Nonlocal Scalar Damage Models

[maxnezdyur]

Reason

MOOSE has local damage models but they can exhibit mesh sensitivity. Nonlocal damage models allow for less mesh sensitivity while enabling the use of the same models. Damage averaging approach would be an easy first step that wouldn't require any change to existing damage models.

Design

Create new damage model that would work with any existing damage model. It takes in a some form of averging user object and creates new damage_index for ComputeDamageStress.

Impact

Would allow every current scalar local damage model to be used as a nonlocal version of itself.

[bwspenc]

We discovered that the initial version of this fails parallel tests for higher processor counts because we are not communicating data from processors that have elements that fall within the averaging radius, but are not the owning processors of the current element's neighbors. @dschwen has an idea on a pretty easy fix for this.

[loganharbour]

We need an open issue for this for SQA

[dschwen]

We need an open issue for this for SQA

There ya go :-)