Large Concrete Tank and STAAD Model 1

[vincentpa]

I am working on a project with a very large concrete tank: 46’ x 46’ x 24’ high. (I know the tank is too large to be rectangular and should be round but the process guys would not budge.) The tank walls are 2 feet thick at the base and the base slab is 2 feet thick. The subgrade modulus is 150 pci. We performed a STAAD model with spring supports based on the subgrade modulus at every node on the plates comprising the base slab. The nodes at the corner plates of the base slab are fixed in connectivity to the nodes of the perpendicular plates of the wall. However, the spring support allows rotation. The results of the fixed connectivity nodes, spring supports and the subgradge modulus give results and wall moments closer to a wall modeled with pinned supports at the bottom (as if I had pinned supports at the edge nodes at the tank base instead of spring supports). This gives me some concern since all of the literature I have tells me that the moments in the wall should be closer to a fixed support condition at the bottom (as if I had fixed supports at the edge nodes of the tank base instead of the springs). I am only getting a maximum vertical deflection of -0.37 inches at the edge in the center of the wall. At the middle of the tank base slab, the vertical deflection is -0.17 inches due to the weight of the water. There is no rotation between the bottom plate of the wall and the edge plate of the base slab. What this is showing me is that the wall and the slab are rotating as one without any differential rotation between the bottom plate of the wall and the edge plate of the base slab to cause high moments in the base of the wall. Should I believe this? I know that spring supports work well for mat foundations where the loads are only vertical. But, what about a situation where tank walls impart not only moment from the horizontal water pressure but a high vertical load from the weight of the concrete to the mat edges? The tank example in STAAD shows every base node support as pinned. I’m not sure that is correct either. Do you have any advice?

In the past for concrete tank design, I have used PCA’s Rectangular Concrete Tanks by Javeed A. Munshi. He and others I have talked to recommend using design moments for the wall based on a fixed support for the base negative vertical moment and a pinned support for the vertical positive moment. This is conservative. But, since we have STAAD and can model spring supports with the subgrade modulus, I thought to try modeling the tanks in STAAD and get a more “accurate” response. Is the spring support allowing too much of a rotation that will not happen in real life? Is this the reason I am getting wall moments closer to a pinned condition?

 

[miecz]

I have modelled a few squatter tanks with finite elements, 24 ft tall, 50 to 120 feet long, half as wide as long. My results have been closer to the fixed end results than the pinned end results from Munshi's tables, but that could be because of the squat geometry.

I determine wall bending moments two ways, and design for the more consevative result.

1. Munshi's tables for a rectangular tank.
2. Finite element model on springs.

I find Munshi's tables are somewhat more conservative for base bending, tension and shear, while finite elements are more conservative for corner bending, tension and shear.

Perhaps your springs are too soft or your concrete too stiff. What are your spring constants? Are you allowing the concrete to deform in shear? Did you adjust your concrete stiffness to account for cracked concrete?

 

[vincentpa]

The springs I am using are 100 k/sq. ft./ft. That is between 150 and 200 pci. I will check the other things in the model. I didn't do the model. I am just checking it. But I have found many inconsistencies with it. I tried fixing the base nodes and then pinning them to compare against the spring model as a check. I haven't done very many tanks yet so I don't have enough experience to make a judgement on a computer model. I have only designed just for the two worst cases in the past.

 

[miecz]

For 150 pci, I get 260 k/sf/ft. My finite element program (I don't use STAAD) has the spring constant units in k/in. To simulate cracked concrete, you can factor down the elastic modulus of the concrete.

 

[kslee1000]

I suggest to check the deflected shape to see if that makes sense per loadings. You may find big surprises/modelling errors much quicker this way.

 

[vincentpa]

The deflected shape is fine. I have checked it. It appears that the springs on the edge are just too weak and allow the entire wall and slab to rotate as one. This basically simulates a pin condition. I might try stiffening the springs at the edge. I need to read a little more on ACI 336 Mat Foundations. It gives some recommendations. I am using a 2'x2' plate grid. I believe that is a fine enough mesh. It doesn't appear to be a meshing problem. I increased the subgrade modulus to 500 ksf/ft and I am getting a little better results. I'm just a little concerned how it is going to act in the "real world." Sometimes we can get to caught up in our fancy computer programs which model very nicely in theory. It is an interesting problem though. I wonder if there is any research on the topic.

