Novel machine learning investigation for Buongiorno fluidic model with Sutterby nanomaterial

Abstract

The nanomaterials are frequently employed in a variety of heat transfer applications arising in energy generation, engine cooling, extrusion procedures, heat exchanger, thermos-chemical systems, manufacturing structures, powered plants etc. Experts and researchers in these fields often encounter such materials, leading to discernible impacts on velocity, temperature and concentration profiles. The objective of this study is to represent the mathematical structures for Sutterby nanomaterial fluidic model involving porous medium BFM-SNM using nonlinear autoregressive exogeneous networks backpropagated with Levenberg-Marquardt technique (NARX-LMT). The mathematical design of the model is originally presented by PDEs, which are converted into ODEs system through suitable modifications with alternative transformation incorporating numbers, i.e., Deborah, Prandtl and Reynold, as well as parameters, i.e., magnetic thermophoresis, radiation and temperature. The synthetic data is produced numerically by simulating Adams numerical method for BFM-SNM and the supervised computing paradigm of NARX-LMT is applied to the obtained datasets, and the results of NARX-LMT consistently align with numerical observations for each variant of the presented model, exhibiting negligible errors. The NARX-LMT performance on exhaustive experimentation is effectively illustrated through iterative convergence curves on mean squared error, control metrics of the optimization, distribution of error on histograms and regression outputs for Sutterby nanomaterial fluidic model.