| dc.contributor | 統計學系 | |
| dc.creator (作者) | 楊素芬 | zh_TW |
| dc.date (日期) | 2014 | |
| dc.date.accessioned | 5-Aug-2015 12:08:55 (UTC+8) | - |
| dc.date.available | 5-Aug-2015 12:08:55 (UTC+8) | - |
| dc.date.issued (上傳時間) | 5-Aug-2015 12:08:55 (UTC+8) | - |
| dc.identifier.uri (URI) | http://nccur.lib.nccu.edu.tw/handle/140.119/77378 | - |
| dc.description.abstract (摘要) | 對連接器（connector)的端子金膜厚之變異導致成本的增加是連接器的業者在 面對全球的競爭下迫切要解決的重要問題。實務上，製程工程師對生產製程和生產 條件完全是主觀判斷而無科學的生產製程調整方法。此外，品質工程師對製程上金 膜厚之監控是採用傳統的X-bar和S管制圖追蹤金膜厚之平均數和變異數變化， 完全忽略金膜厚會隨著金濃度消耗和時間而降低。本二年期研究計畫，乃針對金膜 厚之變異導致成本增加此重要問題提出改善和解決的科學方法。在第一年計畫期 間，我們(1)將就如何使製程之六面金膜厚維持穩定（即金濃度不嚴重偏離特定的目 標值），決定調整製程金鹽（或金濃度）的時間及調整金鹽量（或金濃度）的最適動態調 整決策方法。接著，預測此最適狀況下的六面金膜厚品質及平均損失，並和現況比 較以了解績效。（2)建立六面金膜厚偏離目標值的損失函數(loss function)，再就如何 使製程之六面金膜厚平均損失最小化提出調整製程金鹽（或金濃度）的時間及調整金 鹽量（或金濃度）的最適動態調整決策方法。接著，預測此最適狀況下的六面平均損 失及對應的金膜厚品質，並和現況比較以了解绩效。此外，（1)和(2)的結果將被比較， 以選擇最適動態調整決策方法及結果。在第二年計畫期間，我們將(1)以類神經網路 技術(ANN)決定六面金膜厚在製程中隨著金濃度消耗及時間的最適時間數列函數關 係(profile)，再以bootstrap方法找出其信賴帶和變異數管制圖。此信賴帶和變異數 管制圖即可用以追蹤未來金膜厚profile是否因製程失控(out-of-control process)而發 生改變，（2)製程在統計管制下，以類神經網路技術(ANN)決定六面金膜厚的損失在 製程中隨著金濃度消耗及時間的最適時間數列函數關係，再建立其損失信賴帶和變 異數管制圖。此損失信賴帶和變異數管制圖即可用以追蹤未來損失profile是否因製 程失控而發生改變，（3)第一年計畫，製程在(1)下，製程中金膜厚之變異被減少並維 持穩定，而損失亦然。唯製程仍需要做追蹤以儘速偵測出失控之製程。製程中管制 金膜厚損失應該優於管制金膜厚，是以我們將提出追蹤金膜厚損失之新管制方法。 亦即第二年計畫在提出有效監控金膜厚之變異及成本之統計製程管制方法。 | |
| dc.description.abstract (摘要) | The variation of the film thickness and associated increase in cost of the halo-gold are vital problems to computer connector producers. However, no scientific halo-gold adjustment policy and appropriate process control approach are currently available. Thirty samples of size four of six-side thickness on the film and the corresponding gold concentration are measured and currently collected every thirty minutes in twenty-four hours of a day. The correlation among the six-side thickness and the relationship between the six-side thickness and corresponding gold concentration are investigated. In the proposed two-year study, we will study methods to (1) control the film thickness as close as possible to the target and to reduce the cost of halo-gold consumption by proposing an artificial neural network (ANN) approach and the optimization technique for both the individual six-side thickness and the corresponding loss to determine the optimal adjustment time and the adjustment amount of halo-gold (or gold concentration), which is our control factor; (2) maintain the film thickness and loss of six-side thickness in statistical control and detect out-of control conditions by using the methods of multivariate analysis-PCA, ANN and bootstrap approaches to construct the variance control charts and confidence region of film thickness profile and loss profile; (3) to detect the occurrence of special causes of variation in (2) above, monitoring the loss for six-side thickness should be preferred to monitoring the individual six-side thickness, and we will propose the distribution-free control charts. The data analyses of the film thickness and loss using the proposed approaches help improve the performance of film thickness quality and reduce cost. | |
| dc.format.extent | 144 bytes | - |
| dc.format.mimetype | text/html | - |
| dc.relation (關聯) | NSC102-2118-M004-005-MY2 | |
| dc.relation (關聯) | PA10301-0236 | |
| dc.title (題名) | 金膜厚製程的最適動態調整決策與剖面追蹤 | zh_TW |
| dc.title.alternative (其他題名) | Optimal Dynamic Adjustment Policy and Profile Monitoring for Film Thickness Process | |
| dc.type (資料類型) | report | en |
