Neural network-based study of thermal behaviour in radiative magnetized Casson micropolar nanofluid flow between two disks

This work examines thermally radiative Casson nanofluid flow through the variable porous space between two circular disks. The flow is subjected to the micro-rotation theory with effects of magnetohydrodynamics in the normal direction of the flow system. The governing equations are first solved using the bvp4c numerical solver to generate accurate reference solutions. These solutions subsequently serve as the foundational training dataset for the artificial neural network (ANN), providing a reliable basis for model training, prediction, and performance validation. This work shows that error histograms and regression plots demonstrate excellent agreement between predicted and target values, confirming high accuracy, strong generalization capability, and the absence of overfitting. The performance analysis achieves optimal validation errors of 1.521 × 10⁻⁸, 2.184 × 10⁻⁸, and 1.3366 × 10⁻⁸ for the three cases, confirming rapid convergence and outstanding predictive performance. Axial velocity and micro-rotational velocity decline with growth in magnetic parameter and Casson parameter, while augmented with a surge in Reynolds number. Thermal profiles augment with growth in Reynolds number, radiation parameter, magnetic factor, and heat source parameter. Current work is validated through comparative analysis of current results with an established dataset. This study signifies the advanced thermal management strategies for high-speed rotating machinery, including turbines and electric motors, by harnessing the combined effects of magnetic fields and nanoparticles to precisely regulate fluid flow and heat transfer. The proposed approach contributes to enhanced cooling performance, optimized lubrication systems, improved energy efficiency, and the design of high-performance biomedical devices, thereby offering significant potential for next-generation thermal engineering and industrial applications.