Comparative Analysis of Advanced Machine Learning Methods for Heat Transfer and Thermal Efficiency in a Non-Newtonian Nanofluid with Joule Heating and Lorentz Forces: Dual Solutions and Stability Analysis

Abstract

This research explores the unsteady stagnation point flow of modified second-grade fluid by incorporating magnetized cobalt ferrite (CoFe2O4) nanoparticles across a heated movable plate. In addition, the radiation and Joule heating are also provoked. Since the cobalt ferrite particles are very important in biosensing, drug delivery, magnetic purification/separation, etc. The leading equations are changed into ordinary differential equations by employing similarity factors and then utilizing the bvp4c solver to obtain dual numerical solutions. Stability analysis confirms that the upper branch solution is stable and physically reliable. In addition, this research is further analyzed through advanced machine learning by employing artificial neural networks in conjunction with Levenberg- Marquardt. Moreover, this research deals with an important query by computing the problem through Gaussian Process Regression (GPR).The substantial outcomes indicate the velocity increases while the temperature declines due to the viscosity factor for the upper solution. In parallel, the machine learning outcomes show that the GPR gets an excellent R2 of 1 and, for the Nusselt number prediction, delivers a predictive error within the same order of magnitude as the ANN-LM benchmark. This confirms GPR as a high-fidelity tool capable of achieving near-perfect accuracy,making it a powerful choice where both precision and predictive confidence are essential.

Document