Reseaech On Data-driven Fault Diagnosis Methods For Unmanned Aerial Vehicles

With the continuous promotion of science and technology,driven by the rapid development of computer control technology,electronic information technology,and aviation technology,UAVs have developed into a highly integrated and automated systems that can perform difficult tasks in various unknown environments.Therefore,people’s requirements for the safety and reliability of UAVs are getting higher and higher,making the fault diagnosis technique for UAVs systems attract more and more scholars’ attention.Simultaneously,with the rapid development of sensor technology and storage technology,more and more historical data can be accurately measured and stored completely during the operation of the UAVs.Therefore,how to use the data-driven method to detect the fault of the UAVs system so as to ensure the safe and stable operation of the UAVs is of great research value.The main research content of this subject is the fault UAVs diagnosis method of data-driven,which mainly includes two parts: a fault detection method based on subspace and a fault location method based on residual signals.Firstly,this article introduces two commonly used description forms of linear and time-invariant system,and gives the definition of the coprime factorization technique.On this basis,the definition of the stable nuclear expression SKR of the system is introduced,which gives another description of the residual signals.Secondly,this article uses the residual signals as an indicator to measure whether the UAVs system fails.Because it is difficult to accurately describe the complex system through modeling.Therefore,this topic focuses on the research of data-driven system residual signals.By processing the collected system historical data,the system stable kernel expression SKR based on the data is obtained,and then the residual signals is obtained.Then,according to the obtained residual signals,the evaluation function is constructed through statistical methods,and the threshold is set,the two are compared to determine whether there is a failure,so as to realize the failure detection of the UAVs system based on the subspace technology.Furthermore,this article uses the collected signals of 20 residual generators to construct a logical coding table,in response to the fault location requirements of the UAVs system,After the fault is detected,the coding table is compared to realize the fault location.Due to the limitations of actual conditions,it is difficult to locate all faults through this table.In order to solve this problem,this subject proposes a BP neural network based on momentum gradient,and integrates these two methods,and finally realizes the localization of common body faults,sensor faults and actuator faults of UAVs.Finally,in this article,the application and verification of the above research content was realized by using a fixed-wing UAV.The simulation results show that the fault diagnosis algorithm proposed in this paper realizes the detection and location of common faults of UAVs,and both have good application effects.