Research On Bearing Fault Diagnosis Method Based On Improved Extreme Learning Machine

Rolling bearing is an important component of mechanical equipment.Its high frequency.operation in complex working conditions,the quality of the operation directly affects the safety and stability of the overall operation of the mechanical equipment.Therefore,the fault diagno-sis of rolling bearings is the research focus in the field of mechanical fault diagnosis,and has important research significance and practical value.The vibration signal of rolling bearing has noin-linear and non-stationary characteristics.Firstly,after comparing and analyzing the limitations of short-time Fourier transform and wave-let transform in the processing of bearing vibration signals,the wavelet packet transform is used to denoising and feature extraction of bearing vibration signal.Secondly,aiming at the problem that fault feature information can not directly and accurately identify the fault type.In thesis,based on the extreme learning machine algorithm and the global search advantage of particle swarm optimization,a PSO-ELM fault diagnosis model based on the extreme learning machine of particle swarm optimization is constructed.In this model,the initial input weights and hidden layer thresholds in the extreme learning machine are regarded as the spatial positions in the particle swarm,and the parameters in the global range are optimized.This optimization method can effectively solve the problem of randomly generated input weights and threshold values in extreme learning machine,which affect its generalization ability.However,the PSO-ELM model has certain feasibility in bearing fault diagnosis,but there is still room for improvement in its network performance.Therefore,in order to solve the problem of local optimization and premature convergence of particle swarm optimization extreme learning machine parameters in PSO-ELM model,thesis improves the genetic operators such as selection,crossover,muta-tion and so on in genetic algorithm.An improved genetic particle swarm optimization IGA-PSO hybrid optimization algorithm is proposed.By introducing improved genetic operators into particle swarm optimization algorithm,the population diversity is enhanced,the local and global search ability of particle swarm optimization algorithm is balanced,and the IGA-PSO-ELM fault diagnosis model based on IGA-PSO algorithm is proposed.Through the IGA-PSO hybrid algorithm,the input weight and threshold parameters of the extreme learning machine are optimized,and the diagnostic performance of the extreme learning machine is further im-proved.In thesis,the MATLAB experiment proves that the wavelet packet transform can effec-tively extract the feature vector of the bearing vibration signal.Through the comparison and analysis of multiple model examples,it is verified that the IGA-PSO-ELM fault diagnosis model has faster convergence speed,better convergence accuracy and higher classification per-formance,and has better bearing fault diagnosis results.