Decision-making in chemical engineering and expert systems: application of the analytic hierarchy process to reactor selection

Abstract

Expert systems have been used to model and automate decision making in engineering. Often the problem is cast as a decision tree which is then translated into a set of rules. However, all decision-making problems cannot be cast into this mold. Decisions are usually based on a number of criteria (and possibly subcriteria) which are accepted or established as important to the particular decision. The experience, expertise and accuracy with which the criteria are rated and combined into an overall rating determine the success of the decision. Experts often find it difficult to express their decisions in clear-cut terms. Two experts may not completely agree on the answers to a given problem. The question then is how to capture and represent this uncertainty in decision-making? Current implementations of expert systems employ probability theory, certainty factors and fuzzy set theory. These methods are often found inadequate or unsuitable for the problem leading to less than satisfactory implementation and performance. Within the social science and business fields, a method called the analytical hierarchy process (AHP) has been utilized on a number of different problems. AHP offers a mathematical methodology based on pairwise comparisons which is well-suited for a variety of different problems in both the chemical engineering domain and the expert system domain and it is easily understood by the experts. The AHP structure is applicable to the chemical engineering domain and it is easily implemented in an expert system. This paper will present the AHP methodology and illustrate its use for the engineering problem of chemical laboratory reactor selection. Also, the results of the implementation of AHP into an expert system for laboratory reactor selection will be discussed.