This is mainly a bug fix version of the 3.1 series. All tickets solved are described below.

Tickets fixed

Ticket #131: Plane stress support in Abaqus/Explicit is broken

The documentation of the vumat interface has been badly understood in plane stress. In some cases (namely using the Native finite strain strategy), the conventions about tensors has not been transcribed appropriately (there have been a confusion with the Abaqus/Standard conventions).

Note

The Native finite strain strategy in Abaqus/Explicit (based on a corotational formulation) is not supported in MTest, so there is no unit-tests associated.

For more details, see: https://sourceforge.net/p/tfel/tickets/131/

Ticket #130: The size of arrays of parameters can’t be declared using integer constants

The following statement would not work in previous versions:

@IntegerConstant Nss = 2;
@Parameter tau0[Nss] = {100e6, 100e6};

To solve this issue, a new method readArrayOfVariablesSize has been introduced in the DSLBase class. This method is now shared by all methods which reads the size of an array of variables.

For more details, see: https://sourceforge.net/p/tfel/tickets/130/

Ticket #129: Arrays of parameters are not supported in simple precision

The trouble boils down to the fact that, du to intended limitations of the tvector class, downcasting from double precision to single precision is not allowed by the assignment operator.

The call to the assignment operator has been replaced by a call to tfel::fsalgo::copy which allows downcasting.

For more details, see: https://sourceforge.net/p/tfel/tickets/129/

Ticket #128: Generation of MTest files throws an exception which can’t be caught by Cast3M in case of forced convergence

With the Cast3M interface, the generation of MTest files during a forced convergence sequence throws an exception which can’t be caught by Cast3M, leading to a severe crash.

The trouble comes from the fact that during the forced convergence sequence, the time step is null, which is not handled by the MTestFileGenerator class.

The trouble has been circumvented by introducing a minimal time step of 1.e-50 (the unit depends on the specific case treated).

For more details, see: https://sourceforge.net/p/tfel/tickets/128/

Ticket #127: Substepping in the Cast3M and Cyrano interface may lead to a invalid convergence du to integer overflow

Using substeping with a high number of substeps may lead to a to integer overflow in the counter associated with the remaining steps that have to be performed.

This counter was declared as an unsigned short. For more than 14 substeps, this counter would overflow and was automatically set egal to 0, leading to an erroneous convergence. This counter has been changed to unsigned int to avoid the problem.

For more details, see: https://sourceforge.net/p/tfel/tickets/127/

Ticket #126: Jacobian error in DDIF2Base.ixx

There was a mistake in the jacobian computed in treatFracture method of the DDIF2Base class du to the complex branching made.

For more details, see: https://sourceforge.net/p/tfel/tickets/126/

Ticket #125: Compiling on gcc 8.0.1 (Fedora 28) fails

Compiling TFEL\ withgcc8.0.1 (Fedora 28) fails inCastemGenericPlaneStressHandler.hxx`.

Reported errors a priori look like regressions in the compiler rather than actual bugs in the code. However, as the incriminated code is quite ambiguous, it was worth fixing it.

For more details, see: https://sourceforge.net/p/tfel/tickets/125/

Ticket #124: StandardElasticity brick: the elastic prediction is wrongly computed in plane stress if the stiffness tensor is computed (@ComputeStiffnessTensor) or given (@RequireStiffnessTensor)

For more details, see: https://sourceforge.net/p/tfel/tickets/124/

Ticket #123: Implementation mistakes in various behaviours based on Implicit DSL and using a numerical jacobian (use of perturbated values to update auxiliary state variables)

Various behaviours in the MFront’ tests base used local variables to compute the increment of auxiliary state variables (in order to safe some computational times and reduce the implementation size). This trick is quite often used with analytical jacobian but can cause various problems with numerical jacobian if the consistent tangent operator is requested (in this case, the system is perturbated once after convergence to get the upper part of the inverse of the jacobian, as needed by the getPartialJacobianInvert method).

The issue can be circumvented using the perturbatedSystemEvaluation flag introduced in TFEL 3.1.1 (see http://tfel.sourceforge.net/release-notes-3.1.1.html and https://sourceforge.net/p/tfel/tickets/111/ for details ).

The concerned implementations were the following:

For more details, see: https://sourceforge.net/p/tfel/tickets/123/

Ticket #122: Missing file in the distributed tar ball (mfront-query/include/CMakeLists)

The file mfront-query/include/CMakeLists.txt was not included in the distributed tar ball. This led to a warning when compiling with cmake, but it did not hinder the compilation.

For more details, see: https://sourceforge.net/p/tfel/tickets/122/

Ticket #121: Modifying the reference values read in a file does not work in MTest

In MTest, the following lines was not parsed appropriately:

@Test<file> @reference_file@ {
  'SXX' : '$8*1.e6'
} 1.e3;

For more details, see: https://sourceforge.net/p/tfel/tickets/122/

Ticket #120: Direct call to Cast3M’ umat behaviours in MTest is broken

Since TFEL 3.1, MTest is able to call raw Cast3M’ umat behaviours (not written using MFront). However, the handling of external state variables was broken.

For more details, see: https://sourceforge.net/p/tfel/tickets/120/

Ticket #119: Material laws are not declared in the @AdditionalConvergenceChecks code block

In the generation of the additionalConvergenceChecks method, the material laws were not declared.

For more details, see: https://sourceforge.net/p/tfel/tickets/119/

Ticket #118: The BehaviourDescription::areMaterialPropertiesDependantOnStateVariables method always returns true

The BehaviourDescription::areMaterialPropertiesDependantOnStateVariables method always returned true. This bugs led to unnecessary computations but did not affect the results.

For more details, see: https://sourceforge.net/p/tfel/tickets/118/

Ticket #117: Generated source fails to compile if one ask MTest file generation with array of internal state variables

A using namespace std; statement has been removed in 3.1.0 in the generated code, which led to a compile time error (ostringstream undeclared) when:

A new test case has been introduced for testing all the cases.

For more details, see: https://sourceforge.net/p/tfel/tickets/117/

Ticket #115: The stress free expansion handler shall be a nullptr for true finite strain behaviours

For more details, see: https://sourceforge.net/p/tfel/tickets/115/