Sequence-dependent production scheduling using thep-median integer linear programming model and hierarchical clustering techniques: An empirical study

Heuristic algorithms are developed to specify production sequences based on product families. Actual data is used in cluster analysis to form product families using five measures of similarity and three measures of dissimilarity. The p-median integer linear programming model is compared to the hierarchical clustering techniques of the single linkage, average linkage, and complete linkage algorithms. Families formed by the clustering techniques become input for the heuristic sequencing algorithms.;The clustering techniques and measures of similarity/dissimilarity are evaluated using three measures of effectiveness. These measures are minimization of setup time, minimization of the change in number of workers required for consecutive products in the production sequence, and maximization of the number of common component parts between consecutive products in the production sequence. Setup time is predicted using an ordinal linear regression model. A reduction in setup time leads to increased capacity and lower production costs while reducing the change in number of workers between consecutive products leads to fewer worker scheduling difficulties. Increasing the number of common component parts between consecutive products should lead to a reduction in material handling costs.;The hierarchical clustering techniques performed better on each criterion used to evaluate the production sequences. Improvements in setup time and change in the number of workers required between consecutive products are found using product families formed using a measure of similarity based on the change in number of workers. Improvements in the number of common components between consecutive products are found using five measures of similarity based on the number of common components.