Inverse solutions on folded plate structures.

摘要

Inverse problems such as force identification, impact location, and damage detection have applications in ballistics, foreign object impact of air- and spacecraft, structural health monitoring, damage assessment, and smart structures among other fields. In the particular case of large folded plate structures, available inverse solution techniques have analytical and computational limitations which hinder their use. In this work we formulate the inverse problems in the frequency domain to gain the benefits of spectral analysis and its inherent parallelism. A spectral element approach is taken to reformulate the dynamics of a general class of stiffened and folded plate structures. The resulting spectral plate elements allow large regions to be modeled exactly using nodes only at the folds. At the same time, the frequency-domain nature of the spectral solution allows efficient solution on massively parallel computers and considerably simplifies deconvolution--the essence of our inverse solutions. The three-dimensional spectral element program developed for this work is implemented on MasPar MP-1 and MP-2 computers with 16,384 processors each. Speed-up is shown to be virtually linear with the number of processors. The combined power of the spectral element method and the massively parallel computers makes practical the solution of inverse problems on folded plate structures using experimentally measured acceleration data. Solutions to the inverse problems of force reconstruction and impact location are given for flat plates and a thin-walled box beam.