Research On Deep Learning-based Fault Diagnosis Method For Rudder Engines

Servos are widely used in aerospace,naval,and other fields as a type of servo drive,playing an important role in pose changes and steering control.Intelligent detection of servo faults is of practical significance to ensure the normal running of equipment.Currently,the automation of servo testing equipment makes measurement parameters more convenient and reliable,but manual analysis of large amounts of data not only consumes time but also cannot guarantee accuracy.This paper introduces deep learning to test and analyze servo parameters,constructs a servo fault diagnosis model based on deep learning,uses this model to extract and classify servo testing data features,and promotes the development of servo fault diagnosis technology towards automation and intelligence.Firstly,in response to the problem of imbalanced samples and poor classification results for small samples in the steering gear data,a feature extraction model for steering gear data based on convolutional neural networks(2DFMCNN)was constructed.The one-dimensional steering gear test data was upgraded to two-dimensional data,which provided spatial information for the data and increased the data feature extraction ability of the convolutional neural network.Three layout methods for upgrading the data to two-dimensional,including direct layout,reverse layout,and random layout,were proposed and validated.The reverse layout had the best performance.Moreover,a local feature learning(FM)module was constructed to enhance the network’s learning of internal data features.Finally,the 2DFMCNN feature extraction model was established for learning and analysis of steering gear data,and experimental results showed that the samples could still be correctly classified even in the case of small samples,and the model outperformed other comparison models,indicating that the model improved the feature extraction ability for the data.Secondly,in response to the problems of the large number of output layer parameters and limited classification ability of the classifier in traditional convolutional neural networks,it was proposed to use SVM instead of the Softmax classifier.The 2DFMCNN feature extraction model was used to perform deep feature extraction on the data,which compensated for the insufficient deep feature extraction of SVM.The Sea Gull Algorithm was used to adaptively optimize the parameters c and g in SVM,avoiding interference of human selection of SVM parameters.Finally,a fault diagnosis model based on deep learning and Sea Gull Algorithm optimized SVM was established.Comparative experiments were conducted from the perspectives of the model’s classification ability and the optimization ability of the algorithm.The experimental results showed that the model had excellent performance,solved the problem of insufficient classification ability of the final classifier in convolutional neural networks and the problem of SVM parameter selection,and improved the accuracy and efficiency of steering gear fault diagnosis.This dissertation constructs a servo fault diagnosis model based on deep learning theory and technology,alleviates the workload of manual data analysis,efficiently completes the diagnosis of servo faults,and improves the efficiency and accuracy of servo fault diagnosis.