Research On Intelligent Fault Diagnosis Technology Of Mechanical Equipment Based On Compression Measurement

Rolling bearings are an important component of mechanical equipment.After long-term use in complex and harsh working environments,rolling bearings will inevitably have a variety of failures,greatly affecting the working performance of the whole machine and causing serious safety accidents.Once an accident occurs,it will cause significant human and financial losses,so it is very important to carry out health monitoring and fault diagnosis of rolling bearings.The vibration signal transmission of the rolling bearing carries a lot of information contained in its working process,and the vibration parameters can directly,quickly and accurately reflect the running state of the bearing than other various parameters.Therefore,based on the vibration signal of the rolling bearing,combined with the newly appeared compression sensing theory and neural network method,the fault diagnosis technology of the rollingbearing is deeply researched.The main research resultsobtained are asfollows:(1)The adaptability of commonly used measurement matrix compression to measure mechanical vibration signals is studied.First,the principles of deterministic measurement matrix,random measurement matrix and partial random measurement matrix are introduced;then,based on different orthogonal basis,the sparseness of mechanical vibration signal is verified and analyzed,that is,compressibility;finally,common The measurement matrix is used to simulate the vibration measurement of the mechanical vibration signal.According to the simulation test results,the relationship between the reconstruction error after the compression measurement using various measurement matrices and the reconstruction is changed with the number of measured values,and the performance of various measurement matrix compression measurement mechanical vibration signals is further analyzed and compared.(2)An optimal OST measurement matrix suitable for mechanical vibration signals is studied.First,the characteristics of the classical deterministic measurement matrix orthogonal symmetry Toeplitz matrix are studied and the mechanical vibration signal is compressed.It is pointed out that the orthogonal symmetry Toeplitz matrix or OST measurement matrix is still to be reconstructed for the mechanical vibration signal.Improvement;Secondly,starting from incoherence,the threshold iterative contraction algorithm is used to optimize the measurement matrix,which greatly reduces the mutuality between the measurement matrix and the sparse basis;finally,in order to improve the column independence of the OST matrix,singular values are used The decomposition algorithm further optimizes the measurement matrix,and finally obtains the optimal OST matrix and carries out simulation experiments.The results show that the optimized OST matrix can reconstruct the original vibration signal more accurately than the OST measurement matrix,the Toplitz matrix and the Gaussian random matrix when the compression rate is fixed and variable.The compression measurement effect of the OST matrix is better than othermeasurement matrices.At the same time,the construction complexity of the optimal OSTmatrix is much lowerthantherandom matrix,which is conducivetoengineering technology.(3)Based on the optimal OST matrix,an intelligent diagnosis method for rolling bearing faults is proposed.First,the original OST matrix is used to perform compression measurement on all the original mechanical vibration signals to obtain compression measurement values;secondly,a rolling bearing BP neural network fault diagnosis model is established based on the compression measurement values and the model parameters are determined;finally,the original time domain The signals and compressed domain signals are input to the fault diagnosis model for comparative analysis.The experimental results show that the use of compressed domain signals for fault diagnosis has a higher recognition rate;the compressed domain signals processed by different measurement matrices are input to the neural network for comparison.The experimental results show that the compressed domain signal obtained by using the optimal OST measurement matrix also has a higher recognition rate.The method in this paper only needs a small amount of compressed measurement values to represent the original signal to realize the intelligent diagnosis of the bearing fault signal,saving the time required to directly process a large numberof original signals.