Research On Parallel Fault Diagnosis Method Of Rolling Bearing Based On SVPSO-BP

Rolling bearings are widely used in various types of rotating machinery such as wind turbines and oil drilling equipment.As an important part and main part of rotating machinery,they not only have to bear various alternating loads during the working process,but also It is very prone to failure due to factors such as processing errors,improper installation and improper operation.When the rolling bearing fails,it will produce abnormal vibration and noise.In severe cases,it may even damage the equipment,causing shutdown and production.The study of accurate and effective rolling bearing fault diagnosis methods has very high economic and social value.As a general supervised machine learning algorithm,BP neural network uses BP algorithm and labeled training samples to optimize randomly initialized network weights to achieve prediction or classification.It is widely used in the fault diagnosis of rolling bearings.When the BP neural network is applied to the fault diagnosis of rolling bearing,the weight and threshold of the network will affect the fault diagnosis accuracy.At the same time,the fault diagnosis method of signal processing to construct input feature vector combined with BP neural network has real-time problems.Therefore,this thesis studies the fault diagnosis method of rolling bearing based on BP neural network.main tasks as follows:(1)The optimization method of rolling bearing fault diagnosis based on BP neural network is researched.Because the relevant parameters of BP neural network and the network input feature vector will affect the accuracy of fault diagnosis,this thesis first uses the method of VMD combined with time-frequency entropy to construct signal feature vector,and solves the problem of extracting fault signal feature by using VMD combined with other information entropy signal feature vector construction methods Insufficient problem: The optimized particle swarm algorithm is used to obtain the optimal weights and thresholds of the BP network to solve the defect that the BP neural network is easy to fall into the local extreme value,so as to improve the fault diagnosis accuracy of the BP neural network.(2)The GPU-based VMD-SVPSO-BP parallel algorithm is studied.Aiming at the problem of poor real-time performance of the VMD-SVPSO-BP algorithm and the existence of bottleneck functions,this article first analyzes the feasibility of parallel VMD decomposition,using multi-threaded parallel computing to accelerate the process of extracting fault signal features;on the other hand,using MATLAB explorer Find the "bottleneck" functions existing in the SVPSO-BP algorithm,and then eliminate these "bottleneck" functions through GPU programming,thereby improving the real-time performance of rolling bearing fault diagnosis.(3)A fault diagnosis system for rolling bearings is designed using MATLAB GUI.The system includes three modules: "data processing","feature extraction" and "fault diagnosis".The "data processing" module realizes the function of analyzing signal time-frequency waveforms and signal decomposition;the "feature extraction" module realizes the function of extracting fault signal characteristics and the visualization of characteristics;"fault diagnosis" realizes the function of diagnosing rolling bearing faults with different neural networks.In the fault diagnosis system designed by GUI,the interface is simple and clear,and users can easily perform the whole process of "data processing","feature extraction" and "fault diagnosis" on the rolling bearing vibration data through the preset function controls and modules.