Performance Investigation of Artificial Bee Colony (ABC) Algorithm for Flexible Job-shop Scheduling Problem (FJSP)

Abstract

Flexible Job-Shop Scheduling Problem (FJSP) allows a job to be processed on any machine of a set of alternative machines. Therefore, it acts as an extension of the classical Job-Shop Scheduling Problem (JSP). The complexity is portrayed by dealing with both routing and scheduling sub-problems.  Major weakness of classical Artificial Bee Colony (ABC) algorithm is shown by trapping in local optimum easily during dealing with complicated multimodal problems such as FJSP despite the proof of ABC’s capability to solve FJSP [1], [2].  Therefore, this study aims to study the nature of ABC algorithm. At the same time, the best parameter setting is investigated by using factorial experiment. Brandimarte benchmark (MK01 – MK10) was adopted in the study for the best makespan. The results indicate that a limit value of 5 provides the lowest average makespan for 70% of the benchmark instances. Additionally, selecting MCN = 1000 reduces computational time by approximately 50% while maintaining comparable solution quality.