502 ニューラルネットワークを用いた多層積層材の層厚及び弾性特性評価

概要

論文の詳細を見る
This paper describes a neural network application for nano-indentation testing to evaluate material properties of multi-layered materials. Load-depth curves are produced by finite element simulation of indentation in different material properties with four indenters having different radii. The trained network using the load-depth curves as learning data is utilized to evaluate material properties of multi-layered materials. It is demonstrated that appropriate estimation is achieved for thickness of 2-layered materials. In this study, a strategy named two step estimation is adopted for other estimations. So that, the final estimation is conducted by another network trained by learning data those range are narrowed down based on the first estimation. It is revealed that two step estimation is effective t identify the each thickness of 3-layered materials and elastic properties of 4-layered materials.
社団法人日本機械学会の論文
2000-07-31