• 沒有找到結果。

行政院國家科學委員會專題研究計畫 期末報告

N/A
N/A
Protected

Academic year: 2022

Share "行政院國家科學委員會專題研究計畫 期末報告"

Copied!
48
0
0

加載中.... (立即查看全文)

全文

(1)

行政院國家科學委員會專題研究計畫 期末報告

限制驅導式排程技術(DBR)於 LED 晶粒製造廠應用之強化模 式研究(I)

計 畫 類 別 : 個別型

計 畫 編 號 : NSC 101-2221-E-216-016-

執 行 期 間 : 101 年 08 月 01 日至 102 年 07 月 31 日 執 行 單 位 : 中華大學企業管理學系

計 畫 主 持 人 : 吳鴻輝

報 告 附 件 : 出席國際會議研究心得報告及發表論文

公 開 資 訊 : 本計畫可公開查詢

中 華 民 國 102 年 10 月 15 日

(2)

中 文 摘 要 : LED 晶粒製造(LED-CM)為 LED 產業中重要的一環,晶粒製造 過程由於製程技術的特殊性,容易造成晶粒光電性規格不均 勻及良率不穩之問題,一般 LED 晶粒產出符合客戶需求規格 的比率約只有 6~8 成左右。因而在接單式生產時,當一張訂 單出現產出不足時,生管必須以補投料方式,補投該訂單所 缺的不足量。雖然客戶能體諒 LED-CM 製程產出不穩之特性,

但會要求補投單必須以急單的趕工方式生產,以減低該訂單 產出不足的影響。因此 LED-CM 廠在大量補投單的趕工壓力 下,如何確保這些補投單的如期出貨以及如何降低這些補投 單對正常單交期的衝擊,即為 LED-CM 廠所面對的挑戰。

本研究提出了一套 DBR 的強化模式以改善 LED-CM 廠所面對的 補投單挑戰。這套 DBR 強化模式主要強化了補投單在 DRUM 的 合理插單及在投料計畫的立即投料需求。其次本計畫亦設計 了模擬實驗,以驗證本模式的可行性及效益。實驗結果證 明,不論在平均流程時間或補投單交期回覆誤差等績效,本 模式都優於 EDD 及 EDD+補投單優先模式。

中文關鍵詞: 關鍵字:限制驅導式排程技術、LED 晶粒製造、補投單、趕 工管理、插單模式

英 文 摘 要 : The LED chip manufacturing (LED-CM) is an important process in the LED supply chain. The make-to-order production is a general model for the LED-DM plants to satisfy the variety requirement of their

customers. Because the unstable production output of the LED-CM plants, the effective order fulfillment is low and variety. Therefore, the production planner will thus confront the issue of an insufficient output of a manufacturing order (MO). Although

material re-issued is allowed by the customers, this re-issued lot is required to be rush order. The LED- CM plant thus confront the issues of on-time delivery of those re-issued lots and the impact on the normal lots.

An enhanced models of Drum-Buffer-Rope (DBR) system for LED-CM plants is proposed in this project to improve those issues mentioned above. This model provides the hot lot schedule in DRUM and immediate material-issued in ROPE. A simulation and experiment model is also designed in this project to demonstrate the feasibility and effectiveness of this model.

(3)

Based on the simulation study, this model surpasses the traditional EDD model or EDD + hot lot first model in mean flow time or the deviation between forecast due-date and actual delivery.

英文關鍵詞: Key Words: Drum-Buffer-Rope(DBR), LED chip

manufacturing (LED-CM), Re-issued lot, Expediting model, Hot lot scheduling

(4)

行政院國家科學委員會補助專題研究計畫 □期中進度報告

期末報告

(計畫名稱) 限制驅導式排程技術(DBR)於 LED 晶粒製造廠應用之強化模式研究(I) Enhanced Models of Drum-Buffer-Rope (DBR) System for LED Chip Manufacturers(I) 計畫類別:■個別型計畫 □整合型計畫

計畫編號: NSC 101-2221-E-216-016-

執行期間: 101 年 08 月 01 日至 102 年 07 月 31 日 執行機構及系所:中華大學 企業管理學系

計畫主持人:吳鴻輝 計畫參與人員:

本計畫除 繳 交成果報告外,另須 繳 交以下 出國報告:

□赴國外移地 究心得報告 研

□赴大陸地區移地 究心得報告 研

■出席國際學術會議心得報告及發表之論文

□國際合作 究計畫國外 究報告 研 研

處理方式:除列管計畫及下列情形者外,得立即公開 查 詢

□涉及專利或其他智慧財 權,□一年□二年後可公開 詢 產 查

中 華 民 國 102 年 10 月 1 日

(5)

中文摘要

LED 晶粒製造(LED-CM)為 LED產業中重要的一環,晶粒製造過程由於製程技術的特殊性,容易

造成晶粒光電性規格不均勻及良率不穩之問題,一般LED 晶粒 出符合客戶需求規格的比率約只有產

6~8 成左右。因而在接單式生 時,當一張訂單出現 出不足時,生管必須以補投料方式,補投該訂單產 產

所缺的不足量。雖然客戶能體諒LED-CM 製程 出不穩之特性,但會要求補投單必須以急單的 工方產 趕

式生 ,以減低該訂單 出不足的影響。因此產 產 LED-CM 廠在大量補投單的 工壓力下,如何確保這些趕

補投單的如期出貨以及如何降低這些補投單對正常單交期的衝擊,即為LED-CM 廠所面對的挑戰。

本 究提出了一套研 DBR 的 化模式以改善強 LED-CM 廠所面對的補投單挑戰。這套 DBR強化模式

主要 化了補投單在強 DRUM 的合理 單及在投料計畫的立即投料需求。其次本計畫亦設計了模擬實驗插

以驗證本模式的可行性及效益。實驗結果證明,不論在平均流程時間或補投單交期回覆誤差等績效,

本模式都優於EDD 及 EDD+補投單優先模式。

關鍵字:限制驅導式排程技術、LED晶粒製造、補投單、 工管理、 單模式趕 插

Abstract

The LED chip manufacturing (LED-CM) is an important process in the LED supply chain. The make-to- order production is a general model for the LED-DM plants to satisfy the variety requirement of their customers. Because the unstable production output of the LED-CM plants, the effective order fulfillment is low and variety. Therefore, the production planner will thus confront the issue of an insufficient output of a manufacturing order (MO). Although material re-issued is allowed by the customers, this re-issued lot is required to be rush order. The LED-CM plant thus confront the issues of on-time delivery of those re-issued lots and the impact on the normal lots.

