It appears that the Norsand constitutive model coding has changes and no longer functions for multi-zone models.
I have modeled the staged construction of a TSF using a plane strain approximation and Norsand successfully in FLAC3D version 7.00.147. The results showed realistic and expected stress paths in p:q:e spaces.
However, when I re-analyse the same model, without any changes in the coding or FISH functions, the model does not converge and stresses and displacements reach infinity resulting in the error “An accumulated value has become greater than the maximum allowed…” This was tested in versions 7.00.148 and 7.00.151, and the error occurs in both. When using the Mohr-Coulomb, or P2PSand constitutive models on the same mesh the same error does not occur.
Curiously, simulated triaxial test calibrations using Norsand, like the examples in the documentation, appear to solve without issue.
In the Software Update Revisions, the changes in version 7.00.148 note a “NorSand model update”. However, the support documentation has not been updated to note any changes.
Has anyone been able to find the changes to the Norsand model or how this error can be prevented? It appears I will need to roll back my version to before the update to the Norsand model.
Hi Paul. could you please send the datafile to Itasca (flac3dsupport@itascacg.com) so that we could duplicate the issue and then have an investigation and possible fix?
Hi!
i am new to the forum and i haven’t quite worked out how to post a new topic yet … but i have a similar problem so i thought that i would ask the question in this thread.
i am re-running an old model and this error appears: *** An accumulated value has become greater than the maximum allowed. This is normally caused by bad property values in a constitutive model or by the model going unstable.
what was the solution in the previous case?
i have just updated FLAC3D 9 to its latest version and rebooted the computer a few times.
the error is appearing in a couple of models and the pattern is random. also it appears mid way through a run of steps. if a zone was missing a property, one would expect the error to appear in the first step.
It is difficult to determine exactly what occurred without the data file. Possible reasons could include:
(a) The model has excessive deformation, resulting in negative volume in one zone (or any of the subzones).
(b) The loading step is too large, causing overshooting that may result in negative effective mean stress, particularly in surface zones with low effective stress.
(c) Unrealistic material parameter inputs.
Possible steps (either individually or in combination) to improve numerical stability are:
(a) Improve the grid quality.
(b) Decrease load steps.
(c) Ensure that fluid-tension is zero if used.
(d) Check NorSand parameters. For example, turning on the flag-inner should help to some extent.
(e) Change the surface zones into a simpler model, such as elastic or M-C.
(f) Add a small virtual pressure to the surface.
In any case, investigation should be conducted on a case-by-case basis.
