| The mine ventilator is a key piece of equipment for safe production in mines.With the gradual development of the country’s industrial equipment in the direction of largescale,integrated and intelligent,the National Development and Reform Commission and other departments have also put forward higher requirements for the intelligent transformation of mine ventilation fans.This thesis takes the mine ventilator as the research object,analyses its common fault mechanism,clarifies the causes of faults and typical characteristics,and proposes a mining ventilator bearing fault diagnosis method based on multi-feature fusion and CPA-BLS with the objective of improving the fault diagnosis accuracy.The thesis mainly includes:Vibration signal decomposition: The signal processing method of variational modal decomposition(VMD)is adopted to transform the signal decomposition problem into a constrained optimization problem,which effectively solves the endpoint effect and modal confusion existing in signal processing;in order to address the problem of poor decomposition accuracy caused by improper selection of VMD parameters,the grey wolf algorithm with the minimal value of envelope entropy as the fitness function is proposed to optimize the VMD algorithm;in order to verify the advantages of the VMD algorithm and the effect of the improved VMD algorithm,the signal decomposition experiments of EMD,VMD and the improved VMD are carried out on the vibration signals respectively,and it is proved that the GWO-optimized VMD algorithm is more robust.Feature extraction: On the basis of signal decomposition,the alignment entropy,which can detect kinetic mutations,is introduced for the initial fault feature extraction;for the incomplete coverage of fault information by a single feature,on the basis of the initial alignment entropy feature,time-domain and frequency-domain features are introduced,and the multi-domain feature screening and fusion is carried out by using the feature evaluation method of intra-class and inter-class sensitivity;for the fused features with high-dimensional mixing characteristics,which are easy to generate For the high-dimensional mixed characteristics of the fused features,which are prone to information redundancy,a local linear embedding method is used to reduce the dimensionality of the non-linear high-dimensional feature data to realize the feature extraction of multi-fusion.The experimental verification shows that the method can show the characteristics of vibration signals at different scales while ensuring the essential features of the original signal,and the effect is better than that of single parameter features.Fault diagnosis classification: A fault diagnosis classification algorithm based on Broad Learning System.(BLS).In view of the fact that the improper selection of the number of enhancement nodes and the number of mapping nodes of BLS can lead to poor generalisation performance of the diagnosis model,the carnivorous plant algorithm(CPA)with strong search capability is used to dynamically seek the parameters of the BLS algorithm;Using the extracted fused features as the input layer,the CPA-BLS fault diagnosis model is experimentally validated to achieve better classification results with a more lightweight model structure;Finally,the CPA-BLS is compared with PSO-BLS,GWO-BLS and WOA-BLS models to demonstrate that the CPA-BLS fault diagnosis model based on multi-feature fusion has a high correct diagnosis rate.Fault diagnosis system: Based on theoretical research,an online fault diagnosis system for mining ventilator bearings has been designed through the joint programming of Lab VIEW,MATLAB and SQL Server software platforms.The system has been applied to the actual coal mine production after several experiments and has realized the vibration data collection and analysis,feature extraction,fault diagnosis and data storage of the ventilator bearings during the operation.This thesis includes 73 figures,15 tables and 110 references. |
PDF Full Text Request