Turbine CFD Analysis
Turbine CFD: Theoretical Foundations
Overview
What's the difference between turbine CFD and compressor CFD?
Turbines are on the side that extracts energy from the fluid. Because the flow accelerates, large-scale separation like in compressors is less likely to occur. Instead, blade cooling, secondary flow losses, and transonic shock waves become the main challenges.
Stage Work and Isentropic Efficiency
How is turbine work expressed?
Output and efficiency are defined from the Euler equation.
$h_{02s}$ is the enthalpy after isentropic expansion. For modern gas turbine HP stages, $\eta_{is}=90\sim92\%$, and for LP stages, $88\sim90\%$ is the current design standard.
Blade Loading Coefficient
How do you evaluate the magnitude of blade loading?
The Zweifel blade loading coefficient is the standard.
$s$: pitch, $c_x$: axial chord. $Z_w \approx 0.8$ has been considered the traditional optimum, but in recent high-load designs, $Z_w > 1.0$ is also being researched.
Software Selection
What software is used for turbine CFD?
Ansys CFX + TurboGrid is the most widely used among aero-engine manufacturers. NUMECA FINE/Turbo is efficient for multi-stage turbine setup and is used by companies like Rolls-Royce. STAR-CCM+ has strengths in CHT (Conjugate Heat Transfer) analysis for turbine blade cooling.
Beyond Bernoulli — 100 Years of Hydraulic Turbine Theory
The fundamental equations for hydraulic turbines date back to Euler's turbomachinery equation (1754). This is a simple principle: "The change in angular momentum of the fluid determines the turbine output." However, in actual design, it's necessary to consider the viscous boundary layer around the blades, the centrifugal field of the runner, and the diffusion efficiency of the draft tube. Design cannot converge with 1D theory alone. Before CFD became widespread, designers combined potential flow theory with empirical loss correction factors (loss maps), repeatedly performing hand calculations and model tests to find the optimal airfoil. The process of replacing that empirical knowledge with CFD continues today.
Computational Methods for Turbine CFD
Importance of Blade Cooling
How is turbine blade cooling handled in CFD?
HP turbine inlet gas temperatures reach 1500–1800°C, far exceeding the heat resistance limit of blade materials (about 1000°C for Ni-based superalloys). Internal cooling passages and film cooling are used to lower the blade surface temperature.
Cooling Model Hierarchy
How do you incorporate cooling into CFD?
There are multiple levels depending on the trade-off between accuracy and cost.
| Level | Model | Computational Cost | Accuracy |
|---|---|---|---|
| L0 | No cooling flow (adiabatic wall) | Lowest | Baseline evaluation without cooling |
| L1 | Source Term (mass/energy injection) | Low | Rough estimate for film cooling |
| L2 | Discrete Hole (individual cooling hole BC) | Medium | Quantitative evaluation of film effectiveness |
| L3 | Resolved Cooling Holes (holes meshed) | High | Highest accuracy but high effort |
| L4 | CHT (fluid + solid coupling) | Highest | Predicts internal blade temperature distribution |
Are L3 and L4 practical?
L3/L4 for a single blade row is practically used as a Singleton calculation. STAR-CCM+'s CHT is highly rated for this application. L3/L4 for multi-stage is currently at the research level.
Film Cooling Effectiveness
How do you evaluate the effectiveness of film cooling?
It is defined by the adiabatic film cooling effectiveness.
$T_g$: Mainstream gas temperature, $T_{aw}$: Adiabatic wall temperature, $T_c$: Cooling air temperature. $\eta_f = 0$ means no cooling, $\eta_f = 1$ means perfect cooling. In CFD, it's calculated by outputting the adiabatic wall temperature on the blade surface.
Numerical Methods for Rotor-Stator Interaction — The "Interface" Problem in Hydraulic Turbines
One of the most difficult problems in hydraulic turbine CFD analysis is handling the hydrodynamic interaction between the rotating runner and the fixed guide vanes. The Frozen Rotor method is fast but ignores unsteady effects of rotor-stator interaction. The Sliding Mesh (moving mesh) method is accurate but computational cost can reach 5–10 times that of Frozen Rotor. A practical decision criterion is "the unsteadiness of the physical quantity being evaluated"—for predicting efficiency maps, Frozen Rotor is often sufficient, but for pressure pulsation or fatigue evaluation, Sliding Mesh is essential.
Turbine CFD in Practice
Turbine Blade Row Mesh
Is the mesh for a turbine blade row the same as for a compressor?
The basic structure is the same, but there are turbine-specific points to note.
- Trailing Edge Thickness: Turbine blades have very thin trailing edges (0.3–0.8mm). Sufficient cells are needed around the trailing edge in the O-grid.
- Cooling Holes: Local refinement around cooling holes is necessary for L2/L3 models.
- Transonic Regions on Blade Surface: Resolving the supersonic patch on the suction side and the trailing edge shock wave.
If the trailing edge is 0.3mm, the mesh must be quite fine, right?
For the trailing edge O-grid, at least 10 cells in the radial direction, and also refine the mesh in the wake region immediately behind the trailing edge. TurboGrid's trailing edge cutoff function can control the trailing edge shape.
Transonic Turbine Blade Row
Does turbine flow become supersonic?
In HP turbines, the blade-to-blade Mach number reaches 1.1–1.3. After accelerating to supersonic speed on the suction side, oblique shock waves are emitted from the trailing edge. Accurate prediction of this Trailing Edge Shock System, where the shock wave impinges on the adjacent blade, is key to CFD accuracy.
How much mesh is needed to resolve shock waves?
Cell size orthogonal to the shock wave direction should be less than 0.5% of the chord, with at least 10 cells before and after the shock wave is recommended. Adaptive Mesh Refinement (AMR) to concentrate mesh at the shock wave location is also effective. AMR functions in Fluent or STAR-CCM+ can be used.
Performance Prediction Accuracy
What is the accuracy of turbine CFD?
| Metric | Accuracy |
|---|---|
| Stage Efficiency (multi-stage) | ±0.5–1.5 points |
| Blade Surface Pressure Distribution | Good (qualitatively matches experiment) |
| Blade Surface Heat Transfer Coefficient | ±10–20% (depends on turbulence model) |
| Trailing Edge Shock Wave Location | ±2% of chord |
Related Topics
Experience the theory firsthand with interactive simulators in this field
All Simulators