[en] This paper presents a methodology to take into account model uncertainties related to the load-end shortening curve of stiffened panels. This method is a part of a research activity carried out at University of Liège. His goal is to propose a reliability based model to assess hull girder ultimate strength using a progressive collapse algorithm. The numerical results presented here concern the load-shortening model chosen for this research, which is based on Bureau Veritas rules. Model uncertainty is quantified as a parameter considered as a random variable. Four statistical moments (mean, standard deviation, skewness and kurtosis) of this parameter are calculated and analyzed using a data-base of stiffened panels test results published in recent years. In order to increase the number of "actual" values for the statistical assessment, the data-base is completed with non-linear finite element analysis results. The paper contains also some recommendation for the implementation of the proposed method in ultimate strength reliability based analysis.
