The design of modern turbomachinery components is based on numerical methods which solve the Reynolds averaged Navier-Stokes equations. Statistical turbulence models are employed to close the system of equations. Within this context, linear eddy viscosity model such as the Menter SST k-ω model are the industrial standard. Since the efficient use secondary flows is becoming more important in the design, the turbulence models need to meet more challenging requirements. To correctly account for secondary flow effects, it is necessary to resolve the anisotropy of turbulent stresses. For this purpose Reynolds stress models in explicit algebraic or differential formulation are employed. They are based on modelling assumptions which are validated and calibrated by comparison with experimental data or results of direct numerical simulation. To this end mostly simple geometries and flow topologies are investigated. Many complex flows, however, are still too expensive to compute in a direct numerical simulation and the level of detail of the measured data is often too low for a comprehensive validation of the model predictions.
Within the DLR internal project TurboVaLd, complex flows with relevant effects for turbomachinery are investigated in depth using optical Laser diagnostics. In detail, the following three experiments are conducted:
The results are used to validate and improve Reynolds stress turbulence models. A special advantage is the direct cooperation of experimental and numerical departments which guarantees optimal communication and exchange of the data.
Contribution by the numerical methods department
The numerical methods department contributes the development and implementation of models as well as the simulations of the mixing channel and the vortex tube.
To improve the prediction of flows with stream line curvature, the implemented and validated Hellsten Explicit Algebraic Reynolds Stress Model is extended by a curvature correction. Furthermore, differential Reynolds stress models are implemented in TRACE within TurboVaLd. Their predictive capabilities are evaluated and if necessary improvements will be suggested concerning different pressure-strain correlation and dissipation models. The focus will be on the representation of Reynolds stresses and their anisotropy in addition to the prediction of mean flow quantities. The aim is to provide Reynolds stress models which exceed previous modelling strategies in predictive capability. This ensures an improved design of modern turbomachinery components within the context of statistical turbulence models.