Comparative Metaheuristic Performance for the Scheduling of Multipurpose Batch Plants

Two recent publications by Woolway et al. (2018, 2019) [1], [2] proposed a novel metaheuristic framework to optimise the scheduling of Multipurpose Batch Plants. This initial framework implemented three metaheuristic methods to solve the problem with a Genetic Algorithm (GA) showing superior performance over the others. Two notable opportunities for improvement in the current solution are improving the spread/confidence intervals of the percentiles of the solutions discovered by repeated executions of the GAs and faster convergence. This work considers two adaptations of the GA to an attempt to improve overall spread and speed on the application to two well-known literature examples. We have replicated the work in the original papers in a completely new Julia framework along with our extensions. Results show that our modifications to the GAs can, in fact, lead to tighter spread as well as faster convergence,