Quantitative assessment of random field models in finite element buckling analyses of composite cylinders

Abstract

Large scatter characterizes the collapse load of thin-walled structures due to their imperfection sensitivity. Random fields can be used to include imperfections in a finite element model, which can highly increase the credibility of the finite element (FE) model. Principal component analysis (PCA) and analytical covariance functions are matched to available correlation information of geometrical imperfection measurements. The random fields are realized by Karhunen-Lo`eve expansion combined with Monte Carlo methods. The results of the different covariance models are compared with the deterministic collapse loads of the measured imperfections in the FE-model. This approach isolates the effect of the different covariance models from other inaccuracies such as boundary conditions, loading imperfections, etc. The results show that random fields can be used to improve predictions and the understanding of the structural behavior of thin-walled structures, especially when these predictions are based on imperfection data, using PCA. Caution is recommended when using analytical covariance functions since they may fail to capture the complete behavior of the structure.