[ledelman]

I am working on developing an experimental setup in a wind tunnel. To do so I am attempting to use SU^2 to determine the geometry of a channel floor hump that will have boundary layer separation. To do so I am first trying to approximate the BL behavior in the tunnel with a 2D unstructured simulation. I am meshing using Pointwise and then passing it through to SU^2 for simulation.

I have been having trouble getting the flow conditions to work properly with a total conditions inlet BC. The simulation runs perfectly but I am unable to get the desired tunnel flow rate using this BC definition. I recently discovered the Mass Flow inlet BC definition but this now results in the simulation crashing with a "CSysSolve: :modGramSchmidt: w[i+1] = NaN" error. I assume this has to do with a non-realistic property occurring (i.e. negative density or pressure). The iteration the simulation crashes on changes as I tweak the BC's (density and velocity magnitude) but it inevitably crashes.

I have not been able to find any examples or documentation of the Mass Flow inlet type in the SU^2 user guide, tutorials, or github. I have attached a text version of my .cfg (forums wouldn't permit the direct upload of the .cfg) and a .jpg which shows the grid and setup. I can provide the grid via email if you want since I can't upload it here.

I am fairly new to CFD and SU^2 and have managed to pickup the basics of SU^2 however this issue has me stumped. Any and all assistance would be greatly appreciated.

[ledelman]

Attached is my most recent attempt. It may be closer or further from correct but it is still exhibiting the same error.

[ledelman]

Here is the run window for the attempt associated with that .cfg. It doesn't get to a point where it writes any results to a history file or otherwise.

The point at which it crashes like this varies when I tweak the density and velocity magnitude on the inlet BC. (~20 - 280 iterations).

Thanks in advance for any help. I am still unable to upload the mesh do to file type restrictions.

[liujmljm]

Originally Posted by ledelman

I am working on developing an experimental setup in a wind tunnel. To do so I am attempting to use SU^2 to determine the geometry of a channel floor hump that will have boundary layer separation. To do so I am first trying to approximate the BL behavior in the tunnel with a 2D unstructured simulation. I am meshing using Pointwise and then passing it through to SU^2 for simulation.

I have been having trouble getting the flow conditions to work properly with a total conditions inlet BC. The simulation runs perfectly but I am unable to get the desired tunnel flow rate using this BC definition. I recently discovered the Mass Flow inlet BC definition but this now results in the simulation crashing with a "CSysSolve: :modGramSchmidt: w[i+1] = NaN" error. I assume this has to do with a non-realistic property occurring (i.e. negative density or pressure). The iteration the simulation crashes on changes as I tweak the BC's (density and velocity magnitude) but it inevitably crashes.

I have not been able to find any examples or documentation of the Mass Flow inlet type in the SU^2 user guide, tutorials, or github. I have attached a text version of my .cfg (forums wouldn't permit the direct upload of the .cfg) and a .jpg which shows the grid and setup. I can provide the grid via email if you want since I can't upload it here.

I am fairly new to CFD and SU^2 and have managed to pickup the basics of SU^2 however this issue has me stumped. Any and all assistance would be greatly appreciated.

I think your grid is not suitable for viscous computation. I am afraid your grid is worse.

[ledelman]

The y+ values are all <=1 which makes it suitable for resolving a BL. Also, that wouldn't explain why the same grid works fine (and produces reasonable results) for the total conditions BC.

Though I may be ignorant some some detailed fact. Why isn't the grid suitable and what would make it suitable?