Tampere University of Technology

TUTCRIS Research Portal

The target reliability of the eurocodes

Research output: Chapter in Book/Report/Conference proceedingConference contributionScientificpeer-review

Details

Original languageEnglish
Title of host publicationSafety, Robustness and Condition Assessment of Structures
PublisherInternational Association for Bridge and Structural Engineering (IABSE)
Pages202-208
Number of pages7
DOIs
Publication statusPublished - 2015
Publication typeA4 Article in a conference publication
EventInternational Association for Bridge and Structural Engineering Conference -
Duration: 1 Jan 1900 → …

Publication series

NameIABSE Symposium Report
ISSN (Print)2221-3783

Conference

ConferenceInternational Association for Bridge and Structural Engineering Conference
Period1/01/00 → …

Abstract

The target reliability of the Eurocodes is given clearly: One-year reliability is 4.7 and 50-year reliability 3.8 correspondingly. However, the implementation of the direction is unclear in many ways: The reliability calculations for the Eurocodes are made sometimes by using one-year reliability 4.45 or 4.2. The Eurocodes does not instruct for which reference time the reliability is calculated. Normally, the reliability is calculated for the service time, 50 years, but sometimes for one year. The paper concludes that the reliability must be calculated for the service time. The independent versus the dependent load combination results in different reliability. The independent load combination results in higher reliability with fixed safety factors and up to about 10 % less safety factors with fixed target reliability when two loads are combined and with three loads even less. The loads are combined in the Eurocodes sometimes independently and sometimes dependently. Arguments are given here that the loads must be combined dependently. The variable load distribution is generally assumed Gumbel. However, this distribution is excessively safe as it has a robust upper tail which unrealist ically affects the reliabilit y. Normal distribut ion is one possible alternative, however obviously somewhat unsafe. In the paper, the safety factors are given based on Gumbel and normal distribution. The combination of 20 % Gumbel and 80 % normal distribution is one feasible option. In the current reliability calculation 50-year return load, i.e. 0.98 fractile of the load distribution is usually set at the characteristic load with the target reliability of the service time. This means that one-year loads are only considered in the reliability calculation. For this reason the variable load safety factors are unrealistically low as the target reliability corresponds to the service time loads. Gumbel distribution partly counterbalances the unsafe error but the overall effect is unsafe. The material factors of the Eurocodes are given based on the current calculation and modified calculation. The paper concludes that the reliability should be calculated for the service time loads with the distributions set at the service time location and the reference reliability should be 3.8. The issue of partial factors and design values is shortly addressed and concluded that the design value code is simple with better reliability accuracy than the current partial factor code when the characteristic variable load is made variable.

Keywords

  • Code, Design value, Gumbel distribution, Normal distribution, Partial factor, Reliability

Publication forum classification

Field of science, Statistics Finland