Research On Life Prediction And Fault Diagnosis Of Rolling Bearings Based On Adaptive VMD And Optimized Neural Network

Rolling bearings are widely used in many fields in the machinery industry due to the advantages of high speed,high efficiency and low noise.The working environment of rolling bearing is usually harsh,resulting in a large life span and a high failure rate.The use of its vibration signals produced during the working process for fault diagnosis and life prediction can not only reduce the probability of accidents in mechanical equipment,but also provide reliability decision support for equipment maintenance plans,which has very important practical significance.At present,the variational modal decomposition(VMD)method is widely used in signal decomposition,but its parameters are still not easy to accurately select.For the existing life prediction and fault diagnosis models,it is also difficult to find a great balance between training accuracy and complexity,many of which still stay in the theoretical stage and have not yet been promoted and applied systematically.In response to these problems,in this paper,taking the rolling bearing as the research object,new methods of VMD processing,life prediction and fault diagnosis based on the vibration signals were studied.And a system was designed for bearing vibration signal collection and fault diagnosis.Then,the experimental verification of the main methods and system was carried out.This research may provide help for the health maintenance of machinery and equipment.The specific work can be concluded as follows:(1)A method for VMD parameter optimization was explored and applied to the decomposition of rolling bearing vibration signals and the extraction of fault features.Aiming at the problem that the number of modal components and the penalty factor in VMD are difficult to determine,the whale optimization algorithm was introduced to make it self-optimizing.Then the VMD with optimized parameters was used to decompose the fault simulation signal and the measured signal of the rolling bearing,and the obvious fault frequency was extracted from the envelope spectrum of the best modal component,which proved the feasibility of the method.(2)In order to ensure the high accuracy of the life prediction results and the low complexity of the training process,a rolling bearing remaining life prediction model was constructed based on multi-dimensional features and an improved BP neural network.Taking the bearing life vibration signal as the research object,firstly,the noise was reduced by using singular value decomposition,and then the multi-dimensional features were extracted and reduced by the correlation coefficient method.Besides,the kernel principal component analysis method was used to reduce the dimensions of the reduced features and construct a health index that can characterize the bearing degradation trend.Then the remaining life prediction model based on BPNN was established,and the weights and thresholds of BPNN were jointly optimized by particle swarm optimization and simulated annealing algorithm.Finally,each model was adopted to test the three bearings,and the root mean square error,mean absolute error,correlation and final prediction error percentage were used to evaluate the results.The results showed that the optimized life prediction model had the best prediction effect.(3)The application of improved cuckoo search algorithm and Elman neural network in the identification of rolling bearing faults is studied.Firstly,the rolling bearing vibration signal acquisition experiment was carried out based on the mechanical transmission fault implantation diagnosis test platform.Then,the optimized VMD and multi-scale permutation entropy extraction were combined to construct feature vectors,and they were input into the fault diagnosis model for simulation testing.Finally,the open source data was used to verify the model again.The results showed that the method proposed in this paper had a more excellent and stable fault classification effect.The average classification accuracy rates in the two tests were as high as 97.22% and 97.14%,which were 7.91% and10.32% higher than the original network.(4)The Lab VIEW and MATLAB software were adopted to design the rolling bearing vibration signal acquisition and fault diagnosis system.It mainly included the functions of user login and management,signal monitoring and acquisition,signal processing and analysis,and fault diagnosis.After that,the main modules of the system were tested,and the results verified the effectiveness of the system.