A heuristic approach to synchronize production and transportation planning in a mineral water industry

—Planning problems arise in different circumstances with a set of parameters, decision variables and constraints, particulary capacity constraints. Synchronizing production plan and transportation plan at item level in medium-short term, however, has received little attention in the prior literature. This paper aims at solving a production planning problem that depends not only on internal process and forecast data, but also on subsequent transportation planning. For a multi-product industrial case over a finite planning horizon, an adapted model is established and an improvement-based heuristic is proposed to find a near-optimal solution in a reasonable amout of computation time, with the data from a French mineral water company. The results are compared against those of Cplex to solve an MIP model and those of a combined heuristic-Cplex procedure. The comparison demonstrates that the proposed heuristic algorithm works efficiently. I. INTRODUCTION Planning problems are frequently encountered in supply chains and influence directly or indirectly facilities' performance. To appropriately make use of limited resources to meet industrial goals, manufacturers have to make decisions on allocating resources and schedule their activities over a planning horizon. It is known that two primary goals in both transportation and production planning are satisfying customer demands and resulting in the lowest cost. This paper focuses on simultaneous optimization of production and transportation plans for a mineral water company, with a cost-based combinatorial optimization model. The first studies on production planning started with constant demand, continuous time and infinite horizon. With no capacity constraints, the classical economic order quantity model ([1], [2]) is useful for a single-level production process. On the other hand, models with dynamic demand , discrete time and finite horizon, generally referred to as dynamic lot sizing, seek a minimum cost plan that meets all constraints over a finite horizon. They can often be formulated into mixed integer programs (MIP). The Single Item Lot Sizing Problem (SILSP) is the simplest one among dynamic lot sizing problems. After Wagner and Whitin [3] described a dynamic programming (DP) recursion for the uncapacitated SILSP, various versions of production planning problems have been investigated. Further algorithms were proposed ([4], [5]) to reduce the computational complexity in comparison with the initial