Machine Learning Approach to Predict Stress in Ceramic/Epoxy Composites Using Micro-Mechanical Raman Spectroscopy
Abstract
Micro-mechanical Raman spectroscopy is an excellent tool for direct stress measurements in the structure. The presence of mechanical stress changes the Raman frequency of each Raman modes compared to the Raman frequencies in absence of stress. This difference in Raman frequency is linearly related to stress induced and can be calibrated to stress by uniaxial or biaxial tension/compression experiments. This relationship is not generally linear for non-linear behavior of the materials which limits its use to experimentally study flow stress and plastic deformation behavior of the material. In this work strontium titanate ceramic particles dispersed inside epoxy resin matrix were used to measure stress in epoxy resin matrix with non-linear material behavior around it. The stress concentration factor between stress induced inside ceramic particles and epoxy resin matrix was obtained by non-linear constitutive finite element model. The results of finite element model were used for training a machine learning model to predict stress in epoxy resin matrix based on stress inside ceramic particles. By measuring stress inside ceramic particles using micro-mechanical Raman spectroscopy, the stress inside epoxy matrix was obtained by predetermined stress concentration factor.
Degree
M.Sc.
Advisors
Tomar, Purdue University.
Subject Area
Mechanics|Analytical chemistry|Artificial intelligence|Atomic physics|Chemistry|Materials science|Optics|Physics|Thermodynamics
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.