An enhanced models of Drum-Buffer-Rope (DBR) system for LED-CM plants is proposed in this project to improve those issues mentioned above. This model provides the hot lot schedule in DRUM and immediate material-issued in ROPE. A simulation and experiment model is also designed in this project to demonstrate the feasibility and effectiveness of this model. Based on the simulation study, this model surpasses the traditional EDD model or EDD + hot lot first model in mean flow time or the deviation between forecast due- date and actual delivery.

Key Words: Drum-Buffer-Rope(DBR), LED chip manufacturing (LED-CM), Re-issued lot, Expediting model, Hot lot scheduling

(6)
(7)

一. 前言

LED 晶粒製造(LED-CM)為 LED產業中重要的一環,晶粒製造過程由於製程技術關係,容易造成晶 粒光電性規格不均勻及良率不穩,加上市場上客戶應用差異對電性規格的要求多元,因此在接單式生

型態下的

產 LED-CM 廠中生 與銷售之間的管理是相當有挑戰性的產 [1,2]。一般 LED 晶粒 出符合客戶產

需求規格的比率約只有6~8 成左右。生 管理者產 (以下稱生管)則是這 銷協調的重要橋梁。圖產 1 為 LED- CM 廠之訂單履約管理流程,生管經由訂單得知客戶需要的品名、光電性規格、數量與需求日等資訊,

生管必須先確認庫存是否能滿足客戶需求,如否則須參考產品的”產出分布”及”良率”安排投 計畫,產

並回覆客戶交期。當計畫擬定完成,則建立工單並確認投產產能限制及安排適合磊晶片投料。投料完成

後則需監控生產線產出符合工單狀況,以確保客戶需求能如期達交[2,5,16]。

晶粒製造的前段製為蒸鍍、顯影、蝕刻、熔合及研磨等加工,為決定電性的重要製程,但電性資料

卻是要到點測後才能得知,此時生管必須確認產出規格是否有符合當初所預期的 出分佈,進而確認產

是否會影響達交數量。

圖1 LED-CM 廠之訂單履約管理流程

LED 晶粒生 過程中會造成光電性規格的不均勻,圖產 2為磊晶加工成LED 晶粒時的亮度電性分

佈示意圖,通常磊晶的內部的晶粒均勻性較高,也愈能符合預期的規格。以圖2為例,目標是要打

Grade4 及 Grade5,因此中心區域 Grade4、Grade5 的比例會較高,愈往外圍亮度愈暗, 出數量也較少產 ,

以規格等級與數量的分布看來,較接近常態分配關係,如圖3 所示,同時也會製程的不穩定而造成分

配峰形的改變或產出平均(μ)的偏移。

訂單 依標準良率及需求量擬訂出分布&

計劃

安排投料

&

確認投限制

1. 預計合工單數量 2. 預計入庫日期

目檢/入庫

前段製程 切割 點測 晶粒分類 QC

實際出分布 回覆交期、數量

實際合工單數量

<

預計合工單數量

建立工單

動補料法則

Y

更新交期、數量

確認庫存是否 有足數量

1. 品名 2. 需求規格 3. 數量 4. 需求日期

Y

N

N 監控生產合工單狀況

訂單 依標準良率及需求量擬訂出分布&

計劃

安排投料

&

確認投限制

1. 預計合工單數量 2. 預計入庫日期

目檢/入庫

前段製程 切割 點測 晶粒分類 QC

實際出分布 回覆交期、數量

實際合工單數量

<

預計合工單數量

建立工單

動補料法則

Y

更新交期、數量

確認庫存是否 有足數量

1. 品名 2. 需求規格 3. 數量 4. 需求日期

Y

N

N 監控生產合工單狀況

(8)

圖2 LED 晶粒亮度分佈示意圖

圖3 LED 亮度規格 出分佈產

由於此 出特性,生管在監控 出時所面臨的問題,除了製程良率外還需確認規格、數量是否如計產 產

畫所預期的分佈。當產出分布符合預期分佈時,工單將能順利進行接下來的製程,如過產出分佈不如

預期,導至目標規格數量不足時,生管必需進行補投計畫並更新交期。此特性引申出的另一個問題是

庫存容易因產出不符工單規格,無法出貨導致庫存上升,因此庫存的調配對於產銷平衡是相當重要的。

現有LED-CM 廠的生管都是以其經驗在運作,在排程機制上常常會以交期法則 (EDD)來決定投產

先後,針對上述問題較難有顯然的改善。例如吳與李[2]從生 管理的角度,歸納了四點實務上可能面產

臨的問題:1. 出分布不穩定與客戶需求多變,易造成庫存上升與未能順利達交的風險。產 2. 庫存與

WIP 管理的重要,必須靈活的調度庫存與 WIP,減少不必要的浪費是很重要的。3.逐批 WIP 確認達交 狀況困難,生管必須有良好的管理方法來確保交期的順利達交。4.. 生管人才訓練困難,生管面對問題 總是依經驗處理,對新人來說是需要時間培養的。

因此LED-CM 廠若要輔助生管克服上述問題,進而提升整廠的合單率、訂單達交率、縮短製造時間

以及降低存貨水準等,勢必要引進先進有效的管理方法或系統。在業界有許多可用來改善接單式生產 交期品質的方法,例如及時化生產(JIT)、先進生 排程系統產 (Advanced Production Scheduling System)或 限制理論(TOC)等。但其中以架構在 TOC 上的限制驅導式排程技術(Drum-Buffer-Rope,DBR)[3,12]是

最可行方法之一,是本計畫所要探討的管理方法。

DBR 是由高瑞博士(Dr. E. M. Goldratt)於 1986 年所提出的現場排程與管理技術[11,12],是一套建

立在TOC 管理哲學上的生 管理方法或解決方案。這套技術已成功的應用在愈來愈多的國內外 業產 產

Grade 1 Grade 2 Grade 3 Grade 4 Grade 5 Grade 6 Grade 1 Grade 2 Grade 3 Grade 4 Grade 5 Grade 6 Grade 1 Grade 2 Grade 3 Grade 4 Grade 5 Grade 6

數量

50~100 Grade1

100~150 Grade2

150~200 Grade3

200~250 Grade4

250~300 Grade5

>300 Grade6

亮度( mcd)

磊晶外圍磊晶中心

50~100 Grade1

100~150 Grade2

150~200 Grade3

200~250 Grade4

250~300 Grade5

>300 Grade6

亮度( mcd)

磊晶外圍磊晶中心

(9)

