Compressed dynamic mode decomposition을 통한 Kirchhoff-Love 판의 효율적인 수치해 예측

초록

Modal analysis of a vibrating plate can be effectively performed by dynamic mode decomposition (DMD). Theadvantages of DMD have been proven in various fields including image processing, modal analysis of dynamic systems,haptic localization of large display panel. Meanwhile, one of the disadvantages of DMD is that when the resolution ofeach sample in dataset is too high, the size of the matrix representing the dataset becomes too large. To overcome thisdifficulty, compressed DMD (cDMD) was recently developed. cDMD applies a compression operator before performingsingular value decomposition for a matrix representing dataset, thereby lowering the resolution of the samples at once. Byemploying a low-resolution dataset, cDMD can reduce CPU time. In this work, we perform modal analysis of theKirchhoff-Love vibrating plate using cDMD and propose an algorithm to predict the numerical solution. The performanceof cDMD was almost similar to that of DMD in terms of accuracy. On the other hand, time cost of cDMD was reducedto one-tenth of that of DMD.

키워드