[Pauli]

I am processing CFX results using CFD-Post. I want to query the forces acting at the nodes on a wall boundary. (Actually it is the torque about an axis I ultimately desire.)

I tried using the probe (Tools->Probe) with variable Force X. I displayed the desired surface patch with mesh lines and zoomed in tight to the desired vertex. When I click on the vertex I get a dialog box reporting "ERROR Force can only be calculated on vertices of a surface".

Next I tried creating a Point then changing the point from coordinates to the reported nearest node. I then created an expression "probe(Force X)@Point 1". This produces the same dialog box "ERROR Force can only be calculated on vertices of a surface".


Any ideas?

Thanks,
Pauli

[Antanas]

Originally Posted by Pauli

I am processing CFX results using CFD-Post. I want to query the forces acting at the nodes on a wall boundary. (Actually it is the torque about an axis I ultimately desire.)

I tried using the probe (Tools->Probe) with variable Force X. I displayed the desired surface patch with mesh lines and zoomed in tight to the desired vertex. When I click on the vertex I get a dialog box reporting "ERROR Force can only be calculated on vertices of a surface".

Next I tried creating a Point then changing the point from coordinates to the reported nearest node. I then created an expression "probe(Force X)@Point 1". This produces the same dialog box "ERROR Force can only be calculated on vertices of a surface".


Any ideas?

Thanks,
Pauli

force_x()@your_surface

[Pauli]

Antanas,

Thanks but that command is for a surface. I want to query at a node.

Regards,
Pauli

[ghorrocks]

Force does not exist at a point, so you cannot get the force at a point. Use the integration functions (eg areaInt()) to generate the function you are looking for.

You say you are going to eventually calculate the torque about an axis - so why not just use the torque() function and calculate it directly?

[Pauli]

Hello Glenn,

A couple years ago I created a process where I export the wall force data to a .csv file then inside Excel I compute the desired moment. This was part of a larger analysis system performing multiple data reduction strategies.

I am currently working on streamlining this analysis system. One aspect is to perform the moment calculations inside CFD-Post using torque_z()@my_surface.

When I compare the torque computed by the two methods I get similar trends and small magnitude differences. I am working to understand the source of the discrepencies.

I realize that CFD-Post has to be performing some sort of area based calculation. The forces at nodes are computed for export. I was attempting to access the same information directly inside the GUI. But I keep getting constrained by the requirement to operate on an entire surface. That makes for too many nodes to keep track of the details. And I suspect my discrepancy is buried in the details.

Regards,
Pauli

[ghorrocks]

There will be discrepencies between calculating torques by summing the nodal forces (from face area and pressure and/or wall shear), the torque() function in CFD-Post and the torque() function in the CFX solver.

The CFX solver will calculate the torques from torque() using the integration points in the finite volumes. These integration points are a higher resolution than the finite volumes so this approach will resolve the torque at full simulated accuracy.

The CFD-Post approach does not have the integration point data (from my understanding) but will use an integration scheme which I think allows linear variations between finite volume. This will not be as accurate as the solver using the integration points, but should be very close for most applications.

The approach of exporting "nodal forces" (which I assume is the nodal pressure and/or wall shear times the face area) does not account for linear variations. This approach should not be as accurate as the previous approaches.

[Pauli]

I finally found the source of my error. As is usually the case, it was the user.

The local CSYS I use for torque evaluation inside CFD-Post were slightly misaligned. Now the torque I compute with CFD-Post & Excel match.

[ghorrocks]

Did you get your calculated value in excel to exactly match the CFD-Post value? I did not expect that (ref: my previous post).

[Pauli]

The Excel spreadsheet was formatted to present results to two decimal places (4 significant digits). At that level of precision, the numbers were exactly the same.

The CFD model has a resolved boundary layer mesh - 1st layer Y+=1. I expect that helps limit the dicritization and extrapolation to the wall errors you outlined.

For now I am off to a new project. When I get back to this work I will try to remember to monitor solver computed torque in a future CFD run.

[ghorrocks]

OK, thanks. And yes, in future you should look at the torque calculated by the solver (and probably output as a monitor point) as that should be the most accurate way to calculate forces and torques.