Modern approaches to modeling of elastic-strength properties of polymer composites

<p><strong>Context and relevance.</strong> Improvement of approaches to predicting the elastic-strength characteristics of composites is an urgent task in the mechanics of deformable solids, since polymer composite materials are widely used in modern technology, and the speed and accuracy of modeling the characteristics of such structures play an important role in the design and operation of structures made of composite. <strong>Objective.</strong> Having considered the existing approaches to PCM modeling, to identify the most promising ones. <strong>Results.</strong> The article provides analysis of existing approaches to modeling materials with heterogeneous structures: from early phenomenological and semi-empirical to modern multiscale ones that take into account the mutual influence of processes occurring in composites at the micro-, meso-, and macroscale levels. <strong>Conclusions. </strong>Modern multiscale finite-element approaches to modeling the elastic-strength properties of a composite material are widely used, however, a detailed description of the processes occurring at the micro- and meso- levels of the PCM requires significant computing power, which makes such approaches inapplicable in practice. To solve this problem, the multiscale approach is complemented by machine learning methods using surrogate neural networks.</p>