Sample size dependent probabilistic design of axially compressed cylindrical shells

The fact that the buckling load of cylindrical shells depends on imperfections raised the idea of applying probabilistic design methods for these structures. Whether a probabilistically motivated design load may be regarded as a representative for a type of shell depends, among others, on the samples that build the data basis for the probabilistic methodology. In the current paper a methodology is presented that takes into account the samples size within a fast probabilistic design. The method presented leads to a lower design load, as the sample size decreases. The methodology is applied to a set of 33 beer cans, which have been measured and tested at the Delft University of Technology. The 33 cans are subdivided into groups which are then analyzed probabilistically in order to observe how the probabilistically motivated design load varies for different samples sizes. The results indicate that the method presented provides a useful tool for designers to ensure that a small data basis does not yield a unsafe probabilistic design.