透過您的圖書館登入
IP:18.118.150.80
  • 學位論文

貝氏神經網路的構建及其在可靠度工程上的應用

The Implementation of Bayesian Neural Networks Applying to Reliability Engineering

指導教授 : 陳雲岫
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


本研究運用貝氏神經網路(BNN)的方法來製作可靠度工程中之故障模式 影響及危害度 分析(FMECA),並且將結果與傳統倒傳遞神經網路( BPN)相比較。從實驗中我們發現,平均而言,BNN的正確性不但較BPN來 得高,學習收斂的速度亦較BPN來得快。由於BNN在 本研究中是屬於非監 督式學習的架構,為了提高網路分類的效果,建議可在選擇混合層 贏家 神經元的方法上加以改進。由於歐洲市場已經提出ISO 9000的品質規範, 所以可靠 度工程勢必日益受到重視,本文所提之整合性FMECA架構,是 神經網路在可靠度工程上另 一新的應用。

並列摘要


This thesis has used a Bayesian neural network (BNN) approach to implement to the Failure Modes Effect and Criticality Analysis (FMECA) in the reliability engineering area, and the implemented results have also compared with those of using a traditional Back-Propagation Network (BPN). From the comparison, it has been found that the accuracies of the failure mode classifications and the criticality value generalization using the BNNs are higher than those using BPNs, and the BNN learning convergence speed does the same result. The learning type of the BNN is unsupervised, and the improvement of the unsupervised learning is proposed by the determination of a winner neuron in the hybrid layer of the BNN so as to increase the classification accuracy. The novel application of using neural networks to FMECA can be one of realization of ISO 9000 quality assurance.

被引用紀錄


周湘蘭(2002)。類神經網路在多重產品需求預測上之應用〔碩士論文,元智大學〕。華藝線上圖書館。https://www.airitilibrary.com/Article/Detail?DocID=U0009-0112200611291267

延伸閱讀