Design methods of on-line computer control system for a fixed bed reactor with highly exothermic reaction have been investigated with two different objectives: first, regulation of both product yield and bed temperature at certain desired values and secondly, on-line optimization of profit while controlling the bed temperature. In addition, the control methods proposed have been designed to operate in fully adaptive manner in order to overcome the modelling difficulties and time-varying characteristics inherent in the fixed bed reactor.
To achieve the first objective, extended discrete-time multivariable adaptive control scheme has been developed using long-term predictor. This control scheme has been designed to solve unboundedness problem of control inputs raised when the existing multivariable adaptive control is applied to a fixed bed reactor. From simulation studies, it was revealed that the extended multivariable adaptive control scheme could show a good control performance.
In order to attain the second objective, on-line optimizing control scheme for a nonadiabatic fixed bed reactor has been presented. The control scheme has been devised to perform two functions simultaneously, adaptive control of a bed temperature and on-line steady state optimization by recursively identifying the reactor dynamics. To investigate the performance of the on-line optimizing control scheme, experiments were made with a pilot-scale fixed bed reactor where a partial oxidation of n-butane to maleic anhydride occurs. The reactor system was connected with a microcomputer and control and optimization were done by the microcomputer. With an objective function which consists of a net profit by producing maleic anhydride plus a penalty term on the bed temperature, it was clearly shown that the reaction conditions moved to a supposedly optimum point.
Finally, an improved algorithm for implementing optimizing control scheme is proposed by connecting the above two control methods ...