An intelligent parallel control system structure for plants with multiple operating regimes Arthur Kordon a, *, Prasad S. Dhurjati a , Yuris O. Fuentes b , Babatunde A. Ogunnaike b a Department of Chemical Engineering, University of Delaware, Newark, DE 19716, USA b E. I. du Pont de Nemours and Co., PO Box 80101, Wilmington, DE 19880-0101, USA Abstract A parallel control system, consisting of a bank of concurrently running virtual control loops and a concurrency coordinator, is proposed for plants that operate in multiple, often radically dierent regimes. The virtual control loops acting in parallel and exchanging information among themselves, are coordinated to obtain good closed-loop performance in each operating regime, and bumpless transfer during transitions. The basic strategy, overall structure, and the individual elements of the parallel control system are discussed; the design, analysis, and performance are illustrated via simulation of a chemical reactor designed to manufacture a product in three dierent regimes. # 1999 Elsevier Science Ltd. All rights reserved. Keywords: Parallel control system; Plants; Multiple operating regimes 1. Introduction Many processes in the chemical industry use the same processing equipment for manufacturing dierent product types. Because such processes must, therefore, operate in multiple, often radically dierent, regimes, designing eective control systems for them can be par- ticularly challenging. On the one hand, a single, linear controller will often not perform well in all the operat- ing regimes; on the other hand, a global nonlinear con- troller that can perform well in all the operating regimes is currently extremely dicult to design and implement in practice. A popular option in industrial practice is to design separate individual controllers for each operating regime, based on appropriate local linear models. These local controllers are, however, only activated when the process is in a prede®ned (usually small) neighborhood of the appropriate operating regime; the process oper- ates open-loop during transitions between operating regimes, leading to signi®cant transition losses due to unacceptable product quality. The concept of parallel control using local linear controllers has also been used in the development of fuzzy control systems [1, 2]. More eective transition control techniques have been pro- posed recently (for example: [3±5]. In Banerjee et al. [3] the nonlinear plant is represented by N simple models, corresponding to the N distinct operating regimes. An uncertainty region is associated with each model and the individual model predictions are weighted by speci®c probabilities estimated using Bayesian techniques. It is assumed that each local controller has been designed to perform satisfactorily within its associated operating regime but not necessarily during transition. The pro- blem of smooth transfer between dierent local con- trollers is solved by introducing a scheduling strategy which uses probabilities as scheduling variables. The eectiveness of this strategy depends on the number of models used and the quality of probability estimates. Schott and Bequette [5] use multiple-model adaptive control (MMAC) for control of chemical reactors with input multiplicities and non-minimum phase behavior. In this technique, a weighting function evaluates which model or combination of models best represents the plant behavior. Each model has a corresponding con- troller with tuning parameters based on the parameters of the model. The control implemented on the plant is a weighted-sum of the outputs from controllers in the controller bank. The approach developed in Kosano- vich et al. [4] is based on the switching strategy pro- posed by Morse [6], where a supervisor selects the best feedback controller from a set of linear controllers. A Journal of Process Control 9 (1999) 453±460 0959-1524/99/$ - see front matter # 1999 Elsevier Science Ltd. All rights reserved. PII: S0959-1524(99)00006-2 * Corresponding author at current address: Dow Chemical Com- pany, Freeport, TX 77541, USA.