Gerald R. Sullivan

Generic model control (GMC)

Abstract This paper presents a general framework for process controllers that rely upon a model to approximate plant behaviour. By careful selection of a performance index and an approximate plant model, it is shown that single-loop PI control, feedforward and decoupling control, multivariable regulators, time horizon matrix controllers, internal model control and process model based control can all be derived. Furthermore, nonlinear process models can be imbedded directly into the controller without resorting to linearization. This unifying framework is illustrated with a number of examples to highlight the utility of such an approach.

Process model-based engineering

An integrated approach to the design of computer systems for process analysis, control and optimization called process model-based engineering is presented. The approach is based on the use of the same steady-state nonlinear model for all process functions. For process control use where dynamic models are required, the generic model control structure of Lee and Sullivan is extended to permit the use of nonlinear steady-state models in conjunction with approximate dynamics. The PMBE approach is shown to have several advantages over existing multiple-model approaches, and to have improved control performance over linear controllers with accurate dynamic models in controlling nonlinear processes.

A NEW MULTIVARIABLE DEADTIME CONTROL ALGORITHM

This paper presents a new method for multivariable dead-time compensation for a wide class of chemical engineering process control problems. The method utilizes the powerful performance-following p...

Robust stability analysis of generic model control

In this paper, the robust stability of Generic Model Control (GMC) is analyzed under the condition that the explicit control law is available. This anslysis is performed by finding a strict Lyapunov function for the nominal process and applying a perturbation theorem. Based on the passivity theorem, a procedure to synthesize a robust stable GMC controller is proposed for a given set of orocesses. The significance of this aonroach is discussed as well as its disadvantages.

AN ADAPTIVE STRATEGY FOR CONSTRAINED GENERIC MODEL CONTROL

The performance of control systems on industrial processes is often constrained—constraints on the process inputs and outputs. Effective control algorithms must be cognizant of the presence of these constraints. Generic Model Control (GMC) is a model-based control framework for both linear and nonlinear systems without explicit constraint handling. In this paper, it is shown that an adaptive approach can be incorporated within GMC to accommodate the constraints by adapting one of the two GMC parameters during the control procedure. Adaptation is determined to be necessary when the predicted process state and output variables as calculated by the process model violate their constrained values. The adaption is achieved through assessing the sensitivities of the constraints to the GMC parameters. Two non-linear examples are presented which demonstrate the efficiency of the approach.

IMPROVING CONTROL PERFORMANCE OF MULTIVARIABLE SYSTEMS BY MODIFICATION OF THE PROCESS DEADTIME STRUCTURE

A new multivariate deadtime control algorithm was designed in a previous paper. This algorithm, while very efficient, was developed under the restriction that the deadtimes on the diagonal of the transfer function matrix are less than the others in each row. In this paper, we further investigate this restrictions effect on the performance of the control system, and show that this restriction can be achieved by rearranging the order of the input variables and adding artificial delays. Several examples demonstrate the significance of the results.

Constrained generic model control of a surge tank

Abstract Control in the face of process constraints is of great practical importance in the processing industries. This paper examines the use of generic model

Heat transfer in a coke oven

Abstract Equations describing heat transfer in a combustion flue were developed and incorporated into a coke oven simulation programme, The model can be used to predict the oven temperature history, and the evolution rate and composition of the volatile matter. Unlike other coking models, the present package does not require the temperature history of the heating wall to be specified. It is calculated by the combustion flue model. This significantly extends the number of applications which can be studied. The model has been implemented in a computer program which requires only basic data on the charge properties, fuel gas composition, and oven dimensions as input. The model was tested by comparing predicted temperature profiles with data from two industrial batteries. Good agreement was obtained in both cases.

A coking model for battery-wide applications

Abstract The development of battery-wide fuel scheduling strategies has been impeded by the lack of a suitable coking model. Existing models are either very simplistic and of limited value or quite complex and thus, their application to any sort of large scale problem would require prohibitive computational effort. In light of this, a new coking model was developed. It accurately predicts coking time and final wall temperature, given the fuel history of an oven. For a wide variety of fuel histories, the coking time predicted by this new model is within 20 min of the coking time predicted by one of the complex models. In the case of wall temperature, the majority of the predictions are within 10 °C of the values predicted by the complex model. The computational effort required to solve the new model is significantly less than that necessary for the solution of the complex models, hence it is particularly useful in the development of battery-wide fuel scheduling strategies.