Skip to Main content Skip to Navigation

Optimisation de la modélisation RANS d'écoulements cavitants

Abstract : Turbulent cavitating flows occur in many engineering practical applications such as pumpsand propellers. In these devices, the collapse of the cavitation bubbles combines with instabilities atmultiple scales produce major detrimental effects like flow rate fluctuations, noise, vibrations, anderosion. It is thus essential to accurately predict the behavior of unsteady cavitation, thereby reducingtheir consequences for the machinery. To simulate the turbulent cavitating flows, the most commonlyused approach is still the Reynolds-averaged Navier-Stokes (RANS) method coupled with homogeneouscavitation models, due to its computational tractability. However, it is a consensus that the RANS modelsare not accurate for the complex flows in the presence of adverse pressure gradients leading to flowseparation and recirculation. This limitation leads to the poor prediction on the interactions betweencavitation and turbulence in cavitating flows. Hence, it is necessary to quantity and reduce theuncertainties in the RANS model and thus improve the predictive performance, either with an empiricalapproach or data assimilation (DA) methods. In this thesis, we investigate the applicability of suchmethods for turbulent flows with the objective of introducing the data-driven method into cavitatingflows. Specifically, we first apply the hybrid DA method, ensemble based variational method, toreconstruct the flow field in convergent-divergent channel, through inferring optimal inlet velocity andmodel corrections from observation data. Further, we proposed a regularized ensemble Kalman methodcapable of enforcing the regularization constraints for ill-posed inverse problems. Also, variousensemble-based DA methods are evaluated for uncertainty quantification in CFD applications. Finally, anew empirical modification of the turbulent viscosity is proposed for cavitating flows based onexperimental measurements.
Document type :
Complete list of metadata
Contributor : ABES STAR :  Contact
Submitted on : Monday, May 18, 2020 - 5:16:08 PM
Last modification on : Friday, August 5, 2022 - 2:54:01 PM


Version validated by the jury (STAR)


  • HAL Id : tel-02611790, version 1


Xinlei Zhang. Optimisation de la modélisation RANS d'écoulements cavitants. Autre. Ecole nationale supérieure d'arts et métiers - ENSAM, 2019. Français. ⟨NNT : 2019ENAM0060⟩. ⟨tel-02611790⟩



Record views


Files downloads