Compressor CFD Analysis — Numerical Choke and Back Pressure Control Tips
You can find more comprehensive content at turbocharger-cfd.html
Numerical Choke Issue
What is "numerical choke" in compressor CFD?
A phenomenon where the mesh throat area does not match the actual geometric throat area, causing choke flow rate to deviate. Especially with TurboGrid H-type topology, if the leading edge corners are too smooth, the throat becomes effectively narrower, leading to underpredicted choke flow.
How do I address it?
Make the O-grid around the leading edge sufficiently fine, and accurately reproduce the blade surface curvature. Changing to J-type or L-type topology can also improve the situation. Compare the calculated choke flow rate with the theoretical value from the geometric throat area to confirm the deviation.
Back Pressure Control Techniques
Are there tips for adjusting back pressure when obtaining the characteristic curve?
There are several techniques.
1. Restart from the previous converged solution: Do not raise back pressure all at once; use the solution from the previous operating point as the initial value
2. Back pressure ramping: Gradually increase back pressure over timesteps (Expert Parameter: pressure ramp)
3. Switch to mass flow control: Near surge, stabilize by setting the outlet to mass flow specification (but the flow-pressure relationship is lost)
4. Throttle model: Set up a virtual throttle valve at the outlet to indirectly control back pressure
What exactly is a throttle model?
A virtual orifice with opening area $A_{th}$ is placed at the outlet, and the relationship between flow rate and back pressure is explicitly given. In CFX, it can be implemented as a combination of Opening BC and user functions. It can simulate the system volumetric effect, yielding behavior closer to transient surge phenomena.
Convergence Criteria
How should I judge convergence in compressor CFD?
Residuals alone are not enough; monitoring physical quantities is crucial.
| Monitor Quantity | Convergence Criterion |
|---|---|
| RMS Residuals | Below $10^{-5}$ (steady-state), below $10^{-4}$ per timestep (unsteady) |
| Mass Flow Inlet-Outlet Difference | Within 0.1% |
| Stage Pressure Ratio | Stable with fluctuation within 0.1% |
| Stage Efficiency | Stable with fluctuation within 0.1 percentage point |
| Torque | Change rate within 0.1% over final 100 iterations |
Is it acceptable if residuals do not fully decrease, as long as physical quantities stabilize?
At the Mixing Plane interface in turbomachinery, residuals often plateau at a certain level. In such cases, it is practical to judge convergence based on the stability of physical quantity monitors.
Compressor Wheel — Tracking the "Whee" Sound via CFD
The most commonly reported turbocharger malfunction is abnormal noise occurring at specific RPM. To track this NVH (Noise, Vibration, Harshness) issue with CFD, steady-state analysis is insufficient; unsteady (time-dependent) analysis is required. When analyzing pressure fluctuations near the blade passage frequency (BPF = RPM × number of blades), characteristic resonance patterns emerge when splitter blade placement is inadequate or blade count ratios match particular configurations. At one manufacturer, changing the configuration from 6 main blades + 6 splitter blades to 7+7 eliminated the problematic sound — the unsteady pressure spectrum from CFD became the decisive factor for design modification.
Related Topics
Detail
Error