TOWARD AUTOMATIC PROCESS SIMULATORS - PART II - AN EXPERT-SYSTEM FOR PROCESS SYNTHESIS

This paper presents an explicit mapping of the conceptual activities t hat constitute a ''process design task'' into a series of well-posed, complete, and general formal simulation procedures. Part I of this ser ies of two papers dealt with numerical procedures for process simulati on and showed that structure independence and modularity are two prere quisites for a general-purpose simulator. Part II approaches the probl em from a completely different point of view and considers the questio n: Is it possible to derive a general set of design guidelines that ca n be implemented into a knowledge-based system and result in an automa tic, computer-assisted process design procedure? This problem is diffe rent in character from that tackled in part I. First, it is by its own nature qualitative, i.e., it requires conceptual rather than numerica l answers. Second, it is formulated at a higher level (in Artificial I ntelligence terms, at a metalevel). Its solution is clearly in the dom ain of the logic of process design and, therefore, embeds (contains) a ll possible quantitative numerical schemes and does not depend on any of them for either its position or its solution. IS the answer to this question is affirmative, the resulting code would be a sort of ''Expe rt Assistant'' to the engineer in the sense that it would suggest what process can be best(1) suited for the particular application under co nsideration. The study proceeds by trying to detect conceptual similar ities in different design procedures, to construct a suitable knowledg e base, and to implement a general macro-procedure that could automati cally assist the engineer in the largest possible number of process de sign operations. The contention here is that the most recent developme nts of AI-based methods make it possible to extract from human experts all the essential knowledge that pertains to ''engineering design,'' with the final goal of transferring this body of knowledge-in a form s uitable to machine communication-to an ''Expert System for Process Des ign,'' which can then be applied (interactively or on a stand-alone ba sis) with a high degree of confidence to a variety of particular proce ss simulations. A prototype version of an Expert System Assistant is b riefly discussed, and an application is analyzed in detail. The code i s called COLOMBO and is available as a research tool from the author, Finally Part II builds on Part I of this series of papers. In particul ar it is assumed that a general, modular, numerical Process Simulation Package exists and that it is capable of executing the quantitative m ass and energy balance operations described in Part I.