Intelligent Fault Detection Of Rotary-wing Drones For Oil And Gas Pipeline Inspectio

Pipeline transportation,as the main method for transporting oil and gas resources,is crucial for the development of a country.In China,most oil and gas pipelines pass through sparsely populated areas such as mountains and swamps.Traditional manual inspections not only require a large amount of work but also have low efficiency.With the development of semiconductor technology,control theory and schemes,and computer programming technology,unmanned aerial vehicle(UAV)technology has gradually matured,and rotor-wing drones have been widely used to assist in oil and gas pipeline inspections.However,as the integration of rotor-wing UAV systems becomes increasingly complex in structure and function,frequent component failures,including sensors and actuators,pose a significant threat to the flight safety of rotor-wing drones.Once a rotor-wing drone crashes due to a malfunction during a task,it can affect work progress and,in severe cases,cause damage to the lives and property of ground personnel.Therefore,dynamic fault detection methods of sensors and actuators of quadrotor UAV and reconfigurable flight array used for oil and gas pipeline inspection are studied in this thesis.The main research contents of this thesis are as follows:(1)The kinematic and dynamic models of the quadrotor UAV and the reconfigurable flight array were studied,and their respective simulation models were built based on physical models.Next,common types of rotor-wing drone faults were introduced,including mathematical models and causes of occurrence.Several faults considered in this thesis were injected into the simulation models to explore the impact of flight parameters when malfunctions occur in a simulated environment.(2)To solve the sensor fault detection problem of quadrotor UAVs in complex flight modes,a fusion algorithm of FCM-(PSO-DF)was proposed.The algorithm integrates fuzzy C-means clustering algorithm(FCM),particle swarm optimization algorithm(PSO),and Deep Forest algorithm(DF)to improve the accuracy of sensor fault detection in complex flight modes.To verify the effectiveness of the proposed method,real flight data were used for experiments.The experimental results showed that the proposed algorithm had good performance in sensor fault detection for quadrotor UAVs.(3)For the sensor fault detection problem of reconfigurable flight arrays,a multi-regression method based on Adaptive Deep Forest was proposed to achieve sensor fault detection.This method obtained better prediction accuracy by adding an enhanced cascade layer structure to the standard Deep Forest and redesigning the interlayer loss function.To verify the effectiveness of the proposed method,real flight data from two types of flight arrays were used for sensor fault detection experiments.The experimental results showed that compared with the method based on standard Deep Forest,the proposed method increased the average ACC and AUC by 3%and 2.3%,respectively.(4)For the motor fault detection problem of quadrotor UAVs and reconfigurable flight arrays,a motor fault detection method based on Adaptive Deep Forest was proposed for actuator fault detection.The method integrated the Fast Fourier Transform algorithm and the Adaptive Deep Forest multi-classification model to improve the performance of actuator fault detection.The experimental results show that the accuracy of motor fault detection for quadrotor UAVs and reconfigurable flight arrays reached 95.1%and 94.7%,respectively.