Foundation settlement in layered soil

[pelelo]

Hello,

I am working on a foundation project were I need to compute the predicted settlements.

My soil profile is composed of 10' of Sandy silt (MLS), 20 of Silty sand (SM), 15 of silty gravel (GM) and another 15' layer of MLS.

I don;t have much of experience dealing with foundation settlements but I know there are bunch of methods available for computing elastic settlements. At least the most common I have seen are Theory of Elasticity and Schmertmann's method. To my knowledge Schmertmann's method is better than Theory of Elasticity because it takes into account several Elasticity Modulus values for layered soils, such as my case.

For the ones who like to compute settlements using the Theory of Elasticity, I have seen few examples were it is required to compute settlements for several layers and I have noticed the same formula (S = q * B * (1-u2)/E) is applied for each layer. What I noticed is those authors don't take into account the change (decrease) of stress (instead they used q), using boussinesq's chart, as the soil profile goes deeper.

Also I am not sure if the same footing width "B" I am supposed to keep it when analysing each layer. Probably the deeper the soil profile is, some reduction factor of the Base should be taken into account.

Can anyone give me some imput about this?

On the other hand, can anyone let me know what are the limitations of Schmertmann's method?, Or is this one simply the ideal for solving multilayered profiles?

Thanks a lot.

 

[fattdad]

For any given loading condition, there will attenuation of stress with depth. If you conceptualize the soil as an elastic media, then you need to assign an elastic modulus to each soil layer. (If you can't conceptualize the soil as elastic then you'd be using 1-d consolidation characteristics - i.e., especially if you are exceeding Pp.)

In the absence of stress increase, there is no settlement (duh?). If there is stress increase, there will be settlement, based on the soil's elastic modulus. So, plot your change in stress with depth (i.e., using Bousinesq or Westegaard) and overlay your soil layers onto this stress envelope. (You may have to make soil sublayers in the upper interval to account for rapidly attenuating stresses.)

Let's say you have a sublayer that's 5 ft thick and the stress increase at the top is 1,000 psf and the stress increase at the base is 600 psf. Let's say the soil modulus is 100 tsf. If you integrate the stress profile with respect to depth, you'd have [(1,000+600)/2]*5=4,000 #/ft. If you divide that by the soil modulus you'd have:

(4,000 #/ft)/(100*2,000)=0.02 ft (1/4 in)

Do this for each layer.

Multiply by 1.2 for long term "creep" affects.

This is essentially the approach of Schmertman, just more fundamental. My university professor (J.M. Duncan) liked my solution to these problems, so it's not like I'm on some "flyer."

Enjoy!

f-d

¡papá gordo ain’t no madre flaca!

 

[pelelo]

Thank a lot fathead for your reply.

I will try your way to see how it goes.

For your elasticity modulus values, which values you use?, the ones based on the SPT - N value or the standard values that are in the geotechnical books (which by the way, from book to book they are far away).

Do you think theory of elasticity formula is not applicable for this (S = q * B * (1-u2)/E), or better said, it would be applicable for one soil layer?

Thanks a lot again

 

[fattdad]

getting soil modulus is the problem. . . There are correlations to N-value and there is local experience. Dilatometers are a great tool and when used with N-value program allow for site correlations.

hard to further help from here.

f-d

¡papá gordo ain’t no madre flaca!

 

[pelelo]

thanks again!

 

[Ron]

I agree with fattdad's approach. Keep in mind that sandy silts will sometimes act like silty sands (elastic), so using that approach is reasonable.

Schmertmann had a lot of experience in exactly these types of soils (much of his research was in Florida coastal plains soils of this type), so his data are relatively good. Schmertmann was a bit more theoretical than practical though, so if your soil is highly stratified (relative to the depth of influence of the foundation), be a bit careful with his method.

 

[BigH]

I'll take fattdad's approach a bit differently -
for stresses imposed in different layers of differing moduli values, the higher modulus value layers will "attract" more stress than the lower modulus value layers. There are several ways to determine the stresses with depth - use layered charts (see Poulos and Davis' "Elastic Solutions for Soil and Rock Mechanics").
Another way is to "increase" the layer thickness of the higher modulus values so that you have a uniform layer of the same value (Ron knows this from pavement design methods). Then you can compute the settlements from that - using chart or using elastic equations. One such method is Palmer and Barber (1940). They replace the upper layer thickness h having the elastic parameters E1 and mu1 by an equvialent thickness heq having elastic parameters of the lower layer material E2 and mu2 by:
heq = third root of (E1(1-mu2squared)) divided by (E2(1-mu1squared))
for those in India, Shenbaga Kaniraj's book "Design Aids in Soil Mechanics and Foundation Engineering" has very good and very detailed discussions of elastic theory and presents nearly all the charts found in Poulos and Davis - sometimes a bit clearer.

 

[fattdad]

Ah, the "expanded layer technique." My professor didn't like it. Maybe that's why I don't use it. It does make sense though (i.e., providing I'm fully understanding your approach).

f-d

¡papá gordo ain’t no madre flaca!

 

[pelelo]

Fattdad,

Have you ever tried to compute settlements using Burland and Burbidge's method (1984): "Settlement of foundations on sand and gravel, Proceedings of the Institution of Civil Engineers, Part 1, 1985, 78, Dec., 1325-1381."

