Nonlinear Process Model Based Control

Multi-Model Decoupled Generic Model Control

Authors: James B. Riggs and Peter L. Lee


The control performance of GMC (Generic Model Control) using steady-state models parameter systems can be significantly undermined by the lack of appropriate dynamic information. A decoupling strategy has been integrated into the GMC framework in order to compensate for dynamic mismatch. The approach (Multi-Model Decoupler, MMD) is based upon using a separate model inverse for the calculation of each manipulated variable. The additional model inputs provide the extra degrees of freedom that allow for dynamic compensation. Because the same model inverse is used for the calculation of each manipulated variable, the model inverse must be evaluated more often than conventional GMC would require. MMD-GMC has the advantage over conventional decouplers since the MMD decoupling elements have unity gain and thus take full advantage of the nonlinear prediction capability of the model controller model. An alternative approach using more traditional External Decouplers (ED-GMC) is developed and compared to MMD-GMC.

Both approaches are demonstrated on a simulation of a high relative volatility binary column. MMD-GMC is shown to offer significant control performance advantages over both ED-GMC and GMC based on steady-state models and no dynamic compensation.


Publication Information: Submitted to Journal of Process Control, June 1995

Corresponding Author: James Riggs


Tuning GMC Controllers Using ATV

Authors: Scott E. Flathouse and James B. Riggs


A procedure is developed for applying ATV tuning to GMC controllers. A nonlinear two input/two output test problem, which resembles an exothermic CSTR, is used to test the effectiveness of the new tuning procedure.

The effect of model mismatch on the tuning for the GMC controller model is studied using dynamic and steady-state controller models. In both cases, a GMC controller tuned with the ATV tuning procedure is compared with a GMC controller with optimal tuning. In the case of GMC based upon dynamic models, the ATV tuning procedures with on-line detuning is shown to provide an excellent approach to the optimal performance over a significant range of model mismatch. While the ATV-tuned GMC controllers which use steady-state models yield control performances that are clearly inferior to the ones based upon dynamic models, ATV tuning provides an adequate approach to the results of the optimally tuned controllers.

Publication Information: Submitted to Computers and Chemical Engineering, Dec. 1994

Corresponding Author: James Riggs


Nonlinearity Measures for a Class of SISO Nonlinear Systems

Authors: Debin Sub† and Karlene A. Hoo*
Alcatel, Inc. Toronto, Canada
* Department of Chemical Engineering, Texas Tech University


A new measure of nonlinearity, that is captured by two linear systems, is proposed to quantify the size of the nonlinearity in the input/output behavior of SISO nonlinear systems. As such, the controller design for the nonlinear system is simplified to the design of a controller for these two linear systems. It is also shown that when Hammerstein and Wiener models are involved, the controller design for these two structures are equivalent if the linear dynamic term is stable. Two examples, a numerical example and a nonlinear system characterized by a time-delay and a variable gain are used to demonstrate these concepts.


Publication Information: Int. Journal of Control, Vol. 73, pp 29-37, 2000.

Corresponding Author: Karlene A. Hoo