An intelligent machine learning paradigm for gyrotactic micro-organism-induced squeezing flow of Eyring–Powell hybrid nanofluids with entropy generation

A novel unsteady squeezing flow of a non-Newtonian ternary hybrid nanofluid is investigated by focusing Eyring–Powell fluid rheology in an infinite channel. The nanofluid consists of AA1072, AA1075, and ZrO2 nanoparticles suspended in blood; the analysis incorporates the effects of slip; and diffusion parameters motile micro-organisms and chemical reaction. High-fidelity solutions obtained by employing computational technique in MATLAB and solution is validated by machine learning (ML) technique. The artificial neural network demonstrates excellent agreement with numerical solutions, achieving consistently low MSE and MAE values and coefficients of determination R2≈1 across all flow regimes. From an ML standpoint, the network effectively learns the nonlinear physical scenario concerning governing parameters and flow responses, including strong magnetohydrodynamic damping, slip-induced mobility, and diffusion-dominated transport. Error analyses confirm stable learning without numerical oscillations, while architecture studies identify optimal neuron configurations for accuracy and efficiency. An increase in the magnetic parameter suppresses convection due to the Lorentz force reducing, whereas higher thermal radiation enhances heat diffusion and strengthens wall heat transfer. The entropy generation number increases with the Eckert number due to viscous dissipation converting kinetic energy into thermal energy, thereby enhancing irreversibility. Magnetic parameter intensifies Lorentz-force-induced resistance, further increasing entropy generation and thermodynamic losses in the system. The proposed ML-based framework offers a fast, reliable, and computationally economical surrogate for complex coupled boundary-value problems, ensuring rapid optimization and real-time prediction in investigated blood transport.