Review and improvement of void fraction models for two-phase cross-flows

This work assesses the accuracy of existing void fraction predictive methods for two-phase cross-flows in tube bundles and proposes a new, physically-motivated void fraction model. All experimental measurements from the open literature are gathered and reviewed. The resulting aggregated database contains a total of 1830 experimental measurements, forming the largest dataset todate, and includes seven different mixture compositions. The liquid superficial velocity spans almost three orders of magnitude (0.01 - 6.5 m/s), and the gas superficial velocity spans four orders of magnitude (0.01 - 110 m/s), meaning some data points are representative of industrial operating conditions. When measurements were performed locally, corrections are proposed to account for the non-uniform void fraction spatial distribution. The existing predictive methods are assessed both qualitatively and quantitatively. Building on the most accurate method and starting from a basic force equilibrium equation, a new void fraction model is proposed. Excellent accuracy is achieved, with an overall mean absolute error of 0.038 and a prediction error below 0.1 for 95 % of measurements. The accuracy and robustness of this new model are validated using state-of-the-art statistical methods, including bootstrap and leave-one-out cross-validation.