Track 1 Session 9
8:00 to 9:00 a.m. Friday March 27, 2009

Reliability Prediction Models Based on Accelerated Testing and In-Service Field Return

It is currently more and more difficult to validate the reliability models of electronic components. Indeed, an intrinsic antagonism appears between a high reliability level and the information weakness issued both from the test results and the experienced returns. In order to overcome this problem, we propose to reduce the model uncertainty by using diverse available information: physical (model structure), experimental (accelerated tests) and statistical (data processing). This methodology is applied to an electronic sensor having two failure modes. Its resulting reliability model is validated by a comparison between operation and expected failure rates observed and calculated for the mission profile of a civil aircraft. Taking into account the parametric uncertainties, such a comparison must be considered from a probabilistic point of view.

Key Words: Accelerated Tests, Reliability Prediction, FIDES, Mission Profile, NHPP, Weibull

Franck Bayle
Thales Avionics
France