Research On Fault Diagnosis Method Of Rolling Bearing Based On Feature Fusion And ELM

With the development of industrialization,more and more industries use large-scale automation equipment for production,and rolling bearings are one of the commonly used components of large-scale automation machinery and equipment.If it fails during operation,it will have a normal life and production.It has a huge impact and may even threaten our personal safety.Therefore,research on fault diagnosis of rolling bearings has extremely important significance.This paper focuses on the problem of signal feature extraction,feature fusion and fault identification of rolling bearing faults.The main research content includes the following parts:(1)Aiming at the problem that it is difficult to extract the fault features of rolling bearings under the background of noise,a rolling bearing fault diagnosis method combining generalized morphological difference filtering and time domain feature extraction was studied.This method first applies the generalized morphological difference filtering method to the original signal to filter out the noise component in the original signal;secondly,calculates the time domain index of the signal after noise reduction to obtain a series of characteristic index values;finally,the extracted characteristic index As an input of an extreme learning machine(ELM)model,an ELM fault classification model is established to identify the different running states of rolling bearings.From the experimental results,it can be concluded that the characteristics of the vibration signal become easier to distinguish after being processed by the generalized morphological difference filtering.(2)To address the problem that multi-domain feature sets can reflect fault characteristics more comprehensively and accurately than single-domain features,but will lead to the performance degradation of pattern recognition algorithms,multi-feature extraction and kernel principal component analysis(KPCA)are proposed Combined with the fault diagnosis method,this method first uses intrinsic time scale decomposition(ITD)to decompose the vibration signal,and separately calculate the time domain,frequency domain and entropy value of each inherent rotation(PR)component Features;Then,KPCA is used to extract the obtained features a second time,and the secondextracted feature set is used to construct an ELM fault diagnosis model to realize the recognition of various states of rolling bearings.The experimental results of the bearing show that this method can effectively remove the redundant information between the features,realize the rapid and accurate diagnosis of rolling bearings,and have good practical application value.(3)For the time domain,frequency domain and entropy value features belong to a single scale feature,it is difficult to fully reflect the running state of the bearing,resulting in a low fault recognition rate,a bearing fault diagnosis based on multi-scale feature extraction and kernel principal component analysis is proposed model.Based on multifeature extraction and KPCA-based rolling bearing fault diagnosis,the model uses the advantages of multi-scale information extraction,uses multi-scale feature values as feature vectors,then fuses the feature vectors with KPCA,and finally uses the fused features The vector establishes the ELM fault diagnosis model to realize the identification of rolling bearing status.Compared with the traditional single-scale feature extraction method,multi-scale feature extraction adds a scale factor on the basis of timedomain features,and the addition can greatly improve the method.Advantage.