Particle Swarm Optimization Algorithm for a Batching Problem in the Process Industry

An improved particle swarm optimization (PSO) algorithm is proposed to solve a typical batching problem in a batch processing plant of the process industry. The batching problem (BP) is to transform the primary requirements for products into sets of batches for each task with the objective of minimizing the total workload. On the basis of some preliminary properties, a novel particle solution representation is designed for the BP. Unlike the ordinary idea of taking an objective function as the fitness function for PSO, the original objective function incorporated with a constraint function is to act as the fitness function of the PSO where the constraint and the objective functions are evaluated successively. Such a fitness function, together with a forward repair mechanism, makes it possible for a faster convergence. Further, for each iterative generation, a local search heuristic is used to improve the global best particle found so far. To verify the performance of the proposed PSO algorithm, the well-known benchmark batching instances are tested. The relatively large-scale instances are also added to evaluate the algorithm. The computational results show that the improved PSO may find optimal or suboptimal solutions within a much shorter run time for all the instances.