Soil modelling - Algorithm

[Jhno]

Hi,
following discussion with skamalzare we decided to make specific thread on soil modelling to involve more people. Your input on the subject will be greatly appreciated!
To start with a first subject, I would like to discuss experience with algorithm choice in the model.
I uses the nDmaterial PressureDependMultiYield Material for my soil deposit along with quad element. The material does have different stages (here for details : http://opensees.berkeley.edu/wiki/index ... erialStage).

For a start, in the elastic domain, I observe no difference between the algorithms. In the plastic domain thought, I did obtain very variable model stability and also final result depending on the choice of the algorithm I use (I guess this is normal).
I found the BFGS algorithm to attain more easily convergence and to solve more rapidly.
Of all the algorithm I test, BFGS seems to be the fastest and the more stable. But this made me wonder about the result in itself. Is there a "better" algorithm (not in a mathematical sense, but in a physical sense?).

The script for Site Response Analysis do uses the Newton algorithm to solve the model. http://opensees.berkeley.edu/wiki/index ... _Analysis)
I haven't check the whole script but it looks like a single column of soil so the model isn't very " big " ..so the time issue might not be important here.

To answer my question I have downloaded Edu-shake and will be trying to see if I can match the result.
In the meantime, I was curious about your experience with algorithm choice. Do you have a algorithm you prefer or found the be more reliable? BFGS's speed really saved me a lot of time so far... Still wondering about the result thought. Any1 tested this or remarked this ?