Cavitation Simulation with Consideration of the Viscous Effect at Large Liquid Temperature Variation

The phase change due to cavitation is not only driven by the pressure difference between the local pressure and vapor saturated pressure, but also affected by the physical property changes in the case of large liquid temperature variation. The present work simulates cavitation with consideration of the viscous effect as well as the local variation of vapor saturated pressure, density, etc. A new cavitation model is developed based on the bubble dynamics, and is applied to analyze the eavitating flow around an NACA0015 hydrofoil at different liquid temperatures from 25℃ to 150℃. The results by the proposed model, such as the pressure distribution along the hydrofoil wall surface, vapor volume fraction, and source term of the mass transfer rate due to cavitation, are compared with the available experimental data and the numerical results by an existing thermodynamic model. It is noted that the numerical results by the proposed cavitation model have a slight discrepancy from the experimental results at room temperature, and the accuracy is better than the existing thermodynamic cavitation model. Thus the proposed cavitation model is acceptable for the simulation of cavitating flows at different liquid temperatures.     

改进的物理融合神经网络在瑞利?泰勒不稳定性问题中的应用

基于相场法的物理融合神经网络PF-PINNs被成功用于两相流动的建模, 为两相流动的高精度直接数值模拟提供了全新的技术手段. 相场法作为一种新兴的界面捕捉方法, 其引入确保了界面的质量守恒, 显著提高了相界面的捕捉精度; 但是相场法中高阶导数的存在也降低了神经网络的训练速度. 为了提升计算训练过程的效率, 本文在PF-PINNS框架下, 参考深度混合残差方法MIM, 将化学能作为辅助变量以及神经网络的输出之一, 并修改了物理约束项的形式, 使辅助变量与相分数的关系式由硬约束转为了软约束. 上述两点改进显著降低了自动微分过程中计算图的规模, 节约了求导过程中的计算开销. 同时, 为了评估建立的PF-PINNS在雷诺数较高、计算量较大的场景中的建模能力, 本文将瑞利?泰勒RT不稳定性问题作为验证算例. 与高精度谱元法的定性与定量对比结果表明, 改进PF-PINNs有能力捕捉到两相界面的强非线性演化过程, 且计算精度接近传统算法, 计算结果符合物理规律. 改进前后的对比结果表明, 深度混合残差方法能够显著降低PF-PINNS的训练用时. 本文所述方法是进一步提升神经网络训练速度的重要参考资料, 并为探索高精度智能建模方法提供了全新的见解.