Research And Implementation Of Industrial Equipment Intelligent Fault Diagnosis System Based On IHHO-DBN

In recent years,in order to pursue high performance,high efficiency,low cost and safe and reliable mechanical rotating equipment system,intelligent fault diagnosis of mechanical equipment has attracted more and more attention.Rolling bearing is an important part of mechanical equipment.Once a fault occurs,it may lead to accidents or even catastrophic events.In order to reduce maintenance costs and avoid major losses,rolling bearing health monitoring and fault diagnosis are very necessary.With the continuous progress of artificial intelligence technology based on deep learning,data-driven model has been widely used in intelligent fault diagnosis due to its strong feature extraction and generalization ability,and has become a hot research direction.After the research and analysis of the existing fault diagnosis system,an intelligent fault diagnosis system based on improved Harris Hawk Optimized Deep Belief Network（IHHO-DBN）is designed.Because the number of nodes in the hidden layer of DBN-ELM model is generally determined by human experience and there is a contingency,an improved HHO method is proposed to optimize the number of hidden layer nodes in the DBN model.On the one hand,the traditional HHO is easy to fall into local optimum,so the differential mutation operator is added to the exploration stage to increase the population diversity and improve the optimization accuracy.On the other hand,the escape energy factor is improved from linear to nonlinear in order to balance the exploration and development ability and solve the problem of unbalanced global and local search ability of HHO.Finally,the IHHO algorithm is applied to the DBN-ELM model to construct the fault diagnosis system.The system uses the bearing data set of the University of Caessis Reserve in the United States.The software system is designed as five modules : registration login module,user management module,original data processing module,model building module and intelligent fault diagnosis module.In the original data processing model,the maximum second-order cyclostationary blind deconvolution algorithm（CYCBD）is used to reduce the noise of weak fault or low signal-to-noise ratio pulse signals to enhance the periodic impact components in the signal,and then the time domain feature vector is selected as the model input.In the model construction module,the improved HHO algorithm proposed in this paper is used to determine the DBN-ELM network structure and construct the diagnosis model.Finally,the type of unknown fault is judged in the intelligent diagnosis module to obtain the diagnostic accuracy.Finally,the experimental results show that the proposed IHHO-DBN can extract effective fault features from original time domain signals and has high recognition accuracy.The designed intelligent fault diagnosis system can reach the expected standard in function and performance,and has a certain theoretical basis and industrial practical application prospect.