FEA on conrete slab - use real loads or factored loads? 2

[designsimple]

I am designing a concrete slab with many small circular openings using FEA from STAAD. I am wondering two questions.

First, what loads should be applied in FEA? Real dead and live loads (D+L) or factored loads (1.2D+1.6L)?

Second, after the FEA analysis, how to compare the compressive/shear stresses from the FEA results to the design stresses of concrete? For example, if I get a maximum shear stress of 120 psi from FEA by 1.2D+1.6L, the design shear stress is 2 x 0.75 x sqrt(fc, 5000 psi)=106 psi. This means the shear capacity is N.G. But I just don't think it is correct. I think the maximum shear stress of 120 psi in FEA may be replaced with an average shear stress say 120 psi/1.5=80 psi < 106 psi, OK.

Warmly welcome any comment.

 

[kslee1000]

1. Apply non-factored loads on the model, then use combination function to incorporate load factors to get design loads. By doing so, you can exam the service load deflections, thus have better handle on the behavior of the slab.
2. Unless specifically called for (max/min stresses/forces output), the analysis results are usually (by default) average forces lumped in the center of the elements (please read the manual to understand the commends and outputs).

localized stress concentration tends to occur around floor irregularities/openings, if the excessive stress on an element is within 10-15% of the design strength, justification may be made by inclusion of adjacent element (in a different bend/strip), which may have ample strength remains available to assist its neighbor, and averaging the forces/stresses by judgement. If this still does not resolve the stress issue, you will need to modify your design by suitable means. I recommend to consult with a senior engineer before proceed farther.

 

[JAE]

kslee1000's points are good...agree.

Also, keep in mind that your FEA would initially be done using the full thickness/stiffness of the concrete and using the service loads. In the final design state, this may not be accurate.

The service load results should be studied to determine whether the concrete cracks under the various load combinations (compare with 7.5(sqrt(f'c) = cracking stress) and include axial stresses which can drastically affect the cracking moment Mcr.

Once you determine if and where the concrete cracks, you can then adjust the elements that cracked by reducing their stiffness and re-running the model. This iterative approach will let you chase the cracks where they would occur and allow a more accurate modeling of the deflected shape and the distribution of forces (which depend upon relative stiffness).

With a finalized model, you can then apply load factors and run the model under factored load combinations to determine design forces/moments. The ACI design using load factors does not utilize stresses but rather direct shears and moments (don't know if you are using ACI or not).

Our software allows output for Vx, Vy, Mx, My and Mxy which can be directly used to determine load vs. capacity and required reinforcing.

I'll repeat kslee1000's comment: " I recommend to consult with a senior engineer before proceed farther."

 

[GalileoG]

JAE,

If I understood you correctly, you would adjust the stiffness for the portions of the slab that would crack under service loads only? I would think CRACKING on the slab would be identified under factored loads, for shear, moment AND deflection? (Though for calculating the actual deflection, service loads should be used)

I'm really confused.

Clansman

 

[JAE]

This might be a philosophical question - the concrete element you check (whether it be a beam, slab, whatever) will normally experience loads, theoretically, up to the level of the service loads. The cracking check tells you what is cracked under service levels and then would more accurately distribute loads throughout the structure and better estimate service deflections.

When you design the concrete member, you are designing it assuming a totally cracked section using moments from a structure that are distributed only via a service-cracked structure. I know this doesn't sound right but it's the way we've done it.

If you check for cracking at factored load levels, it seems to me that you are over-doing what "really" happens in real life...the factors are for safety - not for determining behavior through a stiffness matrix. The service level loads represent what is the "real life" maximum load condition.

 

[RWF7437]

As a very plain and very old Civil Engineer may I suggest an amendment to this sentence ?

" The service level loads represent what is the "real life" maximum load condition."

It may be said that the service level loads represent a good estimate of the probable "real life" maximum load conditions. That condition can only be known after the structure has been designed and built and put into use. Sometimes it can only be known after the structure has failed. Occasionally it can be known in a laboratory.

 

[kslee1000]

RWF7437: Beautiful.

 

[haynewp]

ACI allows for an elastic distribution of factored loads through the structure when designing the strength of the members. The load factors are probability derived and the structure may really see overloads (factored loads) and changes in properties in varying areas and therefore deflection and (possibly?) sizeable force re-distributions.

But with concrete, as long as the total load carrying (total positive and negative capacity) of a member is not exceeded, enough re-distribution is possible that I think an elastic analysis of forces has historically been good enough. (I may get hung here for that..)

Kind of related, I have also thought about the wind serviceability drift of concrete frames that have been cracked due to for instance, a past seismic event. I usually see wind drift based on full sections but after it has cracked, it seems the actual wind drift the structure produces should be significantly changed.

 

[kslee1000]

I believe both "Service Load" & "Ultimate Load" are derived from statical means, but determined in a different manners.

The "Service Loads" are derived through examing samples in the population of specific interests, and the higher bounds are conservatively adopted with a certainty closer to 1 - they are highly likely to occur in real life.

In the other hand, the load factors are determined by what-if scenarios at imminent failure (ultimate) stages, the factors incorporate considerations of, and representing the combination of, un-certainties about materials, construction practices, loadings...etc (the higher the uncertainty, the higher the factor). This implies in the mind of code authorities that in order to get to that (imminent failure) stage, a series of events need to happen.

I hope the above are clear, and welcome comments, or further clarifications.

 

[designsimple]

Thanks for everyone's comments, especially kslee100's and JAE's.

So I summarize the answers from all the inputs above if I understand them correctly.

1. Stregth check: 1.2D+1.6L even if using FEA method. Shear stress <=(0.75x2)(sqrt(f'c) to check shear strength and control cracks
2. Servicibility/deflection check: D+L.

Basically there is no difference between the hand calculation and FEA method anyway.

 

[rapt]

designsimple,

Remember that the deflections need to take into account
- cracking as described above by JAE and which can then be modelled as he suggested
- creep and shrinkage effects. To get this right you need to do a lot of very complicated calculation that STAAD cannot do for you or use the approximations in the code (and they are very approximate). STAADS results are not the final results.

Otherwise, yes, it is the same as designing by hand in terms of combinations, all ist does is an analysis, except that you now have Mx, My AND Mxy. Mxy cannot be ignored and must be combined with Mx and My using something like the Wood Ahmer approach to determine final design moments. I think STAAD has an option to do this for you.