ENBIS: European Network for Business and Industrial Statistics
Forgotten your password?
Not yet a member? Please register
ENBIS9 Goteborg
20 – 24 September 2009 Abstract submission: 1 February – 31 May 2009Engineering reliability assessment
22 September 2009, 11:40 – 12:10Abstract
- Submitted by
- Thomas Svensson
- Authors
- Thomas Svensson, Pär Johannesson, and Jacques de Maré.
- Affiliation
- SP, FCC, and Chalmers
- Abstract
- A second moment reliability method is presented for applications influenced by many sources of uncertainty. It is developed from specific problems in mechanical fatigue at variable amplitude. The method is a statistical approach, which means that the load and the strength need to be identified as random variables.
In case of fatigue at variable amplitude we must be first find a scalar representative for the multidimensional load time history experienced in service and find a corresponding scalar for the experimentally obtained fatigue strength.
Since the knowledge about the variation and uncertainty of the load and strength variables is highly limited, the method is based on only the means and the variances of the uncertainty sources that influence load and strength.
In fatigue, typical variational influences are scatter within material, manufacturing and users. The main uncertainties in variable amplitude fatigue are unknown model errors introduced in the projection onto scalar representatives, uncertain usage profiles, and unknown bias in sampling of material batches for strength and in sampling of user profiles for load.
The main problem in a certain application is to identify the important sources of variation and uncertainty, and assign comparable measures for their combination. This task needs skills both from engineering and statistical modelling.
Variances are assessed from observations or judgements, sensitivity coefficients are found from variational studies, usually easily obtained by simple sensitivity studies.
The vague of knowledge about the extreme behaviour of the influencing sources makes reliability assessments at low failure probabilities doubtful. Therefore, there is a need for an extra safety factor, based on engineering judgement.