Hier finden Sie wissenschaftliche Publikationen aus den Fraunhofer-Instituten.

Reliabilty prediction with the aid of bayesian statistics

: Kempf, Michael

Fernandez, J.:
12th Annual International Conference on Industrial Engineering - Theory, Applications and Practice 2007 : Cancun, Mexico, November 4-7, 2007
Cancun, 2007 (International Journal of Industrial Engineering)
ISBN: 978-0-9654506-3-8
Annual International Conference on Industrial Engineering <12, 2007, Cancun>
Fraunhofer IPA ()
Bayesian statistic; Bayes-Entscheidungstheorie; Produktlebenszyklus; reliability; Zuverlässigkeitsvoraussage

The classical methods analyzing the reliability of a technical system use operating and test data respectively. In industrial application this is not possible anymore, since there are not enough data available and running an appropriate number of tests is not feasible or too costly. Obviously there is a need to incorporate additional knowledge in the reliability analyses like experience from previous development projects or information on similar products. The methods of Bayesian Statistics seem to be suitable to achieve this goal. They allow the combination of test data on the one hand with additional sources of knowledge, which are relevant for reliability analyses, on the other hand. This paper presents a method for pooling different pieces of knowledge from several human experts and calculating a probability density function reflecting both the available data and the prior knowledge of the experts. Thereof estimations of reliability parameters can be derived to predict a product's potential lifetime.