API TANK vs. Wind using Algor

[jcoots]

I've been trying to model wind effects on an API 650 above ground storage tank. The model is simple enough and runs fine, however the results I'm getting are way off the mark. Has someone done this successfully before?

I've got modeled a 60' diameter x 60' tall open top tank. I've got the bottom of the shell fixed representing about 3ft of water in the tank. I have no stiffening rings anywhere. I've used thicknesses 1/4" and 3/8". The deflections I'm getting are of the order x10^-2 (0.001). I know this is incorrect, otherwise API wouldn't have us use stiffening rings, aside from what I've seen in person. The wind pressure in this case 35 psf, just above 100mph.

I think the problem with my model is trying to buckle inward a perfect cylinder. Any advice? I think that adding a dummy load to get the buckling started may work, or adding imperfections to the shell may work, but at this point how confident can I be in my results if I'm guessing and adding non-real initial loads?

 

[GBor]

Can you upload an image of your model? How fine is your mesh and have you performed a mesh sensitivity? Are you running linear static analysis, because if buckling is an issue, you won't see this in linear results? I'm not sure your fixity is correct, either...I might pin it instead of fixing it because the water is going to provide a minimum of support.

 

[GBor]

Some additional thoughts have occurred to me:

Are you using feet for all of your units?
Are you applying your pressure as a traction so that it acts more like wind with a global direction? (adding "pressure" makes the vector perpendicular to the shell element)

If you input 0.25 for the thickness, but modeled the tank in feet, you have 3" thick walls, which may not deflect quite as much as you expect.

 

[jcoots]

Attached is a screen shot of the mesh I am using. This particular analysis I had a rectangular door cut out of the bottom. In this case there was a max of 3" deflection at the top, but still nothing close that I am expecting to see.

I have not done a mesh sensitivity study.
I have tried non-linear with much of the same results.
I have only used fixed bottom conditions so far. I will try pinned, non-linear as soon as I can get back on it.
I have been using tractions in psi. The main wind force is about 35 psf = 0.243 psi. If my understanding is correct for an open top tank I should have a windward pressure, about the same pressure, same direction on leeward side, and pressures acting perpendicular the windward wind, pulling the middle away from the tank. I tried to mimic this roughly having 4 different surfaces.

jcoots

 

[GBor]

I doubt you are getting convergence in your answers with that coarse of a mesh. I know the software may run and "converge" within your criteria, but that isn't necessarily the type of convergence to which I'm referring. I would double the mesh and see if you get a similar answer...which I suspect you will not. Continue to double the mesh until your stress doesn't vary more than about 10%.

If you get a chance, upload an archive of your model without results and I'll look for any other issues.

Oh, and when you look at your results and show the load vectors, are they all pointing the correct direction?

Not sure I completely understand what you are saying about the leeward vs. windward side. What I think you are saying doesn't sound right to me, but you seem to understand your problem statement, so I'm assuming I'm just not picturing your statement correctly. If you can upload a picture of the results screen (no displacement or stress...just the FEA objects turned on) that would probably be helpful as well.

 

[jcoots]

Thanks for your advice. I have uploaded the archive model. I haven't had time yet to try your suggestions. I hope to get to it this week.

I do think all the load vectors are in the correct direction.

 

[GBor]

I think you have to rename the file to .zip instead of uploading the .ach file. Each way is probably just to zip up the archive. The way it is now, I don't think there is anything to download.