Automated classification of linear bifurcation buckling eigenmodes in thin-walled cylindrical shell structures

Computational problems in structural engineering are growing ever larger and solutions must increasingly be based on correspondingly large datasets obtained from detailed parametric sweeps. However, the acquisition of computational useful size is also becoming unfeasible without extensive use automation. In shell buckling studies, particularly those thin-walled shells under complex loading conditions, an important qualitative piece information class mode which reveals dominant destabilising membrane stress components. Unfortunately, diversity geometries that can encountered studies truly vast, there currently no way to rapidly assess laborious direct human observation model output. This paper presents automated classification tool for linear bifurcation eigenmodes cylindrical such as found wind turbine support towers, chimneys, silos, tanks, piles pipelines. It a convolutional neural network implemented using PyTorch machine learning framework. The adopted architecture widely image recognition tasks, chosen stratified five-fold cross-validation exercise. trained purposefully generated basic dataset 13,392 modes encoded chromatic signatures .jpg images (enhanced 25,726 by transformations). An example sweep unsymmetrical illustrates performance classifier. A GitHub repository offers Python scripts instructions how download network.