DEVELOPMENT OF FAULT DIAGNOSIS AND FAULT-TOLERANT CONTROL METHODS FOR ROBOT MANIPULATORS

Abstract

This dissertation arm to theoretically develop fault diagnosis and fault-tolerant control methods for robot manipulator systems to keep the robot operating with good tracking performance in the presence of uncertainties and faults. The fault diagnosis in industrial processes is a challenging task that includes fault detection, isolation, and estimation problem. Early fault detection and compensation in the next stage, which is called fault-tolerant control, help robot manipulator systems are still operating in a controllable region, can help to avoid events progression and reduce the number of productivity losses during abnormal events.

In this dissertation, the fault diagnosis methods are mainly developed based on high-order sliding mode observers. Thanks to their ability not only to estimate the lumped uncertainties and faults but also to approximate the system velocities, the requirement of tachometers in robot manipulator systems is eliminated. In addition, the developed fault diagnosis method provides estimation information with fast convergence speed, high precision, low chattering phenomenon, and finite-time convergence of estimation errors.

Along with the fault diagnosis methods, fault-tolerant control schemes are developed. The proposed controllers are designed via an active fault-tolerant control method by combining the developed fault diagnosis schemes with novel non-singular fast terminal sliding mode controllers to accommodate not only system failures but also uncertainties. This combination provides robust features in dealing with the lumped uncertainties and faults, increases the control performance, reduces the chattering phenomenon, eliminates velocity measurement requirement, guarantees finite-time convergence, and provides faster reaching sliding motion. Especially, both two periods of time, before and after the convergence process takes place are carefully considered.

The stability and the finite-time convergence of the proposed controller-observer techniques are demonstrated using the Lyapunov theory. Finally, to verify the effectiveness of the proposed controller-observer methods, computer simulations on robotic manipulator systems are performed.