21.01.2015 Views

COMSOL Multiphysics™

COMSOL Multiphysics™

COMSOL Multiphysics™

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

femnlin<br />

components fixed, the augmentation components are assumed to fulfill a linear<br />

equation. The property Augtol control the tolerance for the augmentation<br />

components and the property Ntol the tolerance for the main components in the<br />

convergence criteria for the combined iteration (two substeps each). The<br />

convergence criteria is that the relative increment, from one iteration to the next,<br />

for the augmented components and the main components must not be larger than<br />

their tolerances. The maximum number of combined iterations is controlled by the<br />

property Augmaxiter. The linear system solver used for the solution of the<br />

augmentation components can be controlled by the property Augsolver. For a<br />

more detailed description of this solution procedure, see the “Nonlinear Solver<br />

Settings” on page 381 in the User’s guide.<br />

Setting Hnlin to on causes the solver to treat the problem as being highly nonlinear.<br />

This option can be tried if there is no convergence with Hnlin set to off.<br />

Depending on this parameter, certain standard values are selected for the Initstep<br />

and Minstep properties. Moreover, certain internal control structures are adapted.<br />

Especially, the error control is biased from a more absolute norm towards a relative<br />

norm. So this parameter is also useful if a solution with components of highly<br />

varying orders of magnitudes are present. In the context of parameter stepping, you<br />

can also try this option if the step sizes in the parameter seem to be too small.<br />

Initstep is the initial damping factor for the step length. The default is 1 if Hnlin<br />

is off and 1e-4 if Hnlin is on.<br />

Maxiter and Minstep are safeguards against infinite Newton iterations. They<br />

bound the number of iterations and the damping factor used in each iteration.<br />

Minstep defaults to 1e-4 if Hnlin is off and 1e-8 if Hnlin is on.<br />

The tolerance Ntol gives the criterion for convergence, see “Nonlinear Solver<br />

Settings” on page 381 in User’s Guide.<br />

The property Out explicitly sets output variables and their order. The output<br />

variable fem means the FEM structure with the solution object fem.sol added. The<br />

solution object sol has a field sol.u, which is the solution vector for the FEM<br />

formulation of the PDE problem. The solution vector u is a column vector with one<br />

component for each degree of freedom of the discretized problem. If the parameter<br />

variation feature is used, then sol.u is a matrix, and there are additional fields<br />

sol.pname and sol.plist. The field sol.pname is the name of the parameter, and<br />

sol.plist is a row vector with parameter values for which a solution was<br />

computed. The corresponding solution vectors are stored as columns in the matrix<br />

sol.u. The output variable Stop is 0 if a complete solution was returned, 1 if a<br />

137

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!