TOWARD AUTOMATIC PROCESS SIMULATORS - PART I - MODULAR NUMERICAL PROCEDURES

This paper presents a general design approach involving automatic, int elligent process simulation procedures. The aim is to derive a general set of design principles and methodologies that can be developed into computer-assisted procedures. This first part deals with numerical, q uantitative calculations i.e., with what commonly goes under the name of ''Numerical Process Simulation.'' It is argued that the existing de sign methods can result in computer-codes or packages that perform exa ctly (and deterministically) the numerical operations an engineer woul d perform. It is also shown that modularity in these codes is dictated by the necessity of automatically implementing numerical procedures t hat depend on the structure of the process under examination rather th an by user's convenience and ease of maintenance. An example of a modu lar, structure-oriented code (CAMEL) is given and discussed in detail, while numerical applications are discussed elsewhere [4]. The second part deals with the more complex qualitative approach to process desig n, i.e., with the possibility of implementing automatic ''expert'' pro cedures that perform the same conceptual tasks as human process engine ers. It is shown that by means of Artificial Intelligence techniques i t is possible to mimic (to an extent) the ''thinking patterns'' of a h uman expert, and to produce process schemes that are both acceptable a nd realistic. A general process synthesis package (COLOMBO) is describ ed and some of its applications discussed. The main goal of the two pa rts of the paper is to show that the very complex activity of process design can be executed automatically, not only in principle, but in ac tual applications, and that both qualitative synthesis and quantitativ e calculations are possible with the present state of the art of our c omputational facilities.