Study On Rolling Bearing Fault Diagnosis Using Improved Multiscale Sample Entropy And Parameters Optimization SVM

As a key component of rotating machinery,the performance of rolling bearing is directly related to the operation of the equipment.Therefore,the type of bearing fault need be diagnosed timely and accurately.In addition,the corresponding maintenance measures should be taken.These have important practical significance to ensure the safe operation of equipment and reduce the occurrence of major accidents.In this paper,a novel fault diagnosis model of rolling bearing is established.Firstly,the vibration acceleration signals of rolling bearing under different working conditions are collected by sensors.Secondly,the vibration signals are decomposed and reconstructed using Variational Mode Decomposition(VMD)to de-noise.Naturally,Refined Composite Multiscale Sample Entropy(RCMSE)is employed to extract fault features,and Weighted Isometric Mapping(W-Isomap),a supervised manifold learning algorithm,is introduced to reduce the dimension of original fault features.Finally,the low-dimensional features are input to the Gray Wolf Optimization Support Vector Machine(GWO-SVM)for diagnosis and recognition.In addition,engineering experiments are used to verify the effectiveness of the proposed rolling bearing fault diagnosis model.The main work of this project is as follows:(1)In view of the noise interference of the rolling bearing vibration signals,VMD is employed to de-noise of signals.Through two groups of simulation experiments and one group of rolling bearing engineering experiment,its advantages are verified.(2)Arming at the difficulty of rolling bearing fault feature extraction,RCMSE,a novel nonlinear technology,is utilized to obtain the fault features.Therefore,an original high-dimensional fault feature set is constructed.Two engineering experiments analysis results show that the feature extraction effect of RCMSE is better than MSE.(3)In consideration of the fault feature matrix extracted by VMD + RCMSE has information redundancy.If this matrix is directly input to the classifier for recognition,it will affect the recognition performance.Thus,W-Isomap is applied to reduce the dimension.Two groups of engineering experiments show that W-Isomap is batter than Isomap,LPP and t-SNE in dimensionality reduction.(4)Due to the recognition effect of Support Vector Machine(SVM)is easily affected by parameters,grey wolf optimization is applied to parameters optimization process.Then,the GWO-SVM is established.Two groups of engineering experiments show that GWO-SVM is better than existing SVM,PSO-SVM and GA-SVM in pattern recognition.(5)A new fault diagnosis technology of rolling bearing,based on VMD,RCMSE,W-Isomap and GWO-SVM,is established,also the fault model is verified by engineering experiment case analysis.