Hybrid Algorithms for Redesigning MRP

In this project, the general topic of production planning is addressed whereby improvements concerning the integration of constrained resources are the main focus. Production planning can be split into long term, medium term and short term planning. On the long term planning level, decisions related to resource investments (e.g. invest in a new machine or not), workforce (e.g. how many new workers to hire), and approximate sales are taken. The main results of medium term planning are production orders that specify for each material how many pieces to produce, when to start production, and at what date the order should be finished. In short term planning, the sequence of orders is decided (e.g. the next order to be produced after finishing the current order is defined for each machine) and additional short term workforce capacity is provided if necessary. In practice, in medium term production planning a number of simplifications are usually made. Among these are the assumptions that lead times as well as demand are known with certainty and that resource limitations can be ignored. Since production resources, such as machines or workers, are usually limited and lead times as well as the actual demand are rarely known with certainty, we focus in this project on the medium term planning level and develop methods to better integrate capacity restrictions and parameter uncertainty into the respective planning models. The aim is to obtain production plans which better fit to the current shop floor situation, concerning available resource capacity and currently processed orders on the shop floor. This will lead to a better performance of the production system with respect to inventory needed and service level provided to the customers.

This project addresses production planning, with a focus on improving medium-term planning by integrating capacity constraints and addressing production system as well as demand uncertainty. Production planning typically operates on three levels: long-term, medium-term, and short-term. Long-term planning involves strategic decisions like resource investments and workforce planning. Medium-term planning produces production orders specifying what to produce, when to start, and when to complete. Short-term planning determines order sequencing and allocates additional workforce as needed. Traditional medium-term Material Requirements Planning (MRP), despite its widespread use, relies on simplifications, assuming infinite resource capacity, deterministic demand, and constant lead times. These assumptions often lead to production plans that do not reflect real-world constraints or uncertainties, such as fluctuating demand and resource availability. This project aimed to overcome these limitations, focusing on developing methods to create production plans that align with actual shop floor conditions. A key objective was to improve MRP by combining simulation-based and optimization-based approaches. A simulation model was developed to replicate stochastic production system behavior, enabling studies on uncertainty's impact on MRP performance. Findings showed that higher uncertainty in demand and production processes increases planned lead times and safety stocks. However, using fixed lot sizes partially mitigated the negative effects of high demand variability. To enhance practical applications of MRP, heuristic models were developed for implementation in Enterprise Resource Planning (ERP) systems. One example, a safety stock relaxation algorithm, demonstrated significant improvements, particularly when forecasts were updated regularly. The project also addressed production logistics by integrating an optimization model into the simulation framework. This approach produced load-dependent production order releases, significantly outperforming traditional MRP in simulations. For more complex scenarios, optimization models were designed to find near-optimal solutions for lot sizes and production orders under uncertainty. By integrating these models with the simulation framework, insights were gained into rolling horizon planning and the impact of uncertainty. While the methods performed well for small production systems, further research is required to scale them to larger, more complex systems. In summary, this project advanced medium-term production planning by addressing capacity constraints and uncertainty through simulation models, heuristics, and optimization techniques. These approaches demonstrated the potential to enhance production system efficiency, reduce inventory, and improve service levels in dynamic and uncertain environments.