Hello all, I am performing nonlinear dynamic effective-stress SSI analyses in FLAC2D 9.0 using PM4Sand and PM4Silt. The model includes two buildings represented by elastic beam elements. The earthquake motion is applied at the base using a compliant-base formulation, where the input stress is calculated from a velocity time history.
The compliant-base input is applied using the following commands:
[global _x_basement = 0, _y_basement = -50.0]
fish define _stress_amp
local _pz, _Vs, _rho
_pz = zone.near(_x_basement,_y_basement)
_Vs = zone.extra(_pz,5)
_rho = zone.extra(_pz,12)
return -2.0*_Vs*_rho
end
program call "Free_Field_w_PM4Soils.f2dat"
[$construct_FF_w_PM4Soils]
zone face apply stress-xy [_stress_amp] table 'vel' range group 'Bottom'
zone face apply quiet range group 'Bottom'
zone face apply velocity-y 0 range position-y [M_yMin]
model dynamic timestep automatic
model solve time-total [table.x('vel',table.size('vel'))]
At the end of the earthquake record, I would like to activate the Post_Shake parameter in the PM4Sand and PM4Silt models and allow excess pwp dissipation. My problem is related to the transition from the strong-shaking phase to the post-shaking phase.
Initially, I tried to remove the dynamic input, initialize the gridpoint velocities to zero, and bring the model to static equilibrium before starting the post-shake phase. However, when I suddenly initialize the velocities to zero, I observe a large acceleration spike at the bottom of the model.
Then I tried another approach: instead of suddenly removing the input, I extended the velocity record with a smooth decay/taper to zero and allowed the dynamic analysis to continue for a short time. However, I still observed a similar acceleration spike at the base.
What is the recommended procedure in FLAC2D for ending a compliant-base dynamic input and starting a post-shaking consolidation phase with PM4Sand/PM4Silt? Should I keep the quiet boundary active and simply stop applying the stress input? Should I allow the remaining velocities to decay naturally without resetting them?
Have you made any changes to the damping (from dynamic to static)? You may also need to remove the free field in post-shake phase and convert it to the static boundary conditions.
Are you turning dynamic off? When I do reconsol, I do not turn it off - I bring the model to rest (either by letting all velocities come to zero, or by forcing the ‘calm’ command, or by applying an equal and opposite velocity) and solve with post-shake on along some dytime of x seconds (the time history input is zero anyways). Perhaps, if you are switching to the static solver, FLAC’s ‘dampers’ penalize you?
Thank you very much for your response. It is truly valuable to hear directly from the developer. Your comment made me reconsider switching to the static solver, which was likely causing the spike. I revised the procedure. First I extended the velocity record with a smooth taper to zero and let the dynamic analysis continue until gridpoint velocities naturally decayed. Then I removed the compliant-base boundaries, fixed the bottom, and run a brief static equilibrium(dynamic off, fluid off). After the static eq, I activated Post_Shake for PM4Sand and PM4Silt zones and run a consolidation with mechanical on, fluid on, and dynamic off. This eliminated the acceleration spike and the transition is now smooth.
I have two remaining questions:
Would you recommend keeping dynamic on during the consolidation phase rather than switching to the static solver, even when the input is zero?
Is it expected that ru values may initially increase at the beginning of the post-shaking consolidation phase before starting to dissipate? We observe a sudden jump in ru at several monitoring points right after fluid flow is activated, which we attribute to pore pressure redistribution between zones. Is this physically reasonable or does it indicate a numerical issue?
What I do is stick to the dynamic module. I am always concerned about the static solver’s approach. The way I understand it, FLAC2D’s static solution procedure uses extremely high damping values that can carry significant shear and normal stresses during shearing, which can cause problems for highly nonlinear, stress-dependent materials if the loading is not applied appropriately. As long as you have now fixed your base and have no input, I don’t see a reason why the dynanic scheme would be prohibitive. It’s counterintuitive because you really are operating under static conditions, but numerically, as long as you don’t have any residual inertia in your system, it should not really matter.
On the pore pressures, I am not sure what you are modeling and what you are validating against, but a “jump” in pore pressure doesn’t sound like a physical response… if it was due to diffusion/redistribution, then I don’t think it would be a spike. It sounds like the turning on/off of the flow and the dynamic scheme is penalizing your pore pressure response.