Reliability-based code calibration for timber in fire

Abstract

For the design of structures, codes such as the Eurocode guarantee the required structural safety, both for the permanent and the fire design. In timber construction, the fire resistance is usually verified with the standard fire curve, since Eurocode provides simple and accurate design models for this exposure. The current Eurocode for timber in fire (EN 1995-1-2:2004) has evolved historically and uses the 20% fractile value for the strength in case of fire instead of the typical 5% fractile value. This has recently been questioned and is therefore thoroughly investigated in this thesis. In principle, there is no evidence that the reliability level of the current code is insufficient. Therefore, this thesis investigates the influence of the chosen fractile value on the homogeneity of the reliability levels of arbitrary structures when designed in accordance with the calibrated code. For this purpose, several reliability-based code calibrations for timber structures exposed to standard fire were conducted. To assess the accuracy of the code format, the used reliability model must be as accurate as possible. For this purpose, an analytical resistance model was developed, which is based on the observed charring behavior of timber. This model was fitted to results of finite element simulations that were considered accurate. The simulations consisted of bending beams with varying geometry and fire exposure. Then, the analytical model was fitted based on the ultimate bending moment of those simulations. Considering the uncertain parameters of the analytical model as random variables, the model can be used for the reliability analysis. In order to achieve a homogeneous reliability level with arbitrary structures it is useful to introduce partial factors for parameters with large scatter. For this reason, the code format for timber in case of fire has been extended by a partial factor on the charring rate. In total 770 different bending beams were considered in the code calibration, in all of which the fire design is decisive over the permanent design. In the calibration, the partial factors on the charring rate and the strength were optimized so that the resulting reliability index of the structures deviated as little as possible from the target reliability index. It could be shown that the partial factor of the charring rate should be chosen slightly above one. At the same time, the partial factor for the strength needs to be chosen below one so that the mean reliability level is not changed. A partial factor below one is equivalent to an increase of the fractile value, which proves that a reduction of the fractile to 5% would not be meaningful since it decreased the homogeneity of the reliability indices resulting from the calibrated code format. With reliability calculations and code calibrations for subsets of the considered structures, it was shown that the resulting reliability is significantly dependent on the fire exposure. Therefore, it would be advisable to differentiate the fire exposure in the code format. Furthermore, different optimization criteria were compared. Slight differences in the calibrated partial factors were found. However, the different results should not be understood as deficiencies of the criteria, but as optima for different optimization targets.