| The navigation system is responsible for navigation positioning and attitude determination,its accuracy and reliability play an important role in aerospace,aviation and navigation.As the core device of the navigation system,the gyroscope can provide accurate navigation and positioning parameters for the vehicle,such as attitude angle,angular velocity and angular acceleration.It has been widely used in the navigation and positioning of various carriers.In the long-term operation of gyroscope,due to different factors of the failure to produce this or that failure.Therefore,the key parameter prediction and fault diagnosis of the gyroscope is of great significance for improving the accuracy and reliability of the navigation system.This paper mainly studies the functional faults of gyroscopes,and proposes corresponding fault detection and fault prediction algorithms.The research content is mainly divided into three parts:the first part is the gyroscope data prediction,the second part is the gyroscope fault detection with redundant configuration,and the third part is the gyroscope fault prediction.In the gyroscope data prediction algorithm,the structure and algorithm of BP network,SVM network and LSTM network are analyzed firstly,and the evaluation index of prediction effect is proposed.Secondly,for the long-term and short-term memory(LSTM)network,a gyroscope data prediction algorithm is established.The three prediction methods mentioned above are compared and simulated by a gyroscope measured data,which verifies that the LSTM algorithm has high prediction accuracy and effectiveness;In the gyroscope fault detection algorithm with redundant configuration,the three-degree-of-freedom gyroscope is taken as the object.Firstly,the three gyroscope redundancy configuration forms are studied,and the combined gyroscope output signal simulation platform is built.Secondly,the built-in combined gyroscope is mounted on an unmanned aerial vehicle for flight test,and a certain gyroscope is interfered to simulate a possible fault,such as the sensor cannot supply power,the sensor deviates from the original measurement position,etc.The gyroscope normal and fault data are obtained by the upper monitor,the fault gyroscope is detected and identified according to the parity check method,and the gyroscope fault detection method based on the redundant configuration information and the LSTM network is proposed,and the fault data is replaced;In the aspect of gyroscope fault prediction,the overall block diagram of gyroscope fault prediction is given firstly.The fault simulation model of typical fault is built under MATLAB software,and the corresponding sample information is collected.The evaluation index of fault prediction effect is proposed,and the BP network,SVM network and LSTM network are used to predict the fault and compare the prediction effects of the three methods.The main innovations of this paper include the study of the three-degree-of-freedom gyroscope redundancy configuration form,and the construction of a gyroscope output data simulation platform with redundant configuration.At the same time,the fault detection algorithm based on redundant configuration information and LSTM network is proposed,and the gyroscope fault data is replaced and predicted.This research provides a practical technical tool for gyroscope redundancy fault diagnosis,abnormal change detection and fault prediction.This study was funded by the National Natural Science Foundation of China(91646108,61473222). |
