整合性供應鏈績效評估與最佳化分析

Abstract

本論文以系統整合角度提出多階物流供應鏈績效評估及最佳化模式。一般來說，多階物流供應鏈可概分為零物料供應、製造、配銷及運輸等供應子系統。以往學者在分析類似問題大多零散不完整，且傾向個別求解。例如已知市場需求，希望求解最適製造策略；或是假設配銷系統供應無限來求解最適庫存水準。本論文假設產品需求及製造環境為隨機分配，各供應子系統產能有限，以矩陣分析方法求出在長期穩定狀態下此供應鏈網路系統績效，即長期穩定作業成本及服務水準。探討各種可能市場需求及供給不確定因素型式對系統績效之影響，提出以馬可夫調適卜瓦松過程建構在此情況下各供應子系統隨機模式，並以矩陣幾何方法求解系統績效。本論文最後運用各種古典及現代尋優理論，配合所提出系統評估模式找出供應最佳庫存及產能水準，並考慮上游供貨不穩定對系統績效之影響，配合實例說明以此整合分析及最佳化模式找出不穩定供需環境下最穩健系統資源規劃。

An integrated matrix analytic model is proposed to evaluate a multi-echelon supply chain (SC) from a systematic viewpoint. Generally speaking, a multi-echelon supply chain consists of four basic subsystems: raw material supply, production, distribution and transportation. In the past, most of the analyses of complex supply chain models seemed to be fragmented and tended to pursue local analysis. For example, some studies assumed market demands were known and tried to find the most adequate manufacturing strategies. Others assumed that supplies were unlimited and tried to find optimal distribution strategies to fight against demand uncertainty. In the first phase of this study, the market demand and supply process were assumed stochastic. Each SC contributor was then treated as a single server in a tandem queueing network. We developed an integrated matrix analytic model for the steady state performance evaluation in terms of total operating cost and customer service level for a fabricated SC problem. We also took upstream unavailability into the modeling process. Next, we investigated deeply into the impact of supply/demand uncertainties on SC performance. We modeled various uncertainty scenarios as Markov modulated Poisson processes (MMPP). Using matrix geometric method we obtained the steady state performance measures and proposed solution measures for a test SC problem with make-to-order mode. Finally, based on the above-developed SC evaluation model, we investigated various optima seeking procedures. Our objective was to solve the optimal buffer and capacity allocation problems simultaneously such that the chain wide operation cost was minimized under pre-specified service level. Especially, we discussed the impact of unreliable supply on system performance.