Rolling horizon production planning with forecast evolution based stochastic optimization
Sprache des Vortragstitels:
Englisch
Original Tagungtitel:
International Conference on Operations Research (OR) 2023
Sprache des Tagungstitel:
Englisch
Original Kurzfassung:
Production planning for multi-level manufacturing systems is a challenging problem in industry. In practice, optimization approaches that tackle this problem are integrated into planning frameworks in a rolling horizon manner, to consider parameter fluctuations and operative disruptions. Uncertainty is usually present in customer demand, as well as in system related parameters, such as setup time or resource capacity. Despite the reactive nature of rolling horizon planning, it is necessary to explicitly incorporate this uncertainty into the planning method. This is especially true for stochastic demands, because in multi-level systems, release of production lots for components needs to be anticipated several periods before the actual due date.
In our work we integrate two-stage scenario based stochastic optimization into an event-based simulation environment, resulting in a rolling horizon simulation-optimization framework. We iteratively solve stochastic multi-item multi-echelon capacitated lot sizing problems considering current system state and demand forecast information. Customer demands are frequently updated, following the forecast evolution approach. We evaluate the stochastic approach and compare its performance to deterministic optimization, as well as to classical MRP. We perform an extensive simulation study testing several lead times, resource utilizations, safety stocks, demand updates and variation factors. We report important key performance indicators, such as the service level and overall production costs. Our approach outperforms both, deterministic optimization and classical MRP, on several of these indicators, which shows that it can help to improve production planning in industry. This research is supported by Austrian Science Fund (FWF): P 32954.
Forschungs-