高科技產業資源規劃與產能分派最佳化之研究

Optimization of Resource Planning and Capacity Allocation for High-Tech Manufacturing Industry

王識閔(Shih-Min Wang)

指導教授 ： 王孔政 ； 陳建良

中原大學/工學院/工業工程研究所/博士(2007年)

https://doi.org/10.6840/cycu200700280

摘要

台灣的高科技產業如晶圓製造與封裝、筆記型電腦製造、主機板製造與液晶螢幕顯示器在世界上均居於領先地位。完善的資源規劃與產能分派對於企業發展與獲利扮演關鍵性的角色。本論文探討的高科技產業資源規劃與產能分派問題包含一系列研究議題，如短暫的產品生命週期、複雜的混合產品生產、時間價值下有限資金對資源的巨額投資、快速的資源技術革新與衰退性的產品價格、具特定搭載限制的同步性資源、潛在的資源取得方式與不確定性需求等。
本研究探討的高科技產業資源規劃與產能分派主要可被區分為兩種層次:單廠規劃層級與跨廠規劃層級。本論文首先針對訂單生產環境，發展一個多資源規劃與產能分派模型。接著，本研究建立存貨生產製造環境下用以平衡多規劃期間產能負荷的模型來解決問題。此模型有助決策者發展最具經濟效益的資源投資組合與規劃計劃。本研究除了根據上述兩種製造環境發展具混合整數特性的確定性數學模型外，亦分別發展一系列對應的限制規劃為基礎的基因演算法。本研究並以實驗展示求解方法的可行性與效率性。此外，本研究針對需求不確定生產環境下之資源規劃與產能分派問題，發展一個能於風險性需求環境下進行生產獲利與風險取捨的數學規劃模型。此一隨機最佳化模型與其對應的基因演算法— 隨機規劃為基礎的基因演算法，成功的解決高科技產業中的隨機性需求下資源規劃與產能分派問題。
至於跨廠規劃層級，本研究提出一個以媒介者為基礎的架構，以解決兩廠之間的產能交易問題(例如:以最經濟的方式撮合產能買方的剩餘訂單與產能賣方的剩餘資源產能)。更進一步，本研究考量多廠但各方資訊不通透的環境，發展一個更精緻的產能協商模型及使用協商決策函數的產能協商架構。本研究並驗證各廠於不同協商態度下，對跨廠資源規劃與產能分派的影響。
本論文的主要貢獻可歸納如下。(1)在單廠規劃層級:本研究針對不同的實務環境發展一系列的數學模型，其中分別考慮了訂單生產、存貨生產與隨機性需求特色。本研究也已發展對應的柔性計算方法來求解上述模型，演算法的良好績效也經由敏感度分析獲得證實。(2)在跨廠規劃層級:本研究的貢獻在於成功的提昇多廠間有限資源的資源利用率，同時也有效的減少了資訊不通透環境下，廠與廠之間所造成的衝突。

關鍵字

產能協商 ； 隨機規劃 ； 基因演算法 ； 資源規劃與分派

並列摘要

Taiwan is the leading country of the world in several high-tech industries, such as wafer fabrication and packaging, notebook manufacturing, electronic mother board manufacturing, and LCD monitor manufacturing. Resource planning and capacity allocation of the industries is one of key strategic decisions and impacts on profit significantly. This dissertation addresses the resource planning and capacity allocation problems in the high-tech industries that possess issues including short product lift cycle, sophisticated product mix, heavy investment capital for resources considering time value of limited budget, rapid technology innovation velocity of resources and declined product price, simultaneous resource with confined configurations, potential resource acquisition alternatives, and risky demands.
The study for resource planning and capacity allocation in high-tech industry is conquered by dividing it into two levels: an individual-factory level and an inter-factory negotiation level. In the dissertation, an individual-factory multiple-resource planning and capacity allocation model first have developed for the make-to-order manufacturing environment. After that, a model of solving a production resource portfolio problem is built to balance the capacity loading so that an economical resource planning plan can be obtained under a make-to-stock manufacturing environment. The two problems mentioned above are formulated as deterministic mathematical constrained optimization models with mix-integer characteristics. Thus, a set of constrained-programming based genetic algorithms (GA) are developed to solve the problems and the feasibility and efficiency of the solving methodologies are also justified by comprehensive experiments. Furthermore, in order to incorporate risk in demands, a mathematical model considering tradeoff between the profits and risks of production is developed to empower resource planning and capacity allocation decision under risky-demand environment. This stochastic optimization model and the corresponding GA algorithm, sampling based stochastic programming, are developed to tackle successfully the stochastic problem in the high-tech industry.
In terms of capacity negotiation at the inter-factory capacity trading level, a framework is proposed through a mediator to solve the capacity trading problem between two parties, i.e., a capacity supplier and a capacity buyer, to economically coordinate remaining orders and capacity of resources. Moreover, a more sophisticated capacity trading framework using negotiation decision function is developed to determine the inter-factory resource planning and capacity allocation among factories with different preferred attitudes and asymmetric local information.
The contributions of this dissertation mainly can be summarized as follows. In the individual factory domain, several mathematical models are formulated for practical purpose by considering several different characteristics of make-to-order manufacturing, make-to-stock and stochastic demand. The corresponding soft programming models are also developed to solve the proposed model and have been demonstrated through sensitivity analysis. In more high level of resource planning, this dissertation makes a contribution in successfully improving the utilization of the limited resources among the factories and efficiently reducing the violations of the asymmetric information environment.

並列關鍵字

Stochastic programming ； Capacity negotiation ； Resource planning and allocation ； Genetic algorithm

參考文獻

Jiang, A. X., 2000, Capacity Trading Among Semiconductor Manufacturing Factories, Master Thesis of Department of Industrial Engineering, National Taiwan University, Taiwan.

Chang, K. H., H. J. Chen, and C. H. Liu, 2002, A Stochastic Programming Model For Portfolio Selection, Journal of the Chinese Institute of Industrial Engineers, Vol. 19, pp. 31-41.

Alonso, A., L. F. Escudero, and M. T. Ortuño, 2000, Theory and Methodology— A Stochastic 0-1 Program Based Approach for the Air Traffic Flow Management Problem, European Journal of Operational Research, Vol. 120, pp. 47-62.

Bard, J. F., K. Srinivasan, and D. Tirupati, 1999, An Optimization Approach to Capacity Expansion in Semiconductor Manufacturing Facilities, International Journal of Production Research, Vol. 37, No. 15, pp. 3359-3382.

Bashyam, T. C. A., 1996, Competitive Capacity Expansion Under Demand Uncertainty, European Journal of Operational Research, Vol. 95, No. 1, pp. 89-114.