Research On Fault Diagnosis And Life Prediction Of Rotating Machinery Based On Deep Learnin

With the development and application of rotating machinery,it plays a vital role in wind power,coal mining,transportation and many other fields.Rolling bearings and gears are extremely important components of rotating machinery.They operate in a complex environment for a long time and are often impacted by external loads,resulting in failures during operation,which can damage mechanical equipment and even cause personal and property safety problems.Therefore,it is of great significance to carry out fault diagnosis and remaining life prediction research on rolling bearings and gears.Many rotating machines usually operate under variable operating conditions,and changes in operating conditions will cause changes in the fault data at the feature level,which increases the difficulty of feature extraction.In addition,in the field of bearing remaining life prediction,the current research method predicts that the degradation curve does not fit well with the actual degradation curve,and the robustness of the prediction model is poor.In view of the above problems,this dissertation mainly conducts in-depth research on the following two aspects:(1)A fault diagnosis method based on Information Fusion Convolutional Neural Network(IFCNN)is proposed.IFCNN adopts multi-domain feature fusion and attention mechanism to improve the convolutional neural network.Firstly,the vibration signals collected by the acceleration sensors at different positions are converted into frequency domain and time-frequency domain information,which are input into the model at the same time;then the convolutional neural network is used to Feature extraction is performed on the frequency domain and time-frequency domain information of the fault signal,and then multi-domain feature fusion is carried out in combination with the attention mechanism,so as to achieve the purpose of selecting important features through network learning.The analysis is verified on the data of the power transmission fault diagnosis comprehensive test bench,and a large number of experimental results show that IFCNN can effectively extract the fault characteristics of the rotating machinery vibration signal.(2)An end-to-end rolling bearing remaining life prediction method using multi-level feature aggregation to improve Transformer(Multi-level Feature Aggregation Convolutional Transformer,MFACT)is proposed.First,the collected original vibration signal is converted into a frequency domain signal,and then its dimensionality is reduced and input to the encoder;in view of the feature that the self-attention module focuses on extracting global information,a convolution module is added to the encoder to extract local feature information;Then aggregate the output features of multiple coding layers to obtain multi-level degraded features,and use the Bagging algorithm to train the test model,thereby reducing the prediction variance and making the life prediction result smoother.The effectiveness of the MFACT life prediction method is confirmed in the PHM2012 bearing dataset.The research in this dissertation effectively solves the shortcomings of the current fault diagnosis research and bearing remaining life prediction research under variable working conditions.The fault diagnosis model based on IFCNN effectively improves the accuracy of gearbox fault diagnosis under variable working conditions.The average recognition accuracy rate in 12 groups of variable working conditions experiments is 98.38%,which is better than the comparison method in this dissertation.The MFACT-based rolling bearing residual life prediction method has an average root mean square error as low as 0.197 in the full-life test set experiment,and an average absolute percentage error of 50.08% in the truncated test set experiment,which is superior to the comparison method in the dissertation.In summary,both the IFCNN-based fault diagnosis method and the MFACT-based rolling bearing remaining life prediction method can play a huge role in fault prediction and health management,thereby reducing personal and property safety issues.