Neural Network Model Using Levenberg Marquardt Backpropagation Algorithm for the Prandtl Fluid Flow Over Stratified Curved Sheet

Artificial intelligence has proliferated across numerous fields of study as a result of its rapid and accurate response times. Its application in fluid dynamics for optimization and computational training of numerical outcomes has been remarkably significant. In the current investigation, a neural network model is constructed to analyze the propagation of Prandtl fluid across a curved, stretched surface that is subjected to radiation and comprises porous media. The interaction between a homogeneous and heterogeneous chemical reactions, magnetic field, thermal stratification, and second-order slip at the surface boundary are all factors under consideration. The numerical solutions of resulting equations are trained and validated using a neural network fitting tool and the Runge Kutta Fehlberg 4th to 5th order method. A total of 15% of data is designated for testing and 15% for validation, whereas 70% of it is utilized for training. Levenberg Marquardt training algorithm is employed to assess the functioning of network through regression analysis and mean squared error. In addition, mean square error is estimated to be 10-8 to 10-6. Additionally, the elastic parameter has decreased velocity while the Prandtl fluid has increased velocity. The thermal relaxation parameter has reduced the temperature, and the non-linear stretching index has enlarged the solute profile. As a result, this research shows that artificial neural networks can serve as a substitute for long-term computation prediction. However, the model's fluid flow structure can serve as a guide for creating an optimal industrial design.