https://doi.org/10.1140/epjp/s13360-025-06646-w
Regular Article
Multi-layer perceptron approach for fostering heat transfer in nanofluid thin-film flow
1
Department of Mathematics, PSG College of Arts and Science, 641014, Coimbatore, Tamil Nadu, India
2
Faculty of Engineering, Kuwait College of Science and Technology, 35004, Doha District, Kuwait
a
priyadharshinip@psgcas.ac.in
Received:
11
February
2025
Accepted:
11
July
2025
Published online:
1
August
2025
Magnetohydrodynamic nanofluid thin-film flows over stretching surfaces pose significant heat transfer challenges in advanced thermal systems. Advanced computational frameworks excel at capturing the intricate nonlinear interactions between momentum-energy equations under variable thermal conductivity and viscous dissipation, enabling precise predictive models for industrial calibration. Governing equations undergo transformation through similarity variables and are solved employing the BVP4C numerical method. A Multi-Layer Perceptron with optimized architecture (2 hidden layers: 100–50 neurons, ReLU activation 5,401 parameters) processes five key parameters: M [1.0–2.0], Ec [0.0–0.6], S[0–1.8], Pr[0.7-6], Rd[0.2–1.4]. The data set comprises 80% training and 20% testing with 1000-epoch optimization. Magnetic forces (M: 1.0
4.0) significantly reduce thin-film velocities through Lorentz resistance. Unsteadiness parameter S contracts boundary layers and simultaneously increases surface resistance. The radiation parameter Rd expands the thermal layers, enhancing the heat flux but reducing the cooling rates. The Eckert number Ec affects temperature in an inverse relationship through energy conversion, and the Nusselt number correlation effectively captures the heat transfer characteristics. BVP4C validation against literature shows 
deviation for skin friction and 
for heat transfer coefficients. MLP achieves exceptional accuracy: Dataset A (
) – 
(training), 
(testing); Dataset B (
) – 
(both phases). MAPE values: 
(training), 
(testing). Hyperparameter optimization confirms an optimal accuracy of 
with a 50–50 neuron configuration. Framework enables real-time optimization in microelectronics cooling (31% temperature reduction), heat exchanger design (45% film thickness control) and solar cell manufacturing. Industrial implementation achieves 97.07% prediction accuracy for thermal management systems, coating processes and energy device applications establishing a robust computational paradigm for nanofluid heat transfer augmentation.
Copyright comment Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
© The Author(s), under exclusive licence to Società Italiana di Fisica and Springer-Verlag GmbH Germany, part of Springer Nature 2025
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
