An expert multivariable controller for distillation process control

The availability of inexpensive digital computers makes advanced control algorithms attractive for process control applications. However, many of these control techniques cannot handle abnormal operating conditions lie sensor failure, valve saturation, and process constraints. An operational control system must account for these problems. Design of such a system requires quantitative as well as qualitative information about control theory and the controlled process. Recent advances in the field of Artificial Intelligence (AI) and the area of Expert Systems (ES) facilitate the encoding of symbolic reasoning and make possible the creation of control systems where algorithmic knowledge and human experience can interact, resulting in more efficient and reliable control systems.; In the present study, a methodology to design control systems that provide safe and effective operation for multivariable processes during on-line closed loop control has been developed. This methodology has been implemented in a control system called Expert Multivariable Controller (EMC). The EMC determines and implements on-line the normal, abnormal (sensor failure and valve saturation) and constrained (process constraints) structures to be used under normal and abnormal operating conditions. This requires selecting controlled and manipulated variables, determining the level of operation, pairing controlled with manipulated variables, selecting controller types, and designing controllers.; The structure and methodology of the EMC are generic and independent of the controlled process and the control theory used. The EMC has been implemented in ART using facts, frames, and rules. Its performance has been tested using linear and nonlinear dynamic models of two product distillation columns, as well as a three product sidestream distillation column. The successful application of the EMC to the control of these processes under a wide range of normal and abnormal operating conditions using diagonal as well as multivariable controller structures (IMC) demonstrates its effectiveness.