| Summary: | Manufacturing organizations are keen to improve their competitive position in the global marketplace by increasing operational performance. Production planning is crucial to this end and represents one of the most challenging tasks managers are facing today. Among a large number of alternatives, production planning processes help decision-making by tradingoff conflicting objectives in the presence of technological, marketing and financial constraints. Two important classes of such problems are lotsizing and scheduling. Proofs from complexity theory supported by computational experiments clearly show the hardness of solving lotsizing and scheduling problems. Motivated by a real-world case, the glass container industry production planning and scheduling problem is studied in depth. Due to its inherent complexity and to the frequent interdependencies between decisions that are made at and affect different organizational echelons, the system is decomposed into a two-level hierarchically organized planning structure: long-term and short-term levels. This dissertation explores extensions of lotsizing and scheduling problems that appear in both levels. We address these variants in two research directions. On one hand, we develop and implement different approaches to obtain good quality solutions, as metaheuristics (namely variable neighborhood search) and Lagrangian-based heuristics, as well as other special-purpose heuristics. On the other hand, we try to combine new stronger models and valid inequalities based on the polyhedral structure of these problems to tighten linear relaxations and speed up the solution process.
|
|---|