On the applicability of a nonlinear control strategy

This paper extends the work of Manchanda and Tham (1991) and discusses the results of a preliminary study into the practicability of globally linearising control (GLC). The conceptual basis for GLC lies in explicitly relating the system output to its manipulated input in terms of an ordinary differential equation. The output is successively differentiated with respect to time until the resulting derivative is a linear function of the input signal. Differential geometric concepts are then applied in developing the control law. To obtain the linearising transformation, the GLC technique requires a state-space description of the process. The controller was applied to a simulated continuous stirred tank reactor, and its performance compared with those of an heuristic, nonlinear UA controller, and a standard PID controller. Particular attention has been given to the effects of´real´ plant conditions: constraints on manipulated input; measurement noise; and plant-model mismatch.