Antonio Mejia
Pronouns: he/him
Research Mentor(s): Minh Hoang Nguyen
Research Mentor School/College/Department: Aerospace Engineering / Engineering
Program:
Authors: Antonio Mejia, Minh Hoang Nguyen
Session: Session 5: 2:40 pm – 3:30 pm
Poster: 78
Abstract
Researchers have searched for ways to accurately predict the properties of composites, highly versatile materials, in an attempt to reduce costs and improve the efficiency of these materials for a wide range of industries. Industries that depend on composite materials, like aerospace, are making an effort to increasingly utilize computational tools to predict the material and structural performance and reduce trial-error physical testing. Predictiveness is crucial during aircraft design since modern airplanes are composed of up to 50% composite materials. The goal of the virtual testing principle is to conduct a small number of physical tests for material characterization and model validation, i.e. obtaining input parameters for the computational model. Our research has followed this path, mainly focusing on building a software tool that, given the inherent parameters of a composite, can predict mechanical properties. The first tool we developed is to obtain the fiber-to-matrix ratio in a composite by using image processing methods. The second tool utilizes machine learning methods such as Convolutional Neural Networks (CNNs) to automate the digital image correlation (DIC) process, which computes the strain field of the specimen.
