settlement calculation based lab results.

[Balageotech]

Dear Experts,

How reliable the settlement estimation in clay using consolidation lab results?.

Regards
Bala

 

[Ron]

If you do it properly, it's as accurate as any other geotechnical computation process!

 

[oldestguy]

That's all fine and dandy assuming a uniform compressible layer. The overall amount of settlement may be close, but timing can be way off. In the many calculations I have run as compared to actual experience, only a small fraction ever came close to the timing that occurred. Usually intermittent sandy layers speeded things up drastically. That's where settlement platforms come in handy for job control.

The reverse experience was found when lab hydraulic conductivity tests tried to predict field drainage or seepage rates, due to layers of finer grained material not in the test sample..

 

[fattdad]

it's a law of averages. If you have one consolidation test, it could be way off, if you have three it'd be more accurate. If you have highly variable stress history that'd add another source of error. If you are using wick drains then you have to deal with not just a shorter drainage path, but anisotropic permeability.

So, given a very robust sampling and lab program, you could do pretty well. Cheapen the programs and the source of error goes up.

f-d

ípapß gordo ainÆt no madre flaca!

 

[BigH]

Settlement prediction is hit/miss affair. You do the best you can AND if you are within 30% of the actual settlement, you've had a good day. As OG says, timing is even more hit and miss . . .

 

[DougHole]

recommend comparing your result with..

1. the Cc/1+e0 from a correlation with moisture content (higher moisture content = more compressible)
2. Cc/1+e0 correlation with plasticity (higher plasticity = more compressible) and
3. if you have CPT data, you can correlate the constrained modulus (M) from that. then take 1/M to get mv (coefficient of volume compressibility) and calculate the settlement from s = mv x H x change in total stress.

I would compare these 3 methods with the consol test to get a feel for the range of settlement...

good luck


Doug Hole
Junior Geotechnical Engineer

 

[BigH]

DougHole - he says he has lab test results. So he should have the Cc/1+eo (instead of plotting e-logp', plot strain-logp'. Then it is a matter of how a reliable settlement he can obtain - which, in most cases also requires the Cr/1+eo value as well. Then there is this little problem of finding the past maximum pressure (vide the curve). Oh, and then there is the issue of figuring out the "real" stress distribution (elastic assumed usually). And if one was having sandy/gravely soils, then would one be better using particulate theory rather than elastic theory? Which is why - if you are within 30% of the real settlement, as I suggested, go out and buy yourself a nice bottle of single malt scotch.

One might look up the book by N.N. Som and S.C. Das "Theory and Practice of Foundation Design" (ISBN-81-203-2190-1). They have a nice chapter on settlement analyses and for the problem with the same soils the range of settlements that can be predicted.

 

[DougHole]

BigH - i agree that settlement predictions are extremely approximate but i find you can bound the problem more easily by looking at different methods. The consol test results rely heavily on obtaining an "undisturbed sample" via a shelby/push tube which we know is difficult. So using correlations to get your Cc/1+ e0 (and Cr/1+eo) is recommended. note you can estimate the preconsol pressure using shear vane results (Su) too. Also, look at the geology and aerial photos to see if there is any historical fill and to figure out if a historical higher ground level was possible. understanding the stress history is important as soil has a memory.

In addition, using the CPT based method and a correlation with the constrained modulus work well. The mv method doesn't depend on the preconol pressure.

Some other ideas...

1. could you do an instrumented test fill at the site? (not a preload) would need to install piezos/monitoring rods too
2. don't forget about secondary compression by the way. this can be significant in peat, organic clays and landfill refuse.

 

[Mccoy]

Doughole introduces sample disturbance as another possibly substantial source of error in lab tests.

mv or M' Values from CPTs are only as goods as the correlations laws are... And sometimes the benchmark is the lab tests, so what we have is really a loop. How good and how representative are the lab tests used as a benchmark for in situ tests correlations? What's the loading level?
Sometimes values are backcalculated from settlements measurements in real structures and that's better off.

The DMT test is supposed to be a reliable method to get the edometric modulus. Not without its drawbacks though. Loading level=overburden stress at the measurement depth, so Always lower than underneath the foundation!

The more I practice geotechnical engineering, the more I'm surprised by the degree of approximation of our calculations but the structures which display excessive settlements are not so many after all, a sign that there is some concealed mechanism at work which prevents big trouble. More mechanisms actually.

