A Study on the Reliability Demonstration Test Planning by Bayesian Approach

• Main Author: 洪振勝
• Other Authors: 柯煇耀
• Format: Others
• Language: zh-TW
• Published: 1998
• Online Access: http://ndltd.ncl.edu.tw/handle/64933235264625334042

碩士 === 國防管理學院 === 國防資訊研究所 === 86 === 　　Demonstration is generally the last formal reliability assurance activity required in the procurement contract after design frozen. This thesis proposes a systematic Bayesian approach for reliability demonstration test planning. First, with the merit of Bayes's theorem, prior information from development phases are considered and expressed by a prior distribution of reliability parameter. Secondly, risks are estimated to establish the required Bayesian demonstration test sampling plan. Since producer and consumer risks can be defined in several ways by this approach, practical application of test plans based upon prior distribution demands a selection of particalar risks more meaningful for the individual application. Choice of risk combinations are studied for differences in the relation between the number of units produced and the number procured by the consumer. Besides the mathematical foundation and the basic assumptions, numerical examples are given to illustrate the feasibility and application procedures for the proposed approach. It is shown that Bayesian sampling plan will be more efficient than the classical plans in terms of test time requirements. By the way, it is significant to note that there may be no testing is required by the Bayesian approach, and the equipment is judged accept/reject when the reliability achieved is definitely greater/lower than the specification requirement. These results tend to indicate that the influence of the prior information is indeed capable of meaningful, practical interpretation. Thus, the saving of required testing time and cost of the proposed approach are a very real possibility.