[4,6-10,12-14],而且改善成效非常的快且大[10]。

雖然已有愈來愈多的文獻探討DBR 成功在 業的應用,其次 了在不同複雜工廠的應用,亦提產 為

出了不少的DBR 改善或 化模式,然而並無在類似強 LED-CM 廠這種製程不穩定且兼具迴流複雜性之

應用。本研究將針對LED-CM 廠不穩定製程及經常需要補投工單等特性以 DBR 排程平台 基礎,來建為

立上述LED-CM 廠問題的解決模式。

本 究因此提出了一套研 DBR 的 化模式以改善強 LED-CM 廠所面對的補投單挑戰。這套 DBR強化模

式主要 化了補投單在強 DRUM 的合理 單及在投料計畫的立即投料需求。其次本計畫亦設計了模擬實插

驗,以驗證本模式的可行性及效益。

二. DBR 的強化模式

LED 晶粒生 過程中的由於光電特性 出容易呈現不均勻分布,及良率的不穩,導致合訂單規產 產

格的中BIN 率容易降低,這時如果庫存無法彌補損失的數量,這時即需立即補投訂單並以急件方式趕

工[16]。如圖 4 所示,點測站常常會形成 LED-CM 廠的 頸站,此站同時也是得知合工單量是否足瓶 夠 的重要站別。透過DBR 排程機制,將工單依 頸站的節奏排出瓶 DRUM 順序,在回推 CCR Buffer,排 出投料節奏(RPOE), 頸站按瓶 DRUM 順序進行加工,而非 頸站亦配合瓶 DURM 節奏指派加工順序。

當工單完成瓶頸站加工,即判斷所點測出的光電性是否符合預期之工單規格,如果能滿則工單預成量,

則繼續接下來的制程,反之如果不能滿足工單預成量,即馬上補投工單。本 究針對此特性所提出研

DBR 排程機制,假如捕投單 x 已 生,則將補投單產 x 與原來 DRUM 中未生 的工單做比較,來決定產 補投單的順序,如果補投單x產生時間點+急件工單 BUFFER 小於工單y瓶頸站開始加工時間,即將補 投單X安插工單y前面順序,補投單X投單時間即為補投單x產生時間點。

三. 模擬和實驗

了分析本究所提DBR 排程機制使用在 LED-CM 廠的優劣,因此將使用 VB 來模擬 LED-CM 廠

製程,與現行實務常使用的法則來做比較,分別為EDD 排程機制(如圖 5)及補投單優先之 EDD 排程

機制(如圖 6),比較三種法則在不同的補投水準下所影響工單流程時間及補投單交期回覆之準確度情 況。

LED-CM 模擬廠有八個主要站點,分別 前段的蝕刻、蒸鍍、顯影、黏合、切割,後段的點測、分類、目為

檢,其代號如圖7 所示。

機台的 能狀況及各站加工時間如表一所示,產 E、D、P、B 為 EPI 狀態作業,作業時間是以整批來計算

,C、Po、S、I為 CHIP 狀態作業,因此作業時間以 chip 顆粒數來計算,每一機台一次只能針對一批工單 作業。

同時模擬三種 品分別產 為A、B、C,每種 品有三種規格例如產 A產品有A1、A2、A3 等三種規格,產

品的製程分別如表二,不同規格有不同的中BIN 率。

(10)

產品製程路徑 工

非CCR派工順序

DRUM CCR操作順序

DRUM

蒸 鍍

顯 影

蝕 刻

黏 合

割 點測 分

目 檢

非CCR派工順序 DRUM

是否足 夠訂單 需求量

新增補投工單 N Y

已投產工單 已過瓶頸加工

ROPE DRUM

工單1 工單1

工單2 工單2

工單3 工單3

工單4 工單4

工單5 工單5

工單y 補投單x

工單7 工單y

補投單x 工單7

工單8 工單8

工單9 工單9

. . .

. . .

工單n 工單n

不需補投工單 圖4 DBR 的 化排程機制強

產品製程路徑 工

投單法則

EDD

各站別派工順序 EDD

蒸 鍍

顯 影

蝕 刻

黏 合

割 點測 分

目 檢

是否足 夠訂單 需求量

Y

N 新增補投工單

不需補投工單

圖5 EDD 排程機制

(11)

產品製程路徑 工

投單法則

EDD

各站別派工順序 補投單優先指派&EDD

蒸 鍍

顯 影

蝕 刻

黏 合

割 點測 分

目 檢

是否足 夠訂單 需求量

N 新增補投工單

補投工單即時投產

Y 不需補投工單

圖6 補投單優先之 EDD 排程機制

蒸鍍(D)

(Deposition) 顯影(P) (Photo) 蝕刻(E)

(Etching) 黏合(B)

(Bonder) 切割(C)

(Cut) 點測(Po)

(Probing) 分類(S)

(Sorter) 目檢(I) (Inspection)7 LED-CM廠製程與代碼說明

站別 E D P B C Po S I

機台數 3 2 3 2 1 1 2 2

整批作業時間(天) 2.5 2 2.5 2 - - - -

每粒chip作業(天) - - - - 0.025 0.042 0.035 0.035

Station 1

Station 2

Station 3

Station 4

Station 5

Station 6

Station 7

Station 8

Station 9

Station 10

Station 11

Station 12

A E D P B C Po S I - - - -

B E D P E D P B C Po S I -

C E D P B E D P B C Po S I

表一 站別產能

表二 產品製程

(12)

圖 8 為 LED-CM 廠之模擬架構,工單隨機 生並同時 生六種資訊,來單時間、品名、需求量、交期、產 產 製程、規格,再將工單依不同排程法則排出投料順序且按順序進行投料,而各站的加工順序則依不同 的機台派工法則的決定工單的加工順序,其中當工單通過點測站(Po)加工完成時,立即判斷該工單數

量是否能滿足預期的需求量,如不能滿足則立即產生補投訂單,再透過排程法則安排補投單順序。

Tnow=0 產生訂單i Output(工單i資料) 1.來單時間 2.品名 3.需求量 4.交期

5.製程 6.規格

依投單法則安排投單順序:

Output 1.投料時間j

j:未投料工單數

投料時間i 否已到達

工單i進行投料 i:已投料工單數

工單結束加 工時間i是否 已到達

依製程將工單i送至下一站

工單i依機台派工法則排序

各站是否有 機台閒置

各站依機台派工順序上料

工單i是否位 於點測站

Y N

Y

N

N

合工單i規格 量是否大於 等於需求量

Y

產生補投工單i Output(工單i資料) 1.產生 時間 2.品名 3.補投量 4.更新交期

5.製程 6.規格

Y N

Min{工單i結束加工時間,投料時間j}=Tnow

Tnow=結束 作業時間 i& j=0

Y

N

N

完成作業

Y

圖8 LED-CM 廠之模擬架構 四. 實驗設計

了確定有效實驗的長度為 ,工單產生頻率在不同水準的指數分配下

(λ=0.2、λ=0.21、λ=0.22、λ=0.23),3000 天內隨機以指數分配 生工單,分別 生產 產 665 、699、725、775 張工

單,以EDD 排程機制作先前測試實驗,得到的 CCR 平均使用率如圖 9,圖中可看出時間點在

250~2900 天期間系統是較穩定的,因此可以判定當 生訂單天數產 為3000 天時,統計 250~2900 天的資 料是有效益的。

了比較排程機制對於為 LED-CM 廠常態補投特性的優劣,本實驗將設計不同補投單比例來測試三種

排程機制,但為了建立在公平的水準下,必需要讓瓶頸站使用率在同一水準下做測試。表三為工單產

生頻率與不同補投單比例所造成CCR 使用率,由表可以得知當 λ=0.198 補投率在 30%時,CCR 的使

(13)

