半導體廠房營運維護預算編列模式之分析

Abstract

半導體廠房之營運維護目前實務上多以實際統計維修需求及歷年來營運維護成本為參考數據，此種營運維護費用之估算方式欠缺客觀依據，與實際歷年金額數據顯示差異過大，無法有效預估來年之營運維護預算。 本研究的目的在於依半導體廠務的實際系統分類，建構營運維護費用之模型，進行來年營運維護費用各項目的預估，並針對預估的結果檢討其準確性及適用性；研究方法共分為描述統計及案例操作。 首先針對案例資訊介紹，並說明編制組織及費用分類，將營運維護費用定義的闡述，參考文獻加以比較探討，再利用三座類似廠房案例，各廠分類項目歷年預算費用比較、歷年實際費用的比較、自廠歷年預算及實際費用比較，統計出差異百分比最大最小值，計算出總差額，顯示歷年來的差異。 運用三座廠房案例Excel中之線性、指數、對數、多項式及乘冪迴歸分析，計算來年營運維護各項目的預測值，並求出R2值，由R2可推估各項費用的準確性及適用性。另使用單一廠房案例將其階段分期，尋求平均差異最高的項目，並利用專家訪談，找出營運維護成本關鍵，提出建議的管理策略。 本研究收集實際案例進行迴歸分析，得出之結果具有一定的準確性及可行性，雖然無法適用於每個項目的推估，但在整體總金額是有其正確性的，建議後續研究可針對成本之特性，建立不同之成本預估模型，能讓營運維護費用的預算編列能有所依據也更合理。

Operating and maintenance cost estimations for semiconductor factories are based on actual statistics and reference to previous years estimation. This current practice not only lacks of objectivity but is also error-prone, resulting in incorrect estimation of the operating and maintenance budget for coming the coming year. The purpose of this study is to provide a model of operating and maintenance cost estimation based on the systems classification of the semiconductor factories. Furthermore, the model is reviewed to test its accuracy and suitability to expected results. The methods used are descriptive statistics and real case manipulation. Firstly, we will give a detailed introduction of the cases, classification of budget and description of operating and maintenance costs. Then we will go through a literature review using three semiconductor factories, statistical analysis and comparison of their annual budgets. The average cost over the years is to be derived through a calculation of the minimal and maximal cost value and the total. Excel functions such as linear, exponential, logarithmic, polynomial regression analysis and exponentiation are used to derive a predicted value R2 which is used to estimate the accuracy and applicability of the cost. Also, search for highest average difference between projects, and interviews with experts to identify key operating maintenance costs, and recommend management strategies. The overall result of this study using regression analysis method has a certain amount of accuracy and feasibility even though it’s not applicable to all projects. The study has proposed a reasonable model which can be used for estimating correctly operation and maintenance costs.