[论文解读] A comparative evaluation of turbulence models for simulation of unsteady cavitating flows

Dhruv Apte, Mingming Ge+1 more|arXiv (Cornell University)|Mar 15, 2023

Cavitation Phenomena in Pumps参考文献 49被引用 7

一句话总结

本研究基于文丘里管中的X射线实验数据，评估了标准RANS-LES湍流模型与混合RANS-LES模型（包括k-ω SST、DES、DDES、PANS和FBM）在预测非定常云 cavitation 方面的表现。结果表明，尽管网格加密使标准模型能够捕捉到蒸汽相的非定常性，但湍流量在局部区域仍存在显著偏差，且Reboud涡粘性修正会抑制脉动，从而限制了精度，尽管其改善了空化脱落的预测效果。

ABSTRACT

Cavitation is a complex multiphase phenomenon characterised by vapour bubbles forming due to a sudden pressure drop and is often accompanied by increased hull vibrations, increased radiated noise and decrease in propeller and impeller performance. Although the Reynolds-Averaged Navier-Stokes (RANS) method coupled with a cavitation model is still considered a practical tool to predict cavitating flows owing to its computational efficiency, it is unable to predict the unsteadiness of vapor shedding and over-predicts the eddy viscosity. To improve the prediction, an empirical eddy viscosity correction,[Reboud et al. 1998] was proposed to consider the compressibility effects produced by cavitation. Additionally, a new type of models termed as hybrid RANS-Large Eddy Simulation (LES) models have also been recently introduced in the community, having the ability to behave as a RANS or a LES model in different regions of the flow in order to combine the computational cost efficiency of RANS with the accuracy of LES modelling. However, there exists a lack of a comprehensive review of various such turbulence models like the k-ω Shear Stress Transport Model (SST), k-ω SST Scale-adaptive Simulation (SAS), k-ω SST Detached Eddy Simulation (DES), k-ω SST Delayed DES (DDES), Filter-Based Method (FBM) and Partially-Averaged Navier Stokes Method (PANS) to predict cavitating flows. In this work, we conduct such a review to compare their ability to predict cloud cavitating flows by comparing them with x-ray experimental data in a venturi. It is shown that with mesh refinement, standard models do show the vapor unsteadiness as seen in the experiment similar to that seen when using the Reboud correction. However, on local comparison of turbulence quantities, it is observed both forms of models have huge discrepancies with experimental data that does not improve downstream.

研究动机与目标

• 评估标准与混合RANS-LES湍流模型在非定常云空化流中的预测能力。
• 评估Reboud涡粘性修正对湍流与空化动力学的影响。
• 在文丘里管构型中，将模型性能与高保真X射线实验数据进行对比。
• 识别尽管全局空化非定常性得到改善，但局部湍流预测仍存在局限性。
• 探索数据驱动方法在克服持续存在的建模瓶颈方面的潜力。

提出的方法

• 使用OpenFOAM进行三维非定常模拟，采用结构化文丘里管网格模拟空化流。
• 应用多种湍流模型：k-ω SST、k-ω SST SAS、DES、DDES、FBM以及不同$f_k$值的PANS。
• 引入Reboud涡粘性修正以考虑空化中可压缩性效应的影响。
• 通过非空化湍流流场案例验证模型实现，对比k-ω SST与PANS（$f_k=1$）以及FBM与k-ε模型。
• 开展网格加密研究，评估预测结果对分辨率的收敛性与敏感性。
• 利用实验X射线数据直接对比空化动力学与湍流量（如雷诺剪应力、湍流动能TKE）。

实验结果

研究问题

• RQ1 与实验数据相比，标准RANS模型（如k-ω SST）在预测云空化中非定常蒸汽脱落行为方面表现如何？
• RQ2 Reboud涡粘性修正在多大程度上改善了非定常空化与湍流量的预测？
• RQ3 混合RANS-LES模型（如DES、DDES、PANS、FBM）在空化流中是否显著优于标准RANS模型？
• RQ4 为何在网格加密与模型升级后，局部湍流量的偏差仍持续存在？
• RQ5 模型选择、空化模型及三维效应在导致云空化中局部湍流预测不准确方面起到何种作用？

主要发现

• 网格加密使标准RANS模型（如k-ω SST）能够重现与实验相似的非定常蒸汽脱落行为，其效果与Reboud修正相当。
• Reboud修正降低了过高的涡粘性预测，并改善了空化非定常性，但过度抑制了湍流脉动，导致某些区域雷诺剪应力为零。
• 即使经过网格加密，DES与PANS等混合模型在湍流量上仍存在持续的局部偏差，表明当前建模方法存在根本性局限。
• PANS模拟在某些$f_k$值下成功再现了云空化，而其他值即使在更低空化数下也失败，凸显模型参数的敏感性。
• 本研究识别出所有模型在准确预测局部湍流特性方面存在瓶颈，表明当前RANS及混合RANS-LES模型在高保真局部预测方面仍显不足。
• 利用实验或DNS数据作为训练输入的数据驱动方法被识别为未来克服持续建模误差的有前景方向。