5. Summary of results#

A thermo-elastic calculation is « carried out » on a cracked metal yoke using plane stress modeling (for the mechanical part) and lumped stress modeling (for the thermal part). In accordance with « good practices » for the quality of studies, two distinct meshes are used: linear in thermal and quadratic in mechanics.

First of all, the thermal calculation is carried out (modeling A) on which the \({P}_{1}\) mesh is freely converged with a spatial error indicator map (CALC_ERREUR + “ERTH_ELEM”) /refinement-deraffination (MACR_ADAP_MAIL “RAFF_DERA”) map coupling.

In the second modeling (B), the two meshes are adapted jointly according to the same process during a chained thermo-mechanical calculation. For the free adaptation of the mechanical mesh, the pure residue indicator “ERME_ELEM” is used.

The objectives of this test case are multiple, they are:

  • to become familiar with and put into practice the two dual problems: error indicator map calculation and mesh adaptation strategies. On standard cases, but also on pathological cases and for combinations of calculations,

  • to detail the various settings of the offending operators (CALC_ERREUR, MACR_ADAP_MAIL) and related operators that may be particularly interesting for these problems (INFO_MAILLAGE, MACR_INFO_MAIL, PROJ_CHAMP…),

  • to hammer out « good practice » advice for the quality of studies and the use of tools already available on the subject. We are only interested in aspects such as mesh geometry, meshing itself and the type of finite elements. We do not focus here on the problems of time steps, the calibration of numerical parameters and on the sensitivity aspects with respect to the data,

  • to illustrate the tremendous potential and facilities that the « language ASTER/PYTHON » coupling allows in the command file of a study (test, loop, display, calculation, calculation, personal macro-command, interactivity…). Since the official test cases are calibrated to work in batch, some of these aspects have therefore been « commented » in the command file.

From a computer validation point of view, this test case makes it possible of course to test the non-regression of various combinations of error map calculations/refinement-deraffination procedures in thermo-mechanics, but also the « pre and post-processing » options for these calculations (smoothing of stresses and heat flows at the nodes, transition from an error per element to an error at the nodes per element).