A NOVEL CELL BY CELL ARTIFICIAL NEURAL NETWORKS APPROACH FOR PREDICTING THE TEMPERATURE OF STEADY STATE, INCOMPRESSIBLE, LAMINAR FLOWS

Abstract

A cell-by-cell artificial neural network approach is used to predict the temperature field of steady-state, incompressible, laminar flows in a two-dimensional computational domain. The temperature field is characterized by the initial flow velocity, fluid temperature and the temperature of the wall boundaries. Two types of neural network architectures are developed in this research, namely cascade-forward and feedforward models. Both models are trained using Levenberg-Marquardt and Bayesian regularization backpropagation algorithms. The training data for the models are obtained by solving the Navier-Stokes equations for steady-state, incompressible, heat conducting laminar flow in two-dimensional domain using commercial ANSYS Fluent software. The results show that the predicted values produced by the ANN models are in good agreement with the CFD simulation data. Even though the introduction of artificial neural networks at the cell level increases the complexity of the training process, this drawback is compensated by the increase in flexibility (generality) of the models. More importantly, the results show that the cell-by-cell artificial neural network approach is capable of providing an accurate prediction of the temperature field for the fluid flow investigated in this research, as indicated by the statistical indices used to evaluate the performance of prediction models. The feedforward ANN model trained using the Bayesian regularization backpropagation algorithm gives the most accurate predictions among all models.

Keywords

• artificial neural networks
• computational fluid dynamics
• conjugate heat transfer
• cell-by-cell approach