Research On Fault Diagnosis Of Rolling Bearing Based On Directed Acyclic Graph Relevance Vector Machine

Rolling bearings,as connectors and fixings,become the core components of large-scale mechanical equipment,such as wind turbines,high-speed trains and aircraft engines.However,the rolling bearing is easy to be damaged due to its own characteristics.If there is an unexpected failure,it will not only lead to the shutdown of the equipment,but also cause incalculable economic losses.Therefore,intelligent and timely fault detection and diagnosis of rolling bearing is of great research value and practical significance.In this paper,the basic knowledge of rolling bearing and the data and platform used in the experimental simulation are studied.In view of the rolling bearing vibration signal denoising,feature extraction and classification algorithm,some improvements are made.The specific research contents are as follows:(1)According to the characteristics of non-stationary vibration signal,a new preprocessing method of CL4 multi wavelet optimized by balanced multi wavelet is proposed,and double comparison and verification are carried out at the same time.The GHM multiwavelets and CL4 multiwavelets are optimized by repeated row preprocessing,approximation order preprocessing and the preprocessing method proposed in this paper,and then the signal is denoised.The overall comprehensive verification shows that the CL4 multiwavelet with balanced multiwavelet processing is the best in denoising,and it can also represent the signal completely.(2)Aiming at the feature extraction of vibration signal,an ensemble empirical mode decomposition(EEMD)method is proposed,which is an improvement of EMD method.In view of the problem that EMD method is easy to produce modal aliasing when processing signals,EEMD method can effectively suppress this problem.After de-noising,the vibration signal is decomposed into several modal components by EEMD,and the first six IMFs components with rich fault information are selected as eigenvectors.The EEMD algorithm is more suitable for the field of signal processing and has higher accuracy.(3)A diagnosis model based on DAG-RVM(directed acyclic graph relevance vector machine)is proposed for rolling bearing faults.At the same time,the 0AO-RVM and OAR-RVM fault diagnosis models are constructed.Combined with the methods of this subject,through the comparative analysis of experiments,it is verified that this method has more advantages than the other two in fault classification.Compared with DAG-SVM classifier,RVM classifier model is more suitable.