The Application Research Of Immune Adaptive DCNN Fault Diagnosis

Rolling bearing is a key equipment of mechanical equipment,there are many researches on intelligent fault diagnosis applied to it,but there are still many deficiencies.The time and frequency domain signals of rolling bearings contain important information about operating conditions.In the model of rolling bearings fault diagnosis algorithm based on vibration signal feature extraction,firstly,the vibration signal is pre-processed and feature extracted,finally,the classifier is used to classify faults.This kind of fault diagnosis method depends on human experience in feature extraction,which leads to the incomplete and inadequate of feature extraction.In this paper,the deep convolutional neural network and antibody immunity are introduced into the field of rolling bearing fault diagnosis,and an immune self-adaptive deep convolutional neural network rolling bearing fault diagnosis model is proposed.The deep convolution neural network is used to extract the characteristics of time domain and frequency domain signals of rolling bearing,which realizes direct mapping from raw data to diagnostic result.This makes the fault diagnosis process more intelligent and reduces the interference of humanfactors.Deep learning has been applied in the field of bearing faults,which is mainly the identification of known bearing faults,but rolling bearing operating state is dynamically changing.When an unknown fault occurs,the deep learning model is unable to adjust adaptively,the model must be retrained,resulting in poor model adaptation.When a unknown fault occurs in rolling bearing,the deep convolutional neural network is feature extraction module,the extracted features are mapped to antigens,the unknown fault detectors are generated by immune self learning characteristics and use for quick diagnosis when the fault type reappears,realizing the recognition of the unknown fault.Aiming at the requirement of real time fault diagnosis and the problem of long time for the detection and learning of adaptive algorithm model,a group colony rapid diagnosis model based on deep convolutional neural network is proposed.In the detection phase,the fault types are determined by comparing the time domain and frequency domain diagnosis result.In the learning phase,the learning efficiency of unknown faults is improved by the grouped cloning strategy and the continuous region mutation operation.The experimental uses the rolling bearing data set published by Case Western Reserve University.The experiment result show that the immune adaptive deep convolutional neural network model can identify the known bearing fault types,which is better than the fault diagnosis method based on feature extraction.It can adaptively learn unknown faults,and the secondaryrecognition accuracy reaches 98.32%.In terms of the degree of fault discrimination,the network model based on deep convolutional neural networks can distinguish the degree of fault well and extract the slight difference feature of the signal waveform.The fast fault diagnosis model proposed by the grouped cloning strategy,comparing with the adaptive algorithm model,its detection accuracy is comparable,but the detection and learning efficiency has been significantly improved.