透過您的圖書館登入
IP:18.224.39.32
  • 學位論文

應用支援向量機於自相關製程下即時辨識管制圖非隨機形狀之研究

Real-time Recognition of Control Chart Patterns in Autocorrelated Processes Using Support Vector Machine

指導教授 : 顧瑞祥
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


在現今全球化發展的趨勢下,由於消費者型態的改變及消費者主義的抬頭,品質與檢驗的重要性急速竄升。產品的品質取決於生產技術能力和生產線之控管,適合的品質管制技術方能產出穩定之水準。藉由嚴格且精密的控管,能提早發現產品線上的異常,提升判斷生產作業狀態的精確度。統計製程管制 (statistic process control, SPC) 工具當中屬管制圖品質之監控運用最為廣泛,管制圖對於製程的控管有極佳的功效,但是,在此領域當中幾乎所有的研究皆假設製程資料為常態的情況,在此假設前提下,管制能力皆能滿足製程監控之標準。然而,由於生產型態之快速轉變,使得實務上許多製程資料皆屬具有相關性之情形,造成此製程前提之假設在許多實際的生產業界當中是無法成立的,對於此假設的製程狀態,對管制圖之辨識系統有很大的影響,如此限制了傳統SPC方法於應用上之有效性。近年來,人工智慧之方法已被廣泛應用在許多領域方面,本研究試以利用人工智慧之新穎工具—支援向量機 (support vector machine, SVM) 做為改善品質監控之手法。製程的自相關環境下將有助於現今快速轉變的生產型態。利用支援向量機來辨識製程之異常,更優於傳統統計製程管制手法。因此,本研究應用支援向量機在自相關製程中即時且準確辨識管制圖非隨機形狀的模擬系統。結果顯示在總體正確辨識率上達到 90% 以上之績效,實際模擬證實支援向量機在自相關製程管制圖圖形辨識上,具有良好辨識能力。

並列摘要


Under nowadays globalization development tendency, because of the consumer state's has been changed, the quality and the examination rise rapidly. The quality of product is determined by the production technology and on line control ability. The suitable quality control technology could deliver the standard stably. Control chart pattern is the most widespread and has good effect the system of regarding used to quality control tool. However, nearly all studies in this field assume that the process data are normal distribution. Unfortunately, this assumption is not even approximately satisfy in many manufacturing processes. The situation has limited the traditional SPC method in application of efficiency. Making the regulation under the autocorrelated process environment to be helpful to transformation production state in the nowadays. In the system of regulation using support vector machine exceptionally to surpass the traditional statistics process control technique. Hence, this paper proposes support vector machines system that can effectively recognize CCPs in real-time for the processes which exhibit various levels of autocorrelation. The result has showed that achieves 90% above the achievements in the overall correct identification rate. The simulation result confirms the support vector machine recognizes in the autocorrelation process data has good recognition ability of control charts.

參考文獻


1.王景南 (2003),多類支援向量機,元智大學資訊管理研究所,碩士論文。
2.王駿發、林博川、王家慶和宋豪靜 (2005),「以支援向量機為基礎之新穎語者切換偵測演算法」,計算語言學通訊,第十六卷。
3.呂奇傑 (2005),「支援向量機於資料探勘中分類模型之應用」,產業管理創新研討會,頁687-693。
4.李羿儒 (2003),財務危機預警模式之再探討-運用支援向量機及Logit 模式,國立高雄第一科技大學財務管理所,碩士論文。
5.林維謙 (2007),基於支援向量機之人臉偵測與人臉辨識,世新大學資訊管理學系,碩士論文。

延伸閱讀