Allowable Stresses in PT Slabs

[CURVEB]

I'm somewhat new to PT design, and have a question about service stresses. We use an FEM program to do final calculations (after performing preliminary calcs by hand). I often find that we will end up with a design section or 2 that exceed permissible stresses, just slightly. Our typical condition is a 2-way slab, where the limit is 6*sqrt f'c in tension (per ACI 318-11, 18.3.3).

In some cases, we are only exceeding the limit by ~5% (IE our allowable might be 464 psi, but we have a section at 488 psi), although sometimes is it more - up to 15% or so.

1) Do you ever ignore these over-stresses? Is there a limit under which you would consider it more of a local anomaly?
2) What would be the practical issue associated with exceeding the allowable stress? I assume cracking of the slab and potential re-distribution of moments, but if the slab has adequate strength isn't this just a serviceability issue?
3) I have found that adding tendons does not often help this issue significantly. I have to add a group of 3-4 tendons to have even a small impact. From a design standpoint, how do you typically make adjustments to get the slab stresses back within limits?

 

[rapt]

If you are designing 2 way slabs by the ACI318 "average moment design" method then, if the tension stress is 6*sqrt f'c in tension, the slab is actually cracked in reality as the real tension stress in the zone over the column will be significantly higher than that limit.

The ACI's misguided limit of 6*sqrt f'c in tension is an absolute maximum. It should be much lower to ensure that the section is not cracked in the "column strip width".

You should never exceed this limit.

And you should be really worried about actual deflections as the calculated values you are getting from your software are meaningless if uncontrolled cracking and redistribution of stresses has occurred at service stress levels. The FEM expected calculations will be meaningless.

Further, if your support pattern is not orthogonal, using FEM to analyse for the moments and stress can be very misleading. You are basically basing a yield line design on a completely unrelated elastic analysis. There is no logical way they can be related if the column layout is random.

PS if you are new to PT design you should not be using software until you understand the design logic.

 

[canwesteng]

From my understanding OP is not designing the floor using column strips, but using FEA and Wood Armer moments. If this is the case, depending on how fine your mesh I would have no problem with localized stress 5% over cracking moments, given that the design live load should only be temporary and after removal of live load the prestress should close the cracks back up.

 

[CURVEB]

canwesteng/Rapt: you are correct in my intent with the question. But I'm guessing that won't make much of a difference in Rapt's response. I have followed several of his posts on this issue, and found these to be pretty descriptive:

http://www.eng-tips.com/viewthread.cfm?qid=38479

http://www.eng-tips.com/viewthread.cfm?qid=430280

Thanks for the feedback.

 

[rapt]

CanWestEngCurveB

If you are checking the maximum stress at any point in the slab, then no, my comments do not apply. But that is not what 18.3.3 in the 2011 code applies to.

I assume as the design is based on ACI318 that the stresses are being checked over the full panel width, so an average value over the full width of the slab (not sure why you would use FEM in that case, 2D will give the same results) for the total panel.

In that case, the real stresses near the support are much higher than the average design value. I do not care how you analyse it.

And no, you should not allow over-stress above these values, you should actually be using about half of the value in support regions to ensure it remains un-cracked if you are using the average stress in the panel.

Also, make sure you are including Wood Armer moments. If using RAM Concept it does not include them automatically (unless they have changed the defaults recently). You have to tell it to. Not Sure about Adapt FEM these days on this.

 

[CURVEB]

We use Adapt. There is an option to turn on or off Wood Armer moments, but the technical note below clarifies how they determine average moments, and it sounds like it differs from some other software packages in that respect. I'm not experienced enough with Concept to comment on that program.

http://www.adaptsoft.com/forum/bijan/TN315_safety_consideration_floors_091708_2.pdf

 

[rapt]

CurveB,

Good that Adapt is using a method that automatically allows for twisting in elements. Too bad some others do not, as it should not be optional.

But, you are still missing the point. You are calculating really accurate elastic stresses at every point in the slab. That is great. Then you are averaging them over a panel width of 25' or 30' or 40' or whatever. The result is no different for a rectangular grid than I would get from my 2D program.

At the critical section at the face of the support over this panel width, if the average stress is 500psi, the stress near the support could easily be 1000psi and the stress at midspan about 200psi. But the average is ok according to the code. How does the concrete now it is supposed to be acting like the average? The slab experiences the elastic stresses and cracks where they are too high.

The problem is that the ACI PT flat slab approach is using a pseudo yield line design to satisfy ultimate, one that requires a proper load balance tendon layout, either banded/distributed or col/mid strip for the 2 directions. Distributed in both directions is not allowed.

But service stresses are dependent on the elastic stress distribution, not the yield line result. And designers are basing the service design on the pseudo yield line result which requires enormous redistribution. This redistribution can only occur through cracking and deflection. The only way to stop the redistribution is to keep the maximum (not average) concrete stress well below the cracking stress, whatever that is. That is why British TR43 1st edition limited the stresses near the support to 0 based on the average stress over the full width. That is NO tension.

The second part of my discussion is that the pseudo yield line method in ACI cannot be based on elastic analysis. It works ok if the column pattern is rectangular and regular in both directions. If the columns are offset, FEM elastic analysis should not be used. You have to loo at each design strip and its loading in each direction independently, ie in 2D.

 

[CURVEB]

Rapt - I understand your point about average vs. maximum stress. The design strip method doesn't show you the maximum stress - just the average, so you would need to review a different result to determine the maximum stress.

Question: In your 2D analysis do you do anything to account for 2-way action, or do you take the load over an entire strip?

 

[rapt]

We divide into column/middle strips, so yes there is still an average but over a column strip width, not the whole slab width. It is impossible to design for the extreme peak that FEM shows (and does not really occur). No one would suggest that.

And then 25% of the total negative capacity should be concentrated within D either side of the column, accounting for the even higher moment there.