Gönczi, Dávid2026-01-152026-01-152025-10-18International Journal of Engineering and Management Sciences, Vol. 10 No. 4 (2025) , 57-66https://hdl.handle.net/2437/402479This paper deals with the numerical analysis of functionally graded spherical bodies subjected to combined thermal and mechanical loads. A method is presented to train deep neural networks to approximate the important solutions. We outline two approaches for generating the training dataset for a deep neural network, followed by a method for creating the neural network itself. Then, through a numerical example, we investigate the axisymmetric problems of radially graded spherical bodies (e.g., ideal spherical pressure vessels). Based on the results obtained, we evaluate the accuracy of solving the outlined problem using the proposed neural network.This paper deals with the numerical analysis of functionally graded spherical bodies subjected to combined thermal and mechanical loads. A method is presented to train deep neural networks to approximate the important solutions. We outline two approaches for generating the training dataset for a deep neural network, followed by a method for creating the neural network itself. Then, through a numerical example, we investigate the axisymmetric problems of radially graded spherical bodies (e.g., ideal spherical pressure vessels). Based on the results obtained, we evaluate the accuracy of solving the outlined problem using the proposed neural network.application/pdfFGM SphereNeural NetworksFEMThermoelasticityFGM SphereNeural NetworksFEMThermoelasticityfolyóiratcikkOpen AccessDávid Gönczihttps://doi.org/10.21791/IJEMS.2025.19International Journal of Engineering and Management Sciences4102498-700X