Research On Fault Tolerant Control Of Multi-DOF Manipulator Under Single Joint Failure

At present,the application of the manipulator is extensive.When it breaks down in aerospace,high-risk environment,nuclear environment and others,it is difficult for human intervention and maintenance.It is urgent for the manipulator to carry out tasks with faults.Fault-tolerant control problems came into being.Due to the suddenness of the fault and the uncertainty of the location of the fault,the traditional fault-tolerant methods are difficult to apply to the strongly coupled manipulator systems.Aiming at the above problems,in this paper,the faulttolerant control of single joint fault of multi-dof manipulator is studied.Firstly,The research background and significance have been presented.The advantages,disadvantages and the research status of passive fault tolerance,active fault tolerance and intelligent fault tolerance control methods have been summarized.Besides,the research objectives and contents have been established and the corresponding research key points have been refined.In addition,this paper have analyzed common failures of manipulators.Taking a typical 6-dof tandem manipulator as an example,the spatial problem has been converted into a planar problem by the motion decomposition based on the physical structure.Considering the value of practical application and the unity of algorithm,the sensor lock fault has been transformed into actuator lock fault.Secondly,the mathematical models of the manipulator have been established.The forward kinematics and differential kinematics of the manipulator have been established by the postposition D-H parameter method.The dynamic model has been established by Lagrangian method.They provide the model foundation and theoretical support for the thesis.In order to judge whether the manipulator can be fault-tolerant,Monte Carlo method has been used to analyze the workspace and degraded workspace of the manipulator.In order to measure and improve the motion ability of the mechanical arm after failure,according to the existing three kinematics performance indexes,the integrated joint maneuverability has been designed.Thirdly,according to the different effects of actuator and sensor on the characteristic parameters of the manipulator,the velocity residual observer has been designed to detect and diagnose joint faults.Based on reinforcement learning theory,a fault-tolerant control algorithm of deep reinforcement learning has been designed,and a comprehensive joint maneuverability has been added to the algorithm to improve the motion ability of the manipulator after the convergence of deep network.Finally,the effect of the residual observer has been simulated,and the fault-tolerant control method is simulated and verified by real object.The results of repeatability experiment and multidimensional analysis show that the velocity residual observer and fault-tolerant control method designed in this paper can effectively complete the fault diagnosis and fault-tolerant control of single-joint fault manipulator.