Modelling of a sheet pile wall in Plaxis

[BC93]

Hello,

I'm using Plaxis to model a sheet pile wall between a quay and a future railroad.
The soil is mostly fill material above thick layers of soft clay followed by non-cohesive materials (such as sand).

I'm modelling the clay layers with undrained Mohr-Coulomb, i.e. the input I gave is cu (from vane shear tests) and phi=0°; but I left the material behavior as drained.

The first stage of the calculation is a gravity loading in order to establish the in-situ stresses. I'm wondering if the calculation yields reasonable results, given that Plaxis only has the undrained shear strength (hence it has no way to derive effective parameters, and using 'Ignore undrained behavior' would have here no effect).

Would it be better to use undrained behavior combined with effective parameters (i.e. c' and phi')? In which case the undrained shear strength derived by Plaxis in the following stages might differ from the field results.

Thank you for any good idea.

 

[TertiaryStresses]

Hello BC93.

Before I get into follow-up questions, a word of advice. Modeling soil behaviour in FEA is a tricky thing to do. Further, the FEA is usually done to determine the movements of the soil mass and structures. Our abilities to input the correct soils parameters, problem geometry and construction sequence is orders of magnitudes more crude than the "precise" movements that the FEA will give. If this is your first time running such an analysis, I recommend hiring someone first, and running your own analysis on the side as a learning program.

Now, for some follow up questions:
1. Have you done the hand calcs first? You should always have an idea of what to expect prior to running FEA. This will give you a baseline of how to judge the results.

2. In PLAXIS, if you are using Mohr-Coulomb soil model to model undrained behaviour, you must use c=c_u and phi = 0. Using UnDrained M-C with effective stress parameters in PLAXIS results in under-conservative results. What you've done originally is fine.

3. Why are you using Gravity Loading as the first step? Have you checked with simple hand calcs to see if the results from the Gravity Loading are acceptable? If not, it might be worth while to start with a K_o analysis and add a few steps to get to the starting conditions of the project.

Best of luck!

 

[BC93]

Hello, thank you for your answer.
I have done some FEA previously with Plaxis and Abaqus but I don't have a very broad experience.

Regarding question #3 I used gravity loading because the K0-procedure is supposed to give correct initial stresses only if everything is horizontal, which is not the case here. That's at least how I learned Plaxis. Is it worth trying to compute the initial stresses with a K0-analysis?

 

[SixDegrees]

I always use Plaxis Ko computation for initial stresses. You're correct if the problem is not fully horizontal then the resulting stresses will not be in equilibrium. For this condition I always follow up the initial stresses computation with a "Null" step (finite element computation for equilibrium) followed by resetting displacements to zero.

Using Phi =0 you will know what the strength is as it will not be stress dependent. Using Drained material behavior will cause Plaxis to not generate pore pressures, which keeps the stiffness as the one that you enter.

 

[AussieGeoEng]

I haven't used plaxis much previously but I am pretty sure that Mohr-Coloumb model is not the appropriate model to use for thick layers of soft clay.


See

http://en.wikipedia.org/wiki/Nicoll_Highway_collapse

 

[AussieGeoEng]

http://www.plaxis.nl/files/files/AJWhittle_RVDavis_2006.pdf

 

[SixDegrees]

It all depends on what you're using Plaxis )or any finite element program for). If all you're interested in is stability you can get a lot more information out of a finite element analysis using the mohr-Coloumb criteria than you can get from a slope stability, toppling or basal heave analysis.

If you're intrested in undrained/drained response and strength gain with pore pressure disspation, (SHANSEP kind of reponse), and accurate predictions of deformations then you may need a better model the mohr-Coloumb.