On this page:
11.1 newton (beta)

11 nonlinear_equation_solvers (beta)

Defines parameters for nonlinear equation solution strategies.

Name

 

Type

 

Description

 

Default

label (beta)

 

STRING

 

The user defined identifier for this nonlinear solution method.

 

newton (beta)

 

SUB OBJECT

 

See linked documentation

 

Example Usage:

nonlinear_equation_solvers: [ { label: "newton_raphson", newton: { target_relative_residual: 0.001, linear_equation_solver: "direct_multifrontal", maximum_iterations: 8 } } ]

11.1 newton (beta)

Defines options for the Newton nonlinear solver method.

Name

 

Type

 

Description

 

Default

target_relative_residual [optional] (beta)

 

DOUBLE

 

The target relative residual for the nonlinear iteration that, upon reaching, signals successful convergence of the iteration. The relative residual is computed as the relative decrease in the current residual norm from the initial residual norm. The default value is overly conservative for most problems, and can usually be increased to improve speed and robustness.

 

1e-8

linear_equation_solver (beta)

 

STRING

 

The user defined identifier of the linear solver to use. When a direct linear solver is used, the Newton method is often called a “Newton-Raphson” method. When a Krylov iterative linear solver is used, the Newton method is often called a “Newton-Krylov” method.

 

maximum_iterations [optional] (beta)

 

NONNEGATIVE INTEGER

 

Set the maximum number of iterations.

 

8

use_line_search [optional] (alpha)

 

true|false

 

This enables a backtracking line search, which dynamically reduces the step size within Newton iteration to avoid divergence. However, it is often overly conservative in its step size reduction, and may slow down analysis significantly, even on steps that would otherwise converge quickly. It is recommended to try adaptive time stepping and review other aspects of the problem setup before enabling this option. This option typically requires a higher-than-default value for maximum_iterations to converge.

 

false

Example Usage:

newton: { target_relative_residual: 0.001, linear_equation_solver: "direct_multifrontal", maximum_iterations: 8 }