Modeling of a Tank supported by a ring girder sitting on a tower (Von-Mises Stress) 2

[Bridge_Man]

Hi,

I am struggling with this one. I have this very old tank (100+ yr) that is still functioning, so we know its OK. I need to perform some assessments on it. Am modeling it using SAP2000

The model consists of the tank modeled as shell plates, a riser that goes up to the tank (also shell plates) a ring girder that surrounds the tank at a certain elevation (frame members connected perpendicularly to the tank through rigid links), and a tower that supports the ring girder and the tank along with it.

Global analysis and reaction all made sense and were within a small % of expected reactions from Water, Dead, wind , etc...

Here is my problem, according to one code am following, I do need to use the Von-Mises stresses and compare them to the allowable stress, the problem is my steel is 30 ksi and the stresses am getting are relatively high (goes up to 60 ksi close to concentration locations where the tower legs meet the ring girder and where the tank bottom meets the riser) I know these stresses are not realistic, otherwise, this tank wouldn't be here today. how should I be looking at this.

Appreciate the help.

 

[JStephen]

"still functioning, so we know its OK". Well, no you don't, generally things that fail are fine until they fail.
It is entirely possible that minor yielding has taken place in areas in the structure where you're calculating high stress.
The original design was likely done by pretty crude methods, and I wouldn't expect a finite element analysis to duplicate those results.
In later years, strain-gauge surveys might have been included, but I don't know that they did that 100 years ago.
I am curious as to which code specifies criteria for 100-year-old tanks, though.

 

[bones206]

Rigid links connecting at a single node to plate elements are expected to present stress concentrations in FEA. You can try using a very fine mesh at the connection regions, and/or averaging out the plate stresses over a reasonable area adjacent to the connection node.

Another thing to try is a rigid beam element parallel to and coinciding with the plate edge, to mimic the actual length over which the load is being delivered. Just a way to distribute the concentrated load from the single rigid link connection node out to multiple plate nodes.

 

[Geoff13]

Wow.  100+ years.  Perhaps you've got your money's worth and it's time for a new tank?  (I'm a tank builder, so I'm biased in this direction.)  Love to see a photo of what you've got.  I'm imaging one of our water towers from 100 years ago.

Technically you don't "know its OK" merely because its standing after 100 years.  All that shows is it hasn't failed.  For me to "know its OK" would mean determining its got 10 or 20 years of life after a very detailed inspection.

Applying FEA to a design done with a slide rule and the builder's previous experience seems inappropriate.  I'm almost surprised you've found so few problem areas, but of course if there were lots of problem areas in the tank you wouldn't be asking this question 100 years later!

I assume you're attempting to justify continued use even though there's a notable amount of corrosion.  But in that case I'm back to my first comment.

Your FEA may be correct for day one, but perhaps the plates yielded or the riveted joints slipped a little upon first loading, and the tank has assumed an optimized shape for the past 100 years.

Good luck.

 

[HTURKAK]

I agree with JStephen (Mechanical) and Geoff13 (Structural) and BPS for their posts..

WOW to (100+ yr) old tank still functioning.. But this does not mean the tank has still substantial remaining life.. It is a long procedure to check for fitness for service and this tank IMO, deserves the right to stay at a Industry Museum for the remaining life.

My suggestion will be , abandon this tank and construct a new one complying the latest AWWA codes .

 

[canwesteng]

We would have to see where the stress concentrations are to give some feedback, but if you are modelling the tower as beam elements, with a fine enough mesh your stresses will approach infinite.

 

[rb1957]

high stresses ? from a linear FEA ? who'd've thunk it ? run non-linear ...

else ... highly localised high stresses ? make an argument about local plasticity.

do a hand calc ?? with a bunch of reasonably conservative assumptions

another day in paradise, or is paradise one day closer ?

 

[bones206]

Here's a quick example of what I was describing as a way to distribute the rigid link reaction into the shell. It's a little crude but you can use your judgement on how well it matches the real load distribution at the boundary conditions.