The Fault Diagnosis Of The Rolling Bearing Based On Local Mean Decomposition And Support Vector Machine

With the advancement of human society,the scientific and technological content of modern mechanical devices is getting higher and higher,and the perfect pursuit of the efficiency and safety of mechanical devices is increasing.As a common component in mechanical devices,the quality of rolling bearings are closely related to the overall operational safety of the equipment.Therefore,the diagnosis and reasonable maintenance of the bearing can not only maximize its working efficiency and performance,but also adapt to the energy saving and environmental protection requirements.The foothold of this paper is based on acoustic signals to diagnose the bearing.Firstly,some faults and principles of frequent occurrence of bearings and the calculation of frequency calculation under different fault locations are described.At the same time,the mechanism and propagation characteristics of the vibration and noise of the bearing are expounded,and the relationship between the two is summarized,which lays a foundation for the next step to build the experimental platform and measurement test.Secondly,according to the characteristics of the bearing acoustic signal,the existing test platform is combined with the existing test equipment to improve,and the single vibration signal detection is changed to the vibration signal detection.Based on the characteristics of the existing experimental platform,the reconstruction and experiment are carried out according to the measurement characteristics of the acoustic signal,and the vibration signal and the acoustic signal are measured,and the comparison test is performed to ensure the accuracy and rationality of the experiment.In the signal processing method,we selected the waveform analysis method,and related the principle and basic steps of the Local Mean Decomposition method.Then,we combined with the characteristics of the acoustic signal with the kurtosis criterion,the Maximum Correlated Kurtosis Deconvolution method is introduced,and the fault bearing signal is combined with the method.Feature extraction,and the envelope spectrum analysis of the extracted feature information to find the fault frequency.At the same time,the permutation entropy algorithm is used to quantize the decomposed components,and an improved method for extracting fault feature information of rolling bearings is proposed.This method is validated by a variety of different bearing signals to prove the practicality of this method.Finally,the support vector machine algorithm is described.By using the principle of manufacturing the optimal classification plane and the advantages of good classification under small sample size,a support vector machine is designed to verify its classification reliability.According to the characteristics of the data,the kernel function is selected and its parameters are adjusted,and the support vector machine will achieve the best classification effect.In the data processing,the pros and cons of the bearing fault classification result is based on the clear input vector.Only the support vector machine thus trained will have a good diagnostic effect.