透過您的圖書館登入
IP:18.223.160.61
  • 學位論文

利用客退資料作雲端產品壽命分析-以KVM 切換器為例

Lifetime Assessment of Cloud Service Products by RMA Data– KVM Switch as Example

指導教授 : 黃乾怡
若您是本文的作者,可授權文章由華藝線上圖書館中協助推廣。

摘要


資訊產業繁榮發展,以致於電腦設備、廠房中機台以及資料中心保存之資料量同時增長。企業相關之硬體設備的建置日趨繁多,控制人員無法準確地掌握每一台伺服器的運作。並衍生辦公空間成本提高與耗電量增加等問題。因此,透過雲端系統連線與相關設備使用,即鍵盤/螢幕/滑鼠切換器(Keyboard /Video display /Mouse Switch;KVM Switch),連接控制人員在遠端同時進行多台設備操作。於是能夠精簡伺服器架設空間,並達到有效率管理及維護的目的。本研究建構存活分析法應用於客退資料分析,經由韋伯配適失效函數得期望壽命為22,459天,並估計產品使用一年將有1%的產品發生失效。其次,藉由Naive失效數量預測模型驗證,得知選擇越長期間為訓練樣本之預測能降低預測誤差,且欲進行預測之期間不宜長於訓練樣本資料蒐集期間,以確保準確之預測結果。因此,研究中利用所有資料蒐集期間(共22個月)之客退資料為訓練樣本,預測未來12個月及24個月之失效數量上限,結果分別為269及517。最後,利用決策樹方法分析產品各屬性對其使用壽命之影響,得知KVM切換器產品於生產線上製程控管之優劣將顯著影響產品壽命,此外,特定代工廠所生產產品之平均壽命較短,乃由於其所選用之電子元件及印刷電路板(Printed Circuit Board, PCB)供應商品質較差所致。

並列摘要


In recent year, the rapid growth on industry of information technology leads to the demand of server is used by manufacturer or information-center growing. The requirement of server increased every year. The operators are difficult to maintain all servers on the short time. And the space cost and exaltation of energy become a problem. The keyboard /video display /mouse switch (KVM Switch) is used to resolve this problem. The KVM switch system links the operator and the server via cloud system to control the multiple servers by remote connection. With this system, the information center can achieve the purposes of efficient management and maintenance. The study applies survival analysis on return merchandise authorization (RMA) analysis. We find that the failure function can be fitted on Weibull distribution. The expected life is evaluated 22,459 days. And we note that the KVM switch that be used one year on customer side will occur failure at a rate of 1 percentage. And then, the verification is performed via Naive quantity of failure forecast model. We find that the forecast error will decrease with the time period on collecting train data increasing. Therefore, this study collects RMA on 22 months as the training sample to forecast the maximum number of failure on 12-24 months coming. The last, the decision trees be used to analyze the effect of the product attribution on lift time. The result shows that (1) the attribution “process control” impact the life time of KVM switch; (2) the product come from some specific contract manufacturer have lower life time. The reason may ascribe to material inspection.

參考文獻


5. 劉淑媛,握力和癌症分期在老年癌症病人的存活分析,碩士論文,國立台灣大學護理學研究所,台北,2011。
1. 朱道鵬,「TFT-LCD面板壽命研究」,品質學報,第十七卷,第三期,2010。
2. 李信昱,利用類神經網路預測組織型客戶退貨數量之研究-以TFT-LCD面板廠為例,碩士論文,國立成功大學工業與資訊管理學系,台南,2009。
7. 賴禹廷,以蒙地卡羅模擬評估型I設限預測模型並推估保固風險-以DOM產品為例,碩士論文,國立臺北科技大學工業工程與管理研究所,台北,2009。
9. B. Boukai, "Bayes sequential procedure for estimation and for determination of burn- in time in a hazard rate model with an unknown change-point," Sequential Analysis, vol. 6, no. 1, 1987, pp. 37-53.

被引用紀錄


王鈺琪(2017)。蜂蜜品牌創新策略之研究〔碩士論文,國立臺中科技大學〕。華藝線上圖書館。https://www.airitilibrary.com/Article/Detail?DocID=U0061-1801201713373700

延伸閱讀