AI-driven analysis of EMHD Carreau–Maxwell electroosmotic nanofluid flow incorporating activation energy effects

This study focuses on forecasting the drag force, energy transfer rate and mass transport of Carreau–Maxwell hybrid nanofluid composed of [Formula: see text] and [Formula: see text] nanoparticles suspended in blood, flowing over a moving wedge. Due to the growing relevance of these nanoparticle mixtures in industries, automobiles and energy applications, the model includes several external physical effects like electroosmosis, electro-magnetohydrodynamic forces, heat generation, thermal radiation, activation energy, thermophoresis and Brownian motion. The governing partial differential equations are transformed into simpler ordinary differential equations using similarity variables and solved numerically through the shooting method. The influence of various dimensionless parameters on velocity, temperature and concentration profiles is analyzed through graphical and contour representations. It is found that the electric [Formula: see text] and magnetic (M) field parameters in electro-magnetohydrodynamic forces, the electroosmosis [Formula: see text], the Helmholtz–Smoluchowski velocity (Uhs) and electric field strength [Formula: see text] parameters in electroosmotic force show an increasing pattern in the velocity profile. Additionally, the parameters of Eckert number (Ec), thermophoresis (Nt), Brownian motion (Nb), and the volume fraction of [Formula: see text] and [Formula: see text] nanoparticles significantly elevated the heat transfer rate of [Formula: see text]. To enhance prediction capabilities, three machine learning models including Artificial Neural Network (ANN), Zeroing Neural Network (ZNN) and Adaptive Neuro-Fuzzy Inference System (ANFIS) are trained using the generated numerical data. The model performance is evaluated using the regression correlation coefficient (R), which approached 1, indicating a strong agreement between the predicted and actual values. Among the models, ANN demonstrated the highest accuracy, with an average absolute error of less than [Formula: see text] in predicting the fluid properties, outperforming both ZNN and ANFIS models.