Support reactions randomly reversing

[Anonymous]

Hi,

I've created a very simple diaphragm that I'm trying to represent a large rectangular roof on a pitch of about 8 degrees. I've been given a whole lot of wind loads to a completely separate grid system to ours and I want to use a finite element mesh to distribute these loads to a support system that I've defined (set of nodes).

I've modelled a dummy 100mm thick concrete slab (with no self weight case) and added a whole bunch of separate wind cases. 

When I observe the reactions, I make sense of 90% of them as just a tributary area x normal force to get the vertical component (as it's at 10 deg pitch). For the remaining 10% (seems to be at one of the edges for about 3 to 4 panels, the loading magnitude seems correct although the reaction is in the same direction as the applied force!?? Which is obviously impossible. 

Before you ask if there is a large load in the central bays and light at the ends (generating an uplift) then the answer is No. The loading is quite large at the ends comparitevely. 

When I observe deflections, the maps seem fine as well and the deflection is occuring in the direction you'd expect.

My panel directions are all identical as well.

Has anyone seen this before?

[Artur.Kosakowski]

Any ignored instability?

[Anonymous]

Not at all. Runs like a dream.

---

[Artur.Kosakowski]

Could you attach the file and indicate support node numbers and load cases to investigate please ?

[Anonymous]

Sure...

There are 32 wind load cases and they all seem to bring up similar results (reversed reaction) for nodes at the edge such as 8303, 8304, 8305 etc. The rest you can probably see visually.

Loads were input from a spreadsheet and pasted as there are so many of them.

The file is 50MB and I can only compress to 25. Do you have an email and I can westransfer it to you?

Thanks.

[Artur.Kosakowski]

Is the load direction as intended?

If you find your post answered press the Accept as Solution button please. This will help other users to find solutions much faster. Thank you.

[Anonymous]

Sure is. All load directions are as intended.

If you look at load case 23 at the nodes I've specified, you will see something funny going on.

Regards,

[Artur.Kosakowski · Accepted solution]

After running some tests my assumption is that the issue is caused by the fact that the directions of local Zs of surface elements for panels where you can see the wrong reactions are reversed comparing to the definition of local Zs of these panels themselves. The solution I found is to run local meshing for them and then calculate the model again as shown on this movie:

https://screencast.autodesk.com/main/details/64cb9e1e-70b0-44bf-88cb-e223439eb5c0

If you open your file on Robot 2015 the program will automatically correct this issue changing and change the status of the results to out of date.

If you find your post answered press the Accept as Solution button please. This will help other users to find solutions much faster. Thank you.