Artificial intelligence-based analysis of flow and heat transfer in blood-inspired Casson hybrid nanofluids via recurrent neural networks with Bayesian regularization

A recurrent neural network with a Bayesian regularization algorithm (RNN-BRA) is employed in this study to model and predict the flow, heat transfer, and thermal characteristics of a Casson hybrid nanofluid. The nanofluid is a blood-based fluid comprising gold and silver nanoparticles. Thermal radiation and convective boundary equations can accurately represent thermal processes. The partial differential equations are transformed into ordinary differential equations by similarity transformations. In Mathematica software, the Adams method is used to generate numerical data. After training with these datasets, the RNN-BRA framework accurately approximates and predicts system behavior. The results show that the fluid velocity decreases with the Lorentz force as the Hartmann number increases. Increasing the Casson parameter and the inverse Darcy number increases resistance to flow and reduces the velocity profile. Higher values of the Biot number, radiation parameter, and heat source parameter raise the fluid temperature. An increase in the thermal relaxation parameter reduces the temperature distribution by increasing the response time to the heat flux. RNN-BRA is tested using regression, error histogram, mean squared error, and fitness curve analysis. These tests demonstrate the structure's validity and dependability in discovering the properties of the Casson hybrid nanofluid model. The study demonstrates that the RNN-BRA algorithm is computationally more efficient than conventional numerical approaches in computational fluid mechanics and is also highly accurate. The framework can be used with high probability in engineering applications, biomedical heating, solar energy systems, electronic cooling, and energy storage. • AI-driven computational scheme is developed for the Casson hybrid nanofluid model. • The Adams method is used to generate data sets for multiple parametric scenarios. • The results of the proposed scheme show excellent agreement with the Adams method. • The validity of applied scheme is assessed through different statistical analysis. • The effects of key parameters on velocity and thermal fields are investigated.