Optimization of Long-Term Production Planning for the Shipbuilding Industry

Abstract

Production planning is important for efficient use of production resources in the shipbuilding industry where lead time of products is long and multiple products are simultaneously produced. Among various types of production plans, long-term production planning determines the type and number of ships to be produced in several years and decides the highest-level production schedule called key events for ships. Despite its significant impact on the profitability of shipyards, it is manually determined. This takes long time to establish and makes difficult to optimize the plan. Long-term production planning is divided into three problems and solved sequentially due to the hierarchical and complex ship production process: product mix plan, dock plan, and capacity plan. The first determines the type and number of ships to be produced and the second decides key event schedules for ships. The third verifies feasibility of the previous plan in terms of various production resources and supports to determine capacity of production resources. In this dissertation, dock planning problem and block assignment problem which belongs to capacity planning are studied. In a dock planning problem, ships are assigned to docks to be processed and key event schedules of ships are determined. Dock planning problem considers not only scheduling constraints but also spatial constraints in docks since multiple ships can be simultaneously constructed in one dock to increase the turnover rate of dock. A mixed integer linear programming model and an efficient large neighborhood search heuristic are developed for the problem. In block assignment problem, blocks are assigned to assembly shops and some blocks must be assigned together within a region, which is a collection of assembly shops. Each assembly shop has its target work level. The objective of the problem is minimizing the sum of the workload violation of the assembly shops during the given planning period. Block assignment planning problems was formulated as a mixed integer linear programming model but it could not be solved in a reasonable time. Therefore, a two-stage matheuristic algorithm was proposed as a solution method. The proposed methods in this dissertation were tested using real-world ship order data and block data provided by the shipyard and the experimental results demonstrated their efficiency.