This paper aims to propose a structured method to support decision-making in complex operational contexts by improving the efficiency of multi-objective simulation optimization (MOSO). The focus is on helping managers and analysts handle large-scale decision problems with high-dimensional search spaces, often present in production and logistics systems.
The proposed method integrates Latin hypercube design (LHD) and data envelopment analysis with variable returns to scale (DEA-VRS), including super-efficiency analysis, to identify promising regions in the search space. The approach was applied to two real-world case studies in logistics and manufacturing environments.
The proposed method achieved a substantial reduction in the search space, ranging from 70% to 89%, and reduced the number of optimization experiments by up to 31%. In both case studies, the reduced search space led to improved outcomes across most optimization profiles. In the logistics case, costs decreased by up to 10%, and the quantity shipped increased by up to 219%. In the manufacturing case, lead time was reduced by up to 26% while maintaining the same production output, demonstrating enhanced computational efficiency without compromising solution quality. These results confirm that the method enhances computational efficiency without compromising solution quality in complex MOSO scenarios.
The method enabled the identification of high-quality solutions with significant operational benefits. These improvements were achieved using fewer simulation runs, up to 31% less, demonstrating the method’s ability to accelerate decision-making and reduce computational effort. Its integration with existing simulation platforms and consistent performance across diverse optimization profiles make it a valuable tool for supporting data-driven decisions in complex operational environments.
This study introduces a novel combination of LHD and DEA-VRS to enhance the performance of simulation optimization methods. It contributes to both the fields of operations research and operations management by offering a robust, interpretable and computationally efficient framework for solving complex MOSO problems in industrial applications.
