Tactical planning deals with the elaboration of production plans in order to meet the customer's request, with the lowest costs. Usually, tactical planning is divided into three main plans : the Sales and Operation Planning (S &OP), the Master Production Schedule (MPS) and the the Material Requirement Planning (MRP). In order to elaborate such plans, mathematical models, called "lot-sizing problems", have been developped. However, the complexity of the tactical planning problematic has considerably grown thanks to the mechanisms of fusion and acquision between enterprises by integrating multi-site aspects, according to Supply Chain concept and, in our mind, there is no knowledge and mathematical models which can be seen as references for this problematic. In this thesis, we propose a generic knowledge model for the tactical planning and a generic mathematical model based on the latter which allows to find, by applying it, the main mathematical models of literature. We also propose efficient optimization methods in order to elaborate production plans (S&OP, MPS and MRP) in mono-site and multi-site contexts : - In mono-site context, we proposed to use metaheuristics in order to solve the Capacitated Lot Sizing Problem (CLSP) which is the model dedicated to the S&OP and MPS elaboration problematics. Lower Bound are also provided. By using these methods, we improve some best solutions known in the literature. - We propose a mathematical model, derived from our generic mathematical model, for the tactical planning of a supply chain whose network looks like an Hybrid Flowshop, and an efficient optimization method in order to elaborate S& OP and MPS for such Supply Chain. Next, we deal with operationnal feasibility of tactical plans determined by lot-sizing models by using different sorts of combination beetwen mathematical an simulation models., in order to insure the vertical synchronization of tactical production plan. Finally, in an industrial context, we deal with the setting up of a rationing stock policy. We first analyse the feasibility of such setting up in an industrial context and then, we have conceived algorithms and software which allow to compute rationing thesholds in order to test this policy in a real industrial case.
