Research On Intelligent Fault Diagnosis Of Rolling Bearing Based On Improved NLM

With the development of modern industry and the progress of science and technology,the status of rotating machinery in machinery and equipment industry is increasing day by day,but the failure rate increases as well.As an important part of rotating machinery,the working status of bearing directly affects the performance of the whole equipment,so study the method of diagnosising bearing fault has a great practical significance.This paper starts from three aspects including noise reduction,feature extraction,feature selection and feature classification of rolling bearing fault signals.Several key problems existed in the diagnostic methods are deeply studied.The main innovations of this paper are as follows:The components of the rolling bearing faults vibration signals are complex.The fault features of the bearing are usually difficult to extract under the interference of the uncorrelated frequency components and noise.According to the shortcomings that the EMD method can not accurately extract the features under strong noise,this paper proposed a new extraction method based on the non-local means and EMD algorithm,this algorithm can preserve the detail features of the signal to the maximum degree.Then the denosing signal is decomposed into multiple components by EMD and remove the redundant signal to get more accurate fault signature information.In order to reduce the computational complexity and improve efficiency,a fast NLM algorithm was proposed,and then we used a new Wave-exponential weight kernel function,which solved the problem of insufficient weighting of exponential function in FNLM algorithm.Experiments show that this method can identify the characteristic frequency of bearing fault quickly and effectively.In the stage of feature selection,a two-phase feature selection method is adopted,and a new hybrid feature selection technique combining DET and PSO is used.The DET can effectively filter out irrelevant features,and the PSO algorithm was used to get a superior subset of features.In the classification stage,we compared two popular classification methods K nearest neighbor and Support Vector Machine to identify the fitness to the proposed method.Finally,one of the most suitable classification methods is found out.According to the experimental results of bearing vibration signals,the effectiveness of the proposed fault diagnosis scheme is verified.