A Plant Design Heuristic Considering the Eventual Measurement of Currently Unknown Variables

Abstract

It is common practice for chemical plants to be sized using estimated parameter values that are uncertain at the design stage, but whose true values will be known once the plant is in operation. Moreover, not all design decisions are fixed once the plant is built, as some may be adjusted during operation. In this paper, we present a heuristic method for plant design under uncertainty that takes these characteristics into account. The problem is framed as selecting the best from a set of candidate designs, where each candidate design results from optimizing the plant for a set of possible values of the uncertain variables. Decision trees are used to select the best-performing alternative given the probability distribution of the uncertainties. A working example is presented that relates to the design of a heat-integrated reactor with uncertainty in the plant inlet composition. Candidate designs and optimal operation for different compositions are found by using the Solver add-in of MS Excel. It is concluded that decision trees allow post-construction operational adjustments and parameter uncertainties to be easily and clearly incorporated into the design process.