半導體是一新興高科技產業,因產業的特性使得產品製造時程很長;雖然一般顧客總是提前在半年或是一年前下訂單,可是過長的交期,不論對於廠商或是顧客在資金上的調度,是一大考驗。在傳統的交期指定模式中,大都使用工件本身及現場的資訊來做交期預測,但對於半導體產業製造流程的複雜性,很多的影響因素並非呈現線性關係,所以在原有的交期模式下,無法有效利用這些影響因素之資訊;因此,本研究即針對廠商與客戶之間發展一有效的交期指定模式(due date assignment model),以避免廠商與客戶之間不必要的機會成本之損失。 本研究的方法以動態現場狀況(dynamic shop condition)為基礎,透過電腦模擬與統計分析過濾出影響交期指定的因素,經由類神經網路,發展一適用於半導體產業的交期指定模式,期望能夠有效預測出實際交期。本研究工作第一階段為結合電腦模擬與統計分析方式,找出影響交期指定的因素;第二階段針對影響因素,經由類神經網路建構一有效之交期模式,來精準預測半導體產品之交期。
Semiconductor is a new hi-tech industry. One characteristic of hi-tech industry is the long manufacturing lead times. Although clients usually order the product half or one year before, the over-long due date assignment may make the factories and clients have financial difficulties. In the original due date assignment, a company often predicts the due date according to the product and shop condition. But in regard to the complicatedly technological processes of semiconductor industry, there are many effective factors that do not present correlation; so, the conventional due date assignment model cannot use the information of these effective factors reasonably. Therefore, this research is to develop an operative due date assignment between factories and clients for them to avoid losing unnecessarily prime costs. The research is based on dynamic shop condition to identify the factors that effect due date assignment through computer simulation and statistic analyzes, then develop a due date assignment model that is useful for semiconductor industry by way of neural network, and hope to predict practical due date validly. The first step of the research is to combine computer simulation and statistics analyzes, and use it to find out the effective factors of due date assignment. Second, establish an efficient due date assignment according to effective factors through neural network to predict the due date of semiconductor products precisely.