The optimization of the lean supply chain management using meta-heuristic approach

Abstract

Since the early 1990s, Lean Manufacturing (LM) has received worldwide attention from both scholars and practitioners due to the tremendous success of Toyota. Witnessing the fruitage of LM’s founder, enterprises have been attempting to implement LM within the factory and then the supply chain (SC) under the form of lean supply chain (LSC). Over time, lean supply chain management (LSCM) is now being considered as the ideal model for companies to gain competitive advantages and also hedge against threats. Inspired by this, the thesis here entitled “The optimization of the lean supply chain management using meta-heuristic approach,” studies the opportunities of improving the performance of LSCM through solving its problems in an optimal manner. The thesis is developed based on the findings from four articles conducted in this field. The thesis begins with a systematic review of the most relevant areas from LSCM in order to build up the necessary background, thereby orienting the research direction. Ensuing these bases, a novel quantitative framework of optimizing the design of pure LSC is introduced. This work applies LM as a dual filter to eliminate waste on both SC function and SC structure. The problem is illustrated through a numerical example and solved by priority Genetic Algorithm meta-heuristic (APPENDIX I, p.161). In the development of LSCM, the lean model was integrated with an agile paradigm through a decoupling point to form the leagile supply chain (LA SC). This hybrid SC was widely evaluated as the most advanced and intelligent model in modern management. Following this progression, the thesis coins the concept of leagile bill of material (LA BOM) to add agility into the above new-designed LSC. In this LA BOM, LM tools are employed to simplify the structure of a product family and to amplify the combination of components. The joint design of the product family through LA BOM and its LA SC is conducted and optimized simultaneously. The joint design also takes into account the placement of decoupling points to define the best configuration of the chain and its product allocation. The framework is illustrated by a case study in the furniture industry and solved by Genetic Algorithm MH. The framework is validated by comparing it with the exact solver LINGO. In an era of globalization, SC facilities may scatter in different regions to meet business goals. To save costs, the plants tend to select local suppliers who are aggregated into the milk-run delivery (or milk-run) within a relatively narrow region. Bearing in mind the facts, this thesis uses the aforementioned LA SC in case study to build up a leagile resilient green (LARG) SC. It aims at contemporaneously gaining a raft of benefits of cost reductions, responsiveness, and environmental reputation, while improving resilience to disruptive risks. To attain this goal, this work focuses on both design and management stages. In product design, the thesis adds the green factor in the mentioned LA BOM. Next, to enhance the system’s resilience to threats, besides inventory, two resilient practices—‘dual sourcing’ and ‘supplier's reliability’—are employed in the supplier selection. The lean-green practice is also applied through the implementation of ‘milk-run delivery’ in the sourcing network. These works are formulated into a bi-objectives mathematical model with the objective of minimizing the total purchasing cost as well as the ‘Miss-the-Target’ value of suppliers. In the management stage, two robust measures namely ‘capacity reserve’ and ‘surplus capacity of supplier’, are employed. Moreover, the thesis shares one practical robust practice, the socalled 70/30 rule, which is currently applied in the above case study. Here, the problem in the design step is optimized by weighted goal programming. However, it confronts the complex nature of the milk-run, the NP-complete problem, which is hard to optimize by the exact method. This has inspired the thesis to implement a meta-heuristic approach. Specifically, the thesis tries to develop a novel hybrid meta-heuristic (HMH), namely HAT, which is hybridized from the two meta-heuristics of Ant Colony Optimization (ACO) and Tabu Search (TS). HAT is tested in one milk-run case study of a small-sized automobile LSC, which was solved by ACO. The thesis also qualifies the HAT in large-scale milk-runs through random data by comparing it with ACO, TS and LINGO. In the former, HAT proves superior to the original results although it has yet to reach the global optimum. In the latter, the HAT’s solution is quite promising when it surpasses both those of ACO and TS on quality search and outperforms LINGO at processing time. From such promising results, the thesis proposes a new method to optimize the milk-run delivery problem, which uses exact method for small cases and HAT for large milk-runs. Finally, this HMH HAT is applied to optimize the proposed LARG SCD. The solutions demonstrate how a real company can simultaneously develop LARG model in reality.

Date	14 Sept 2020
Original language	American English
Awarding Institution	École de technologie supérieure
Supervisor	Thien-My Dao (Supervisor)