Sorry for my introduction into the novel science of geotechnophilosophy!

http://www.mccoy.it

 

[BigH]

Doughole - I have been involved in soils since the mid-1970s and am fully versed with correlations - 'ell, I use them all the time myself so I do agree with you that they are very useful - but I don't think that they are necessarily "recommended" over good quality sampling. They can be used to validate a "range" of possible settlements. As I had indicated earlier Som and Das computed settlements for the "same" soil using stress paths, consolidation theory, elastic theory - and the results are all different. As you and most of us are fully aware, there is this "thing" called judgment that, unlike structural and mechanical engineering, soils engineers must put up with (extreme COVs).
As a general comment, if one uses correlations, though, you really do need to know what "clay mineralogies" were used in determining them. You will get different correlations if you have illite or kaolinite or montmorillonite. Most correlations, if I remember correctly, are are kaolinite soils. And, remember that correlations are fuzzy - not a precise line.
Terzaghi Peck and Mesri mirror your comments that correlations can be used - mainly as they are faster to obtain than good quality undisturbed samples - which they do recommend if you are dealing in "new" environs. The standard Cc = 0.009*(LL-10%) is useful; so is the Figure 22.1c in Lambe and Whitman (SI Version) which is comparable to TP&M Figure 16.3 - both relate Cc (or Cc/1+eo) to water content. I find that the Figure 9.6 in David Muir Wood's book (Soil Behaviour and Critical State Soil Mechanics) is a good one - hopefully the author nor Cambridge University Press will not mind as I've attached it. This figure relates Cc vs plasticity index. He gives a nice discussion (but you need a little infinity towards CCSM) in Section 9.4 of his book and he agrees to that ". . . simple predictive charts and formulae can be produced against which one can test the results of the apparently crude index tests and other data."

 

[DougHole]

Great comments BigH and Mccoy. A couple of questions/thoughts...

1. how do i determine the clay mineralogy? is that the case of using an electron mircoscope? or just knowing the geology?

2. with Cc correlations i have always struggled with what e0 to use. I remember going to a lecture by Paul Mayne where he said you should never "separate" Cc from it's e0! what should one assume for an e0 (if they have gone down the correlations route).

3. i have put another post on Critical state soil mechanics. it has not had any replies. i wonder if you could explain it in simple terms. to this day no one has managed to explain it to me successfully!!

4. As an industry, why don't we push our clients to survey buildings to build up a database of how actual settlements compare to the predicted values. if we don't start doing this, we may never get to the bottom of our settlement prediction headaches. Has anyone done this?

5. A final point i think people haven't made is that local practice is very important in settlement predictions. Experience in the geology and the materials and knowing what has been used before and what has worked in the past in some ways trumps the fanciest of lab tests and the most detailed correlations....

Doug Hole
Junior Geotechnical Engineer

 

[BigH]

As to point 1, it is more likely, from a practicing point of view, to know your geology and what clayey mineralogy comes from the parent rocks (if residual) or from "upland" rocks if lacustrine. You can get an idea by going the activity route.

Paul, a fellow Cornellian at the same time, is quite right - don't separate eo from Cc . . . in lab testing it is an easy process for one only has to plot strain vs log p' rather than e-log p'. In correlations it is a bit more difficult - which is why Lambe and Whitman's figure I alluded to can be used since it is actually a Cc/1+e0 correlation graph. Another way it to estimate eo by knowing the unit weight of the soil, its moisture content, (estimated) specific gravity (usually 2.65 to 2.7 is sufficient) and the phase volume relationships.

Critical state is not an easy subject to get into - as for "normal" practice, it has little application, in my view although Prof Graham (of University of Manitoba) and David Muir Wood and others might/will disagree. However, there are situations where there is no real clear choice as the normal ways just don't make sense. I suggest, as a start, to get either David Muir Woods book or the book by Malcolm Bolton (A Guide to Soil Mechanics) . . . I don't think I've ever used it . . . just has never been in vogue in practice (I started out in the mid-1970s; yes, I knew about it, but the engineering firm(s) I was with never considered it.

We can push the clients to survey buildings to build up a data base - but what clients are willing to do so? If you approach a client and ask him if you could do it (for free or if you can find a gov't agency to give you money to do so), I am sure most client's would be happy to do so. Hydrodams seem to love such instrumentation for the long term - but they have the dam safety boards to appease.

As for your last point - engineers who practice in a specific area do tend to develop their owns sense/feeling of what works and what serviceability movements will occur - it trumps for routine style jobs (one to 4 story commercial or residential buildings, storage tanks and the like) but I'd never bet on them on high profile/complex projects - - 1500 ft high rise or 60 ft deep excavations adjacent to existing buildings and the like.

Geotech practice is much more of a judgment practice than most other civil when it comes to behaviour as these "god given" materials do behave always like one wishes - what stability calcs for footings take into account varved clays, residual soil remnant joint planes, slickensides, etc. With COVs of 30 to 50% vs 3 to 5% for steel and concrete, there is little wonder that judgment is paramount and the structural types cannot understand why geotechs are so conservative. (of course, it is easier to retrofit a superstructure problem than a substructure one.