Artificial neural network analysis of heat and mass transfer in fractional Casson flow

This study applies the Atangana–Baleanu fractional derivative to model free convection flow of Casson fluid under combined thermal and concentration gradients, exothermic reactions, and chemical processes. The governing equations are transformed using the Laplace method, and artificial neural networks with the Levenberg–Marquardt algorithm are trained on 70% of the data, with 15% for testing and validation. Quantitative analysis demonstrates a mean squared error below 1 0 − 4 , indicating high accuracy in predicting flow characteristics. Results reveal that fluid velocity decreases with increasing fractional parameters, while temperature and concentration profiles are significantly affected by chemical and thermal parameters. Graphical and numerical analysis validate the model’s effectiveness in capturing the flow dynamics under fractional calculus.