베이지안 기법을 이용한 품목 수준의 개발 신뢰도 최신화 방안 연구 (A Study on Updating Development Phase Components Reliability Using a Bayesian Technique)

서지정보

· 발행기관 : 한국신뢰성학회
· 수록지 정보 : 신뢰성 응용연구 / 21권 / 1호 / 90 ~ 98페이지
· 저자명 : 나일용, 김헌길

초록

Purpose: This study proposes the Bayesian model that can update the development phase reliability using the small field failure data.
Methods: A Gaussian distribution with the development phase reliability as one parameter, μ, was used as the prior distribution of reliability. Moreover, the posterior distribution was estimated by Markov Chain Monte Carlo. The standard deviation was assumed to be proportional to the development phase reliability. That optimal proportional value, m, was derived from the simulation.
Results: The effectiveness was verified by applying the proposed method to real cases. The proposed method involves only moderate fluctuations depending on the amount of data considering that only a fraction of data was used. Comparatively, the results of the existing methods showed large fluctuations.
Conclusion: We expand the reuse area of the development phase reliability, where a high cost and effort was invested, Moreover, we propse a procedure to calculate meaningful reliability with a small amount of data. Moreover, an intuitive and familiar Gaussian distribution can double its effectiveness.

영어초록

Purpose: This study proposes the Bayesian model that can update the development phase reliability using the small field failure data.
Methods: A Gaussian distribution with the development phase reliability as one parameter, μ, was used as the prior distribution of reliability. Moreover, the posterior distribution was estimated by Markov Chain Monte Carlo. The standard deviation was assumed to be proportional to the development phase reliability. That optimal proportional value, m, was derived from the simulation.
Results: The effectiveness was verified by applying the proposed method to real cases. The proposed method involves only moderate fluctuations depending on the amount of data considering that only a fraction of data was used. Comparatively, the results of the existing methods showed large fluctuations.
Conclusion: We expand the reuse area of the development phase reliability, where a high cost and effort was invested, Moreover, we propse a procedure to calculate meaningful reliability with a small amount of data. Moreover, an intuitive and familiar Gaussian distribution can double its effectiveness.

참고자료

· 없음