Machine learning-based prediction methods for flow boiling in plate heat exchangers

The present paper describes novel machine learning-based prediction methods for local heat transfer coefficient and local frictional pressure gradient of flow boiling within plate heat exchangers. Plate heat exchangers are becoming a valuable alternative, and in most cases a preferred solution, compared to traditional tube-in-tube and shell-and-tube heat exchangers, as they provide higher flexibility, compactness and lower weight. Using a consolidated dataset from the open literature, the proposed machine learning-based prediction methods take system geometry and flow conditions, such as mass flow rate, saturation temperature, heat flux, etc., as input parameters to predict the local heat transfer coefficient and local frictional pressure gradient. Compared to non-linear regression-based methods previously reported, the new methods significantly improve the prediction accuracy and can be easily adopted to design and analyze thermal performance of two-phase cooling systems. Reliable, accurate and validated prediction methods allow critical two-phase cooling systems to be readily designed given geometric parameters and operating conditions.