用率為95.55%,λ=0.21 補投率在 20%時,CCR 的使用率為 95.93%,λ=0.22 補投率在 10%時,CCR 的

使用率為95.79%,這三種條件下 CCR 的使用率水準相當,因此本實驗將此三種條件設定 實驗因子。為

圖9 工單 生頻率在不同水準指數分配下產 CCR 使用率狀況

訂單到達頻率 補投單所占比例

指數分配(λ) 10% 20% 30%

λ=0.198 87.49% 90.50% 95.55%

λ=0.21 92.88% 95.93% 98.69%

λ=0.22 95.79% 98.79% 99.23%

4.1 實驗環境與基本假設

綜合上述整理出實驗環境及基本架設如下:

1. 一共三種產品A、B、C,每種 品三種規格產 A1~A3、B1~B3、C1~C3,工單隨機 生品名。產 2. 每張工單需求量為Normal(80K,5)隨機 生。產

3. 需補投量為Normal(15K,3)隨機 生。產

4. 一般補投單回覆交期的設定為各產品平均急件作業流程時間1.5 倍。

5. DBR 法則補投單回覆交期的設定為 DRUM+Shipping Buffer。

6. 每一機台一次只能針對一批工單作業。

7. 各站針對每工單加工時間固定,如表一。

8. 工單按表二製程順序加工。

9. LED-CM 模擬廠假設無當機及人員異常問題。

10. 3000 天內工單 生間距符合指數分配產 ,當所有工單都完成,實驗即停止。

11. 模擬系統 簡化版為 LED-CM 模擬廠。

4.2績效評估因子

表三 訂單到達(產生)頻率與不同補投單比例所造成 CCR 使用率

(14)

本實驗績效評估因子一共有二個,分別 平均流程時間為 (MFT)、補投單交期回覆誤差(回覆交期與實

際完成時間的差異),MFT為產品從投入線上到完成的平均時間,越短代表產線對於WIP 與 能的控產

管能力愈佳。而補投單交期回覆誤差目的是要比較何種機制對時交期回覆的準確性較佳,LED-CM 廠

在LED 供應鏈 業中屬中上游,而接單式生 的模式中交期的準確性是相當重要的,因 這會一產 產 為 併

影響下游客戶的生產計劃,如果生產節奏太快容易影響雙方的庫存,而節奏太慢又會影響客戶斷線,

因此回覆交期的準確性是相當中的,尤其在已經跳票的補投單更是重要。

4.3 實驗結果

圖 10 到圖 12為模擬實驗結果之比較圖,由圖10 可以看出補投單優先之 EDD 排程機制與 DBR 排程

機制在任何條件下,都能有效地縮短補投單的MFT,但補投單優先之 EDD 排程機制確會因過於照顧

補投單而使一般的MFT 延長,而 DBR 排程機制,不但能使補投單的 MFT 維持在最短的水準,還能

有效的同時縮短一般單的MFT,在相同的生 系統中當產 MFT 愈短意味著線上的 WIP 也越低。

補投單10%的MFT&FTStd

41.09

22.60 22.91

82.25 83.21

62.54

- 20.00 40.00 60.00 80.00 100.00

EDD EDD(補單優先) DBR

補投單FTStd 一般單FTStd 補投單MFT 一般單MFT

補投單20%的MFT&FTStd

44.30

22.35 22.95

82.85 84.80

71.55

- 20.00 40.00 60.00 80.00 100.00

EDD EDD(補單優先) DBR

補投單FTStd 一般單FTStd 補投單MFT 一般單MFT

補投單30%的MFT&FTStd

46.17

22.45 22.94

83.77 87.11

77.00

- 20.00 40.00 60.00 80.00 100.00

EDD EDD(補單優先) DBR

補投單FTStd 一般單FTStd 補投單MFT 一般單MFT

圖10 在不同補投單比重下各法則的 MFT(單位:天)

圖 11 目的是要比較補投單比重對法則的影響,由補投單的角度來看,當補投單增加時,補投單優

先之EDD 排程機制與 DBR 排程機制兩法則皆能有效縮短 MFT 且維持補投單 MFT 於最短的狀態,而

EDD 排程機制確會因補投單數量越多而使 MFT 越長。由一般單的角度來看,EDD 排程機制與補投單

(15)

優先之EDD 排程機制會因補投單數量越多而使 MFT 越長,尤其是補投單優先之 EDD 排程機制,而 DBR 排程機制雖然也會因補投單越多而使 MFT 加長,但比較之下卻是三種機制中能讓一般單 MFT 最 短的機制。

不同補投單比重對補投單MFT之影響

41.09 44.30 46.17

22.60 22.35 22.45

22.91 22.95 22.94

- 10.00 20.00 30.00 40.00 50.00

補投單10% 補投單20% 補投單30%

補投單MFT (EDD) 補投單MFT (EDD(補單優先)) 補投單MFT (DBR)

不同補投單比重對一般單MFT之影響

82.25 82.85 83.77

83.21 84.80 87.11

62.54

71.55 77.00

- 10.00 20.00 30.00 40.00 50.00 60.00 70.00 80.00 90.00 100.00

補投單10% 補投單20% 補投單30%

一般單MFT (EDD) 補投單MFT (EDD(補單優先)) 一般單MFT (DBR)

圖11 比較補投單比重對法則的影響 (單位:天)

