[CIG-SHORT] benchmarking results so far
Brad Aagaard
baagaard at usgs.gov
Thu Aug 31 17:56:01 PDT 2006
Hi all:
The benchmarking results so far are given below.
NOTE: PyLith and GeoFEST output displacements for nodes on the fault
as the average of the two sides of the fault, so these values don't
agree very well with the analytic solution. This is very evident in
plots of the local error. Femlab outputs the displacements as one
expects (offset across the fault) so its error is much less.
Attached image shows the local (element) error (difference) for PyLith
relative to GeoFEST at t=10 yrs (only elements with errors > 0.001
are shown).
Error: L2 norm integrated over the volume and normalized by the
volume, which can be considered to be the average difference in
displacement over the volume. I also calculate the L2 normal
integrated over each element normalized by the element volume, which
can be considered the average difference in displacement over the
element.
I'll try to get some more plots put together by early next week. The
differences between PyLith and GeoFEST are small but there is
definitely a little difference between the two in the viscoelastic
solution.
Brad
======================================================================
REVERSE SLIP, NO GRAVITY
======================================================================
1000m UNIFORM RESOLUTION
Time Step: 0
PyLith-0.8, error wrt analytic: 4.5162e-02
GeoFEST, error wrt analytic: 4.5165e-02
Femlab, error wrt analytic: 2.7168e-03
Time Step: 0
PyLith-GeoFEST, difference: 2.8474e-04
Time Step: 10
PyLith-GeoFEST, difference: 9.6947e-04
Time Step: 50
PyLith-GeoFEST, difference: 8.0500e-04
Time Step: 100
PyLith-GeoFEST, difference: 5.6430e-04
500m UNIFORM RESOLUTION
Time Step: 0
PyLith-0.8, error wrt analytic: 3.2159e-02
GeoFEST, error wrt analytic: 3.2163e-02
Time Step: 0
PyLith-GeoFEST, difference: 3.0927e-04
Time Step: 10
PyLith-GeoFEST, difference: 9.7460e-04
Time Step: 50
PyLith-GeoFEST, difference: 8.3009e-04
Time Step: 100
PyLith-GeoFEST, difference: 5.8660e-04
-------------- next part --------------
A non-text attachment was scrubbed...
Name: pylith_geofest_tet4_0500m_t10.png
Type: image/png
Size: 49856 bytes
Desc: not available
Url : http://geodynamics.org/pipermail/cig-short/attachments/20060831/cc1c1ef1/pylith_geofest_tet4_0500m_t10-0001.png
More information about the CIG-SHORT
mailing list