Shaking table test simulation using FLAC3D

FLAC3D
1

Hi everyone,

I am using FLAC3D 7.0 to simulate the shaking table test conducted by Ko and Li (2020) to investigate the seismic response of a pile group in liquefiable soil. In this study, a large laminar shear box was utilized as the soil container, and a 1/25 scale model of a 2 x 2 pile group for an offshore wind turbine was tested. The input motions were 2-Hz sinusoids with various acceleration amplitudes. For simulation comparison, I chose a 50 gal acceleration amplitude. Figure 1 shows the experimental layout and the results for the excess pore pressure ratio (ru).
image
In the numerical model, I created a 2-meter high brick model and used the SANISAND constitutive model. Figure 2 shows the pore pressure distribution at the end of the seismic event. Most of the pore pressure within the model is zero and less than static condition.
image
Figure 3 presents the time histories of the excess pore pressure ratio (ru) at different heights, compared with the experimental results in Figure 1. It is unreasonable for the pore pressure to become zero during dynamic analysis. I’m unsure if this issue is related to the boundary conditions. How can I adjust the model?
image
Additionally, if I want to present the behavior of pore pressure dissipation after the seismic event in the simulation, how should I conduct the subsequent analysis? I attempted to activate the fluid flow, but it didn’t work. Does anyone have any advice?

Codes and figures are attached. Thank you.
test01.dat (5.3 KB)
veldata-base.txt (15.7 KB)

2

Hi Wei,

I looked at your “test01.dat” file, and I noticed that you are using free-field conditions on the side boundaries. I would not recommend using SANISAND adjacent to these boundaries, as the model’s shear and bulk modulus are calculated using an unrealistic Poisson ratio. Here are a few suggestions:

  1. Try using rigid boundaries or given the symmetry in your model,
  2. maybe try attaching your side nodes.
  3. Ultimately, you can add an additional elastic elastic column adjacent to the FF-boundaries, with a shear and bulk modulus calculated using a more realistic value for poisson.

I’m not sure which implementation of SANISAND you are using, but I would highly recommend performing a sensitivity analysis while using a constant δt​ time step (e.g., 1.0x10^{-5}) versus using a multi-step approach during the dynamic analysis.