As part of the OC7 Phase III WP 3.1 project, I'm simulating a 1:70 scale rotor. The system rotates at 240 rpm and the rotor diameter is 2.38132 m. The mean wind speed considered is 4.19 m/s and I use a turbulent wind speed representative of the wind tunel measurements.
I observe the expected rotor thrust force and torque, but somehow, after 32 seconds I observe a significant drop in the rotor loading.
I performed a large number of sensitivity analysis:
- Changed the number of near-wake panels.
- Changed the distribution of free and frozen panels in the near wake.
- Ran with and without far wake.
- Changed the wind fom unsteady to steady.
- Included/excluded the tower shadow.
- Included/excluded the unsteady aerodynamics in the polars.
- Used constant polars instead of interpolating between Reynold numbers.
- Ran with/without parallel solver (OMP).
- Ran with single and double precision solver.
However, the behavior was always the same: after 32 seconds, the rotor loads dropped. For reference, I always use a time step of 0.002 seconds that corresponds to 2.88 deg azimuth in the rotor. This is already pretty demanding from a computational point of view.
Looking at the wake behavior, I can see that the near-wake clearly behaves different after 32 seconds.
The interesting part is that if I reduce the time step in half (from 0.002 s to 0.001 s), the same phenomena appears but at 16 s instead of 32 s. This indicates that, for some reason, the phenomena is triggered when the step count is close to 16,000 (32/0.002=16/0.001).
@ebranlard: do you know what could be triggering this?
For reference, I attach the comparison between OLAF (orange) and BEMT (blue) below:
As part of the OC7 Phase III WP 3.1 project, I'm simulating a 1:70 scale rotor. The system rotates at 240 rpm and the rotor diameter is 2.38132 m. The mean wind speed considered is 4.19 m/s and I use a turbulent wind speed representative of the wind tunel measurements.
I observe the expected rotor thrust force and torque, but somehow, after 32 seconds I observe a significant drop in the rotor loading.
I performed a large number of sensitivity analysis:
However, the behavior was always the same: after 32 seconds, the rotor loads dropped. For reference, I always use a time step of 0.002 seconds that corresponds to 2.88 deg azimuth in the rotor. This is already pretty demanding from a computational point of view.
Looking at the wake behavior, I can see that the near-wake clearly behaves different after 32 seconds.
The interesting part is that if I reduce the time step in half (from 0.002 s to 0.001 s), the same phenomena appears but at 16 s instead of 32 s. This indicates that, for some reason, the phenomena is triggered when the step count is close to 16,000 (32/0.002=16/0.001).
@ebranlard: do you know what could be triggering this?
For reference, I attach the comparison between OLAF (orange) and BEMT (blue) below:
