parameter selection. Signal classification is carried out on preprocessed bearings by using a stacked autoencoder, fault diagnosis is realized, and the fault diagnosis accuracy is up to 96%.

Description

technical field [0001] The invention relates to the technical field of mechanical equipment fault diagnosis, in particular to a rolling bearing fault diagnosis method based on random resonance and an automatic encoder. Background technique [0002] With the rapid development of industrial mechanization, the power and efficiency of mechanical equipment have also been continuously improved, and the working status of equipment has become complex and changeable, which has caused many difficulties in fault diagnosis. Rolling bearings are widely used in machinery industry and other fields due to their advantages in strong load capacity and small friction coefficient. At the same time, they are also one of the most vulnerable parts in rotating machinery. In addition, the characteristics of weak bearing fault signals are not obvious and are easily affected by factors such as noise and human interference, making detection difficult. If bearing faults can be diagnosed and repaired as...

AI Technical Summary

Problems solved by technology

SR is affected by the adiabatic approximation theory. When the frequency of the driving signal increases gradually, the peak of the signal spectrum is far away from the low-frequency region where the noise energy is concentrated, so that the particles lack the support of noise energy when they transition between potential wells, and finally the noise energy cannot be transferred through stochastic resonance. give signal

Therefore, SR theory can usually only detect signals with lower frequencies, which seriously affects the popularization of SR methods in industrial applications.

In recent years, some scholars have made some optimizations to the stochastic resonance method. These methods only consider the optimization of a single parameter, but ignore the interaction between parameters, and fail to fully exert the ability of the stochastic resonance method to extract weak signals.

In addition, for the weak fault signal extracted by stochastic resonance, there is also a lack of verification in actual engineering diagnosis

Images