Title:
Analysis of Micro Mechanical Properties of Geopolymer Gel Using Machine Learning Technique
Author(s):
Roshan Arachchige
Publication:
Web Session
Volume:
Issue:
Appears on pages(s):
Keywords:
DOI:
Date:
10/23/2022
Abstract:
Machine learning is a widely used statistical tool that is now common in most fields. Trained machine learning models can easily replace sophisticated and expensive methods used for material characterization. Here, we utilize the latest development in machine learning to predict micromechanical properties such as hardness and modulus of elasticity of Geopolymer paste (GP) from statistical nanoindentation results along with energy dispersive X-ray spectroscopy (EDS) mapping. We analyzed nanoindentation data using K-means clustering and the Gaussian mixture model to cluster and identify the phases present in the alkali-activated pastes. Furthermore, EDS mapping was done on the samples where nanoindentation was performed. The EDS maps are later converted to RGB photographs where the color intensities represent the major elemental intensities (Si, Al, Na). Nanomechanical data and the color intensities from EDS maps are trained using Multiple Linear regression (MLR), Support Vector Regression (SVR) and Gaussian Process Regression (GPR) to build up models to predict the nanomechanical properties for given elemental compositions. The GPs are produced using calcined clays, ground bottom ashes, volcanic ashes, and fluidized bed combustion ashes as pre-cursors and sodium silicate solution as the activator. The statistical nanoindentation and the EDS maps were collected from 28-days old ambient cured paste samples. The predictive capabilities and advantages of the method will be assessed by direct comparison with experimental results for different geopolymer gels.