Predictor System

Predictors, as the name suggests, predict the future solution based on the previous solutions. They, currently, act on the entire nonlinear solution vector. They are applied right before the beginning of the nonlinear equations solve.

A predictor is an optimization to apply on top of time integration schemes, or may already be part of some explicit or multi-step schemes. As a predictor extrapolates based on earlier solutions, it will not be able to predict changes due to threshold phenomena, and fares best when the evolution of the variables is smooth.

commentnote

Using a predictor in an implicit time integration scheme will only update the starting guess for the next nonlinear iteration solve. It should not modify the result, unless additional intricacies such as a predictor-corrector scheme is used.

commentnote

Using a predictor changes the value of the solution on each time step. As such, it can lower the accuracy of explicit higher order time integration scheme if it does not work tightly within that scheme.

Available Objects

  • Moose App
  • AdamsPredictorImplements an explicit Adams predictor based on two old solution vectors.
  • SimplePredictorAlgorithm that will predict the next solution based on previous solutions.

Available Actions