An informative column generation and decomposition method for a production planning and facility location problem

This paper develops an informative column generation and decomposition method for a capacitated multi-item lot sizing and facility location problem with backlogging. The method hybridizes column generation to achieve a relaxed linear solution and a lower bound and a decomposition method (i.e., relax-and-fix) to achieve a feasible solution. The two solutions are used for a neighborhood search procedure that fixes a subset of setup decision variables to iteratively reduce problem sizes. The relax-and-fix method is applied again to solve these smaller-size restricted problems with a purpose of progressively improving solution qualities. To show the effectiveness of the method, a number of computational tests are performed using newly-generated benchmark problems. Computational comparisons with a commercial solver (CPLEX) show that the proposed method provides competitive solution results.