A matheuristic algorithm for block assignment problem in long-term production planning in the shipbuilding industry

Abstract

This study introduced the block assignment problem in the long-term production planning of a very large real-world shipbuilder and formulated it as a mixed integer linear programming model. For the problem, blocks are assigned to assembly shops for processing; some blocks must be assigned as a group to a region, a collection of assembly shops, because of their mutual dependency. Each assembly shop has its target work level. The objective is to minimize the sum of workload deviations of the assembly shops during a specific planning period. Because the mathematical model could not be solved within a reasonable time, a two-stage matheuristic algorithm was proposed. The algorithm determines the region of blocks in the first stage and the assembly shop for each block in the second stage. Computational experiments were performed using a real-world instance provided by the shipbuilder, including instances of different sizes generated from real-world data. The experiment results demonstrated that the algorithm yielded remarkable reductions of 10.5% and 35.3% in assignment costs and factory workload violations compared with manual planning, respectively. The proposed decomposition-based algorithm also demonstrated superior effectiveness in solving the problem than the full mathematical model. The proposed algorithm is currently embedded in the planning system of the shipbuilder.