# Component-Wise Newton Raphson Strategy

From KratosWiki

(Difference between revisions)

(Created page with "''' Constructor Main Data ''' Solution strategy for Non-Linear analysis. It is derived from the Residual-Based Newton Raphson strategy but it sets automatically the Builder...") |
|||

Line 1: | Line 1: | ||

− | ''' | + | ''' Definition ''' |

Solution strategy for Non-Linear analysis. It is derived from the Residual-Based Newton Raphson strategy but it sets automatically the Builder and Solver as | Solution strategy for Non-Linear analysis. It is derived from the Residual-Based Newton Raphson strategy but it sets automatically the Builder and Solver as | ||

Line 7: | Line 7: | ||

Which is going to build the linear system by components. | Which is going to build the linear system by components. | ||

That makes the building of the system slower than the "Residual-Based" one, where the all components are build and assembled together. | That makes the building of the system slower than the "Residual-Based" one, where the all components are build and assembled together. | ||

+ | |||

+ | ''' Constructor Main Data ''' | ||

The usual common data is required in the | The usual common data is required in the |

## Revision as of 11:08, 23 March 2016

** Definition **

Solution strategy for Non-Linear analysis. It is derived from the Residual-Based Newton Raphson strategy but it sets automatically the Builder and Solver as

ComponentWiseBuilderAndSolver(LinearSolver);

Which is going to build the linear system by components. That makes the building of the system slower than the "Residual-Based" one, where the all components are build and assembled together.

** Constructor Main Data **

The usual common data is required in the constructor for the Newton Rapshon Strategy:

- Model Part
- Time Integration Scheme (Newmark, Bossak, ...)
- Convergence Criterion (Residual, Displacements, ...)
- Linear Solver (Direct, Iterative, ...)
- Maximum Number of Iterations (default 30)
- Common Flags: (false by default)
- CalculateReactionFlag (compute reactions)
- ReformDofSetAtEachStep (check and change the dofs at each step)
- MoveMeshFlag (update the coordinates of the nodes at each step)

If the maximum number of iterations is reached, the last result is taken as converged and the analysis keeps on running. When this happens and the analysis does not crash, the accuracy of the results will be very low and we recommend to not trust it.