基于SISSO算法的混凝土细观模型压缩行为分析OA

The mechanical properties of concrete under external loads are influenced by its mesoscale compo-nents.Due to their heterogeneity,experimental and numerical methods struggle to reveal the impacts of me-soscale structures on the macroscopic mechanical behaviors of concrete.To effectively predict the peak stress of a 3-phase(aggregate,mortar and voids)mesoscale model of concrete under uniaxial compression,a frame-work for mesoscopic concrete was established with PYTHON and ABAQUS,to generate a dataset of models with varying aggregate volume fractions,porosities and peak compressive stresses.The sure independence screening and sparsifying operator(SISSO)machine learning algorithm,combined with the K-fold cross valida-tion for hyperparameter optimization,was employed to derive a formula describing the effects of the aggregate volume fraction and the porosity on the peak stress.The formula accurately describes the peak stress variation trend,thereby achieving precise predictions and offering physical interpretability.Compared to traditional ma-chine learning algorithms,the SISSO demonstrates advantages of maintaining precision while reducing computa-tion costs and improving interpretability.It overcomes the"black box"limitations of conventional methods,of-fering new insights for multiscale mechanical analyses of composite materials.