Research On Fault Detection And Fault-tolerant Control Of Robot Manipulators Based On Sliding Mode Observer

With the expansion of the application scope and the gradual complexity of the application environment,a higher requirement is proposed for operating safety and reliability of the robot manipulators system.The robot manipulators system is a class of complex nonlinear systems,actuator as an important part of the robot manipulators control system,whose fault is main cause of the control system failure.At present,the analytical model based fault detection method is widely used.However,due to the uncertainty of model,external disturbance and parametric perturbation of the actual system,it is difficult to improve the speed and accuracy of fault detection and reconstruction currently.That the fault information is used to design fault-tolerant controller to improve the security and reliability furtherly of the system is of great significance.On the basis of the research on the fault detection and fault-tolerant of the robot manipulators at home and abroad,the problem of low speed and precision of fault detection and chattering of the control system is studied in this thesis.Based on quasi-continuous high-order sliding mode observer(QC-HOSMO)principle,a method of fault detection and reconstruction for robot manipulators actuator is proposed,and a fault-tolerant controller is designed according to reconstructed fault signal.Firstly,the dynamic characteristics of a class of n-joint rigid manipulator system are analyzed,and the dynamic equation of the manipulator system under normal condition is established.The causes of the actuator failures and the main types of faults are analyzed,and a fault modeling is built by establishing the parameters such as effective factor and external disturbance,then the dynamic equation of the manipulator system under fault state is established.Secondly,the quasi-continuous high-order sliding mode observer(QC-HOSMO)is established to solve the issues such as low accuracy,large chattering and asymptotic convergence that existed in conventional sliding mode observer.The continue control feature out of sliding surface of quasi-continuous high-order sliding mode is used to eliminate chattering and ensure the system converge to the sliding surface in finite time.Radial basis function(RBF)neural network is used to approximate the model uncertainties and unknown disturbances.Then a nonsingular terminal sliding surface is selected so that the state estimation error and the output estimation error are stable bounded in finite time.The equivalent output error injection method is used to complete the accurate reconstruction of the actuator actual faults.Thirdly,the active fault-tolerant control strategy is proposed for the robot manipulators system with actuator failure,and the fault model is compensated by the online reconstructionfault signal.A nonsingular fast terminal sliding mode manifold is defined to ensure the convergence in finite time,and to avoid the singularity.Then active fault-tolerant controller is designed by definiting the trajectory tracking error.Finally,several kinds of fault types are simulated and verified.By comparing the simulation results of high-order sliding mode observer and QC-HOSMO,it is verified that the quasi-continuous sliding mode mothed can eliminate chattering and improve the accuracy and rapidity in fault reconstruction.And the improvement of fault-tolerant performance is verified by simulation results of active fault-tolerant control.