Peristaltic transport and thermodynamic analysis of hybrid nanofluids in porous media using physics-informed neural networks.

This study presents the wavelet-based physics-informed neural networks (PINNs) simulation to analyse entropy generation in hybrid nanofluid peristaltic flow through a curved porous channel. The flow and heat transfer characteristics of a water-based hybrid nanofluid containing multi-walled carbon nanotubes (MWCNTs) and magnetite [Formula: see text] nanoparticles are examined in a Darcy porous medium. The model accounts for nonlinear thermal convection, nonlinear radiation, and entropy generation, providing a comprehensive assessment of transport and thermodynamic irreversibility's. The governing equations are formulated in curvilinear coordinates, transformed into dimensionless form, and solved using wavelet-PINNs. Unlike conventional numerical schemes, the PINN approach embeds boundary conditions and physical constraints directly into the training process, ensuring accuracy, mesh-free implementation, and generalization across parameter ranges. The results indicate that variations in key physical parameters, including buoyancy forces, nonlinear convection, porous medium resistance, thermal radiation, curvature, and viscous dissipation, significantly influence the flow, thermal and entropy generation characteristics. The velocity, temperature, and entropy generation profiles exhibit non-uniform behaviour across the channel, with notable variations near the walls due to stronger gradients. Entropy generation is observed to be higher in regions of intensified velocity and temperature gradients, while relatively lower values occur in the core region. The present PINN framework provides a stable and efficient approach for analysing hybrid nanofluid transport in complex geometries, with potential application in the design and optimization of thermal systems involving peristaltic transport in porous channels.