An Actuator Fault Diagnosis Method Combining Multi-Channel Residuals and Deep Learning

Abstract

The invention belongs to a method for diagnosing mechanical equipment faults, and provides a method for diagnosing actuator faults combined with multi-channel residuals and deep learning. Use the input and output data of different sensors in the normal state of the system to train multiple neural network observers, then generate multi-channel residuals according to the actual input and output data of the system, and perform feature extraction and fusion on the multi-channel residuals, and finally use deep learning to train and diagnose The model implements fault diagnosis. The multi-channel residual simplifies the data structure while retaining the fault feature information, and reduces the dependence of traditional model methods on expert knowledge and experience. In addition, multi-source information, multi-channel residual feature extraction and deep learning fault diagnosis can make full use of multi-sensor data feature information and the complex data processing advantages of deep learning, and can realize multi-redundant structure actuator fault diagnosis and improve the operating efficiency. Accuracy of actuator fault diagnosis.

Description

technical field [0001] The invention belongs to a method for diagnosing mechanical equipment faults, in particular to a method for diagnosing actuator faults combined with multi-channel residuals and deep learning. Background technique [0002] Electric hydrostatic actuators have many advantages such as large output force, high precision, and fast response speed. They are widely used in many fields such as aviation, ships, and transportation. They are important motion execution components to ensure the safe and reliable operation of these major equipment. . [0003] The performance reliability of the actuator directly affects the task quality of the equipment. At present, most equipment uses electrical or mechanical redundancy to improve the reliability of the actuator system. However, failures are still unavoidable. In order to realize fault identification and reduce maintenance costs, fault diagnosis technology is widely used to improve the reliability of actuators. [0...

AI Technical Summary

Problems solved by technology

Due to the complex structure of the actuator system, the analysis based on the signal processing method is difficult

The deep learning algorithm has strong feature extraction ability and nonlinear fitting ability, and can realize the processing of complex data. However, because the original data obtained from multi-source sensors contains too much interference, if it is directly used to train the model, it cannot be obtained. Better diagnostic results are not conducive to giving full play to the advantages of deep learning

[0006] To sum up, none of the current fault diagnosis methods are well applicable to the multi-source and complex system structure of actuators, and the accuracy of fault diagnosis needs to be improved.

Images