I was simulating a pile in a fully drained soil subjected to horizontal loads using ABAQUS (as shown in Figure 1) and noticed from the monitor that the calculation did not continue for a long time and when I interrupted the job I got a very large “.log” file (100-300g).
I have looked at the “.log” file and found that it is a repeated error message (as shown in figure 2), what is the meaning of this message and is there any way to avoid it?
Thank you in advance for your help!
Sorry I got the order of the pictures wrong. Figure 1 is the “.log” file that reported the error, and Figure 2 is the piled soil model
substepsize is too low < 1e-18
problem cannot been solved.
perhaps model is wrong
Thank you so much for helping again! Does this error mean that the analysis step is too small?
I’m curious why my ABAQUS doesn’t break the calculation but keeps repeating the attempt
It do not know which units you have used, but if the displacement is in meters the displacement is large (20cm), and it lookes there is a gab between pile and soil. IF you have used fully drained and cohesion is zero this is unstable and because you have used an implicit methode it probably not converts. Try to increase cohesion (p-shift).
The units in the simulation are m. I am simulating the horizontal displacement of a pile in a sandy soil.
Shouldn’t the cohesion c=0 and parm18 be taken as 0 for sand?
Also, is 20cm a large displacement? I have seen in many papers that the horizontal displacement of the mud surface in the simulation results reaches D/10, for my simulation (pile diameter D=10m) this displacement should reach 1m.
Thanks again for your reply!!!
To add to this, the monitor shows that the incremental step size is not reduced, it stays at the maximum step size I set, but no further calculations are performed, just the machine idles (100% CPU usage) while generating a large LOG file.
If convergence is a problem, ABAQUS should show 1U,2U… Until the calculation breaks isn’t it?
Why does ABAQUS not show 1U.2U etc. after calling SANISAND’s UMAT, but stays at one analysis step and repeats the LOG file writing?
I’m running a full drained simulation, i.e. can’t c take 0 in the case of full drained?
There is tension in the soil: because c=0 it is not possible.
DT_k is reduced to DTmin and DTMin is very low.
Send .inp, .cae, .for in a zip-file to firstname.lastname@example.org and I can see.
c … if pp_2 lower than zero
DT_k= one – T_k
write(*,*) ‘mario_DT – pp_2<0 – mario9=one, pp_kk=',pp_kk
Thank you very much! I’ve noticed what this code does.
After reading your reply yesterday, I set parm18=0.1 and ran the simulation again, but the simulation did not achieve the target of 1m horizontal displacement of the mud surface either.
I have send my .inp, .cae and .for to your email, Thank you in advance for your help!
use dynamic implicit instead of static general for displacement
take 1 step after gravity.
time for example 100sec and minimal increment 1 sec
Thank you for your advice! I have tried to replace the steps from static to dynamic implicit. However, after the geostress step is completed, ABAQUS begins to idle rather than continue the dynamic implicit step.
The error is still the same as the Fig.1 in my original post.
I am now looking for which step of the UMAT run will cause pp _ kk to become NAN.
geostep is not correct: only soil stresses are calculated: state var are zero (included void)
for geostep you need umat or sigini+sdvini
if you use umat some code have to be added to add stresses to avoid lot of vertical displacement.
But it will not solve your problem that model gives umat-rrors and not convert.
Thank you for reply, Arie! I will correct GeoStep with the guide of your advice. As for the errors, I am still looking for the code causing this bug. There is no division operation in the calculation of PP_KK, I don’t think the result of it could get the result of NAN. If you know, can you tell me what other cases would cause PP_KK=(SIG1+SIG2+SIG3)/3 to be calculated to get the answer of NAN?
Just a general question: Why you want to use sanisand umat?
I am study in the direction of constitutive models. And SANISAND model is the first sand model that I have been exposed. Since its code is easily available (posted on this website), I chose this model as the first step in learning the constitutive model.