Solvers

Nonlinear Solvers

• NonlinearBlockGS

• NonlinearBlockJac

• NonlinearRunOnce

• NewtonSolver

• BroydenSolver

Linear Solvers

• LinearBlockGS

• LinearBlockJac

• LinearRunOnce

• DirectSolver

• PETScKrylov

• ScipyKrylov

• LinearUserDefined

Linesearch/Backtracking

Backtracking line searches are subsolvers that can be specified in the linesearch attribute of a NewtonSolver, and are used to pull back to a reasonable point when a Newton step goes too far. This can occur when a step causes output variables to exceed their specified lower and upper bounds. It can also happen in more complicated problems where a full Newton step happens to take you well past the nonlinear solution, even to an area where the residual norm is worse than the initial point. Specifying a value for linesearch can help alleviate these problems and improve robustness of your Newton solve. There are two different backtracking line-search algorithms in OpenMDAO:

• ArmijoGoldsteinLS

• BoundsEnforceLS

Modifying Solver Options

A method is provided to allow the user to modify solver options:

• Modifying Solver Options