Abstract
A maximum likelihood analysis method allowing for random censoring has been developed for statistical evaluation of fatigue data. The method is shown to be especially useful when analysing welded structures that have two or more potential failure locations. In those cases where fatigue cracking is observed at only one location, the method allows the non-failures to also be statistically considered. The method is illustrated using a fatigue data set from spectrum loaded welds made from 10 mm thick HTS 390 and VHTS 690 high tensile strength steel with three fabrication procedures. Each of the specimens tested had two or more failure locations and only a few of them showed significant fatigue cracking at both locations. As compared to the “usual” quality welded specimens from high strength steel, the use of a second weld pass followed by TIG dressing improved fatigue strength by about 25%. Full penetration and grinded welds from very high strength steel showed fatigue strength improvement of about 30% at 1×106 cycles to failure but significantly more at longer lives. For both “usual” quality and grinded welds, spectrum loading with alternating tension and compressive mean stresses resulted in about twice the fatigue life as compared to spectrum loading with constant tensile mean stress. This indicates that the portion of the spectrum with compressive cycles was non-damaging.
Article PDF
Similar content being viewed by others
Avoid common mistakes on your manuscript.
References
Maddox, S.J., Statistical analysis of fatigue data obtained from specimens containing many welds, IIW XIII-XV-XXXX-94, April 1994, 3 p.
Marquis, G. 1995. High cycle spectrum fatigue of welded components. Espoo 1995. Technical research Centre of Finland, VTT Publications 240, 83 p. + app. 100 p.
TSG, User Manual and Documentation for WASH (Version W, 1.02) Time History, Technical Software Consultants Ltd., Milton Keynes, UK, 1989, 72 p.
Wallin, K., “The probability of success using deterministic reliability”, Fatigue Design and Reliability: ESIS Publication 23, G. Marquis and J. Solin, Eds., Elsevier Science Ltd., Amsterdam, 1999, pp. 39–50.
Author information
Authors and Affiliations
Rights and permissions
About this article
Cite this article
Marquis, G., Mikkola, T. Analysis of Welded Structures with Failed and Non-Failed Welds Based on Maximum Likelihood. Weld World 46, 15–22 (2002). https://doi.org/10.1007/BF03266361
Published:
Issue Date:
DOI: https://doi.org/10.1007/BF03266361