| Rotor system is a key component of rotating machinery,the occurrence of sudden failures can cause economic losses and even casualties.Monitoring and diagnosing the rotor system can timely eliminate faults,avoid major accidents,and ensure human life safety.As one of the important components of the rotor system,rolling bearings can rapidly deteriorate and affect the overall mechanical system once they fail.However,the complex and volatile operating environment of modern mechanical equipment and weak early fault signals of bearings lead to difficulty in extracting fault features and effectively diagnosing bearings.Therefore,conducting research on fault diagnosis of rotor systems is of great significance for preventing the occurrence of faults.This dissertation takes the bearings in the rotor system as the research object.Aiming at the problems of difficult extraction of bearing fault features and low accuracy of fault diagnosis due to factors such as weak early signal of rolling bearings and distortion of traditional DBN high-level information,a rotor system fault diagnosis model based on ISSAVMD-Feature Fusion-EDBN is constructed to achieve early bearing fault diagnosis.The main research contents are as follows:(1)Aiming at the problem of weak early signal of rolling bearings that makes it difficult to extract fault features,a fault feature extraction model based on ISSA-VMD-Feature Fusion was constructed.The model proposed an ISSA-VMD method to decompose signals by studying variational mode decomposition(VMD)and sparrow search algorithm(SSA),and experiments have verified that the method has better stability and decomposition ability.Then,mutual information methods were used to select sensitive IMF components and calculate their energy characteristics and sample entropy characteristics to form a high-dimensional multi feature set.Finally,t-distributed Stochastic Neighbor Embedding(t-SNE)was introduced to perform feature fusion on high-dimensional multi feature sets.(2)Aiming at the problem of low accuracy of fault diagnosis caused by the distortion of high-level data samples in traditional deep confidence network(DBN)networks,a fault diagnosis recognition and classification model based on extended deep confidence network(EDBN)was constructed.Experiments have verified that the model has higher stability and fault diagnosis accuracy than traditional DBNs.(3)The rotor system fault diagnosis model based on ISSA-VMD-Feature Fusion-EDBN has been established.Using rolling bearing vibration signals as experimental objects,parameters of methods such as VMD and EDBN in the model were set to achieve accurate diagnosis of early bearing faults.Comparison of fault diagnosis results with different decomposition methods,single feature indicators,and different classification methods has verified the superiority of the fault diagnosis model proposed in this article(4)Using MATLAB App Designer development tools to develop a rotor system fault diagnosis system applied to practical processes,making the entire fault diagnosis simpler and more convenient. |
