One of the main issues in the glass industry is the minimization of the trim loss generated when cutting large parts (stocks) into small items. In our application stocks are produced in the plant. Many distinct stock sizes are feasible, and technical constraints limit the variety of cutting patterns to those producing a single type of item per stock. Consequently, the focus is not on seeking an optimal subset of cutting patterns, but rather on choosing an optimal subset of a limited number of stock sizes. In this paper we discuss a 0-1 linear programming formulation for this problem based on a p-median model. Tested on data from the field, the formulation shows an impressive reduction of the trim loss produced in the present plant operation and definitely outperforms traditional exact approaches in terms of computation time.
Technical Report N. TRCS 022/2004, Dipartimento di Informatica,Università degli Studi di L'Aquila