Development of CFD MethodsCopyright: © RWTH Aachen | IST
Within the framework of a partnership with the Institute of Propulsion Technology at DLR, the Institute of Jet Propulsion and Turbomachinery contributes to the further development of the source code. The focus is on the development of numerical concepts and method validation using high-performance computing.
TRACE (Turbomachinery Research Aerodynamic Computational Environment) is a powerful simulation software for calculating three-dimensional unsteady flow in multistage compressor and turbine components. TRACE has been developed for more than a decade by the Numerical Methods Group of the Institute of Propulsion Technology at the German Aerospace Center, or DLR.
TRACE is used both by research institutes and higher education institutions for the scientific analysis of the highly complex flow phenomena in turbomachinery and in the industrial environment for development and optimization tasks in the design process of new turbomachinery components.
The further development is mainly dedicated to the topics aeroelastic and aeroacoustic, aerothermodynamics as well as turbulence and transition modelling with focus on specific turbomachinery aspects. For this purpose, modules have been developed to solve the linearized equations in the time domain and the non-linear equations in the frequency domain.
An adjoint solution module is also available for aerodynamic shape optimization.
Scale Resolving Simulations (SRS-Cascade)Copyright: © RWTH Aachen | IST
Cascade measurements play an important role in the research and development of turbomachinery. However, the quality of these measurement results is reduced by measurement inaccuracies. In contrast, a numerical test-bench based on highly-fidelity CFD simulations (SRS) promises a noticeable increase in the results quality.
SRS promise a significantly improved prediction accuracy, in which at least a part of the turbulence spectrum is resolved.
This project aims at validating the SRS methods currently implemented in the flow solver TRACE on the basis of compressor and turbine cascades. Highly accurate measurement data and reference calculations serve as a benchmark. These data can also be used to improve the turbulence models.
Prediction of Tip Leakage VortexCopyright: © RWTH | IST
The prediction of the tip leakage vortex in axial compressor is for today’s RANS simulations a challenging task.
In comparison to RANS simulations, scale resolving simulations are better fit to capture unsteady flow phenomena (e. g. the tip leakage vortex) correctly as the unsteadiness is fully resolved.
In consequence, scale resolving simulations are used in a research project to investigate the tip leakage vortex in its details. The results of this investigation are then used to improve the RANS setups for TRACE. Therefore, the influence of the different turbulence models, their extensions and other parameters on the tip leakage vortex is investigated.
Therefore, the results of these parameter studies can then be compared to scale resolving simulations as well as experimental data to identify a promising RANS setup for the best representation of the tip leakage vortex.
TRACE – Non-Reflecting Boundary Conditions
For the numerical simulation of turbomachinery flow, non-reflecting boundary conditions at the inlet and outlet boundaries are indispensable for the realistic representation of fluid physics. In the case of stationary simulations, non-reflecting boundary conditions are also required at the coupling planes of the grid rows, also known as mixing planes.
A family of such non-reflecting boundary conditions for turbomachinery simulation has been proposed by Giles [1988, 1990, 1991]. The conditions are based on a characteristic analysis of the flow state at the boundary based on the linearized two-dimensional Euler equations. Due to the wave nature of the conservation equations, incoming and outgoing disturbance waves are distinguished in a modal approach, so that flow conditions can be prescribed at the boundaries in such a way that artificial reflections are prevented.Copyright: © RWTH Aachen | IST
Stationary Non-Reflecting Boundary Conditions
The original formulation of the Giles boundary conditions was done for a vertex-centered solution scheme in which the boundary of the computational domain is formed by vertices of the computational mesh. The flow state is therefore known on the boundary.
For the application of the Giles boundary conditions in a cell-centered solution scheme, as it is used in the CFD method TRACE of the DLR Institute of Propulsion Technology, the theory has to be adapted.
In a research project, a physically correct reconstruction of the flow variables on the boundary, taking into account the characteristic wave propagation, was derived (Robens). In the flow solver TRACE two reconstruction methods were implemented, which differ in how the propagation of the flow between the cell centre of the boundary cell and the boundary itself is modelled: On the one hand, a simplified modelling based on the linearized, one-dimensional Euler equations (referred to as Characteristic Boundary Conditions) and on the other hand a more precise modelling based on the linearized, two-dimensional Euler equations (referred to as Modal Boundary Conditions).
The validation on relevant test cases showed the higher accuracy of the two-dimensional, modal reconstruction.Copyright: © RWTH Aachen | IST
Higher Order Transient Non-Reflecting Boundary Conditions
The transient, non-reflecting boundary conditions described by Giles on the basis of the linearized, two-dimensional Euler equations are non-local due to the involved integral transformations of the flow state: On the boundary, a spatial Fourier transformation in circumferential direction and a temporal Laplace transformation are performed.
Especially the temporal non-locality is a major hurdle for the implementation of these boundary conditions in a time domain solver, because the history of the flow state would have to be stored on the boundary.
By approximating the exact theory, Giles succeeded in localizing the boundary conditions in time. However, the resulting loss of accuracy of the boundary conditions can lead to non-negligible reflections in case of aeroelastic and/or aeroacoustic simulations. In a research project, higher-order non-reflecting boundary conditions were implemented in the time domain solution module of TRACE (Henninger ).
For this purpose, the exactly non-reflecting boundary conditions were localized in time by introducing additional functions on the boundary according to Hagstrom et al . The accuracy order of the boundary conditions and the associated runtime costs can be controlled by the user via the number of additional functions considered.
The validation by means of turbomachine-specific test cases showed the superiority of the higher-valued Hagstrom boundary conditions compared to the approximate Giles boundary conditions for aeroacoustics and aeroelastic problems.
S. Henninger (2019). "Zeitbereichsimplementierung höherwertiger nichtreflektierender Randbedingungen für die Simulation instationärer Turbomaschinenströmungen", Dissertation, Institut für Strahlantriebe und Turbomaschinen, RWTH Aachen, 2019
S. Robens (2015). "Stationäre nicht-lokale Randbedingungen für zell-zentrierte Schemata und integrale Bilanzierung von Casing-Treatments in Turboverdichtern", Dissertation, Institut für Strahlantriebe und Turboarbeitsmaschinen, RWTH Aachen, 2015
M.B. Giles (1988). "Non-reflecting boundary conditions for the Euler equations", Techn. Ber. MIT, CFDL-TR-88-1.
M.B. Giles (1990). "Nonreflecting boundary conditions for Euler equation calculations", In: AIAA Journal 28, S. 2050–2058.
Giles, M.B. (1991). „UNSFLOW: A numerical method for the calculation of unsteady flow in turbomachinery“, Techn. Ber. GTL Report 205.
Hagstrom, T. und J.W. Goodrich (2003). „Accurate radiation boundary conditions for the linearized Euler equations in cartesian domains“, In: SIAM Journal on Scientific Computing 24.3, S. 770–795.