運用迴歸分析分析生產週期時間－以C公司為例

Abstract

在評估企業績效的眾多指標當中，生產週期時間(Cycle time)是許多企業都相當重視且需不斷監控的指標之ㄧ。在各種生產工廠裡，生產環境常會被許多複雜的因素影響而產生變化，因此造成了生產週期時間非常難以準確的預測。一般來說，生產週期時間通常都是仰賴公司資深員工的經驗法則來預測，但這樣除了缺乏客觀的數據外，更會有無法傳承經驗的情形。因此，本研究的主要目的是收集生產工廠裡的各種生產資訊，應用逐步迴歸分析法對重要因子排序並建構一個準確且可信度高的生產週期時間預測模型，再依此模型來分析相關生產問題，最後找出生產週期時間預估值之信賴區間。另外，工程師也可以依據影響生產週期時間之因子的重要性，來控制這些因子，以提高生產週期時間之預測準確性。本研究最後以台灣某家IC封裝測試公司的實例，驗證本研究方法確實能有效的預測生產週期時間與分析相關生產問題。

Cycle time is one of the most important indexes for evaluating the achievement of an enterprise and it must be monitored continuously. In many factories, the producing environments often change by a lot of complicated reasons. Thus, it is hard to predict the cycle time accurately. In general, cycle time is often predicted by experienced engineers. In this situation, however, the prediction of the cycle time maybe subjective and the experience may not be able to pass to other engineers. Hence, the main objective of this study is to constructing a reliable prediction model of cycle time. First, all production information in the factory are assembled and the key factors for cycle time are identified using stepwise regression analysis. The confidence interval of the cycle time is then obtained via the regression model. Moreover, engineers can enhance the accuracy of prediction model by controlling the key factors found through the stepwise regression analysis. A real case from an IC packaging and testing company is used to illustrate the effectiveness of the proposed method.