Date of Award

Spring 2005

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Computational Analysis and Modeling

First Advisor

R. Selmic

Abstract

Common actuator nonlinearities such as saturation, deadzone, backlash, and hysteresis are unavoidable in practical industrial control systems, such as computer numerical control (CNC) machines, xy-positioning tables, robot manipulators, overhead crane mechanisms, and more. When the actuator nonlinearities exist in control systems, they may exhibit relatively large steady-state tracking error or even oscillations, cause the closed-loop system instability, and degrade the overall system performance. Proportional-derivative (PD) controller has observed limit cycles if the actuator nonlinearity is not compensated well. The problems are particularly exacerbated when the required accuracy is high, as in micropositioning devices. Due to the non-analytic nature of the actuator nonlinear dynamics and the fact that the exact actuator nonlinear functions, namely operation uncertainty, are unknown, the saturation compensation research is a challenging and important topic with both theoretical and practical significance.

Adaptive control can accommodate the system modeling, parametric, and environmental structural uncertainties. With the universal approximating property and learning capability of neural network (NN), it is appealing to develop adaptive NN-based saturation compensation scheme without explicit knowledge of actuator saturation nonlinearity. In this dissertation, intelligent anti-windup saturation compensation schemes in several scenarios of nonlinear systems are investigated. The nonlinear systems studied within this dissertation include the general nonlinear system in Brunovsky canonical form, a second order multi-input multi-output (MIMO) nonlinear system such as a robot manipulator, and an underactuated system-flexible robot system. The abovementioned methods assume the full states information is measurable and completely known.

During the NN-based control law development, the imposed actuator saturation is assumed to be unknown and treated as the system input disturbance. The schemes that lead to stability, command following and disturbance rejection is rigorously proved, and verified using the nonlinear system models. On-line NN weights tuning law, the overall closed-loop performance, and the boundedness of the NN weights are rigorously derived and guaranteed based on Lyapunov approach. The NN saturation compensator is inserted into a feedforward path. The simulation conducted indicates that the proposed schemes can effectively compensate for the saturation nonlinearity in the presence of system uncertainty.

Download

Share

COinS