Research On The Intelligent Diagnosis Method Of Bearing Faults Based On Deep Learning

Machine fault diagnosis is essential for optimizing production scheduling and increasing the reliability of intelligent manufacturing.In order to achieve the requirements for efficient machine operation and low maintenance costs,intelligent online monitoring and fault diagnosis has become an integral part of industrial production.By using deep learning algorithms for data mining and analysis,information on the characteristics of bearing operating conditions can be extracted as well as identifying the type and severity of bearing faults,thus helping engineers to carry out maintenance and repair work in a timely manner.At present,in terms of rolling bearing fault diagnosis algorithms,the model extracts a relatively single feature angle and the information extracted is relatively simple,which leads to a certain deviation in the prediction results.In addition,different working conditions have a large impact on the vibration signal,and how to fuse multitask fault features into a deep learning model to achieve a richer fault diagnosis is still a problem of concern.To address these issues,we propose a multi-task deep multi-scale information fusion network model called MEAT(Multi-task-based Deep Multi-scale Feature Fusion Network)for bearing fault diagnosis.The model uses a multi-task mapping decomposition method to classify both fault scales and fault types,and extracts fault features at different scales by multi-scale convolution to obtain more comprehensive and accurate multidimensional information.The model also employs a hierarchical attention mechanism to weight and fuse the features to obtain more accurate results,and a multi-Block structure to improve the prediction accuracy.The experimental results show that the model has high accuracy and robustness and can effectively address the limitations of traditional convolutional neural networks in complex working environments,providing a new solution for rolling bearing fault diagnosis.In addition,some new ideas and methods are proposed in the paper,such as multitask mapping decomposition,multi-dimensional feature extraction and hierarchical attention mechanism,which are of great significance in targeting the problem of bearing fault diagnosis.Meanwhile,the introduction of multi-Block structure further enhances the robustness and generalization ability of the model,which makes the performance of the model more reliable and superior under different data sets and experimental conditions.The experimental related research results are compared with the results of the article in the Chinese Academy of Sciences Partition II,and the results are all better than the comparison article.Therefore,the MEAT model has high accuracy and practicality in the field of rolling bearing fault diagnosis,and can provide important reference and guidance for engineering applications in related fields.