Particle Swarm Neural Network-based Fault Tolerant Control For Nonlinear Systems

Adaptive dynamic programming(ADP)combines the ideas of optimal control,dynamic programming(DP),neural networks and reinforcement learning.It can overcome the “curse of dimensionality”in traditional DP.ADP is a kind of intelligent control methods with strong learning and optimization capabilities,which have great potential in the optimal control and tracking control of nonlinear systems.In order to meet the requirements of increasing the production efficiency,the scale of modern industrial control systems has become larger and larger,and the control process has become more and more complex.In such systems,the faults will occur inevitably in actuators and sensors.If they are not dealt with on time,they may reduce the control performance of systems,and even threaten people's lives.Therefore,designing a fault tolerant control system is particularly important,and the improvement of the reliability and stability of the control system has become a research hotspot.To tackle fault tolerant control problems of continuous-time nonlinear systems,the main work of this paper are summarized as follows:1.To deal with the tracking control problem of continuous-time nonlinear systems with actuator failures,a fault tolerant tracking control method based on ADP is proposed.Firstly,for the fault-free system,an augmented system is constructed by combining the tracking error dynamics and the desired trajectory dynamics.Then,by constructing a performance index function with a discount factor,the tracking control problem is transformed into optimal control problem.Secondly,the particle swarm optimization(PSO)algorithm is introduced to train the critic neural network to avoid the difficulty in selecting initial weight vectors,and a PSO based online policy iteration algorithm is proposed.Thirdly,a neural network based observer is designed to estimate the state of the augmented system and the actuator failure,and the online compensation is employed to compensate the actuator fault.Finally,two simulation examples with comparative results are provided to illustrate the effectiveness of the proposed fault tolerant tracking control method.2.To solve the optimal control problem for continuous-time nonlinear systems with both actuator and sensor faults,an optimal fault tolerant control method based on ADP is proposed.Firstly,the optimal control law is derived by ADP framework,and the PSO algorithm is employed to train the critic neural network,and then the approximate optimal control law is obtained.Secondly,by contructing a filter and an augmented system,the sensor fault is transformed into pseudo-actuator fault,then,the actuator faults are estimated and compensated online,and the outputs of the transformed system are replaced by the corresponding observer outputs to deal with sensor failures.Finally,two numerical simulations are provided to verify the effectiveness of the proposed optimal fault tolerant control method.Finally,the conclusion of this work and the future research are provided based on the experience in the research process.