Evaluation of ANN and ANFIS Methods in Study of the Motion of a Bubble in A Combined Couette-Poiseuille Flow

Abstract

The equilibrium position of a deformable bubble in a combined Couette-Poiseuille flow is investigated numerically by solving the full Navier-Stokes equations using a finite-difference/front-tracking method. The present approach is examined to predict the migration of a bubble in a combined Couette-Poiseuille flow at finite Reynolds numbers of 5, 10, and 15. The related unsteady incompressible full Navier-Stokes equations are solved using a conventional finite-difference method with a structured staggered grid. The purpose of this study is to evaluate ANN and ANFIS methods in study of the lateral migration of the bubble. Evaluation criteria of accuracy in test set derived from ANFIS demonstrates that estimated values of correlation coefficient (r), Mean Absolute Error (MAE), and Root Mean Square Error (RMSE) are 0.97, 0.001, and 0.0014, respectively. The ANN model with RMSE of 0.0007, MAE of 0.0004 and r of 0.99, is better than ANFIS model. It is also demonstrated that the bubble position estimated by the ANN and ANFIS models closely follows the one achieved from front tracking method.

Keywords

• ann
• anfis
• bubble
• front-tracking method
• reynolds number