|
Mathematical model and genetic algorithm for the integrated problem of production network design and inventory positioning |
|---|---|
| รหัสดีโอไอ | |
| Title | Mathematical model and genetic algorithm for the integrated problem of production network design and inventory positioning |
| Creator | Nguyen Trong Tri Duc |
| Contributor | Jirachai Buddhakulsomsiri, Advisor |
| Publisher | Thammasat University |
| Publication Year | 2566 |
| Keyword | Genetic algorithm, Guaranteed-service approach, Mixed-integer linear programming, Inventory positioning, Integrated approach, Network design, Rank-based decoding, Sourcing strategies |
| Abstract | This research presents integrated approaches to address the production network design and inventory positioning problem within multi-product and multi-period contexts. Our objective is to concurrently determine the network structure, safety stock amounts, and their respective locations while accounting for normal demand distribution. The study aims to minimize the overall production network cost. This approach combines traditional network design formulation with inventory positioning concepts utilizing the guaranteed service approach. We apply the proposed model to a numerical study involving office furniture manufacturing, considering the bill of materials. Our approach's effectiveness is demonstrated through comparison with a sequential strategy, where network structure and safety stock decisions are made sequentially. Results show the integrated approach outperforms the sequential one, with cost savings ranging from 1.7% to 3.7%. Furthermore, the integrated approach yields higher optimal cycle service levels. To assess model performance, a sensitivity analysis on critical parameters is conducted, revealing insights into the influence of committed service time and demand variation coefficient on solutions. Notably, longer committed service times result in reduced safety stock costs and higher optimal service levels. This leads to allocating more safety stock to upstream components compared to finished goods. Additionally, an innovative genetic algorithm is proposed for dealing with the integrated problem to tackle the problem's nonlinearity and complexity with different sourcing strategies, where the concept of rank-based decoding within the genetic algorithm is introduced. This decoding technique transforms the network design component into a simplified minimum-cost flow problem. By leveraging mixed-integer linear programming models, we solve the minimum-cost flow and inventory positioning problems. We validate our genetic algorithm solution by comparing it with a commercial solver's solutions, which are optimal for medium-sized instances, or the best found for large ones. Results demonstrate the genetic algorithm's capability to attain optimal solutions for medium-sized problems and outperform the best-found solutions for large-sized problems. |
