Comparison of Artificial Neural Network-Based Fuzzy Logic Model and Analytical Model for the Prediction of Optimum Material Parameters in a Heat-Generating, Functionally Graded Solid Cylinder

This study presents an artificial intelligence-based predictive framework as an efficient alternative to conventional analytical procedures for evaluating elastic–plastic thermal stresses in long functionally graded solid cylinders (FGSCs) subjected to uniform internal heat generation. A hybrid artificial neural network-based fuzzy logic (ANNBFL) model is developed to estimate dimensionless thermal load parameters at both the cylinder center and outer surface by learning from validated analytical reference solutions. The material properties, including yield strength, elastic modulus, thermal conductivity, and thermal expansion coefficient, are assumed to vary radially following a parabolic gradation law. Eight influential material parameters are incorporated as input variables to describe the coupled thermo-mechanical behavior of the FGSC. Multiple ANNBFL subnetworks are trained using analytically generated datasets and subsequently integrated into a unified prediction framework, enabling rapid and accurate stress field estimation without repeated analytical calculations. Model performance is systematically assessed by direct comparison with analytical solutions, demonstrating an overall prediction consistency of approximately 98.2%. The results confirm that the proposed ANNBFL approach provides a reliable, computationally efficient surrogate modeling tool for parametric evaluation and optimum material design of functionally graded cylindrical structures under thermal loading.