Research On Motor Bearing Fault Diagnosis Technology Based On Compressing Sampling

With the continuous development of science and technology,electromechanical equipment such as electric cylinders and other mechanisms are more and more widely used in the field of vehicle equipment and intelligent manufacturing.As a very basic actuator in electromechanical equipment,the failure of its bearings will cause the overall equipment significant losses.In order to carry out real-time monitoring,by collecting vibration signals at key locations of the equipment,large equipment that needs to collect many locations will cause huge pressure on data transmission and storage.The compression sampling technology based on the principle of compression sensing can be used to compress signals while collecting signals.This paper takes motor bearings as a research object,proposes a fault diagnosis method based on a compression sampling model and a shift-invariant dictionary,and the accuracy of fault diagnosis Carried out simulation verification.This paper first analyzes the main fault form and vibration mechanism of the motor bearing,explains the applicability of the vibration signal of the bearing in the compression sensing theory,establishes the actual sampling simulation model based on the compression sensing theory,and derives its mathematics model to extract the projection matrix for signal reconstruction.Then,for the reconstruction of the signal with background noise,combined with a shift-invariant dictionary learning algorithm and a wavelet packet threshold denoising method,the noise reduction preprocessing of the training data,and then through the dictionary learning to obtain the basis with pure fault impact features function,the timeshift expansion of the base function is used to obtain an over-complete dictionary for signal reconstruction.Through simulation analysis,the reconstructed signals under different compression sampling rates are compared.The results show that less data can be used to accurately reconstruct the fault characteristic waveform in the original signal,and the reconstructed signal contains less noise.Then the method of motor bearing feature extraction is studied.In this paper,a method of directly processing the sparse coefficient is proposed.The energy characteristics of the reconstructed signal on different basis functions are calculated,and the Relief F algorithm is used for feature selection.For comparison,the modal decomposition method is used to process the reconstructed complete signal waveform.After obtaining the modal function,the kurtosis value and the fuzzy entropy are used as feature vectors,respectively.The results show that the method proposed in this paper can better distinguish fault.Finally,this paper establishes a motor bearing fault diagnosis model,uses a support vector machine classification algorithm,uses particle swarm optimization to find the optimal value for hyperparameters,compares and analyzes the fault classification accuracy obtained by several commonly used kernel functions.Using the average value of the accuracy of different fault recognition as an indicator,the Case Western Reserve data set was used to verify.The results show that when the compression sampling rate is high,both methods can achieve good results.When the sampling rate is low,the accuracy of the basis function energy feature method is higher,and the average diagnostic accuracy of 98% can be obtained when the compression sampling rate is 0.1.Then verify the situation of the composite fault,and still have higher accuracy through the basis function energy feature method.