Research On Bearing Fault Acoustic Emission Signal Noise And Feature Recognition Method

Rolling bearings,as one of the most important components in rotating machinery,are susceptible to damage due to factors such as load,manufacturing process,and material.Bearing damage can pose serious safety hazards to industrial production.With the advancement of China’s Made in China 2025 strategy,equipment health and fault diagnosis has become top-priority.Acoustic emission testing is an online non-destructive testing technology that can detect bearing faults effectively.However,this technology currently has the following issues: dependence on experience and expert systems,detection signals are often affected by noise interference,and the reliability of its results in practical applications needs to be improved.In addition,conventional feature extraction methods have limitations,while signal features are crucial for bearing fault classification and identification.To address these problems,this article collects acoustic emission signals from the bearing fault experimental platform,studies and improves the relationship between bearing faults and acoustic emission signals,and accurately identifies and classifies rolling bearing acoustic emission signals.The following research contents are conducted for bearing fault recognition:(1)The relationship between rolling bearings and acoustic emission signals is studied.Bearing signal acquisition experiments are carried out on a rotating machinery fault simulation test platform,and the expression of bearing fault characteristic frequency is obtained.(2)Using empirical wavelet transform method for bearing signal feature extraction.By deducing wavelet transform and EMD decomposition,analyzing the shortcomings of EMD decomposition,and proposing empirical wavelet transform theory.This method preserves the characteristics of the target signal more completely and solves the problems of modal aliasing and endpoint effects,avoiding the situation where the target signal cannot be fully reconstructed.The simulation results validate the effectiveness of the empirical wavelet transform method.(3)For the bearing fault classification problem,this article simplified the coding method of genetic algorithm and used an improved genetic algorithm to optimize the BP neural network.This optimization algorithm improved the fitness of the population evolution and improved the stability of global optimization while ensuring the convergence speed of the algorithm.This article also analyzed the spectrum of bearing faults,selected seven typical fault feature parameters for bearing fault identification and classification.After testing with experimental data and bearing fault test bench data,the improved genetic algorithm optimized BP neural network recognition effectively improved the recognition efficiency of rolling bearing faults.Compared with BP algorithm and PSO-BP algorithm,the recognition accuracy increased by about 5%,improved the problem of BP neural network easily falling into local optima,and verified the effectiveness of the algorithm.