12. J modeling#
12.1. Characteristics of modeling#
We use 3D modeling in non-linear mechanics with hyper-reduction of the model. This modeling tests the accuracy of the empirical modes obtained in modeling E on the reduced domain created in the F modeling.
We therefore use the bases produced in modeling D and the reduced domain created in modeling F. This modeling tests how to reconstruct fields on the domain wholemeal
12.2. Characteristics of the mesh#
The mesh contains 27 elements of type HEXA8.
12.3. Tested sizes and results#
12.3.1. Values in the RID#
We test the value of the displacements obtained with a hyper-reduced model compared to those obtained in the complete model:
Location |
Moment |
Component |
Movement ( DEPL ) |
Instant |
Precision |
Node A in \(({1,0,3})\) |
|
DX |
\({0,0696319525128}\,{mm}\) |
0.0015 |
|
Node A in \(({1,0,3})\) |
|
DY |
\({0,199062276741}\,{mm}\) |
0.0035 |
|
Node A in \(({1,0,3})\) |
|
DZ |
\({0,529606351907}\,{mm}\) 0.0015 |
0.0025 |
And the constraints:
Location |
Instant |
Component |
Component |
Constraint ( SIEF_NOEU ) |
Precision |
|
D knot in \(({3,1,0})\) |
|
|
|
0, 20% |
||
D knot in \(({3,1,0})\) |
|
|
|
0, 20% |
||
D knot in \(({3,1,0})\) |
|
|
|
0, 20% |
12.3.2. Values outside of RID#
Values outside of RID cannot be obtained directly from the hyper-reduced calculation with STAT_NON_LINE. It is necessary to rebuild the fields. We are testing the reconstruction of fields « by hand » by retrieving the reduced coordinates and applying the formula with CREA_CHAMP:
Location |
Moment |
Component |
Movement ( DEPL ) |
Instant |
Precision |
Node B in \(({3,3,3})\) |
|
DX |
\(-{0,208992921588}\,{mm}\) |
0.0015 |
|
Node B in \(({3,3,3})\) |
|
DY |
\(-{0,208992921588}\,{mm}\) |
0.003 |
|
Node B in \(({3,3,3})\) |
|
DZ |
\({0,547254495773}\,{mm}\) |
0.0025 |
Location |
Instant |
Component |
Component |
Constraint ( SIEF_NOEU ) |
Precision |
Node C in \(({0,0,0})\) |
|
|
|
|
|
Node C in \(({0,0,0})\) |
|
|
|
|
|
Node C in \(({0,0,0})\) |
|
|
|
|
We test the reconstruction of the fields by gappy- POD using the REST_REDUIT_COMPLET operator:
Location |
Moment |
Component |
Move ( DEPL ) |
|
Node B in \(({3,3,3})\) |
|
DX |
\(-{0,208992921588}\,{mm}\) |
0.0015 |
Node B in \(({3,3,3})\) |
|
DY |
\(-{0,208992921588}\,{mm}\) |
0.003 |
Node B in \(({3,3,3})\) |
|
DZ |
\({0,547254495773}\,{mm}\) |
0.002 |
Location |
Instant |
Component |
Component |
Constraint ( SIEF_NOEU ) |
Precision |
Node C in \(({0,0,0})\) |
|
|
|
0.002 |
|
Node C in \(({0,0,0})\) |
|
|
|
0.002 |
|
Node C in \(({0,0,0})\) |
|
|
|
0.002 |
We are also testing PREDICTION =” EXTRAPOLE “:
Location |
Moment |
Component |
Movement ( DEPL ) |
Instant |
Precision |
Node A in \(({1,0,3})\) |
|
DX |
\({0,0696319525128}\,{mm}\) |
0.0015 |
|
Node A in \(({1,0,3})\) |
|
DY |
\({0,199062276741}\,{mm}\) |
0.0015 |
|
Node A in \(({1,0,3})\) |
|
DZ |
\({0,529606351907}\,{mm}\) |
0.033 |
12.4. notes#
The reconstruction of fields, especially outside of RID, is giving excellent results in terms of precision.