Thermal management in polymer sheet extrusion using physics informed neural networks and non-Fourier viscoelastic modeling

AbstractEfficient thermal management in polymer sheet extrusion is essential for preserving material integrity and ensuring uniform product quality. This study examines steady boundary layer flow of a third grade viscoelastic polymer melt over a stretching surface incorporating magnetohydrodynamic, porous resistance, thermal stratification, and a first order chemical reaction. Finite speed heat conduction is modeled through the Cattaneo–Christov framework to account for thermal relaxation, which is relevant in high-rate polymer processing. Benchmark solutions are first obtained using MATLAB bvp4c solver. A Physics Informed Neural Network is then developed in PyTorch, which embeds the transport equations and boundary conditions into a unified loss structure optimized through Adam and LBFGS. The neural network shows excellent consistency with the benchmark results, capturing the nonlinear interactions inherent in viscoelastic, magnetically influenced, and thermally stratified flows. The analysis demonstrates that viscoelasticity promotes momentum transport, magnetic and porous resistances suppress flow development, thermal relaxation results in the reduction of near wall heating, and chemical reaction reduces concentration levels. The combined numerical and data driven framework provides a flexible tool for predictive thermal control in polymer extrusion, coating, and related manufacturing processes.