Bug description
When running OLAF within standalone AeroDyn with OpenMP parallelization enabled, the simulation results are non-deterministic across runs.
Occasionally, when starting a new simulation after a previous one has already completed in the same console session, the console displays a warning about vortices detected below the ground. This occurs within the very first few time steps. But it's physically impossible for blade-convected vortices to reach the ground in such short time (under 0.002 seconds). The warning reports 60 vortices below ground, and the simulation crashes.
Looking at the output file, I can see that the second time step already writes NaNs in many of the outputs:
However, if I close the Windows command prompt and rerun the identical case in a fresh console, the simulation runs normally with no warnings.
Since the issue only appears when a simulation is run after a prior one in the same session, this seems to suggest a state variable or buffer used by OLAF's OpenMP threads is not being properly reinitialized (e.g., it is keeping old data) between runs.
OpenFAST Version
OpenFAST v5.0.0.
Bug description
When running OLAF within standalone AeroDyn with OpenMP parallelization enabled, the simulation results are non-deterministic across runs.
Occasionally, when starting a new simulation after a previous one has already completed in the same console session, the console displays a warning about vortices detected below the ground. This occurs within the very first few time steps. But it's physically impossible for blade-convected vortices to reach the ground in such short time (under 0.002 seconds). The warning reports 60 vortices below ground, and the simulation crashes.
Looking at the output file, I can see that the second time step already writes NaNs in many of the outputs:
However, if I close the Windows command prompt and rerun the identical case in a fresh console, the simulation runs normally with no warnings.
Since the issue only appears when a simulation is run after a prior one in the same session, this seems to suggest a state variable or buffer used by OLAF's OpenMP threads is not being properly reinitialized (e.g., it is keeping old data) between runs.
OpenFAST Version
OpenFAST v5.0.0.