The multiinput/multioutput pole-placement self-tuning controller (MIMO PPSTC) previously developed by McDermott and Mellichamp is used to control a fixed-bed autothermal reactor with internal countercurrent heat exchange. The performance of the controller is demonstrated using step changes in the set points and in the primary disturbance variables. It is shown that the complete temperature profile can be maintained by controlling two temperatures in the catalyst bed: one just before the hot spot and the other at the exit of the bed. Simulated results using a 36th-order nonlinear reactor model operating
SCOPEControl of catalytic fixed-bed reactors is a difficult and challenging problem. These systems are typically distributed in nature and possess extreme nonlinearities. Due to complex interactions between thermal and kinetic processes, fixed-bed reactors exhibit so-called wrong way (or inverse) response characteristics, which further compounds the control problem.In recent years, several multivariable control approaches have been successfully demonstrated experimentally. Sdrensen (1977), Clement et al. (1980), and Sdrensen et al. (1980) have considered the optimal control of a nonadiabatic fixed-bed reactor. Silva et al. (1979) investigated the linear quadratic Gaussian control of a two-bed catalytic reactor. A multivariable proportional-integral control scheme was investigated by Wallman et al. (1979) on the same two-bed reactor. Jutan et al. (1977a, b, c, d) considered the control of a nonadiabatic butane hydrogenolysis reactor by means of a multivariable linear quadratic feedback controller. Wong et al. (1983) studied the control of a fixed-bed Correspondence wncerning this paper should be addressed lo R. C. Rinker autothermal reactor operating at an open-loop unstable steady state using modal control theory.In most multivariable control techniques, a mathematical model of the system is required in order to design the controller. The usual approach for reactor control is to start with a phenomenological model of the system, discretize the spatial derivatives, and then use model reduction to obtain a low-order model suitable for controller design. Before the model reduction step can be done, off-line parameter estimation must be performed to determine any unknown model parameters. This method of obtaining a model is often very timeconsuming and costly. Furthermore, the resulting model is only valid in the vicinity of the operating point at which it was developed. If reactor operating conditions change, the model and the controller are no longer applicable.Recently, major developments have been made in the area of multivariable self-tuning control. These controllers are attractive because they are easy to implement (no a priori process model is required), and they are applicable over a wide range of operating conditions.
