Proposing a Simulation-based Optimization Model for Determining Optimal Parameters in a Demand-Driven Material Requirements Planning Approach

The demand-driven material requirements planning approach (DDMRP) considers inventory buffers at certain points in the bill of materials to respond more quickly to customer demands and demand fluctuations. In this approach, the presence of a buffer at each point of the bill of material does not necessarily improve the material flow and may lead to an increase in inventory costs. Moreover, this approach includes parameters that are experimentally set by the manager, and inappropriate values for these parameters can result in deficient performance. Therefore, this paper aims to simultaneously determine optimal values for the parameters of this approach at both the strategic level (strategic inventory positioning) and the operational level (planning phase) to minimize inventory cost and achieve a 100% service level within customer tolerance time.

In this research, a simulation-optimization model is introduced to determine the optimal values of three fundamental parameters: strategic inventory position, variability, and lead time factors. To address this, a combination of genetic algorithm and mixed-integer linear programming with a CPLEX solver is utilized. In the genetic algorithm phase, the allowed positions of the buffer are randomly selected in a way that the delivery of the final product is less than the customer tolerance time. In the mixed-integer linear programming phase, the optimal values of variability and lead time factors are determined with the aim of minimizing inventory cost and avoiding stockout issues.

The proposed model is evaluated across 12 randomly generated instances of the bill of materials, each varying in levels and the number of parts. This set includes the specific case studied in the article by Jiang and Rim. The performance of our presented model is then compared with that of Jiang and Rim's model. The results consistently reveal that the proposed model demonstrates superior efficiency across all instances. The comparison of results with the data from the main article proves that the inventory cost of the proposed model has decreased between 82% and 86%, with an average reduction of 83.6%. The improvement percentage of on-time deliveries ranges from 0% to 4%, with an average improvement of 2.2%. Comparison with randomly generated data indicates that the average inventory cost using the proposed model is reduced between 73% and 91%, with an average reduction of 81.8%. Moreover, with 100% confidence, all orders are fulfilled within a time frame shorter than the customer tolerance time.

In the presented model, despite more restrictions regarding the buffer position in the bill of material, the performance results of the proposed model demonstrate a significant reduction in average inventory costs compared to the Jiang and Rim model. Additionally, with full confidence, all orders will be fulfilled. The methodology employed in this research can function as a decision-support tool for managers. It aids in determining the optimal quantity and timing of manufacturing or purchasing orders, minimizing inventory costs while aiming to achieve a 100% service level within the customer tolerance time