 

[kslee1000]

The base of the wall and the edge of the slab (two elements) share the same node (2D), therefore, they rotate at the same rate to maintain the fixdity (90 degrees) - the end rotation for both shall be zero. The rotation of the support node is a function of the rigidity of the supporting material, as you have already noted in your analyses, as well as the moment in each elements. However, talking about "real word", or "exact solutions", we may never know until we can reproduce the "real/exact" on-site soil conditions. Even with that, the old methods still will carry weight as they are time tested and proven.

 

[miecz]

vincentpa-

Wow. That's quite a jump in spring constants! I think you may have gone overboard there. I would stick in the 250-300 ksf/ft range. Try cutting your elastic modulus to simulate cracked concrete. That should send more of your force to the base mat. I think a 2'x2' grid is too coarse. I use 1'x1'. Also, I hope you are using elements outside the tank walls to simulate your extended footing. You'll want at least a foot of mat outside the walls to develop your top bars, and that will stiffen your mat. I share your doubt about the validity of the finite element model. That's why I use the book as well, and design for the greater forces of the two methods. Make sure your springs are compression only.

Finally, a word about tension steel. Munshi's book suggests dividing the tension evenly between the two layers of reinforcing. If one of my layers has twice the reinforcing of the other, I proportion the tension accordingly.

 

[Qshake]

Here in again is the classic myth of performing a FEA....why is anyone tweaking the model to fit a conservative publication from yesteryear?!

I have run a great of FEA and while it's nice to have a reference that allows you some window of maximum or minimum values, but a FEA should be based on as close to proper fixity or boundary conditions as is possible and to use the proper elements along with the appropriate meshing if a h or hp program is being used. P-FEA is not based on meshing but has limitations as well. The FEA should not mimic the reference.

One thing is certain here, you're making modifications to the soil that may make the model soil conditions unreasonable....just to suit a early reference.

Use the reference or the most accurately modeled FEA not both.

Lastly if you don't fully undestand the FEA you should use the reference. And while STAAD is good for many frame problems and multibay frame problems it is not a robust FEA with a large selection of elements to use.

Regards,
Qshake

Eng-Tips Forums:Real Solutions for Real Problems Really Quick.

 

[phuduhudu]

This thread comes at a good time for me as I was about to start one on this topic too. I have an 8m x 8m x 6m high tank and I have done FE models of the walls as slabs with fixed edges on three sides. I end up with a max base moment of 150kNm or so and a corner moment of about 115kNm. From other similar tanks I have seen, the base steel is less than the corner steel, not more. I wonder if fixing the base side is realistic and whether some of that base moment is not redistributed into corner moments further down the corners.

My other question though is on the detailing. For crack control to BS8007 I am ending up with 16mm dia bars at 80mm centres in the base. Would you put a toe on the outside of the tank to help with the congestion of anchorage steel in the bottom corner? Also would you use a chamfer on the inside corners and base to avoid stress concentrations (It will make it harder to construct)? What sort of steel detail would you use in the base of the wall? I recently saw a detail with a bar in the top of the slab which did a loop round the corner and then went up the inside of the wall face which I have never seen before.

Carl Bauer

http://www.bauerconsultbotswana.com

 

[Ussuri]

80mm centres. That is really pretty close 64mm gap between bars. You might have problems getting the concrete placed and compacted properly, and a lot of pokering. Is that getting up around your maximum percentage?

I have seen internal chamfers on reservoirs before but I dont think it is for relieving stress hotspots, concrete is just not that precise a material. The chamfers (quite large) are used to prevent a nook in the corner where debris could build up, but more often than not I dont see them. They are a pain in the a**.