圖 12 補投單回覆交期誤差之比較圖,為 DBR 排程機制與補投單優先之 EDD 排程機制,針對補投單

的回覆交期誤差都在2 天左右明顯的優於 EDD 排程機制。

這表示如果生產管理者用此兩種機制來回覆補投單之交其,較能有效的掌握補投單的達交狀況。

五、結論

LED 晶粒製造(LED-CM)為 LED產業中重要的一環,晶粒製造過程由於製程技術的特殊性,容易

造成晶粒光電性規格不均勻及良率不穩之問題,一般LED 晶粒 出符合客戶需求規格的比率約只有產

6~8 成左右。因而在接單式生 時,當一張訂單出現 出不足時,生管必須以補投料方式,補投該訂單產 產

所缺的不足量。雖然客戶能體諒LED-CM 製程 出不穩之特性,但會要求補投單必須以急單的 工方產 趕

式生產,以減低該訂單產出不足的影響。因此LED-CM 廠在大量補投單的 工壓力下,如何確保這些趕

補投單的如期出貨以及如何降低這些補投單對正常單交期的衝擊,即為LED-CM 廠所面對的挑戰。

本研究提出了一套DBR 的 化模式以改善強 LED-CM 廠所面對的補投單挑戰。這套 DBR強化模式

主要強化了補投單在DRUM 的合理 單及在投料計畫的立即投料需求。其次本計畫亦設計了模擬實驗插

以驗證本模式的可行性及效益。實驗結果證明,不論在平均流程時間或補投單交期回覆誤差等績效,

(16)

本模式都優於EDD 及 EDD+補投單優先模式。

圖12 補投單回覆交期誤差之比較 (單位:天) 五、結論

LED 晶粒製造(LED-CM)為 LED產業中重要的一環,晶粒製造過程由於製程技術的特殊性,容易

造成晶粒光電性規格不均勻及良率不穩之問題,一般LED 晶粒 出符合客戶需求規格的比率約只有產

6~8 成左右。因而在接單式生 時,當一張訂單出現 出不足時,生管必須以補投料方式,補投該訂單產 產

所缺的不足量。雖然客戶能體諒LED-CM 製程 出不穩之特性,但會要求補投單必須以急單的 工方產 趕

式生產,以減低該訂單產出不足的影響。因此LED-CM 廠在大量補投單的 工壓力下,如何確保這些趕

補投單的如期出貨以及如何降低這些補投單對正常單交期的衝擊,即為LED-CM 廠所面對的挑戰。

本研究提出了一套DBR 的 化模式以改善強 LED-CM 廠所面對的補投單挑戰。這套 DBR強化模式

主要強化了補投單在DRUM 的合理 單及在投料計畫的立即投料需求。其次本計畫亦設計了模擬實驗插

以驗證本模式的可行性及效益。實驗結果證明,不論在平均流程時間或補投單交期回覆誤差等績效,

本模式都優於EDD 及 EDD+補投單優先模式。

六、參考文獻

1. 呂紹旭,台灣LED 磊晶片、晶粒廠商募資擴 不斷持續高成長可期,產 PIDA產業分析(2010)。

2. 吳鴻輝與李明峰,「LED 晶粒廠接單模式與生 管理模式探討」,產 2011 兩岸工業工程與管理學術

討會

研 (光碟片),12 月 10 日,明新科大(2011)。

3. 吳鴻輝與李榮貴,限制驅導式現場排程與管理技術,修訂版,全華科技圖書股份有限公司(2001)。

4. 潘志聖、徐潤忠與許家禔,「成功案例分享」,TOC 訂單履約管理機制整體解決方案 討會,研 2004

年5 月 27 日,煙波飯店。

5. 簡禎富、鄭家年與李佶玟,「LED 晶片投片與採購決策及營運決策支援系統」,2011 兩岸工業工程

與管理學術研討會(光碟片),12 月 10 日,明新科大(2011)。

6. Atwater, J.B. and Chakravorty, S.S., “A study of the utilization of capacity constrained resources in drum buffer rope systems,” Production and Operation Management, 11, 2, 259-273(2002).

(17)

7. Betterton, C. E. and Cox III, J. F. , “Espoused drum-buffer-rope flow control in serial lines: A comparative study of simulation models,” International Journal of Production Research, 117, 66-79(2009).

8. Bhardwa, A., Gupta, A. and Kanda, A., “Drum-Buffer-Rope: The Technique to Plan and Control the Production Using Theory of Constraints,” World Academy of Science, Engineering and Technology , 69, 103-106(2010).

9. Chakravorty, S. S., “An evaluation of the DBR control mechanism in a job shop environment,” The International Journal of Management Science, Omega 29, 335-342(2001).

10. Corbett, T. and Csillag, J. M., “Analysis of the effects of seven drum-buffer-rope implementations,”

Production and Inventory Management Journal, Third/Fourth Quarter, 17-23(2001).

11. Goldratt, E.M., The Goal, 2nd revised edition, North River Press, Croton-on-Hudson, New York(1986).

12. Schragenheim, E. and Ronen, B., "Drum-Buffer-Rope Shop Floor Control," Production and Inventory Management Journal, Third Quarter, 18-23(1990).

13. Wu, H. H., Chen, C. P., Tsai, C. H. and Yang, C. J. , “Simulation and scheduling implementation study of TFT-LCD Cell plants using Drum–Buffer–Rope system,” Expert Systems with Applications, 37,8127- 8133(2010).

14. Wu, H. H., Chen, C. P., Tsai, C. H., Tsai, S. C. and Lee, C. J., “Simulation and Implementation Study of Robust Drum-Buffer-Rope Management System to Improve Shop Performance,” International Journal of Humanities and Social Science, Vol.1, No.1 , pp. 14-22(2011).

15. Wu, H.H., Li, M. F. & Hsu, T. F., “An order fulfillment model for the LED chip manufacturing plant,”

Advanced Materials Research, Vols. 694-697, pp 3446-3452(2013).

16. Wu, H.H., Li, M. F. & Yeh, C. H., “A study of the material re-issued models for an insufficient output of the LED chip manufacturing plant,” Advanced Materials Research, Vols. 694-697, pp 3434-3440(2013).

(18)

國科會補助專題研究計畫成果報告自評表

請就 研 究內容與原計畫相符程度、達成預期目標情況、 究成果之學術或應用 研 價 值 (簡要 敘 述成果所代表之意義、價 值 、影響或進一步發展之可能性)、是否 適合在學術期刊發表或申請專利、主要發現或其他有關價 值 等,作一綜合評估 。 1. 請就 究內容與原計畫相符程度、達成預期目標情況 研 作一綜合評估

□ 達成目標

□ 未達成目標(請說明,以 100 字 限) 為

□ 實驗失敗

□ 因故實驗中斷

□ 其他原因 說明:

2. 研 究成果在學術期刊發表或申請專利等情形:

論文:□已發表 □未發表之文稿 □撰寫中 □無 專利:□已獲得 □申請中 □無

技轉:□已技轉 □洽談中 □無 其他:(以 100 字 限) 為

1. Horng-Huei Wu, Ming - Feng Li and Cheng-Hsin Yeh, 2013, “A Study of the Material Re- issued Models for an Insufficient Output of the LED Chip Manufacturing Plant,” Advanced Materials Research Vols. 694-697, pp 3434-3440.

