Text mining approach for Bottleneck detection and analysis in printed circuit board manufacturing

This paper proposes a production scheduling procedure for the production lines of a printed circuit board company. Linearity features of the manufacturing process of this company are characterized and exploited for developing an efficient three-phase scheduling procedure. The production line is formulated as a linear job shop that resembles a flow shop. First, the N-gram modelling approach is adopted to analyse the data sets to detect the machines that would be candidates of the bottleneck in the production lines. Second, according to the candidates, the bottleneck data are extracted from the original data sets and solved as flow shops by a mixed integer programming model. The optimal solutions of the bottleneck flow shop is next extended by incorporating upstream and down stream operations to form approximate solutions of the original problems. We propose three different strategies for forming the approximate solutions and the best one is designated as the final solution. The performance of the proposed heuristic algorithm is tested and compared with the well-known NEH algorithm through numerical instances from real production lines. Statistics indicate that for most instances including up to 80 jobs the proposed method delivers competitive solutions within a much shorter time than the NEH algorithm.