Fault Diagnosis And Fault-tolerant Control For Aircraft Attitude Control System

When a aircraft is performing a mission,if a component fails,it will inevitably affect the flight mission and even cause a flight accident.In order to improve the safety and reliability of the aircraft,the article carries out the following work.Firstly,the aircraft attitude kinematics equations and attitude dynamics equations based on quaternion are established.Then the common faults in aircraft actuators are classified,and the fault model is given.Secondly,fault detection is performed on the startup of the thrust device.The traditional fuzzy C-means clustering algorithm is sensitive to the initial clustering center,and it is easy to fall into the local optimum.Therefore,combine the FCM with the firework algorithm in this article.The explosion mechanism and mutation mechanism in the firework algorithm can ensure the diversity of the population.The initial clustering center of the traditional FCM is obtained through the firework algorithm.Then the membership matrix and the clustering center matrix are updated iteratively.After obtaining the minimum value of the objective function,the best classification result is output.In the fault detection,use the improved FCM to obtain a group of cluster centers under the condition of no fault of the actuator.Then compare the group of cluster centers with the data to be checked to determine whether a failure occurs at this time.After verification by simulation,it is proved that the improved FCM has better detection effect in fault detection.Then,the fault diagnosis of the aircraft actuator is performed by building a neural network.One-dimensional convolutional neural networks can extract certain features in aircraft attitude data through convolution and pooling operations.The LSTM network can better preserve the long-term memory in the timing signal by introducing a "gate" mechanism,and effectively solve problems such as gradient disappearance.Combine the two networks to build a CNN+LSTM diagnostic network.The data set of the training network is obtained by simulating the failure of the aircraft.Diagnose the bias fault of the actuator.The diagnosis results show that the network of CNN+LSTM has a better diagnosis effect than the network of only CNN.Then,the fault-tolerant control method is applied to aircraft with actuator faults.The attitude control process of the aircraft is a finite time process,and the use of a finite time convergence control law has certain advantages.This paper adopts the nonsingular terminal sliding mode theory.Based on the CNN+LSTM diagnosis network,an active fault-tolerant controller is designed.In the ordinary non-singular terminal sliding mode controller,the real-time diagnosis result of the diagnosis network is introduced.The simulation proves that the fault-tolerant controller has better faulttolerant performance.Finally,the full text is summarized,then the further study of the work are pointed out.