Plantwide control involves the systems and strategies required to control an entire chemical plant consisting of many interconnected unit operations. A general heuristic design procedure is presented that generates an effective plantwide control structure for an entire complex process flowsheet and not simply individual units. The nine steps of the proposed procedure center around the fundamental principles of plantwide control: energy management; production rate; product quality; operational, environmental and safety constraints; liquid-level and gas-pressure inventories; makeup of reactants; component balances; and economic or process optimization. Application of the procedure is illustrated with three industrial examples: the vinyl acetate monomer process, the Eastman plantwide-control process, and the HDA process.pelling improved capital productivity. Requirements for onaim product quality control grow increasingly tighter. More energy integration occurs. Improved product yields provide additional plant capacity by lower reactant per-pass conversion and higher material recycle rates. These are all economi-

A simple, practical approach to the problem of finding reasonable controller settings for the N single-input-single-output controllers in an /th-order typical industrial multivariable process is presented. The procedure is a straight-forward extension of the familiar Nyquist method and requires only nominal computing power. The method has been tested on ten multivariable distillation column examples from the literature, varying from 2X2 systems up to 4 X 4 systems. The settings determined by the method gave reasonable dynamic responses that were comparable to the empirical settings reported by the original authors.

No abstract

No abstract

IntroductionNonlinear behavior is the rule, rather than the exception, in the dynamic behavior of physical systems. Most physical devices have nonlinear characteristics outside a limited linear range. In most chemical processes, understanding the nonlinear characteristics is important for designing controllers that regulate the process. Chemical processes contain valves that saturate, connecting lines whose time delays vary with flow rate, reacting mixtures which obey a power law, and separation units which are very sensitive to input changes and disturbances. Besides those nonlinearities, there are phenomena such as incomplete mixing whose effects are not well understood. The combination of such effects can sometimes cause unpredictable behavior of a process. Even though most controllers used in process control are of the PID type, the understanding of process nonlinearities can be utilized to select manipulated variables, to locate sensors, and to use special effects such as antireset windup schemes, gain scheduling and override control.One approach to understand the nonlinear behavior is to form a mathematical model of the process. To achieve this, a mathematical model of each unit operation has to be formed by making some simplifying assumptions, and then these models are combined to obtain a model representing the complete system. This approach has been utilized in modeling many processes and is a very useful one. However, there are some limitations to the mathematical modeling route:Correspondence concerning this article should be addressed to S. H. Johnson The assumptions made in deriving the model can be too restrictive or not very realistic. Therefore, the model might not capture the essentials of the behavior of the process.The parameters in the models such as overall heat transfer coefficients and reaction constants may not be accurate.The models derived can be very large: they can be made up of many differential equations. This limits the use of such models to derive analytical insights.Another approach commonly used is to build the model directly from the observed behavior of the process itself. This kind of empirical model building is called system identification. It might be used along with mathematical modeling in some cases (i.e., estimating the parameters, such as heat transfer coefficients, from empirical data). However, the bulk of the work done in the field of system identification starts with representing the process as a black box. We may have access to the inputs and outputs, but the internal mechanisms are assumed to be totally unknown to us. The problem in system identification is to construct a model which would mimic the 'inner mechanisms' of the system, using the input/output data. The usual procedure is to select a model structure with some unknown parameters and then to estimate the model parameters. The last step is to check whether the model obtained is adequate. Some iterations on this procedure might be necessary to arrive at a model which is good enough for one's purposes.Th...