2. Horng-Huei Wu, Ming-Feng Li and Tzu-Fang Hsu, 2013, “An order fulfillment model for the LED chip manufacturing plant, “Advanced Materials Research, Vols. 694-697, pp 3446- 3452.

3. Horng-Huei Wu, Hung-Hsuan Wu, and Ming - Feng Li, 2013, “A Study of the Shifting MO Model for an Insufficient Output of the LED Die Manufacturing Plant, “ Global Business &

International Management Conference Journal, Vol.6, No.2, pp.46-55.

(19)

3. 請依學術成就、技術創新、社會影響等方面,評估 究成果之學術或應用價 研

(簡要 述成果所代表之意義、價 、影響或進一步發展之可能性)(

值 敘 值 以

500 字 限) 為

(1) 對於學術 究、國家發展及其他應用方面之貢獻:研 ‧建立 DBR 於 LED 晶粒製造廠之 究模式。研

‧建立應用 TOC/DBR 在 LED 晶粒製造廠之模式。

‧將 究結果提供給國內工廠有意導入研 TOC/DBR 方法之業界作 參考。為

(2) 發表本計畫之成果於國內外 討會與國際期刊研 (SCI),提供給學術界與業界參考。

(20)

國科會補助專題研究計畫項下出席國際學術會議心得報告

       日期:102 年 4 月 1 日

       

1、 參加會議經過

2013 The International Conference on Manufacturing Science and Engineering was held in Dalian, China.

The conference served as important forum for the exchange of ideas and information to promote understanding and cooperation among the Industrial Engineering and System Management.

In the conference, I published a paper entitled “A Study of the Material Re-issued Models for an Insufficient Output of the LED Chip Manufacturing Plant”. In addition, some other topics about management and engineering were all impressed me very much.

2、 與會心得

The international conference provided good opportunities for exchanging information and ideas among scholars in relative research fields. And I feel great for that I can also know other countries student's research.

3、 考察參觀活動(無是項活動者省略) no

4、 建議 no

5、 攜回資料名稱及內容

Conference Program: 2013 The International Conference on Manufacturing Science and Engineering 6、 其他

A Study of the Material Re-issued Models for an Insufficient Output of the LED Chip Manufacturing Plant

計畫編號 NSC 101-2221-E-216-016-

計畫名稱 限制驅導式排程技術(DBR)於 LED 晶粒製造廠應用之 化模式 究強 研

報告人姓名 葉承鑫 服務機構

及職稱

中華大學工業管理學系 碩士班研究生

會議 時間

102/3/30~3/31 會議 地點

Dalian, China

會議 名稱

(中文)2013 年製造科學與工程國際 討會研

(英文) 2013 The International Conference on Manufacturing Science and Engineering 發表

論文 題目

(中文) LED 晶粒製造廠 出合工單量不足時的補投單模式之探討產

(英文)A Study of the Material Re-issued Models for an Insufficient Output of the LED Chip Manufacturing Plant

(21)

Horng-Huei Wu

1, a

, Ming - Feng Li

2, b

and Cheng-Hsin Yeh

3, c

1Department of Business Administration, Chung Hua University, Hsinchu City, Chinese Taipei

2,3Institute of Industrial Management, Chung Hua University, Hsinchu City, Chinese Taipei

ahhwu@chu.edu.tw, bu8803042@chu.edu.tw, cm10121003@chu.edu.tw

Key Words: Material re-issue strategy, Expediting model, Rescheduling, Unstable production output, LED chip manufacturing

Abstract. The LED chip manufacturing (LED-CM) is an important process in the LED supply chain. The make-to-order production strategy is a general production model for the LED-CM plants to satisfy the variety requirement of their customers. Because the unstable production output of the LED-CM plants, the effective order fulfillment is low and variety. Under the severely competitive pressure, the production planner will thus confront the issue of material re-issued for an insufficient output of an MO. The purpose of this effort is to meet the required due-date of this MO and to reduce the unnecessary inventory of by-product chips.

Therefore, a study of the material (EPI) re-issued MO model for an insufficient output of the LED chip manufacturing plant is proposed in this paper. Three material re-issued models are proposed and detailed discussed first. A real-life LED-CM case is then utilized to demonstrate and evaluate the application and effectiveness of these models. A simulation model is also designed to complete the further experiment.

Introduction

The major processes of the LED industry compose of four parts: material substrate, upstream, midstream and downstream[1]. Substrate is an important material to produce the Epitaxy wafer (EPI). Two substrate materials are Sapphire substrate and Silicon substrate. The manufacturing process of LED Chip Manufacturing (LED-CM) plant is in the midstream. The production process of a LED-CM plant not only complex but also unstable, and the quality of the LED chips will determine the function of the subsequent applications. Therefore, the LED-CM plants are important in the LED supply chain and the focus of this paper.

The make-to-order production strategy is a general production model for the LED-CM plants to satisfy the variety requirement of their customers. However, the special features of the unstable production output and a product composed of the chips of different feasible Bins exist in the LED-CM plant. The production planner will confront the issue of effective inventory control and exact due-date performance under the severely competitive pressure. An order fulfillment model was therefore presented by Wu et al.[2]. Therefore a re- issued model for order fulfillment is further proposed in this paper to improve the due-date performance of the LED-CM plants.

The specificity and order fulfillment issues of the LED-CM plants are first discussed in the second section.

The material re-issued models are then presented in the third section. A real-life LED-CM case is also utilized to demonstrate and evaluate the application and effectiveness of the proposed models.

The Order Fulfillment Issues in the LED-CM Plant

(22)

Basically, the input material of LED-CM plant is EPI and the final products are LED chips. The manufacturing process of LED-CM can be divided into two parts, i.e., frontend process and backend process.

The major purpose of the frontend process is to manufacture the electrical function into EPI. And the functions of the backend process are chip probing, cut and break, sort (by different bins) and package. An EPI can be cut or broken into several or several ten thousands LED chips depending on its size. Although the specifications of the LED chip are several, such as structure, size, electrical functions, voltage or customized requirements etc., the major specification is the electrical functions, i.e., lightness and wavelength. And these two functions are only utilized as the product specification of the order in the following discuss. The notations utilized in this paper are as follows:

I: The types of the product specification.

i: The index of a product specification, i =1,2,,I.

lli: The lower bound light grade of product specification i , i =1,2,,I.

lui: The upper bound light grade of product specification i ,i =1,2,,I.

