A machine learning technique for computational analysis of EMHD ternary nanofluid with heat generation: Al2O3+CuO+TiO2/water

Abstract

In order to model ternary nano fluid EMHD flow with internal heat generation, we present a machine learning approach. In this case, traditional numerical methods are often computationally difficult to apply on such complex systems. Ternary nanofluids enhance the efficiency of solar thermal systems by a large margin compared to single or binary nanofluids because of the significant enhancement of heat transport in ternary nanofluids. Similarity transformations are applied to the governing PDEs that are solved with the MATLAB bvp4c solver. Effects of key factors on flow and thermal behaviour are identified by the use of numerical simulations; effects of friction factor and Nusselt number are checked by means of tabular and graphical data.

The trained machine learning model estimates parameters of temperature, velocity, and heat transport more accurately than the traditional techniques. The model fits numerical simulations efficiently by taking into account magnetic field effects and heat generation. In terms of prediction, justification, and optimisation, Artificial Neural Networks (ANNs) perform better than conventional regression approaches. Exactness and generalisation are guaranteed by using 80 % of the bvp4c-generated data for training and 20 % for testing. Regression analysis (for both temperature and velocity) reveals a low mean squared error (MSE ∼ 10⁻⁵) and a closely perfect correlation coefficient (R = 1). Strong agreement across all data subsets is shown by regression graphs and error histogram. A potential paradigm for improving thermal system design in industrial, aerospace, and energy applications is presented by this AI-CFD integration.