ENBIS: European Network for Business and Industrial Statistics

Monte Carlo versus second moment evaluation in reliability, a fatigue life example

21 September 2009, 14:25 – 14:45

Abstract

Submitted by

Thomas Svensson

Authors

Sara Lorén Thomas Svensson

Affiliation

FCC, Göteborg and SP, Borås

Abstract

Monte Carlo simulations have become very popular in industrial applications as a tool to study variational influences on reliability assessments. The method is appealing since it can be done without any statistical knowledge and produces results that appears very informative. However, in most cases the information gathered is no more than a complicated transformation of initial guesses, since the statistical distributions of the dominating variational influences are normally unknown. The seemingly informative result may then be highly misleading, in particular when the user lacks enough statistical knowledge.

Instead, in cases where the input knowledge of the distributional properties is vague, it should be better to use a reliability method based on the actual knowledge, often not more than second moment characteristics. This can easily be done by means of the Gauss approximation formula, based on variances, covariances, and sensitivity coefficients.

Here, a specific problem of fatigue life of a welded structure is studied by 1) a Monte Carlo simulation method and 2) a second moment method. Both methods are evaluated on a fatigue strain-life approach and use experimental data showing variation in weld geometry and material strength parameters.

The two methods are compared and discussed in view of the engineering problem of reliability with respect to fatigue damage.

View paper

Return to programme