Machine learning classification of in-tube condensation flow patterns using visualization

Identifying two-phase flow patterns is fundamental to successfully design and subsequently optimize highprecision heat transfer equipment, given that the heat transfer efficiency and pressure gradients occurring in such thermo-hydraulic systems are dependent on the flow structure of the working fluid. This paper shows that with visualization data and artificial neural networks, the flow pattern images of condensation of R-134a refrigerant in inclined smooth tubes can be classified with more than 98% accuracy. The study considers 10 classes of flow pattern images acquired from previous experimental works for a wide range of flow conditions and the full range of tube inclination angles. Although not the focus of this paper, the use of a Principal Component Analysis allowed feature dimensionality reduction, dataset visualization, and decreased associated computational cost when used together with multilayer perceptron neural networks. In addition, the superior two-dimensional spatial learning capability of convolutional neural networks allowed improved image classification and generalization performance. In both cases, the classification was performed sufficiently fast to enable real-time implementation in two-phase flow systems.

Automated summary

Identifying two-phase flow patterns through visualization data and artificial neural networks can achieve high accuracy in classifying flow pattern images, especially when combined with Principal Component Analysis for feature reduction and convolutional neural networks for spatial learning capabilities. This approach enables real-time implementation in two-phase flow systems with improved classification and generalization performance.