Implementation of on-line control in chemical process plants

A range of identification, estimation and control algorithms has been implemented and tested on a chemical process plant/process control computer system which is typical of installations in the process industries. The plants studied are a gas-separating unit consisting of a pair of 9 m high absorption/distillation columns and a two-stage fractional crystallization plant. All on-line estimation and control was performed by a Honeywell 516 computer system. The topics studied on the plants included continuous on-line estimation of states and chemical process parameters using Kalman filtering techniques, the use of these estimates in various control algorithms, the application of optimal control theory to a variety of problems (minimum variance control, adaptive control, time optimal control), and the identification of process models for subsequent use in the design and implementation of multivariable control algorithms. In all cases, the implementation of these aspects of modern control theory on a real process plant was successful, but pointed to several non-trivial complications which must be resolved before these algorithms can be adopted for general use in the process industries.