Bug report--Error when conduct simulation using OLAF & turbulent wind (TurbSim)

[ZL6462]

Dear all,

I met a bug when conduct simulation using OLAF & turbulent wind (TurbSim). The simulation object is the monopile IEA-15MW wind turbine downloaded from "https://github.com/IEAWindTask37/IEA-15-240-RWT/tree/master/OpenFAST". The error informantion is "forrtl: severe (157): Program Exception - access violation10:53:31."

After a preliminary judgment, I think the error may occur when the wake of the rotor beyond the boundary of the wind field generated by TurbSim. I have tried to enlarge the wind field to avoid the error, but with the development of the wake, there are still some vortices enter the ground and therefore cross the boundary of the wind field, so the error occurred again.

Some Information of the executable program is given as follows:

1. OpenFAST version 3.5.0
2. Complied with "Intel(R) Fortran Compiler 1800" with OpenMP on.

Thank you for your attention.

[ebranlard]

Hi @ZL6462
If you know how to compile OpenFAST, can you also try to compile the branch rc-3.5.1 and see if you still have the issue?

If you still have the issue, could you package all your input files (including the input file for turbsim) into a zip file and share them here?

[ebranlard]

You can also compile by activating the traceback in Visual studio.
Right click on FASTLib in the solution Explorer > Properties > Fortran > Run-time > Generate Traceback Information = Yes

This might give us a better idea of where the access violation occurs.

[ZL6462]

Dear @ebranlard,

Follow your suggestion,

1. All the input files are attached here: input files.zip
2. I have activate the "traceback", and the error INFO are shown as follow, it maybe the same as my judgement, namely the wake of the rotor beyond the boundary of the wind field grid.
   
3. I have also compile the version "rc-3.5.1", and the same error occur.

Maybe, we cannot call this error a bug, because under the turbulent wind condition, it is the wake of the rotor is of course somewhat disordered, so it may beyond the wind field boundary (for example when the wake enter the ground). Even though, I still wonder how to avoid this if possible.

Thank for your attention.

[ebranlard]

Dear @ZL6462

Thanks a lot for trying all of this out.

I believe there are some bugs in the recent re-implementation of Inflowind (FlowField submodule).

In the meantime, I think that you can make your box bigger in the "y" direction (I think that's part of your issue in that case). You can also use a gridheight, such that your box starts right above the ground (for instance GridHeight=299.99)

[ebranlard]

Hi again,
I believe this is now fixed in the latest rc-3.5.1 if you want to try it out. Let me know if it works.

[ZL6462]

Dear @ebranlard,

By using the latest rc-3.5.1, the question mentioned above is solved, and maybe it is unnecessary to enlarge the GridHeight to 299.99, 299 is enough.

However, a new error called " stack overflow" occur:

May this is due to the simulation time is relative large. I have set the AnalysisTime of TurbSim to 3200s. Some supplementary of the question is:

1. When the AnalysisTime of TurbSim to 300s, everything is OK, without any error.
   2 The error have not occurred in the version 3.5.0 and the old version of rc-3.5.1(downloaded in 2023.08.25, namely before the "IfW: Bug Fix: compute vel avg profile when exceed allowed IfW: Bug Fix: compute vel avg profile when exceed allowed #1748").
2. The error may occurred during the initialization process of the program, because no .out file is generated when the error occured.

Thank you for your attention.

[deslaughter]

According to https://community.intel.com/t5/Intel-Fortran-Compiler/stack-overflow-even-with-allocatable-array/td-p/1177786, the calculation of VelAvg is causing a temporary array to be allocated which is overflowing the stack. I'm creating a PR that avoids this allocation and will let you know when it has been merged into rc-3.5.1

[ZL6462]

Dear @deslaughter, @ebranlard

Thank you for your attention, please let me know when the PR is created and merged to the rc-3.5.1.

Thank you again for your attention.

[ZL6462]

Dear @deslaughter ,

Sorry to interrupt you, I am wondering whether the bug mentioned before has been corrected?

[deslaughter]

@ZL6462 Sorry for the delay, PR #1759 has been merged to address this issue. Please give it a try and let me know if it works for you.

[ZL6462]

Dear @deslaughter and @deslaughter,

Thank for your effort.

I have using the latest rc-3.5.1, and the bug discussed above maybe corrected.
However, there is a new error, and I am not sure whether that is related to this issue. The new is reported in a new page "#1765".

Thank you for your attention.

[ZL6462]

Dear @jjonkman @ebranlard and @deslaughter ,

Sorry to interrupt you, may be there is still a error, and this issue may be need to be reopened.

By using the latest version of the rc-3.5.1 the error discussed above has been solved. However, a new error occurs after several time steps, and if we use the steady inflow rather than TurbSim FF, there is no such error.

Thank for your attention.

[deslaughter]

@ZL6462 this issue was corrected in the dev branch. I've submitted a PR to integrate the solution into the rc-3.5.1 branch.

[deslaughter]

Merged #1769 into rc-3.5.1 please see if it resolves your issue.