Dear Paul,

Thanks for your email. I'll certainly try out your suggestion
of using a small model, one element might be too small but
4 should be OK.

>We did notice some problems related to item (2) of your comments and to
>get around this we used a routine to calculate principal stresses from
>the "Numerical Recipes" book by Press et al p460 in 2nd edition. 

The algorithm on p.460 performs the calculation of eigenvalues and
eigenvectors of a symmetric matrix. How does that relate to the corner
points on the Mohr-Coulomb hexagon?

>It may be that using the explicit time integration got around many of the
>convergence problems (of course with the penalty of a small time step and
>hence longer run times).

Yes, it may be worth trying. No guarantee that the "solution" is
correct though... How to guarantee that plastic points lie on
the yield surface?

>I hope the above is of use.

Thanks, it is.

Cheers,
Roger.

PS Hope you don't mind me posting your email (below) to the list....
.........................................................................
Roger,

I read your e-mail this morning regarding material models and
convergence.

Several years ago we looked at the problem of seismic loading on tunnel
cross sections using an explicit finite element code LS-DYNA3D
(www.lstc.com).  We had to implement similar material models to yourself
but we did it for solid elements.
  
It may be that using the explicit time integration got around many of the
convergence problems (of course with the penalty of a small time step and
hence longer run times). 

We did notice some problems related to item (2) of your comments and to
get around this we used a routine to calculate principal stresses from
the "Numerical Recipes" book by Press et al p460 in 2nd edition. 

With regard to item (1) on your list, it appears that the most or the
whole model is going plastic and if it is load controlled with no strain
hardening then convergence may be difficult.  

A good check for (1) with the material models is  a single element test,
with various boundary constraints and then applying enforced
displacements and monitoring forces.  This should also be comparable to
hand calculations.  In effect this is like physical testing with various
degrees of confinement.

I hope the above is of use.

Best regards

Paul Jacob
Dames & Moore
9801 Westheimer, Suite 500
Houston, TX 77042
USA

Tel (1) 713 914 6545
Fax (1) 713 780 7414

e-mail [log in to unmask]





%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%