Department of Mechanics: Seminar: Abstract Krysl 2019: Difference between revisions
(Created page with "===== Using a priori model reduction for the FEM to quickly approximate free-vibration response of solids ===== ==== Petr Krysl, University of California, La Jolla, San Diego,...") |
No edit summary |
||
| (3 intermediate revisions by 2 users not shown) | |||
| Line 1: | Line 1: | ||
=== Using a priori model reduction for the FEM to quickly approximate free-vibration response of solids === | |||
==== Petr Krysl, University of California, La Jolla, San Diego, USA ==== | ==== Petr Krysl, University of California, La Jolla, San Diego, USA ==== | ||
Modal expansion is a workhorse used in many engineering analysis algorithms. | Modal expansion is a workhorse used in many engineering analysis algorithms. One example is the coupled boundary element-finite element computation of the backscattering target strength of underwater elastic objects. To obtain the modal basis, a free-vibration (generalized eigenvalue) problem needs to be solved. This tends to be expensive when there are many basis vectors to compute. In the above mentioned backscattering example it could be many hundreds or thousands. Excellent algorithms exist to solve the free-vibration problem, and most use some form of the Rayleigh-Ritz (RR) procedure. The key to an efficient RR application is a low-cost transformation into a reduced basis. In this work we show how a cheap a prior transformation can be constructed for solid-mechanics finite element models based on the notion of coherent nodal clusters. The inexpensive RR procedure leads to not insignificant speedups of the computation of an approximate solution to the free vibration problem. | ||
One example is the coupled boundary element-finite element computation of the | |||
backscattering target strength of underwater elastic objects. To obtain the modal | |||
basis, a free-vibration (generalized eigenvalue) problem needs to be solved. This | |||
tends to be expensive when there are many basis vectors to compute. In the above | |||
mentioned backscattering example it could be many hundreds or thousands. | |||
Excellent algorithms exist to solve the free-vibration problem, and most use some form of | |||
the Rayleigh-Ritz (RR) procedure. The key to an efficient RR application is | |||
a low-cost transformation into a reduced basis. In this work we show how a cheap a | |||
transformation can be constructed for solid-mechanics finite element models based | |||
on the notion of coherent nodal clusters. The inexpensive RR procedure leads to | |||
not insignificant speedups of the computation of an approximate solution to the free | |||
vibration problem. | |||
Latest revision as of 21:28, 26 June 2020
Using a priori model reduction for the FEM to quickly approximate free-vibration response of solids
Petr Krysl, University of California, La Jolla, San Diego, USA
Modal expansion is a workhorse used in many engineering analysis algorithms. One example is the coupled boundary element-finite element computation of the backscattering target strength of underwater elastic objects. To obtain the modal basis, a free-vibration (generalized eigenvalue) problem needs to be solved. This tends to be expensive when there are many basis vectors to compute. In the above mentioned backscattering example it could be many hundreds or thousands. Excellent algorithms exist to solve the free-vibration problem, and most use some form of the Rayleigh-Ritz (RR) procedure. The key to an efficient RR application is a low-cost transformation into a reduced basis. In this work we show how a cheap a prior transformation can be constructed for solid-mechanics finite element models based on the notion of coherent nodal clusters. The inexpensive RR procedure leads to not insignificant speedups of the computation of an approximate solution to the free vibration problem.