It deals with the SPT-N value and depth of influence Zi, in which Zi = B^0.75.

I tried to compare values between your way (similar to Schmertmann's) and Burland and Burbidge's and with the last one I get very very low settlement values.

Using your way I get values of around 5 and 6 centimeters, whereas using Burland's and Burbidge's is get values of around 3 and 4 milimeters. Sorry for using SI units.

Do you know if Burland's and Burbidge's is well accepted in the geotechnical engineering industry?

Please let me know,

Thanks in advance

 

[fattdad]

I don't know and I'm not familiar with their research or publications.

Sorry. . .

f-d

¡papá gordo ain’t no madre flaca!

 

[Mccoy]

Do you know if Burland's and Burbidge's is well accepted in the geotechnical engineering industry?

It sure is accepted, depending on local experience and engineers' habits though.

The difference in output you cite is not acceptable, there must be some mistake in one procedure or the other, i hope not in both of'em!!!

One improvement to B&B's method has been proposed by Berardi And Lancellotta of the Italian geotech school:

BERARDI R; LANCELLOTTA R., Stiffness of granular soils from field performance., GEOTECHNIQUE, pp. 149-157, 1991, Vol. 1, ISSN: 0016-8505

The method becomes more complex though and iterative techniques must be applied.

 

[pelelo]

Mccoy,

Thanks for your help.

Not many people I have spoken with know about B&B's approach.

Yeah, checked several times and using B&B's I am still getting too low settlement values.

So far, to me, using Schmmertmann's method is more realistic.

I will go through BERARDI R; LANCELLOTTA R and will see how more realistic would be B&B's approach.

 

[Mccoy]

peleo,
might it be that in BB there is some conversion problem due to the bilog-scale axes in the N-Ic plot?

What's your average NSPT in the influence thickness?

 

[pelelo]

Mccoy,

The B&B´s approach I am using is the one I got from "Soil Mechanics in Practice Engineering" by Terzagui, Peck and Mesri (p. 396):

Sc = (1/3) * B^0.75 * (1.7/N60) * q

Where B is in meters and N60 is the average of SPT-N values measured within a thickness Zi (Zi = B^0.75). q is in Kpa and Sc in milimeters.

I am not sure what you mean by the N-Ic plot.

My average Nspt in Zi is about 25.

An example:

q = 200 Kpa,
Nave = 23
B = 2 m

Sc = 2.36 mm = 0.236 cm,

I found it very very low. I think it should be somewhere between 1 and 2 cms.

Please let me know If I have any mistake.

 

[Mccoy]

Peleo,
please find the original B&B graph with the Ic = f(NSPT) value plotted inside a spreadsheet file

the value would be Ic = 1.81

 

[Mccoy]

I found that some sources say Ic=1.7/NSPTave like the TPM you cited, but others say Ic=1.7/NSPTave^1.4 (metrics) which would yield 0.018, exactly the same value of 1.81, 2 orders of magnitude smaller as in the original B&B plot.

If you adjust the units (m to cm end so on) you might come up with a reasonable value for the settlement.

May be you want to take a look at the berardi's slide presentation I'm attaching, page 31/55, in the same presentation he also illustrates the Berardi and Lancellotta method

 

[Mccoy]

Peleo,
I took the oppotunity to review the method, with your foundation data the settlement, according to foundation depth and lenght (i assumed D=1 m, L = 10 m)

should be in the region of 8 mm + 4 creep settlement = 12 mm =about 1.5 cm

 

[BigH]

Just thought I would throw in something - McCoy and I have talked about this in the past. Settlement computations are just that - computations based on various methodologies. None are exact. Do they take into account increase in E as the load is applied? etc. Years ago, in Ground Engineering Magazine, there was an article (before B&B) that reviewed about 15 different methods of settlement computations. Their recommendation for the "estimate" of settlement was to pick three methods that you feel comfortable with and use the average of the three for your estimate. If one reviews the book by Som and Das (Theory and Practice of Foundation Design), they carried out settlement computations for a clay using various methods - including stress path method. they got differnt answers - off by as much as 10 mm or so. To think that anyone can compute settlement to the mm (and I have seen numbers like 15.4mm thrown about), I would posit that he/she is ready to buy land in Atlantis. This comment notwithstanding the points made in the recent posts - just a perspective.

 

[Mccoy]

You're dead right, bigH, sometimes we tend to forget how crude settlements estimates can be, Tomlinson is ready to clarify that we are always dealing with (best) estimates.

Many times the choice of the correct transformation law (i.e: NSPT to elastic modulus)is vital (and approximate), much more so than the choice of the algorithm itself.

From this point of view B&B tends to eliminate a source of error, the transformation itself, since it relates directly the settlement to the NSPT, withouth further passages, that would mean less additional error into the output.

 

[YannisChasiotis]

The Burland & Burbridge method, is the only method that is the result of a statistical evaluation of field performance. What they did is to evaluate high quality, reliable data of load and deformation. The method uses sound engineering judgement, but it should be realised that it is based on statistics...
Although the method has many shortcomings, it is the standard for Britain and other countries...