Research On Intelligent Diagnosis System Of UAV Flight Control Fault Based On Machine Learning

In recent years,drone technology has developed rapidly,and the safety and reliability of drones is the most important link in its development.The structure of the drone control system is complex,and the drone does not have the on-site operation of the driver during the mission.Conditions,so the duration will fail,affecting task execution.Under this background,the monitoring technology of the UAV's health status and the efficient and accurate diagnosis technology of the fault are put forward higher requirements.This thesis takes a certain type of UAV flight control as the research object,and conducts indepth research on several key issues in the field of intelligent fault diagnosis: fault mode,fault propagation,diagnostic expert knowledge persistence,fault intelligent diagnosis method,main work The content includes the following four aspects:(1)From the aspects of failure mode,fault propagation,fault diagnosis expert knowledge persistence storage,intelligent diagnosis method,etc.,the research status at home and abroad is investigated,compared and analyzed,and the research objectives and technical routes are formulated.Related fields include the research and analysis of the theoretical content of UAV flight control failure modes and their characteristics,fault propagation rules,fault information anomaly detection technology,and neural network foundation.(2)Traditional fault diagnosis relies on expert knowledge and experience.For this problem,it is necessary to use information technology to study more simple and intelligent diagnostic techniques.How to convert expert knowledge into machine-readable data and perform persistent storage is the primary research goal.Therefore,the third chapter of this paper studies the method of establishing the knowledge map of UAV flight control fault diagnosis.The different expert knowledge extraction schemes are compared and analyzed.A diagnostic knowledge map based on reverse fault tree is proposed.Modeling faults of no less than ten failure modes.(3)Research on the intelligent diagnosis method of UAV flight control fault based on neural network,construct convolutional neural network,replace the traditional convolution kernel with the convolution kernel added with cavity,increase the receptive field,and greatly enhance the deep layer.Feature extraction ability,and establish a double-layer void convolutional layer.By adjusting the expansion coefficient to obtain the optimal value of the network,an intelligent diagnosis method based on cavity convolutional neural network is proposed,which is based on the transformed expert diagnosis knowledge.The network is trained,and then the neural network is optimized from the aspects of network structure,activation function,elimination of over-fitting,etc.Finally,the network is simulated and verified.(4)A detailed analysis of the system requirements of this paper is carried out.Combined with the research on the establishment of the knowledge map of the UAV flight control fault diagnosis and the intelligent diagnosis method,the detailed design scheme of the system software is given.Including the overall architecture,technical route,each module design,and then the software development and implementation,and describes the various subsystem interfaces and functions of the software,and finally use the UAV flight control fault data to test the software.