Fault Diagnosis Method Of Acoustic Emission Signal Mechanical Equipment Based On Compressed Sensing

With the rapid development of the industrial manufacturing industry,the health of machinery and equipment is always affecting the industrial production process,causing serious safety problems and maintenance efficiency issues,so it is of great practical importance to detect early equipment failures and solve them in an effective way.Compared with vibration signal analysis methods,acoustic emission signals are more sensitive to faults,less affected by interference,and more effective for weak faults,but the high sampling frequency and high number of sampling points of acoustic emission signals bring a challenge to analysis and storage.This paper introduces a compressed sensing processing framework and carries out research on techniques and methods for data compression,feature extraction and diagnostic evaluation.The main research content is as follows:1.The current research status of fault diagnosis technology at home and abroad is introduced in detail,its research background and significance are explained,and the current applications of compression sensing in the field of fault diagnosis are investigated.2.Research is carried out in three areas: sparse representation of signals,compressed reconstruction and compressed projection,to investigate the compression characteristics of signals and to minimise the amount of data while ensuring that the compressed data contains sufficient original information.3.Taking rolling bearings as the research object,a wavelet sparse convolutional network is constructed,a wide convolutional kernel based on continuous wavelets is introduced,and an energy pooling layer is designed to strengthen the compression feature depth mining capability;a regularised loss function is introduced to improve the diagnostic accuracy and robustness through feature sparsification.Experiments are carried out on a home-made bearing test rig,and the experimental results demonstrate that the extracted fault features of bearing acoustic emission signals can improve the analysis efficiency and achieve accurate classification of bearing faults.4.A deep compression learning method based on compression perception is proposed using rolling bearings and RV gearboxes as research objects.Without reconstructing the signal,fault features are mined directly from the compressed data and diagnosed and identified.Experiments are conducted on a home-made experimental bench,and the experimental results show that the method not only has advantages in terms of diagnostic accuracy,but also has high computational efficiency and generalisation performance.5.A data analysis system has been developed based on MATLAB,and the methods in this paper are applied to the system to achieve data analysis,data compression,and optimal compressed data component selection.