An identification strategy for stochastic fatigue models of welding joints from structural experiments

• Published in: Engineering structures 2022-11, Vol.270, p.114860, Article 114860
• Main Authors: Guo, Han, Feissel, Pierre, Druesne, Frédéric, Bouzebda, Salim, Limnios, Nikolaos, Patigniez, Alain, Bouyaux, Stéphane
• Format: Article
• Language: English
• Subjects: Cost function; Engineering Sciences; Fatigue life; Fatigue life of structure; Fatigue tests; Materials fatigue; Maximum likelihood estimation; Metal sheets; Parameter identification; Parameter uncertainty; Probabilistic S-N curve; S N diagrams; Statistical analysis; Statistical tests; Stochastic inverse problem; Stochastic parameter identification; Welded joints; Welding
• ISSN: 0141-0296; 1873-7323
• DOI: 10.1016/j.engstruct.2022.114860

The fatigue behavior of automotive structures requires some knowledge on welding joints which connect several sheet metal parts. In the automotive context, the fatigue analysis is performed on a structural assembly. The S-N curve for the materials of those welding joints and the associated dispersion is hardly identified from standard tests on an homogeneous specimen. This paper aims at developing an identification strategy of probabilistic S-N (P-S-N) curve based on the data collected from experimental tests on chassis. The experimental grouped fatigue life data are analyzed with the maximum likelihood estimation method. An identification of P-S-N curve has to be done by an inverse approach taking into account the parameter uncertainty. 6 novel cost functions, including one parametric and 5 non-parametric statistical test based methods, are proposed and investigated for a stochastic parameter identification problem. In addition to a convergence analysis, the performance of the methods has been validated by numerical examples. Furthermore, the proposed non-parametric methods based on Cramér von-Mises statistics show significantly better performance in terms of precision and efficiency of identification. •Identification of stochastic parameters from structural experimental tests.•Identification of P-S-N curve parameters for welding joints.•Maximum likelihood estimation for grouped data.•6 novel stochastic cost functions.