Research On Fault Diagnosis Of Rolling Bearing Based On VMD And Deep Belief Network

The health of rolling bearings has a crucial impact on machinery and even the entire mechanical system.In order to ensure the normal and safe operation of the mechanical system and reduce the personal injury and economic loss caused by mechanical failure,it is necessary to perform condition monitoring and fault diagnosis on the rolling bearings.This paper takes rolling bearings as the research object,takes rolling bearing vibration signals as the starting point,and uses modern signal processing methods and deep learning theory as the basic idea to conduct in-depth research on the existing problems in the field of rolling bearing fault diagnosis.The main research contents are as follows:(1)In view of the difficulty in extracting fault characteristics of vibration signals,Variational Mode Decomposition(VMD)was studied in depth,and the good performance of VMD in resisting modal aliasing and endpoint effects in non-stationary and nonlinear signal decomposition was verified by simulated signal.A fault diagnosis method based on VMD and t-SNE(t-distribution random neighborhood embedding)is proposed.VMD is used to decompose the original signal to obtain several intrinsic mode functions,the time-frequency characteristics of different components are calculated to form a high-dimensional feature vector,and the classification results are obtained by the K-means classifier after t-SNE dimensionality reduction.Experimental verification was performed using rolling bearing vibration signals,the results show that the method can accurately determine the bearing fault category,and unlabeled fault diagnosis is implemented in an unsupervised learning manner,which solves the problem of difficulty to obtain labeled data to a certain extent.(2)Aiming at the problem that the fault diagnosis accuracy of the shallow classification model is not high,the Deep Believe Network(DBN),as a typical deep learning model,is studied in depth.This paper proposes a fault diagnosis method based on VMD timefrequency characteristics and DBN.The original signal is decomposed by VMD to obtain a number of intrinsic mode functions.The time-frequency characteristics of different components are calculated to form high-dimensional feature vectors.DBN classification model is trained with normalized feature vectors as input,and use the trained DBN model to identify faults.The advantages of the DBN model in fault diagnosis are verified through experiments,and the influence of hyperparameters on the performance of the DBN model is discussed.(3)Aiming at the problem that feature type selection and DBN hyperparameter selection generally depend on expert experience,a fault diagnosis method based on VMDCS-DBN is proposed.This method eliminates the step of manual feature extraction,directly uses the reconstructed signal as the input of the DBN model,which is decomposed by VMD.At the same time,the key hyperparameters of DBN is determined by Cuckoo Search(CS)algorithm.The method is applied to the fault diagnosis of rolling bearings,and the results show that the method can adaptively determine the key structural parameters of the model,and can accurately identify the different positions and different degrees of damage of the rolling bearings under different working conditions.