AI-driven neural network for heat transfer optimization in squeezed microfluidic flow

This study develops an artificial intelligence–driven Levenberg-Marquardt backpropagation neural network (AI-LM-BPNN) to optimize heat transfer in squeezed ternary hybrid nanofluid (THNF) flow across a porous microfluidic surface, incorporating viscous dissipation, porosity, magnetic strength, and pressure gradients within a robust mathematical model. The governing partial differential equations are converted into non-linear ordinary differential equations using similarity transformations and solved numerically using MATLAB's bvp4c. The AI-LM-BPNN demonstrates faster convergence and higher accuracy than bvp4c in predicting nonlinear flow behaviour. The THNF, composed of silver Ag , copper Cu , and ferric oxide F e 3 O 4 nanoparticles in water, achieves superior thermal performance owing to the excellent conductivity of Ag and Cu and the magnetic response of F e 3 O 4 . Results show that the ternary nanofluid enhances heat transfer by up to 15.3 % compared to hybrid and mono nanofluids. Increasing porosity from 0.2 → 0.6 raises surface friction by 7.6–12.1 %, while stronger magnetic fields and radiation improve boundary layer heat transfer by 8.5 % and 6.7 %, respectively. These findings highlight the potential of AI-assisted modeling for advancing thermal management in microfluidic chip systems.