Multi-objective Optimization for Supply Chain Management
| dc.contributor.author | Hemici, Meriem | |
| dc.date.accessioned | 2024-02-28T13:20:34Z | |
| dc.date.available | 2024-02-28T13:20:34Z | |
| dc.date.issued | 2024-02-22 | |
| dc.description.abstract | The research activity of this research falls under two major themes: multi objective optimization and supply chain management (SCM). Multiobjective optimization is a branch of combinatorial optimization whose specificity is to seek to optimize several objectives simultaneously for the same problem. As a result, the use of multi-objective optimization methods allows companies to use optimal strategic solutions to optimize and improve the quality of deci sions for SCM. In this thesis, we proposed three new multi-objective evolutionary algorithms to solve three important problems in SCM where our attention was focused on the part related to transportation and distribution problems, namely the vehi cle routing problem with time windows, the multi-depot green vehicle routing problem, and the ambulance relocation and dispatching problem. The first algorithm called External Archive Guided Nondominated Sorting Ge netic Algorithm II (EAG-NSGA-II) is based on improving the NSGA-II algo rithm by integrating the local search method and an external population called “archive” to store the non-dominated solutions to solve the multi-depot green vehicle routing problem. While in the second algorithm, an improved MOEA/D algorithm using simulated annealing (SA) was proposed to solve the ambu lance relocation and dispatching problem. In the last algorithm, a new evolu tionary algorithm was developed by combining the ϵ-MOEA algorithm, local search method, and multi-type crossover operators to solve the vehicle routing problem with time windows. After being tested, the proposed algorithms have shown encouraging results in both solving the vehicle routing problem with time windows, the multi-depot green vehicle routing problem, and the ambulance relocation and dispatching problem, and have been found to be more efficient compared to many multi objective evolutionary algorithms. | en_US |
| dc.identifier.issn | MD/22 | |
| dc.identifier.uri | http://10.10.1.6:4000/handle/123456789/4899 | |
| dc.language.iso | en | en_US |
| dc.publisher | Université de Bordj Bou Arreridj Faculty of Mathematics and Computer Science | en_US |
| dc.title | Multi-objective Optimization for Supply Chain Management | en_US |
| dc.type | Thesis | en_US |
Files
Original bundle
1 - 1 of 1
- Name:
- Thesis.pdf
- Size:
- 27.45 MB
- Format:
- Adobe Portable Document Format
- Description:
- n the domain of supply chain management (SCM), the primary focus is on optimiz ing the supply chain operations while considering multiple conflicting objectives. These objectives encompass various aspects such as cost reduction, lead time minimization, en hanced customer service, inventory optimization, sustainability improvement, and overall supply chain performance enhancement. However, addressing one objective often leads to compromises in other areas. For instance, reducing costs may result in longer delivery times or compromised product quality, while prioritizing customer service may increase costs. Consequently, decision-makers face the challenge of finding a balance and identify ing trade-offs among these objectives to achieve an optimal solution. In this thesis, we have introduced novel multi-objective optimization methods to address three variants of the SCM problem. More specifically, we proposed three algorithms: the G MOEA/D-SA algorithm for ambulance dispatching and relocation problems, the VRPTW MOEA algorithm for the vehicle routing problem with time windows, and the EAG-NSGA-II algorithm for the multi-depot green vehicle routing problem. Chapter 1 provided an overview of the thesis, while Chapter 2 delved into the combinatorial optimization problems, their theoretical complexities, and the class of NP-hard problems, which were fundamental to this project. Moreover, we introduced basic concepts related to multi-objective optimization problems and explored relevant approaches employing evo lutionary algorithms. Various algorithms, particularly Multi-Objective Evolutionary Algo rithms (MOEAs), widely recognized in the literature and utilized in Chapters 4 and 5, were presented. This chapter concluded with quality metrics for evaluating the performance of multi-objective evolutionary algorithms, emphasizing their advantages. Chapter 3 was dedicated to the supply chain management problem and its variants, which formed the basis of the studies conducted in this thesis. After that, we recalled the definitions and basic notions about the multi-objective opti mization problems, the supply chain management problem and its variants, and evolu tionary algorithms; which are critical to understand the scope of the present thesis. We have presented two research works developed in this thesis project: 1. In Chapter 4, we presented an improved MOEA/D algorithm for real-time ADRP, called G-MOEA/D-SA. The proposed algorithm combines the standard version of the MOEA/D with SA and the external archive. In addition, a brief explanation was given as we looked for possible improvements, which opens up several directions to explore as 102 103 prospects for future work. 2. In Chapter 5, we presented a new variant of MOEA, namely VRPTW-MOEA for the ve hicle routing problem with time windows. The proposed algorithm essentially con sists of an adapted local search of ε-MOEA, aiming to maintain a mechanism of con vergence and diversity when dealing with multi-objective optimization problems. In addition, we applied a multi-type crossover operators to accelerate the convergence significantly, and also employed the external archive based on adaptive ε−dominance to prevent the loss of satisfactory solutions once they are found. Experimental results demonstrated that VRPTW-MOEA was effective in solving VRPTW when compared with its variants, the five most well-known MOEAs, and the best-known solutions. 3. In Chapter 6, we presented a new variant of NSGA-II, namely EAG-NSGA-II for the multi-depot green vehicle routing problem (MDGVRP). We have proposed an initial ization algorithm that generates only valid conformations for the initial population of EAG-NSGA-II. This algorithm eliminates reverse moves during solution construction. EAG-NSGA-II consists of using the local search algorithm and the external archive to explore the search space more efficiently. According to our experimental results, EAG-NSGA-II can find the best known solutions and is more efficient than other ex isting algorithms in terms of stability. In terms of future applications, EAG-NSGA-II can be used to solve other optimization problems in the context of combinatorial optimization.
License bundle
1 - 1 of 1
No Thumbnail Available
- Name:
- license.txt
- Size:
- 1.71 KB
- Format:
- Item-specific license agreed to upon submission
- Description: