Neural networks approach for prediction of gas-liquid two-phase flow pattern during convective condensation in microscale channels

Abstract

Paper presented at the 7th International Conference on Heat Transfer, Fluid Mechanics and Thermodynamics, Turkey, 19-21 July, 2010.

In two-phase flow, almost every constitutive relation is flow regime dependent because physical mechanisms that control beat transfer and pressure drop vary with the flow regime. Thus, the identification of flow pattern is an important issue to properly design, and operate two-phase flow systems. In recent years, emphasis has been put on the characteristics of two-phase flow and heat transfer in small and microscale flow passages due to the rapid development of microscale devices, but yet, no general accepted flow-pattern map for microscale channels is available. In this paper, a three-layer, feedforward neural network was designed. Mass velocity, vapor quality and fluid temperature were adopted as input data. The artificial neural network (ANN) was developed based on two-phase flow data for refrigerant R134a in a microscale channel. The validity of the adopted neural network was evaluated by cross validation. The results show that the neural network can provide good flow pattern predictions.