A hybrid fuzzy-neural-based dynamic scheduling method for part feeding of mixed-model assembly lines

Abstract Since the mixed-model assembly lines are highly adopted in the automobile industry, the part feeding process has become a critical challenge. Therefore, in this paper the dynamic part feeding scheduling problem under a Kanban system is studied, which aims to optimize a productivity-related objective (the throughput of the assembly lines) and a cost-related objective (the total delivery distance of the automatic guide vehicles (AGVs)) simultaneously. A comprehensive part feeding process considering the part feeding tasks generation, the loading, sequencing and dispatching problems is analyzed and modeled. To properly solve the problem, this study proposes a hybrid fuzzy-neural-based dynamic scheduling method, integrating the Self-organizing maps (SOM) with the Fuzzy c-means (FCM) algorithm and the knowledge base (KB). The SOM is adopted to pre-cluster the system status and optimize the initial clustering centers, then the FCM is availed to guide the clusters, so as to improve the clustering performance simultaneously. Computational experiments are conducted to evaluate its scheduling performance in a dynamic manufacturing environment and verify its superiority over the benchmark algorithms. The method allows decision makers to select more rational scheduling schemes based on their decision impacts in both productivity and part feeding costs and the real-time status of the assembly lines.