Artificial intelligence-assisted surrogate networks for thermal energy transport in radiative nanofluid model containing oxytactic microorganisms with mass flux-free conditions

Purpose This paper aims to explore the thermal energy transfer in radiative nanofluid flow of oxytactic microorganisms under zero mass flux conditions with potential application in engineering systems. The investigation concerns free convection over an isothermal perpendicular cone in a permeable medium filled with a nanofluid. Design/methodology/approach A deep learning-based nonlinear autoregressive exogenous neural network optimized with Levenberg–Marquardt backpropagation (NARX-DNN-LMBP) is designed for the model of the radiative nanofluidic flow with microorganism (RNF-MO). It includes the effects of Brownian motion, solar radiation, thermophoresis, convective heating, suspended nanoparticles and oxytactic microorganisms along with zero mass flux boundary conditions. The synthetic data set is created using the Lobatto IIIA numerical solver and then randomly split into training, validation and test sets used to develop the neural network. Findings The proposed NARX-DNN-LMBP approach is found to be effective in predicting the RNF-MO model for various physical scenarios with mean squared error as low as 10−10. It is further evaluated with the mean squared error convergence trends, error histograms and error autocorrelation analysis to show accurate and reliable computational performance. Originality/value The novelty is achieved by coupling the NARX-DNN-LMBP deep learning framework with the Lobatto IIIA solver for fluid flow of complex radiative nanofluid with oxytactic microorganisms under zero mass flux condition. By simultaneously considering Brownian motion, convective heating, solar radiation and thermophoresis, a comprehensive computational framework for advanced thermal transport analysis can be provided.