An open problem in manufacturing control is the design of controllers with guaranteed closed loop performance on plants subject to unmeasurable disturbances and changing dynamics. Existing methods for Adaptive Control of these plants typically rely on system identification techniques, which lead to a multi-objective optimisation task that offers no explicit guarantee regarding the stability of the closed loop system or the quality of its output. Robust Control design guarantees stability in the presence of unmeasured disturbances, although at the expense of sub-optimal control performance. We propose a control architecture and methodology for online adaptation of the process behavior that does not require explicit identification of the plant or disturbance dynamics. An arbitrary function approximator is used for control of the plant and the closed loop behaviour of this coupled system is adapted to that of a generic performance target. No exogenous signal is required for excitation of the adaptation, which instead used the inherent variability in the plant output. The use of a closed loop performance target for adaptation is practical, simple to implement and does not require a knowledge of either the plant dynamics or the disturbances acting on the process, since the controller is only concerned with maintaining transient process behavior within acceptable bounds. The resulting controller is capable of achieving a specified closed loop performance in the presence of unmeasurable, bounded disturbances, making the proposed scheme a viable candidate for disturbance rejection tasks.

This content is only available via PDF.
You do not currently have access to this content.