Applied stress in a creep analysis

FLAC
1

Hi all,

I’m performing a creep analysis in which I’d like to apply a series of nstress on a wall. Each applied nstress has to be applied on the wall for only 1 second.
Premise: I’m not performing a dynamic analysis because my idea is to model this applied stress pattern over a very large time window (days, months or years) that can not be easily modeled in a dynamic configuration.
I thus set the creep analysis in this way:

set cr 0
set fobl 10
set fobu 1E9
set lat 100
set lmul 2
set max 1E8
set min 1E3
set um 0.5
set crdt 0.1
set large

This is a tipical “time series” of applied nstress:

apply nstress -0.00 var 0.0 -0.00 from 258, 204 to 257, 234
solve age 1
apply nstress -0.00 var 0.0 -0.00 from 258, 204 to 257, 234
solve age 2
apply nstress -0.00 var 0.0 -0.00 from 258, 204 to 257, 234
solve age 3
apply nstress -0.00 var 0.0 -0.00 from 258, 204 to 257, 234
solve age 4
apply nstress -1694.29 var 0.0 -1694.29 from 258, 204 to 257, 234
solve age 5
apply nstress -15302.50 var 0.0 -15302.50 from 258, 204 to 257, 234
solve age 6
apply nstress -19457.20 var 0.0 -19457.20 from 258, 204 to 257, 234
solve age 7
apply nstress -14706.00 var 0.0 -14706.00 from 258, 204 to 257, 234
solve age 8
apply nstress -6932.76 var 0.0 -6932.76 from 258, 204 to 257, 234
solve age 9
apply nstress -1327.86 var 0.0 -1327.86 from 258, 204 to 257, 234
solve age 10
apply nstress -0.00 var 0.0 -0.00 from 258, 204 to 257, 234
solve age 11
apply nstress -0.00 var 0.0 -0.00 from 258, 204 to 257, 234
solve age 12
apply nstress -0.00 var 0.0 -0.00 from 258, 204 to 257, 234
solve age 13

As you can see, it is a kind of sinusoidal “time series” (of 13 seconds) in which the applied stress starts from 0 Pa, reach a maximum value and then go back to 0 Pa.
If I plot the xdisplacement induced on a specific zone on which I apply the nstress it seems that the displacements are not affected by the increasing and decreasing stress trend.

Cattura(3)
Cattura(3)1051×807 63.9 KB

I would expect, instead, that displacements would also increase and then decrease following the applied stress trend. It appears, in essence, that the applied stresses accumulate with each other instead of following an up-and-down trend.

Otherwise, if I set up a “classic” mechanical analysis (i.e., no creep) in which I define the number of solving steps for each applied nstress the results look different:

apply nstress -0.00 var 0.0 -0.00 from 258, 204 to 257, 234
step 100
apply nstress -0.00 var 0.0 -0.00 from 258, 204 to 257, 234
step 100
apply nstress -0.00 var 0.0 -0.00 from 258, 204 to 257, 234
step 100
apply nstress -0.00 var 0.0 -0.00 from 258, 204 to 257, 234
step 100
apply nstress -1694.29 var 0.0 -1694.29 from 258, 204 to 257, 234
step 100
apply nstress -15302.50 var 0.0 -15302.50 from 258, 204 to 257, 234
step 100
apply nstress -19457.20 var 0.0 -19457.20 from 258, 204 to 257, 234
step 100
apply nstress -14706.00 var 0.0 -14706.00 from 258, 204 to 257, 234
step 100
apply nstress -6932.76 var 0.0 -6932.76 from 258, 204 to 257, 234
step 100
apply nstress -1327.86 var 0.0 -1327.86 from 258, 204 to 257, 234
step 100
apply nstress -0.00 var 0.0 -0.00 from 258, 204 to 257, 234
step 100
apply nstress -0.00 var 0.0 -0.00 from 258, 204 to 257, 234
step 100
apply nstress -0.00 var 0.0 -0.00 from 258, 204 to 257, 234
step 100

Cattura2
Cattura21071×809 57.9 KB

The induced displacements agree with the applied stress trend, first increasing and then decreasing. Therefore, it seems that in this case the applied stresses do not accumulate but are overwritten with each new stress application.

Can anyone explain to me why this difference? Where am I going wrong with the creep analysis?

Thank you in advance

2

Is that a one-element model or something more complicated?

Anyways I would suspect that your creep timestepping is wrong. Without your total script I cannot recreate it, but my guess would be that the creep timestep is so large, that you only do a few steps for each iteration, thereby not transfering the load changes.

Remember that the creep in Itasca-codes is simply a stress correction at each mechanical step and you - in prinicple - have to make sure yourself, that each creep timestep is close to mechanical equilibrium. You can check this in principle, by e.g using a very small fixed timestep (e.g. 1e-4), so that it would take at least 1e4 steps for one “second” and then see, if the displacement behavior is more like the one you expect.

3

Thank you very much Markus for your reply! I am new to FLAC and therefore I’m not an expert…
I’m trying to reproduce the impact of sea-waves on a vertical rocky cliff. The code that I’m trying to develop (1wave_cr.txt) restores a .sav file (VT_j_th_12m_C_divided.sav) on which I carried out a thermal analysis. I assign a new fictitious cvisc model to the domain “locking” the viscosity-related parameters (vis, k_vis, k_shear) by assigning them very high values. As I told you, my intention is not to analyse the viscous behaviour of the model (which is why I set the viscosity-related parameters very high), but I simply want to be able to control the real time of the modelling (i.e. not the mechanical one) without having to perform a dynamic analysis.
I set up the creep analysis configuration and I call a .txt file (apply_1wave_cr.txt) in which I stored the applied pressure “time series”. History 250 is related to the xdisplacement that I already shown to you.

By changing the crdt to 0.0001 as you suggested seems to make no difference :frowning:

I’ll leave you a wetransfer link from where you can download the above files in case you have a chance to help me out (I’m working with FLAC v. 8.00.463):

Thank you!!