Implementation of multi-layer perceptron to gravitational convection of a nano-suspension subject to exothermic reaction: Second law analysis

Background: Interestingly effective cooling is needed in the cavities of complex type geometries like adverse trapezium-shaped chamber in microelectronics, energy storage, heat exchangers, solar collectors, and chemical processing etc. for modern industries. Further, gravitational convection provides low-grade heat transfer applications and titanium dioxide is an effective nanoparticle because of its low cost, high stability and greater thermal conductivity. That is why the present study covers the natural convection of TiO₂-water nanofluid along with second law analysis within an adverse trapezium-shaped chamber subject to exothermic reaction via Arrhenius kinetics and periodic magnetic field. Methods: The numerical solution of the present problem is obtained by implementing Finite element method (FEM). Multi-layer Perceptron (MLP) as a kind of artificial neural network (ANN) is utilized to anticipate the amounts of maximum temperature and mean Nusselt number. Significant Findings: The major new significant findings of the present study include that the streamlines, isotherms, velocities peter out due to the increment of the strength of the magnetic field. In other words, higher strength of magnetic field accounts for the controlling factor of the nanofluid motion and heat transfer within the adverse trapezium-shaped chamber. However, streamlines and isothermal lines upgrade with rise in Frank-Kamenetskii number and Rayleigh number, ratio of the cooler's position to the length of the inclined side of the chamber. Total entropy generation upsurges due to rise in Hartmann number and Frank-Kamenetskii number. Average heat transfer rate shows significant enhancement of 148.35% for rise of Rayleigh number from 10⁵ to 10⁶. In addition, the applied learning algorithm has great potential to predict the values of mean Nusselt and maximum temperature.