Application of computational intelligence methods for the heterogeneous material stress state evaluation

The use of surrogate models providesgreat advantages in working with computer-aided design and 3D modeling systems, which opens up new opportunities for designing complex systems. They also allow us to significantly rationalize the computing power automated response time low energy consumption are critical. This work is devoted creation a model approximating finite element solution problem dispersion–strengthened composite plane sample deformation. An algorithm constructing parametric two–dimensional proposed. calculation created using ANSYS Mechanical analysis program APDL scripting builder. parameters stress-strain state material microstructure processed convolutional neural network. A network based on U–Net architecture encoder-decoder type has been predict distribution equivalent stresses according geometry load values. direct sequence layers takenfrom specified architecture. To increase speed stability training, part changed. consists serially connected blocks, each combines such as convolution, normalization, activation, subsampling, latent space that connects encoder decoder adds data. combine vector, concatenator created, additionally includes Dense, Reshape Concatenate layers. loss function defined root mean square error over all points source matrix, calculates difference between actual value target variable generated by model. Optimization ofthe performed first–order gradient local optimization method ADAM. study learning process illustrated plots functions additional metrics. There tendency indicators coincide training validation sets, indicates generalizing capability Analyzing output andthe metrics, conclusion made about sufficient quality However, values weights after still not optimal terms minimizing function. And also, accurately reproduce (FEM), proposed quite simple requires clarification. comparisonof obtaining results FEM ahead used calculations determine overall approximation problems mechanics this