Reactive schedule modification in multipurpose batch chemical plants

Abstract

A new scheme is described for reactive schedule modification in the face of unexpected deviations in processing times and unit availabilities of a multipurpose batch plant (MBP). Schedule modification is done using at least impact heuristic beam search which proceeds in two levels: creation of a decision tree which makes use of possible reroutings of the product that is causing a conflict, and heuristic pruning of the search space to contain the combinatorial complexity. The heuristic chooses a path among all possible reroutings for a product such that the impact of each decision on the rest of the schedule is kept as small as possible. This approach has been implemented and tested on a number of simulated deviations in a MBP case study with three products. The proposed least impact heuristic was found to perform better than the earliest finishing unit heuristic in all the cases considered.