Version 3.2.2 is mainly a bug fix version of the 3.2 series.

This version was released along with Version 3.1.5 and inherits from all the fixes of this version.

New MFront Features

New algorithm to treat damage in the DDIF2 stress potential

Two improvements were made to the numerical treatment of the DDIF2 stress potential:

• The DDIF2 damage behaviour is now treated by default using an algorithm based on statuses. In each damage directions, the damage state is kept constant during the Newton iterations. Once converged, the consistency of the damage state with the solution found is tested. If the state is inconsistent, the iterations are restarted with a new state.
• The implicit scheme is now divided in two steps by default. In the first step, the time step is set to zero before the prediction stage. This is meant to filter all viscoplastic flows and the convergence is thus performed only on the damage state (unless another rate-independent mechanisms is considered). Once converged on the damage state, the time step is reset to its original value and the implicit resolution is restarted.

The DDIF2 stress potential (and thus the DDIF2 brick) has two new options:

• use_status_algorithm (boolean value, true by default). If false, the algorithm used in previous version is used.
• first_converge_on_damage (boolean value, true by default). If false, the time step is not set to zero at the prediction stage.

Issues solved

Ticket #192: DSL ‘Implicit’ definition

The documentation of the Implicit DSL was wrong about the automatic definition of the elastic strain.

For more details, see: https://sourceforge.net/p/tfel/tickets/192/

Ticket #187: Conversions in Number::getCxxFormula is biased by rounding errors

Number::getCxxFormula did rely on std::to_string which only has 6 significant digits. As an example, the output of std::to_string(1e-14) is 0.000000.

To overcome this issue, the Number structure now holds the original string representation of the number, as given by the user.

For more details, see: https://sourceforge.net/p/tfel/tickets/187/

Ticket #186: “Less than” conditional is not correctly translated to C++ by tfel::math::Evaluator

The tfel::math::Evaluator class allows to translate a formula in C++ since Version 3.2.

Operators < and <= were not correctly implemented, so the following formula a<b ? 0 : 1 was translated a>b ? 0: 1 in C++.

For more details, see: https://sourceforge.net/p/tfel/tickets/186/

Ticket #182: Jacobian error in the SwiftIsotropicHardeningRule

For more details, see: https://sourceforge.net/p/tfel/tickets/182/

Ticket #174: Using formulae with the Voce hardening rule is broken

For more details, see: https://sourceforge.net/p/tfel/tickets/174/

Ticket #173: The axial component of the deformation gradient was not correctly updated in the generic interface in plane stress when using the GreenLagrange strain measure

For more details, see: https://sourceforge.net/p/tfel/tickets/173/

Ticket #157: Remove access members to the jacobian blocks associate with array of integration variables when a numerical jacobian is used

For more details, see: https://sourceforge.net/p/tfel/tickets/157/

References