[sfepy] Re: question about non-linear solver

Hi Patricia,

On 05/19/2017 06:27 PM, Patricia Garcia Cañadilla wrote:

Dear all,

I have some questions about the internal behaviour of Sfepy that I hope you can answer me.

• What is it the convergence criterion of the non-linear Newton solver?

For example, this means:

 'i_max'      : 5,
 'eps_a'      : 1e-10,
 'eps_r'      : 1.0,
 'eps_mode'   : 'and',

max. 5 iterations, the solution is considered converged if the absolute residual l2 norm |r_i| is smaller than eps_a and/or the relative residual norm (|r_i|/|r_0|) smaller than eps_r. The and/or behaviour is determined by eps_mode, 'and' is the default.

• When modelling a Nearly incompressible Mooney-Rivlin hyperelastic material, like the example shown here: http://sfepy.org/doc-devel/examples/large_deformation/hyperelastic.html, how the nearly-incompressibility constraint is stabilised?

It is not stabilized in that example - the bulk modulus is high enough to have a small compressibility and low enough in order not to cause convergence problems.

r.