Fault Diagnosis Research Of Bearing Based On IDA Optimized Extreme Learning Machine

Rolling bearings play an important role in various mechanical equipment,and their health status is closely related to the safe and stable operation of mechanical equipment.The research on the fault diagnosis technology for rolling bearings can not only effectively improve the reliability and safety management level of mechanical equipment,but also have important theoretical application value.In this paper,the research on fault diagnosis methods of rolling bearings is mainly feature extraction and pattern recognition.The research work is as follows:First,in this paper,based on the method of vibration signal processing,the signal preprocessing of the rolling bearing fault vibration signal with non-stationary characteristics.Analyze and compare the commonly used signal processing methods,and finally,the time-frequency domain processing method based on wavelet packet analysis with improved wavelet transform,which has advantages in processing non-stationary signals,is selected.Then,the wavelet packet transform is used to reduce the noise of the bearing vibration signal and extract feature information to construct an effective feature vector of the input classifier.Second,after constructing the feature vector,the diagnosis model based on the extreme learning machine is used for pattern recognition of the bearing state.Aiming at the problems caused by randomly initializing the input weights and hidden layer thresholds of the extreme learning machine,the dragonfly algorithm is used to optimize the input weights and hidden layer thresholds of the extreme learning machine by taking advantage of the optimization speed and global optimization ability.Then construct a diagnostic model based on the dragonfly algorithm to optimize the extreme learning machine,and apply the model to the fault diagnosis of rolling bearings.The experimental results prove the effectiveness of the model,and the modelhas a certain degree of improvement in the speed of testing and diagnosis accuracy of rolling bearing faults compared to the extreme learning machine model.Third,aiming at the problems existing in the optimization of the dragonfly algorithm,an adaptive learning factor is introduced and the differential evolution strategy is integrated to improve the dragonfly algorithm.The improved dragonfly algorithm has better performance such as stability,convergence speed and accuracy.In this paper,the improved dragonfly algorithm is used to optimize the extreme learning machine for input weights and hidden layer threshold parameters,a diagnosis model based on the improved dragonfly algorithm to optimize the extreme learning machine was constructed,and this model was used for the fault diagnosis of rolling bearings to further improve the performance of bearing fault diagnosis of the extreme learning machine.This article uses MATLAB tools to verify the methods used.Wavelet packet transform can effectively reduce noise and extract feature of bearing vibration signal.The diagnosis model based on the improved dragonfly algorithm and optimized extreme learning machine can effectively diagnose the rolling bearing faults.This model not only improves the stability of the diagnosis results of the extreme learning machine model,but also has certain advantages in convergence accuracy and speed and classification accuracy compared with other models.