Model-Based Actuator Fault Diagnosis For Robot Manipulators

The design of the fault diagnosis system is an important research direction of control theory and practice.The strongly coupling nonlinear characteristics of robot system make it play a vital role in the field of fault diagnosis of nonlinear system.The requirements of fault diagnosis for robot manipulators are becoming more and more urgent because of their wide range of application.This thesis presents a fault diagnosis system for robot manipulators which subject to faults of the joint actuators.By the model-based analytical redundancy method,the diagnosis system can achieve the fault detection,fault isolation and fault identification rapidly.Based on the dynamic model of manipulators,an adaptive state observer is given for the robot.The stability proof of the state estimates is also developed.Then a diagnostic observer is built to indicate the occurrence of a fault in the robot system by the residual.Once a fault is detected,a bank of isolation observers is activated to accomplish the fault isolation by a bank of isolating residuals.Using the generalized observer scheme,each isolating residual is maximally insensitive only to one type of fault,at the same time,sensitive to the effect of the other faults.By the way,the fault can also be identified.In response to the problem of simultaneous faults,a basic fault estimator is given according to the dynamic model of robot.Then a modified estimator can be obtained which doesn't need acceleration measurements.Two cases of fault isolation and fault identification have been discussed that a constant fault and a time-varying fault happen to different joint actuators simultaneously.At last,both the methods are experimentally tested on the 6-dof robot manipulator.