Turbulent mixed convection flow and heat transfer properties in a driven cavity with two circular cylinders arranged one above another are analyzed numerically with the incorporation of a finite element scheme, based on the Galerkin method of weighted residuals. A comparison of streamlines, isotherms, and local, average Nusselt number is provided to illustrate how the Richardson number Ri, Reynolds number Re and the eddy viscosity ε affect the transport phenomena inside the cavity. The local and average Nusselt number have been evaluated and it is found that strong convection at the top wall causes the average Nusselt number to progressively rise with respect to Ri and ε, while steady or slightly varying profiles are seen with regard to Re. In order to predict the thermal distribution inside the cavity due to these hot cylinders, Artificial Neural Network (ANN) and Gaussian Process Regression (GPR) have been developed with these model parameters as input and average Nusselt Number as target values. Several plots have been depicted to come to a conclusion that both models have been trained very effectively with the minimal observed Mean Square Error (MSE) values of 0.0026018 and 0.0026428 for ANN and GPR, respectively. In support to these findings, the coefficient of determination R2 suggests that ANN (R2 = 0.89) outperforms GPR (R2 = 0.87) in testing accuracy, hence it is recommended for prediction task.
Read full abstract