A machine learning approach combined with lattice Boltzmann method for efficiently evaluating double-diffusive natural convection with attenuation effect

A large number of scholars have considered the effects of temperature and concentration on fluid diffusion in a square cavity when using the lattice Boltzmann method (LBM) to study fluid motion, few have studied the influence of attenuation effect. In this study, the double-diffusive natural convection with temperature-dependent viscosity and attenuation effect inside a porous cavity was numerically investigated by LBM at representative elementary volume (REV) scale. The influence of different flow governing parameters, including attenuation coefficient, the temperature-dependent viscosity and porosity on heat and mass transfer rates was investigated. The streamlines, isotherms, isoconcentrations, average Nusselt number and average Sherwood number curves for different parameters were discussed in detail. The results indicated that decreasing the viscosity and increasing the porosity could improve the heat and mass transfer, and a certain degree of concentration decay rate contributed to fluid diffusion. When the parameters were changed, the Nusselt number under the corresponding parameters had to be recalculated, which took a lot of computational time. Therefore, we introduced a machine learning method to predict the Nusselt number. The goodness of fit reached 0.968 and the error rates were all within 7%. It achieved a high prediction accuracy and improved the computational efficiency.