Development and comparison of multiple genetic algorithms and heuristics for assembly production planning

Abstract

Assembly jobs with tree-structured precedence constraints in their bill-of-materials structure are a generalized version of traditional jobs involving only line-structured precedence constraints. The assembly job shop scheduling problem deals with assembly jobs, in contrast to job shop scheduling which deals with only traditional jobs. This research explores the ability of different genetic algorithms (GAs) to solve the assembly job shop scheduling problem. The objective is to minimize the makespan (maximum completion time) of a given set of assembly jobs. Random key GAs are proposed which differ using three factors: decoding, schedule justification and individual rearrangement. The three factors have two, seven and two levels, respectively, resulting in 28 different GAs. Specifically, we have conducted a full factorial design of GAs using forward/backward decoding, 0–6 local steps of justification, with/without individual rearrangement. The aim is to test the performance of GAs using different factor settings. As benchmarks, two heuristics have been proposed. Lingo, a software tool for linear and non-linear optimization problems is also used for solution by setting the time limit to 30 min. The experiments have revealed significant effects of the aforesaid three factors on the performance of the GAs.