J: The maximum range of light grade.

j: The index of light grade for a Bin, j=1,2..J.

bj: The chip inventory in the Bin indexed with the jth grade light and kth wave band,j=1,2..J.

ri: The required quantity of an order of the product specification i, i=1,2..I.

gi: The gross required quantity of an order of the product specification i, i=1,2..I.

pij: The planning output distribution of the Bin indexed with the jth grade light, i=1,2,..I, j =1,2,,J.

qi: The issued quantity for product specification i, i=1,2,..I.

eij: The actual output of the Bins for the order product specification i, j =1,2,,J.

ti: The total production output probability of the feasible Bins for a product specification i, i=1,2,..I.

The Feasible Bins for a Product Specification. The major product specifications compose of lightness and wavelength. For example, for the yellow light, the wavelength is between 584nm and 594nm and lightness is between 100mcd and 300mcd. Therefore, besides the color of the LED chips, the specification of order must further define clearly the range of wavelength and lightness. For example, if users want more yellow and light LED chips, their specifications of wavelength will be between 586nm and 588nm and lightness between 200mcd and 250mcd. However, for the yellow LED chips, the wavelength of another user requirement will be between 588nm and 590nm and lightness between 150mcd and 200mcd.

In the LED-CM plant, a user product specification is then transferred to be feasible Bins[1-3]. A Bin is defined by the light grade and waveband. However, waveband is ignored in this paper to reduce the complexity of Bin. The detailed Bins descriptions please refer Wu et al[2]. The lightness of a color LED chip is divided into several light grades (i.e., J grades) depending on the requirements of users. For example, the yellow lightness is between 100mcd and 300mcd and is divided into 6 grades as shown in Table 1. Therefore, the feasible Bins for an order with product specification i are defined in Eq. 1.

FBi ={bj, lli≦j≦lui}, i=1,2, ..I (1)

The shipment chips of an order are then based on the chips of these feasible Bins. For example, the specifications of an order are yellow LED chips and lightness is between 80mcd and 160mcd. Because the quality of Bins must narrow than the specification of order, the feasible Bins for this order are light grade 2~3

(23)

(i.e., lli=2 and lui=3), as shown the shadow boxes in Table 1. Therefore, the inventory chips of these feasible Bins for a product specification are as shown in Eq. 2.

i

i

lu ll j

j

i b

v , i=1,2, ..I (2)

Table 1. An example of Bin definition for yellow LED chips.

Bin # Current Date: 10/19

Light Grade j 1 2 3 4 5 6

Lightness [mcd] 50~100 100~15 0

150~20 0

200~25 0

250~30 0

300~350 Chips in Bin [×K

grains] bj

100 5 5 20 5 50

If the inventory chips of these feasible Bins are sufficient for this order, the shipment of this order is fulfilled by these inventory chips. Otherwise, a production plan or MO of the 1st product specification will be issued to produce for this order.

Production Output Distribution of Bins. Because the status of process stable or not in the frontend process will determine the electrical distribution of those several thousand chips in an EPI, the electrical functions of the production output of chips in an EPI will be gradient distribution. In other words, the electrical function of each chip in an EPI is different. Therefore, these chips with non-identical electrical function are classified into different Bins. Basically, the different Bins distributions will be determined by the different process parameter or EPI. Therefore, the different product specification identifications are provided by the R&D or engineering department to get the optimal Bins distributions, as shown the example in Table 1. The total production output probability of these feasible Bins for a product specification i is thus as shown in Eq. 3.

i

i

lu ll j

ij

i p

t , i=1,2, ..I (3)

When the inventory chips in the feasible Bins are insufficient for an order, a production plan or MO of a feasible product specification is required. The gross required quantity of an order i is first determined by Eq.

4.

i i

i r v

g , i=1,2,..I (4) Then the production quantity of chips for this order is determined by the gi and ti, as shown in Eq. 5.

i i

i t

q  g , i=1,2,..I (5)

Hit Target of a MO. Because the status of process stable or not in the frontend process will determine the electrical distribution of those several thousand chips in an EPI, the actual distribution of the production output of chips in an EPI will not be same as the planned. The Hit Target (HT) for a MO is the ratio of the production output chips which are in the feasible Bins. If the ratio is greater than the value of ti, a higher HT is

(24)

performed for this MO. For example, the lightness specifications of two MOs are both 150~200mcd.

Therefore, if the HT of a MO is greater than the planed level, this order can be delivered on time. However, the planner will get too much excess chips in the Bins. On the contrary, if the HT of a MO is less than the planed level, a shortage will occur for this MO. An expediting replenishment plan must be switched on by the planner to catch the requirement of this MO[4]. The determination of an expediting replenishment is by the remaining time before the due-date of the order of product specification i. The first alternative is to re-create a MO and re-issue it to replenish the shortage quantity. And, the second alternative is to select a shifting MO from the WIP to replenish the shortage quantity. The this paper studies the material re-issued models for the first alternative. Although the order release field has been a popular topics in the literature[5-9], the re-issued model is little studied. The purpose of this paper is therefore to present the material re-issued models for an insufficient output of a MO in the LED-CM plant.

Material Re-issued Models for LED-CM Plants

Three material re-issued models are proposed in this section. The first model is to issue the shortage quantity only (SQO). The second model is to issue the shortage quantity and loss buffer. (SQLB). The third model is to consider the global WIP inventory status after the probing process. The re-issued quantity depends on the global shortage quantity(GSQ).

SQO Model. Let si is the shortage quantity of the order of product specification i. Then si is as:

i

i

lu ll j

ij i

i r e

s (6) Then, the inventory chips of the feasible Bins for this product specification must be reviewed again. And si is revised as:

i

i

lu ll j

j i

i s b

s max 0, (7)

Therefore, if si is greater than zero, the re-issued quantity (rqi) of SQO model is as :

i i

i t

rq s (8)

SQLB Model. The loss buffer (L) is considered in the SQLB model to confirm the enough output of the re- issued quantity. Basically, the loss buffer is the average loss ratio in which the HT is lower than ti in the last some periods. Therefore, if si is greater than zero by utilizing (7), the re-issued quantity (rqi) of SQLB model is as:

i i

i t

L rq s (1 )

(9)

GSQ Model. The global WIP inventory status after the probing process is considered in the GSQ model. The global shortage quantity(gsi) is equal to the required quantity minus the output of the orders with the same product specification and in the process after probing. Let wi is the current process of the order of the product

(25)

specification i, i=1,2..I. The gsi is then as shown in Eq. 10.

 

}

| {

) (

process probing w i i

lu ll j

ij i

i

i

i

i

e r

gs (10) The re-issued quantity depends on the global shortage quantity(gsi). If if gsi is greater than zero, the re-issued quantity (rqi) of GSQ model is as :

i i

i t

rq  gs (11)

A Case of a LED-CM Plant

A LED-CM plant located in the Hsinchu Science Park of Taiwan is utilized to illustrate the above models. The lightness specification is only considered. The lightness is further divided into six specifications, i.e., A~F. as shown in Table 2. The range of light grades and production output distribution are also shown in Table 2. The normal or hot run cycle times (CT) for different process are shown in Table 3. Suppose today is 10/19 and there are 28 lots in production. Table 4 shows the information of these lots. Lot # 001~009 have completed the Probing process and the HT is known. The HT of other lots are still unknown. For example, the HT quantity of lot #1 is 263K and greater than the required quantity 250K. Therefore, this lot can be delivered on time.

