Bayesian regularization artificial neural network for weakly nonlinear stability analysis of multi-diffusive convection in radiative Navier-Stokes-Voigt fluid heated and salted from below

Accurate prediction of convective behaviour under complex physicochemical conditions is important in geothermal systems, chemical reactors, energy devices and cooling technologies. In the present work, the linear and weakly nonlinear stability of triple-diffusive convection in a radiative Navier-Stokes-Voigt fluid layer with chemical reaction is investigated. The governing momentum, energy and concentration equations are expressed using the Boussinesq approximation. Linear stability analysis predicts the onset of stationary and oscillatory convection. The heat and mass transfer characteristics are measured by the Nusselt and Sherwood numbers in a weakly nonlinear analysis based on truncated Fourier series. The obtained Lorenz-type nonlinear system is solved numerically by the odes45 solver in MATLAB. The Bayesian regularization artificial neural network is used as a surrogate predictive model for weakly nonlinear transport characteristics. The results indicate that the Kelvin-Voigt parameter and Prandtl number do not influence the stationary convection and the viscoelastic effects destabilize the oscillatory convection. Thermal radiation tends to stabilize the system and to suppress heat and mass transfer. Chemical reaction tend to enhance transport rates. Excellent agreement between the numerical and neural network predictions is observed with the maximum absolute error in the Nusselt number being less than 4.82 × 10 − 4 , which confirms the accuracy, reliability and computational efficiency of the proposed framework.