A BRANCH&BOUND ALGORITHM FOR SOLVING ONE-DIMENSIONAL CUTTING STOCK PROBLEMS EXACTLY

Many numerical computations reported in the literature show only a small difference between the optimal value of the one-dimensional cutting stock problem (1CSP) and that of the corresponding linear programming relaxation. Moreover, theoretical investigations have proven that this difference is smaller than 2 for a wide range of subproblems of the general 1CSP. In this paper we give a branch&bound algorithm to compute optimal solutions for instances of the 1CSP. Numerical results are presented of about 900 randomly generated instances with up to 100 small pieces and all of them are optimally solved.