However, lot #9 met some troubles. The HT quantity of lot #1 is only 867K and less than the required quantity 1000K. Therefore a re-issued lot is required as soon as possible. Based on the proposed re-issued models, the re-issued lot is shown in follows respectively.

SQO Model. Because si is 133K and the inventory chips of the feasible Bins for this product specification are 25K, si max

0,133K 25K

108K. Therefore, by utilizing (8), the re-issued quantity

. 180 6 . 0 /

108K K

rqi That is an expediting lot # 009-1 is issued and its quantity is 108K and duedate is 11/6. Therefore, the lot #9 is split into two lots. The first lot is still # 009 but its quantity is revised to be 892K.

The second lot is #009-1 and its quantity is 108K and duedate is 11/6.

SQLB Model. The si is same as SQO model. Suppose the loss buffer (i.e., L) is equal to 20% by past experience. Therefore, by utilizing (9), the re-issued quantity rqi 108K(1 0.2)/0.6216K. That is an expediting lot # 009-1 is issued and its quantity is 108K and duedate is 11/6. Therefore, the lot #9 is split into two lots. The first lot is still # 009 but its quantity is revised to be 892K. The second lot is #009-1 and its quantity is 108K and duedate is 11/6.

Table 2. An example of product specifications and their output distribution.

ID of product specification

[i]

Light grade [j]

Feasible EPI

Output distribution of Bin Total production output probability

[ti]

1 2 3 4 5 6

A 2~3 X 15% 30% 30% 15% 5% 60.00%

B 3~4 Y 5% 15% 30% 30% 15% 5% 60.00%

C 4~5 Z 5% 15% 30% 30% 15% 60.00%

D 2~4 X 15% 30% 30% 15% 5% 75.00%

(26)

E 3~5 Y 5% 15% 30% 30% 15% 5% 75.00%

F 4~6 Z 5% 15% 30% 30% 15% 75.00%

Table 3. The cycle time for the different process of LED-CM plant.

Table 4. Lot information in production.

Current Date: 10/19 Unit: × K grains

[days] Front process Cut Probing Sorter QC Die Count

Standard CT 15 2 1.5 3 1 2.5

Hot Run CT 10.50 1.40 1.05 2.10 0.70 1.75

Lot # Hot Run [Y/N] Prod.

Spec. Required Qty Issued

Qty Issued

date DueDate Current

process Remaining

Days [Std] Remaining Days[Hot run] DueDate

[Hot Run] Output probability

[ti] HT Qty of HT

001 N B 250 417 9/27 10/22 Die Count 3 2 10/21 60.00% 63.00% 263

002 N E 400 533 9/27 10/22 Die Count 3 2 10/21 75.00% 75.00% 400

003 N C 300 500 9/27 10/22 Die Count 3 2 10/21 60.00% 62.00% 310

004 N A 500 833 9/28 10/23 QC 4 3 10/22 60.00% 65.00% 542

005 N B 1200 2000 9/29 10/24 Sorter 5 4 10/23 60.00% 61.00% 1220

006 N C 700 1167 9/29 10/24 Sorter 5 4 10/23 60.00% 66.00% 770

007 N A 300 500 10/1 10/26 Probing 7 5 10/24 60.00% 64.00% 320

008 N D 100 133 10/1 10/26 Probing 7 5 10/24 75.00% 79.00% 105

009 N C 1000 1667 10/1 10/26 Probing 7 5 10/24 60.00% 52.00% 867

010 N A 500 833 10/3 10/28 Cut 9 6 10/25 60.00%    

011 N B 500 833 10/3 10/28 Cut 9 6 10/25 60.00%    

012 N E 600 800 10/5 10/30 Front process 11 8 10/27 75.00%    

013 N C 100 167 10/5 10/30 Front process 11 8 10/27 60.00%    

014 N F 500 667 10/7 11/1 Front process 13 9 10/28 75.00%    

015 N E 100 133 10/7 11/1 Front process 13 9 10/28 75.00%    

016 N F 200 267 10/7 11/1 Front process 13 9 10/28 75.00%    

017 N B 700 1167 10/7 11/1 Front process 13 9 10/28 60.00%    

018 N A 800 1333 10/13 11/7 Front process 19 13 11/1 60.00%    

019 N C 600 1000 10/13 11/7 Front process 19 13 11/1 60.00%    

020 N D 1000 1333 10/13 11/7 Front process 19 13 11/1 75.00%    

021 N B 200 333 10/15 11/9 Front process 21 15 11/3 60.00%    

022 N B 300 500 10/15 11/9 Front process 21 15 11/3 60.00%    

023 N E 2000 2667 10/15 11/9 Front process 21 15 11/3 75.00%    

024 N D 100 133 10/18 11/12 Front process 24 17 11/5 75.00%    

025 N A 500 833 10/18 11/12 Front process 24 17 11/5 60.00%    

026 N C 200 333 10/18 11/12 Front process 24 17 11/5 60.00%    

027 N F 700 933 10/18 11/12 Front process 24 17 11/5 75.00%    

028 Y C 130 217 10/19 11/13 Front process 25 18 11/6 60.00%    

參考文獻

相關文件

Since the research scope of industrial structure optimization and transformation strategy in Taiwan is broad and complicated, based on theories of service innovation and

Prioritising improvement of service attributes using impact-range-performance analysis and impact asymmetry analysis. Hospitality and travel

Lin, “Automatic Music Genre Classification Based on Modulation Spectral Analysis of Spectral and Cepstral Features”, IEEE Trans.. on

This study collected consumer expectations and perception of medical tourism industry, with Neural Network Analysis and Mahalanobis Distance Analysis found the key to the

因此,本研究發展一套擁有企業資源規劃(Enterprise Resource Planning, ERP)知識的多人線上遊戲學習系統 (Multiplayer Online Game-based Learning System,

The issue of construction surplus soils can be solved by using it for production of CLSM (Soil-based CLSM, S-CLSM), and the effective reclamation of resources can reduce the

The issue of construction surplus soils can be solved by using it for production of CLSM (Soil-based CLSM, S-CLSM), and the effective reclamation of resources can reduce the

Different from protein domains, molecular interface motifs emphasize on binding interfaces, the cores to perform protein functions1. Furthermore, our results showed