Domain‐concept association rules mining for large‐scale and complex cellular manufacturing tasks

Abstract

PurposeThe purpose of this paper is to provide a novel domain‐concept association rules (DCAR) mining algorithm that offers solutions to complex cell formation problems, which consist of a non‐binary machine‐component (MC) matrix and production factors for fast and accurate decision support.Design/methodology/approachThe DCAR algorithm first identifies the domain‐concept from the demand history and then performs association rule mining to find associations among machines. After that, the algorithm forms machine‐cells with a series of inclusion and exclusion processes to minimize inter‐cell material movement and intra‐cell void element costs as well as to maximize the grouping efficacy with the constraints of bill of material (BOM) and the maximum number of machines allowed for each cell.FindingsThe DCAR algorithm delivers either comparable or better results than the existing approaches using known binary datasets. The paper demonstrates that the DCAR can obtain satisfying machine‐cells with production costs when extra parameters are needed.Research limitations/implicationsThe DCAR algorithm adapts the idea of the sequential forward floating selection (SFFS) to iteratively evaluate and arrange machine‐cells until the result is stabilized. The SFFS is an improvement over a greedy version of the algorithm, but can only ensure sub‐optimal solutions. Practical implications – The DCAR algorithm considers a wide range of production parameters, which make the algorithm suitable to the real‐world manufacturing system settings.Originality/valueThe proposed DCAR algorithm is unlike other array‐based algorithms. It can group non‐binary MC matrix with considerations of real‐world factors including product demand, BOM, costs, and maximum number of machines allowed for each cell.