Abstract
The multilevel lot-sizing (MLLS) problem is a key production planning problem in the material requirement planning (MRP) system. The MLLS problem deals with determining the production lot sizes of various items appearing in the product structure over a given finite planning horizon to minimize the production cost, inventory carrying cost, backordering cost, etc. In a previous study, a particle swarm optimization (PSO) algorithm integrated with flexible inertial weight (WPSO-MLLS) was proposed for uncapacitated MLLS problems by embedding the characteristics of the MLLS problem, a coding system and redefinitions of operators for 'velocity' plus 'velocity', 'position' plus 'velocity', and 'position' subtract 'position'. This research further investigates the suitability of WPSO-MLLS by comparing the solving performance of WPSO-MLLS with a PSO algorithm without inertial weight (PSO-MLLS) and two extended hybrid versions of PSO, which are WPSO-MLLS integrated with crossover and mutation operators of a genetic algorithm (GA) (HWPSO-MLLS) and PSO-MLLS integrated with crossover and mutation operators of GA (HPSO-MLLS). Some benchmarking testing instances are adopted to compare these PSO algorithms.
