Nonlinear model based control of a recycle reactor process

Abstract

Abstract A control configuration, which includes two nonlinear controllers, has been applied to the Tennessee Eastman Challenge Problem (TECP), The approach used in developing this control approach was based upon a phenomenological understanding of the process and was based upon using advanced controls only where necessary. Nonlinear model based control (Generic Model Control) was applied for reactor temperature control while a nonlinear steady-state compensating controller was integrated into the composition controller for the stripper. The various kinetic parameters for the reactor temperature controller model were regressed from filtered steady-state results obtained from the TECP simulator. Further, the reactor model had on-line model parameterization applied to it in order to keep the controller model true to the TECP process simulator. The reactor pressure control problem proved to be the most challenging control problem encountered and was solved by manipulating the process feed rate to control reactor pressure. The resulting variations in feed flowrate to the stripper necessitated the use of the aforementioned compensating controller for the stripper. In addition, over-ride controls were found to be necessary for two disturbances that caused saturation in the control valve for one of the feeds. The controller was tested for all setpoint changes and the disturbances that were built into the simulator for the base case product mix. Excellent control results were obtained for all control objectives while calling for a smooth use of the various feed flowrates. In fact, the variations in the compositions of the primary products could not be seen over the measurement noise from the product analyzer for all but two cases.