The detail you adopt depends on your design, fixed or pinned, and whether you have a net opening or closing moment at the joint. For a full fixed base I usually see 'L' starter bars tied into both the top and bottom mat of the base steel. I have also seen 'U' bars at ninety degrees to each other.

 

[phuduhudu]

It comes to 0.72% steel for the inside face as it is a 350mm wall. Perhaps a 20mm dia bar at 120 centres would be better for congestion but they have longer anchorage lengths. With U- bars that wouldn't be a problem.

Carl Bauer

http://www.bauerconsultbotswana.com

 

[vincentpa]

Qshake,
I was only tweaking the FEA model to see what would happen if the edges of the mat were rotating too much because of weak springs at the edge supports. I inherited a finite element model from an engineer that is on vacation now and I am trying to work it out. I don't agree with all of his assumptions and input. I am probably better off starting over with a new model.

Jmiec,
For buildings in a lateral analysis, I decrease the stiffness due to cracking for slabs by 50% (Multiply EI by 0.5) and beams and columns by 30% (Multiply EI by 0.7). How much do you generally decrease the stiffness for concrete tank walls?

To the others,
Please start another thread if you are looking for answers on a particular topic/subtopic. This thread was started by me to discuss FEA models for tanks not steel details and congestion problems in tanks. - Thanks

 

[phuduhudu]

I would have thought that you do not want to decrease stiffness for water tank design as the crack widths will be limited to retain water. On the contrary you will want to increase the stiffness on account of the steel in the section.

I read a reference recently that said the moments were not very sensitive to the relative soil stiffness and concrete stiffness.

Carl Bauer

http://www.bauerconsultbotswana.com

 

[miecz]

vincentpa-

This is one of those engineering judgement calls like the soil spring constant. A few things things to consider: Your cracked vs uncracked moment of inertia, true vs design concrete strength, tanks crack, and, there is no scientific evidence, laboratory testing or field measurements that have studied this (that I'm aware of) as related to tanks. As such, your guess is as good as mine. I reduce the elastic modulus by 30%.

 

[phuduhudu]

On the distribution of base and corner moments in the wall of my rectangular tank I did a bit of a sensitivity analysis. When I reduced the soil mat spring stiffness by half I only got a 17% reduction in the rotational stiffness at the base of the wall. This in turn gave me only a 4% reduction in the base moment. This is all for the case described in my original post. So it would seem that the model is not very sensitive to the soil mat stiffness. I imagine that larger tanks would be more sensitive.

Carl Bauer

http://www.bauerconsultbotswana.com

 

[miecz]

Carl-

From the top, I think the ratio of base moment to corner moment depends on the height to width ratio of the tank. My tanks are sqatter than yours, and my base moments always exceed the corner moments, but not by much.

You must have differerent crack control requirements than we do in the US. I would try to keep my minimum spacing to 150 mm and find the bar that works at that spacing.

I always use a 300mm to 600mm toe outside the tank to develop (with a hook )the top mat bars. Also, without the toe, I wouldn't count on face bars with 50mm cover to resist the base shear for shear friction. Use diagonal bars across the middle of the joint.

I use U bars (with a diagonal for shear friction) at the corners, and conventional L bars at the base.

I don't chamfer the corners to avoid stress concentrations.

Thanks for sharing your sensitivity analysis. I've noticed that varying the soil stiffness has little effect on bending moments, but I've never taken the time to study it as you have. If you have a reference that backs that up, I would be very interested in reading it.

 

[phuduhudu]

Thanks for the pointers. It is interesting to compare the way things are done.

I read through a number of books in the technical library over the weekend but unfortunately I can't remember the reference for the insensitivity to varying the soil stiffness. Would it not be easy for you to halve your soil spring stiffness and check the effect on the base moment?

On the shear - The base shear is less critical than base moment and with the additional steel for cracking the shear capacity of the base is plenty. Interesting about the toe because my client has had several of these tanks commissioned, and I have seen a few, and they don't have toes on them.

On cracks, we design to a max crack width of 0.2mm for liquid retaining concrete.

Carl Bauer

http://www.bauerconsultbotswana.com