Strange PWP Ratio Response in 1D Analysis Compared to 2D Version

FLAC
1

Hello all,

I am currently performing a 1D site response analysis using FLAC 2D. My model consists of eight soil layers: the top seven layers are defined with PM4Sand and PM4Silt, while the bottom (eighth) layer is modeled using the Mohr-Coulomb (MC) constitutive model. When I monitor the pore water pressure ratio (ru) at the center of Layer 2, I observe a strange transient response that doesn’t align with expected behavior. I’ve attached a figure showing this unexpected fluctuation.

1dvs2d
1dvs2d1280×720 98.1 KB

Interestingly, when I run a 2D analysis using exactly the same soil layering, geometry, and input motion, the ru response in the same location appears much more reasonable and physically consistent.I am trying to understand:

  • What might be causing this transient response in the 1D configuration?
  • Could the transition from the Mohr-Coulomb layer to the liquefiable PM4Sand/PM4Silt layers be creating numerical artifacts or strain localization?

I’ve also plotted the shear strain increment across the profile, and noticed a sudden jump at the interface between the Mohr-Coulomb and the overlying liquefiable layers. Could this be contributing to the irregular ru behavior?

shearstrain
shearstrain3800×2346 90.9 KB

Any insights or recommendations would be highly appreciated!

Below is my FISH code used to calculate ru. I’m wondering if there might be an issue with it as well:

fish define _ru(_x,_y)
local _pz = zone.near(_x,_y)
return (zone.pp(_pz)-zone.extra(_pz,17))/zone.extra(_pz,16)
end

fish callback add [Ru_L2 = _ru(0,-2.5)] -100 interval 100
fish history name ‘Ru_L2’ Ru_L2

; zone.extra(17): initial pore pressure
; zone.extra(16): initial vertical effective stress

2

What are boundary/attach conditions for your 1D and 2D models? Without detailed datafile it is hard to tell what occurred?

Is the localization occurred in Zone 7 or 8, in other words, in PM4Silt or MC?

3

Thank you for your quick response.

Here are the boundary and attachment conditions for both models:

In the 1D model:

  • Horizontal boundaries are attached to each other at relevant gridpoints. (i.e., lateral boundaries are attached to simulate 1D behavior). No free-field boundaries are used.
  • Water boundaries are open laterally and closed at the base.
  • The bottom boundary is quiet boundary with compliant base
    In the 2D model:
  • Both attached and free-field boundaries are applied on the lateral sides.
  • Water boundaries are also open laterally and closed at the base.
  • A quiet boundary with compliant base behavior is defined at the bottom.

Regarding your second question on strain localization:
Strain localization begins just below the top zone of the Mohr-Coulomb layer (Layer 8). The maximum shear strain is observed exactly at the topmost zone of the Mohr-Coulomb layer. This high strain then continues upward into the bottom part of Layer 7, which is modeled with PM4Silt. I have attached a figure showing the distribution of shear strain increment in the profile.
Would you recommend any approach to smooth this transition?

Resim1
Resim12115×1801 297 KB

5

Thanks for the additional information. However, we still cannot determine the details without the datafile. Theoretically, these two methods should produce identical or very similar results. I would suggest the following:

  1. Split Zone 2 into two zones so that the water table does not cut through a single zone.
  2. Regarding “water boundaries are also open laterally and closed at the base”: I’m not entirely sure what is meant by “closed” and “open,” but both the lateral and base boundaries should ideally be free to allow pore pressure to build up. In other words, there should be no fixed pore pressure during dynamic analysis.
  3. Check whether a fixed pore pressure (= 0) is set at the water table.
  4. It seems that the Mohr-Coulomb (MC) zone with large strain has undergone shear yielding in the 2D model. Is this also the case in the 1D model? If this zone behaves very differently (e.g., one localizes while the other does not), it will have a different impact on surrounding zones.
  5. You could consider switching the MC model to an elastic model, since you have already assigned hysteretic damping. This could help avoid localization effects.
1 Like
6

Dear Cheng,
Here is a quick update. The strange ru response was caused by how I initially defined the water table using the zone water plane origin command. Instead, I used the following FISH function to assign initial pore pressure conditions more explicitly:
fish define _Assign_WT
local _x, _hw, _pgp
command
zone face apply pore-pressure 0.0 range position-x [M_xMin], [M_xMax] position-y [w_yMax], [M_yMax]
endcommand

loop foreach _pgp gp.list
    _x = gp.pos.x(_pgp)
    _hw = (w_yMax) - gp.pos.y(_pgp)
    if _hw > 0
        gp.pp(_pgp) = _hw * _gamma_w
        gp.sat(_pgp) = 1.0
    else 
        gp.pp(_pgp) = 0.0
        gp.sat(_pgp) = 0.0
    endif
endloop

end
[_Assign_WT]
Once I used this function (and avoided fixing pore pressure at the water table), the ru behavior became much more stable.
Regarding the shear strain localization issue, I followed your suggestion and replaced the Mohr-Coulomb layer with a linear elastic model. Then I adjusted the hysteretic damping parameters. This significantly reduced the strain concentration at the interface.
Thank you again for your valuable guidance, your comments helped a lot.