Machine learning-based insights into MHD Casson nanoliquid heat transfer with zero mass flux boundary condition: an artificial neural network application

Abstract The present study explores the MHD flow and heat transmission characteristics in a Casson nanoliquid over a nonlinearly stretched sheet, based on the Neural Network Backpropagation approach. The parameters controlling the mathematical formulation of the problem are Casson liquid parameter, power-law number, random motion and thermophoresis of nanomaterials, magnetic strength and Schmidt number. The boundary value problem is solved by the BVP4c numerical method, which is also used for the generation of the reference dataset for Neural Network Backpropagation. Solution graphs and error analyses for different parameter variations are evaluated in MATLAB based on this dataset. Regression analyses, error histograms, and determination of the mean squared error serve as tools to assess the accuracy of Neural Network Backpropagation. The MHD-Casson nanofluid model is worked on using test, validation, and training approaches. The influences of concentration, temperature distribution, and fluid flow on influential parameters are further analyzed. From the results, it is also discerned that with increasing Casson fluid parameter, both the temperature and nanoparticle volume fraction go higher.