Data-driven modelling of mechanical properties

• Main Author: Chang, Eldridge Wen Wei
• Other Authors: Upadrasta Ramamurty
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2022
• Subjects: Engineering::Materials::Non-metallic materials; Engineering::Computer science and engineering::Computing methodologies::Artificial intelligence
• Online Access: https://hdl.handle.net/10356/159156
• Institution: Nanyang Technological University

Computational simulation of mechanical structures is an important tool for the design and optimization of structural components and the control of new processes. However, for simulations to be predictive, it is necessary to generate models that can accurately describe the mechanical properties of materials under different states of stress. These constitutive laws are often determined on an ad-hoc basis, by fitting the results of standard mechanical tests to phenomenological models. In this Final Year Project, machine learning tools including Genetic Algorithm are utilized to develop a data-driven model for materials behaviour. We compare these models with conventionally fitted ones derived for analytical formulations. We test our strategy by fitting these models to experimental data from a historical dataset for rubber. The fitting strategy of our model is to take data from one stress state for training and validate our model by comparing its performance in two other stress states.