Full metadata record
DC FieldValueLanguage
dc.contributor.author唐麗英en_US
dc.contributor.authorTONG LEE-INGen_US
dc.date.accessioned2014-12-13T10:41:39Z-
dc.date.available2014-12-13T10:41:39Z-
dc.date.issued2012en_US
dc.identifier.govdocNSC100-2221-E009-060-MY3zh_TW
dc.identifier.urihttp://hdl.handle.net/11536/98616-
dc.identifier.urihttps://www.grb.gov.tw/search/planDetail?id=2381096&docId=377514en_US
dc.description.abstract由於消費者對於產品品質的要求日趨嚴格,製造商也不斷改善產品品質,進而提升 產品之競爭力。製造商通常會使用線外品質管制(off-line quality control)方法如:實驗設 計(Design of Experiment, DOE)來最佳化製程,以提升產品品質;或使用線上品質管制 (on-line quality control)方法如:管制圖(Control chart)來找出造成製程失控之一些可歸屬 原因(assignable causes),以降低製程變異。此外,由於製程中也常存在一些無法避免的 自然干擾(disturbances),這些干擾往往也會使製程之產出值(output)偏離目標值(target), 此時可以應用工程製程管制(Engineering Process Control, EPC)方法來調整製程參數 值,以使製程之產出值能迅速修回目標值。目前一般工廠在應用EPC 時大多是由工程師 憑其經驗,根據製程產出值主觀地決定是否應調整製程參數值以及調整之幅度,然而此 作法之成效常因工程師個人之經驗而異,無法建立一套有系統之調整製程參數的作法。 有些高科技工廠在應用EPC 時雖有導入國外資訊流自動化控制系統,然導入此系統不僅 須花費鉅額資金,且在維修系統時因無法掌握系統之關鍵流程而受制於國外廠商。一些 中外文獻已指出整合SPC 與EPC 會比分別使用 SPC 或EPC 更能降低至成變異。因此, 本計畫之主要目的是針對生產機台與量測機台之製程數據,整合DOE、SPC 與EPC 方法, 發展一套完整之建構資訊回饋控制系統之方法。由於製程所產生之時間序列資料具有量 少且非線性之特性,本計畫在所提出之回饋控制系統中,特別發建構了一個混合灰預測 模型(hybrid grey model),以便回饋控制系統能迅速準確地預測未來時間點之製程產出 值,而能及時啟動正確之相關管制機制,以有效降低製程之變異。本計畫共分三年完成, 第一年之工作是先利用實驗設計 (如:中心混成設計)及反應曲面(Response Surface Method, RSM)方法找出製程參數水準之最佳組合,這些參數設定值即為回饋系統中製程 輸入變數之設定值。然後利用逐步邏輯斯迴歸分析(Stepwise Regression Analysis)將影響 產出值之顯著製程參數排序,再利用自組性演算法(Group Method of Data Handling, GMDH)建構出顯著製程參數與產出值間之關係式,並將此關係式建置於回饋系統中, 以供回饋系統計算製程參數應調整之幅度。本計畫第二年之工作是探討現有中外文獻上 一些時間序列預測模型之優缺點以及量測機台資料之特性,發展出一個混合灰預測模 型,並將之建置於回饋控制系統中。回饋控制系統經由混合灰預測模型得到下一個時間 點之製程產出值,即可判斷下一個時間點之製程產出值是否超出管制界線(control limits),而發出停止生產作業之警訊;或是判斷下一個時間點之製程產出值雖未失控但 偏離目標值,而啟動調整製程參數之機制,應用本計畫第一年所建構之GMDH 方程式 表 C011 共 3 頁 第 1 頁 來調整製程參數值,以使下一批產品之量測值能夠迅速修正回目標值。本計畫第三年之 工作則是根據本計畫第一年及第二年之成果,利用模擬之製程數據以及一個台灣半導體 製程實際案例來說明本計畫所提出之回饋控制系統確實有效可行,最後並將本計畫所提 出建構生產機台與量測機台間之回饋控制系統之方法寫成標準作業流程,以利工廠電腦 化及導入此回饋控制系統。 目前之中外文獻尚未見過整合線外及線上品質管制方法之回饋控制系統,且本計畫 所擬發展之混合灰預測模型亦可應用於其他領域(如:能源需求量之預測),因此本計畫 應具學術價值。業界應用本計畫成果不僅可節省大量導入回饋系統之經費及製程管制人 力,也可避免人為調整製程參數之錯誤及有效降低製程變異,因此本計畫也具實用價值。zh_TW
dc.description.abstractBecause the demand for product quality from customers is more stringent than before, many manufacturers employ quality improvement or control tools to enhance their products’ quality. Design of experiments (DOE) is of utilized to improve the product quality. Statistical process control (SPC) is employed to reduce the process variability. Moreover, engineering process control (EPC) is implemented to adjust the process input, so that the process output can be revised quickly to the target value while the production process is continuously run. Many studies recommended that integrated SPC-EPC has better performance in controlling process variation than just employing SPC or EPC alone. Therefore, the main objective of this project is to construct a feedback control system between the production and metrology tools using integrated SPC-EPC with a hybrid grey forecasting model developed by this project. This three-year project is divided into three phases. In the first year, DOE is utilized to determine the optimal parameter settings. Stepwise regression is also employed to rank the significance of input variables. The significant input variables are then utilized to construct a prediction model for the output using the Group Method of Data Handling (GMDH). The GMDH model is built in the feedback system. In the second year, this study reviews existing forecasting model and develops a hybrid grey forecasting model and this model is built in the feedback control system to forecast the future process output values. If one sample point falls beyond the adjustment threshold but within the control limits, the feedback system will be automatically directed to adjust the value of the most significant input variable. If the most significant input variable is already reached its upper or lower limits and can no longer be adjusted, the feedback system will adjust the value of the next significant input variable, so that the output value of the next run can reach the target value. In the third year, this study uses a real case from a Taiwanese semiconductor manufacturer and the simulated data to demonstrate the effectiveness of the proposed method. The results of this project have academic contribution as well as practical use for manufacturers.en_US
dc.description.sponsorship行政院國家科學委員會zh_TW
dc.language.isozh_TWen_US
dc.subject實驗設計zh_TW
dc.subject自組性演算法zh_TW
dc.subject統計製程管制zh_TW
dc.subject工程製程管制zh_TW
dc.subject回饋控制系統zh_TW
dc.subject生產機台zh_TW
dc.subject量測機台zh_TW
dc.subject混合灰預測模型zh_TW
dc.subjectStepwise regressionen_US
dc.subjectGMDHen_US
dc.subjectSPCen_US
dc.subjectEPCen_US
dc.subjectfeedback control systemen_US
dc.subjectproduction toolen_US
dc.subjectmetrology toolen_US
dc.subjecthybrid grey forecasting modelen_US
dc.title整合線外與線上品質管制方法建構製程回饋控制系統zh_TW
dc.titleIntegrating Off-Line and On-Line Quality Control Methods to Construct a Feedback Control Systemen_US
dc.typePlanen_US
dc.contributor.department國立交通大學工業工程與管理學系(所)zh_TW
Appears in Collections:Research Plans