This paper presents a novel method for the design of “optimal” (or quasi-optimal) HEN. The method consists of an Expert System (“ES”) based on a small number of powerful and strongly selective heuristic rules. The important contribution of this study does not lie in the formulation of the rules, that have been adapted from the existing literature, but in their expression as logical propositions, and in their subsequent implementation in a prototype ES that performs interactively with the user.
It is not unusual to find chemical processes with as many as 100 interacting streams, and even simple thermal processes, excluding refineries and chemical plants, contain at least a 10-streams-HEN: hence the high demand for an “automatic” (in some sense) Design Procedure that may conveniently be adapted to design-and-optimisation problems. Pinch Technology (“PT”), at present the almost universally adopted design procedure, is very successful in most types of applications (except in cases where mechanical and thermal power must be optimised concurrently), but it constitutes an operative tool, and does not improve its user’s comprehension of the problem: it assumes, rather, that the user is already familiar with the design of HEN. The approach we present in this paper is entirely different: we do not “mask” the thermodynamic and thermo-economic principles that guide the engineer in the path towards the “optimal” HEN configuration, and do not allow concerns about “user friendliness” to impair the necessary participation of the user to the HEN synthesis procedure. In fact, though our ES (which we prefer to call “Expert Assistant”, to underline its peculiarity of constantly interacting with the user) is still lacking many of the capabilities that a good designer possesses, the underlying procedure is, unlike any of the other existing Design-and-Optimisation Procedures, entirely inspectable by the user for what its decision-making rules are concerned. It can be interrogated about its decision making, so that the logical path followed from the design data to the final solution can be inspected at will, and it can be used to directly compare different alternatives in a logically systematic fashion. The paper begins with a brief review of the HEN design problem, followed by a critical discussion of the heuristic rules that form the basis for the Inference Engine of the Expert System. The formalisation of these rules into logical propositions suitable for Knowledge Based Methods is then presented, and the resulting macrocode developed. As a preliminary validation, two examples of application of the code (named Heat Exchanger Network Expert Assistant, HENEA for short) are presented and discussed: since both cases have published, and their “optimal” solutions are known, the performance of HENEA can be assessed by comparison.