期刊
PROCEEDINGS OF THE ROYAL SOCIETY A-MATHEMATICAL PHYSICAL AND ENGINEERING SCIENCES
卷 458, 期 2023, 页码 1593-1620出版社
ROYAL SOC
DOI: 10.1098/rspa.2001.0927
关键词
indentation; thin films; constitutive behaviour; neural networks; parameter identification
The indentation experiment is a popular method for the investigation of mechanical properties of thin films. By application of conventional methods, the hardness and the stiffness of the film material can be determined by limiting the indentation depth to well below the film thickness so that the substrate effects can be eliminated. In this work a new method is proposed, which allows for a determination of the reduced modulus as well as the nonlinear hardening behaviour of both the film and substrate materials. To this end, comparable deep indentations are made on the film/substrate composite to obtain sufficient information on the mechanical properties of both materials. The inverse problem is solved by training neural networks on the basis of finite-element simulations using only the easily measurable hardness and stiffness behaviour as input data. It is shown that the neural networks are very robust against noise in the load and depth. The identification of the material parameters of aluminium films on different substrates results in a significant increase in yield stress and initial work-hardening rate for a reduction of the film thickness from 1.5 to 0.5 mum, while the elastic modulus and the extent of work hardening remain constant.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据