Probabilistic design methods for steel structures using refined models of resistance and loads

Refined probabilistic models have been proposed for steel resistance, dead loads, live loads, snow loads, and wind loads, which can be used in reliability analyses by the FORM and SORM methods. The theoretical distributions were selected using the Anderson–Darling goodness-of-fit criterion. It was found that steel resistance is most accurately represented by a lognormal distribution. The analysis was carried out on the basis of extensive experimental data on the yield strength of steel grade S235. It is shown that the use of the Generalized Extreme Value (GEV) distribution for snow and wind loads provides the best agreement with observed annual maxima obtained from 50-year meteorological records. The results improve the accuracy of probabilistic design calculations and can be applied in the calibration of code-specified (baseline) load and resistance values.