6. Summary of results

The results obtained are generally satisfactory. Whether with a complete mass matrix or with a diagonal mass matrix, we find the frequencies of publications.

The results on the nodal reactions resulting from the calculation by spectral seismic method are for the most part close to the reference results at less than 5%. However, we note that on a few points of support the differences in nodal reactions between*Code_Aster* and ANSYS amount to more than 30% (modeling A). They are explained by differences in the modal deformations calculated by the two software programs (up to 2% on modeling B), as well as in the reactions to the supports of the modes (up to 5% difference in modeling B). These differences can be traced back to distinct implementations on finite beam elements between the two codes. In particular for A-modeling, ANSYS uses « finite element » interpolation for beam elements while*Code_Aster* uses « exact » integration [R3.08.01 – éléments «exacts» de poutre].

However, it is noted that they fall well short of the differences resulting from the choice of method or the « missing mass » effect (effect of high frequency modes).

As far as the so-called Gupta method is concerned, it seems well validated by this external reference.

The advantage of C modeling is to validate the use for a modal base resulting from a calculation by dynamic substructuring as well as option ACCE_ABSOLU in mono-support.

The advantage of D modeling is to validate multi-support and multi-direction calculations with or without taking into account the pseudo-mode.