Applications of adaptive techniques in chemical processes.

摘要

Applications of adaptive techniques in chemical processes have largely remained at an experimental stage, despite the substantial progress in adaptive theory and its increasing use and appreciation in other fields in recent years. To improve this situation, both rigorous understanding of the theoretical underpinnings and their proper applications to real systems are essential. This thesis is intended to contribute in this direction. More specifically, it is concerned with the applications of model identification, adaptive prediction and control techniques in chemical engineering processes.; An on-line recursive least squares algorithm is employed throughout for parameter estimation. Based upon the confidence region approach, this algorithm is modified by furnishing a useful stopping rule to it. Parameter set reduction problem is addressed and a systematic hypothesis testing procedure is developed by utilizing Kolmogorov-Smirnov test. Information criteria are used for model order determination as well. Elucidation of certain mathematical results with regard to the conditions for estimation convergence is provided in terms of concepts in point estimation theory, which makes the theoretical results more accessible to practitioners.; A comprehensive review of some of the available adaptive techniques which have direct consequences in practice is given. Applications of those techniques to a binary distillation column with a continuously changing feed composition are examined. For such nonlinear and highly interactive system, conventional PID controller can hardly be expected to work well due to the ever changing operating conditions. Both SISO and MIMO adaptive approaches are therefore investigated extensively. Practical aspects in the implementation of adaptive control in the underlying system are discussed in detail. Emphases are laid on such issues as choices of process- and performance-related factors, reconciliation of estimation and control, and operator intervention. As a result, this study shows that by carefully choosing these factors with certain specified precautions, adaptive control can indeed provide satisfactory performance.