turbulEnt Dense Gas flow modeling using large Eddy Simulation – EDGES

This Young Researcher Project proposes to introduce Large Eddy Simulation (LES) in the field of Organic Rankine Cycle (ORC) turbine flow modeling.
No prior study has been focused on the accuracy and relevance of current subgrid-scale (LES) or statistical (Reynolds Averaged Navier-Stokes or RANS) models in the context of dense gas flows used in ORCs. This project therefore proposes to :

(i) ----use Direct Numerical Simulation (DNS) to build a database of turbulent dense gas flows,
Turbulence statistics vary from one turbulent flow to the other. The objective in this project is to analyze turbulence using Direct Numerical Simulation (DNS) in canonical flows such has Decaying Homogeneous Isotropic Turbulence (DHIT), Forced Homogeneous Isotropic Turbulence (FHIT), shear layer and pipe flows.

(ii) ----verify "a-priori" the accuracy of current LES and RANS closure models,
In compressible LES, Martin et al.(2000) identify six turbulence closure terms in perfect gases: the SubGrid-Scale (SGS) stress tensor , the SGS heat flux, the SGS pressure-dilatation, the SGS viscous dissipation, the SGS turbulent diffusion and the SGS viscous diffusion. In addition to those terms the influence of the complex Equation of State (EoS) driving the thermodynamic properties of dense gases should also be taken into account. Compressible turbulence closure terms have been modeled by various authors and the research team will (a) identify the most relevant models for LES and RANS and (b) test their accuracy against the DNS database.

(iii) ----propose new LES and RANS models for dense gas flows,
A strategy will be implemented to tackle the issue of turbulence modeling in dense gas flows. This strategy will largely rely on the notions of Optimal Estimator and of Uncertainty Quantification.
--The Optimal Estimator theory will help determine among all the variables available in the LES or RANS approaches which one should be taken into account to form the best turbulence model both in terms of structural error (space/time correlation with the DNS term to be modeled) and functional error (accurate dissipation of energy).
--Uncertainty Quantification principles and methods will help determine the sensitivity of the output results to the variance of the model constants. It will therefore be an instrumental tool for the design of precise and robust turbulence models.

(iv) ----verify "a-posteriori" the relevance of these new closure models in the case of a full ORC turbine stage using High-Performance Computing (HPC) capabilities to perform LES and unsteady RANS simulations,
In the specific region of the trailing-edge of the turbine blade, turbulent structures created at all scales in shear layers, boundary layers and in the wake are essential to the development of losses. Using the same cases as Dura Galiana et al. (2015), the research team will compare the results obtained with traditional turbulence models with the ones gathered using the advanced models developed in the project EDGES. Together with experimental results obtained on the same configurations, this will serve as an a-posteriori validation of those models. A full ORC turbine radial stage including rotor and stator will then be studied in order to analyze the overall impact of the new turbulence models on the stage losses and efficiency.

(v) ----tighten links with experimental research teams to validate the newly developed models, foster collaborations with industrial partners and disseminate results in the industry.
The research team will disseminate the results of the project in the experimental research community in order to foster new long-term collaborations, and in the industrial community where the newly developed RANS models could be implemented and serve to the improvement of practical ORC expansion systems.