[DjCFD]

I am working on a drone propeller CFD project using STAR-CCM+ with an MRF (moving reference frame) RANS (Reynolds-Averaged Navier-Stokes) approach but am getting 30-40% thrust error (overprediction) from experimental data. This experimental data comes from a calibrated test stand that has been tested over and over with consistent results. We compared thrust and torque vs RPM (~1000-3000 for a freestream velocity of 0 ft/s), with results attached (along with STAR-CCM+ scenes). We started also looking into an unsteady RANS simulation but so far are getting even higher thrust results. Is this amount of error typical for this type of simulation or is there something we are missing? Do we need to go back through the 9 tests below or try something else? Thank you in advance and please let me know your thoughts!

Assumptions:

• ~22in (~0.56 m) diameter blades, dual blades attached to a motor

• Neglect the far-field reflections

• Flow is incompressible

• Steady state incompressible model, segregated flow solvers, SST (Menter) K-Omega model with all y+ wall treatment to model the turbulence

• Internal Interface Boundaries – in place

To achieve more accurate results, we conducted the following tests:

1. Various mesh sizes and prism layers

2. Adding a floor

3. Including side walls and floor

4. Enclosing the room

5. Various turbulence models and settings

6. Using a coupled solver and considering compressibility, removing segregated flow assumption

7. Pressure sensitivity

8. Adjusting interface cylinder sizing (MRF region)

9. Comparing and trying two different test stands as well as neither

Tests 1-5 showed thrust and torque results were nearly unchanged (~0.5% difference from previous CFD results) so they do not affect the simulation that much. Test 6 showed some improvement in thrust and torque by lowering the overprediction by a few percent (~39 to 33%) but not enough and required more computational resources. Test 7 showed that pressure, not dynamic viscosity of air, is the primary variable changing the outputs for different altitudes. Test 8 showed us that we needed an interface size big enough to allow the flow field to completely evolve (radius = 0.35 m, start = -0.105 m, end = 0.105 m). Test 9 showed values less than half a percent different. These tests still left us far from our data.

Q&A Images-compressed.pdf