Research On Fault Diagnosis Of Ventilator Bearing Based On Negative Selection Algorithm And Extreme Learning Machine

As a typical rotating machinery,ventilator plays an important role in mine ventilation,boiler ventilation and metallurgical industry.Therefore,It's of great significance to research on bearing's condition monitoring and fault diagnosis.By considering the ventilator as a research object,in this paper,the features of bearings in different positions of ventilator in different fault conditions are extracted,and the fault early warning and diagnosis of ventilator bearings are realized.The main research contents are as follows:In feature extraction part,firstly,Hilbert-Huang Transform(HHT)algorithm is introduced,and support vector machine for regression(SVR)is used to extend signal at both ends to avoid end effect problem in empirical mode decomposition(EMD)algorithm.Meanwhile,experimental results show that SVR can effectively solve the EMD endpoint effect problem;Next,in order to solve the modal aliasing problem in EMD algorithm,a complete ensemble empirical mode decomposition with adaptive(CEEMDAN)is proposed to solve the modal aliasing problem,and the superiority of algorithm is verified by decomposition experiment of bearing fault signal;Finally,CEEMDAN-energy and CEEMDAN-permutation entropy method are used to respectively extract features of bearing fault signal.And for the lack of uniformity within the CEEMDAN-permutation entropy feature extraction algorithm,a weighted permutation entropy feature extraction model based on CEEMDAN decomposition is proposed,which lays a good foundation for fault early warning and fault diagnosis.In fault early warning part,a negative selection algorithm is introduced.At the same time,three fault feature extraction methods based on CEEMDAN-energy,CEEMDANpermutation entropy,and CEEMDAN-weighted permutation entropy are selected in order to respectively combine with fixed radius negative selection algorithm(NSA)and negative selection algorithm with variable sized detectors(V-detector NSA)to implement fault early warning of ventilation bearings.In fault diagnosis part,extreme learning machine(ELM)is used to diagnose bearing faults in ventilators.firstly,in order to solve the problem of unstable classification performance caused by input weights and hidden layer bias in the ELM algorithm,differential optimization algorithm(DE)is introduced to optimize the two parameters in ELM algorithm.Next,CEEMDAN-weighted permutation entropy feature extraction method is combined with DE-ELM to construct a fault classification model,and this model is applied to diagnose fault of ventilation bearings;Finally,experimental results show that comparing with CEEMDAN-entropy and CEEMDAN-permutation entropy feature extraction algorithm,the proposed model has better performance in accuracy and timeliness.