Harnessing Physics-Informed Neural Network for advanced thermal management: Radiative MHD flow of hybrid CNTs past a stretching cylinder with Stefan blowing

This article presents the thermal transport characteristics of the MHD Darcy–Forchheimer flow of Casson hybrid carbon nanotubes (HCNTs) over a stretching cylinder with radiation, viscous dissipation, endothermic/exothermic chemical reaction, heat consumption, Stefan blowing and zero mass flux condition. This investigation is motivated by the emerging demand to optimize thermal control in heat exchange devices, cooling thermal equipment, turbines, chemical synthesis and energy conversion processes. The hybrid CNTs can be created by the suspension of SWCNTs (Single-Walled Carbon Nanotubes) and MWCNTs (Multi-Walled Carbon Nanotubes) in water. In addition, the non-Fourier heat and mass flux theory is incorporated in energy and nanofluid concentration expressions. The system of nonlinear ODEs derived through the implementation of the suitable transformations on the fundamental expressions. These equations are computed by employing the PINN (Physics-Informed Neural Network) and bvp4c solver. The PINN consist of a feed-forward network with one input layer, three hidden layers and one output layer for velocity, temperature and NF concentration profiles. It is trained with 10000 epochs, learning rate of 1 0 − 3 and Adam optimizer. An ANFIS (Adaptive Neuro-Fuzzy Inference System) model proposed with suzeno type, 30 Gaussian membership function and 1000 epochs. The proposed PINN and ANFIS model achieve accuracy with R 2 values is close to 1 . The research outcomes demonstrate that the magnetic field and porosity parameters create a negative effect in the flow velocity profile and a positive effect in the temperature profile. The curvature parameter decline the thermal transfer gradient values. The findings of our research may have applications in many sectors, like drug delivery, cooling electrical equipment, tumour treatments and heat exchangers.