Sheiati, S., Nguyen, H., Kinnunen, P., & Ranjbar, N. (2023). Cementitious phase quantification using deep learning. Cement and Concrete Research, 172, 107231. https://doi.org/10.1016/j.cemconres.2023.107231
Cementitious phase quantification using deep learning
|Author:||Sheiati, Shohreh1; Nguyen, Hoang2; Kinnunen, Päivö2;|
1Department of Wind and Energy Systems, Technical University of Denmark, Frederiksborgvej 399, Roskilde, Denmark
2Faculty of Technology, Fibre and Particle Engineering Research Unit, University of Oulu, PO Box 4300, 90014, Finland
3Department of Civil and Mechanical Engineering, Technical University of Denmark, 2800 Kgs, Lyngby, Denmark
|Online Access:||PDF Full Text (PDF, 17.9 MB)|
|Persistent link:|| http://urn.fi/urn:nbn:fi-fe2023062057017
|Publish Date:|| 2023-06-20
This study investigates deep learning-based backscattered electron (BSE) image segmentation as a novel approach to automatise phase quantification of cementitious materials and estimate their degree of hydration and porosity. The case study was on Portland cement paste that hydrated from 1 day to 2 years. The initial findings suggest that using arbitrary thresholds for phase segmentation, a strong correlation can be established between the results from BSE image analysis, quantitative XRD, and EDS/BSE, particularly for samples with a hydration age >28 days. The second part demonstrates the success of automated image segmentation that relies on learning the material composition from a meticulously analysed image database, which can then predict the content of numerous other images within seconds. This novel approach can turn the analysis of cementitious materials’ phase composition from a tedious process that requires specialised equipment and expertise into a routine test for quality control.
Cement and concrete research
|Type of Publication:||
A1 Journal article – refereed
|Field of Science:||
216 Materials engineering
212 Civil and construction engineering
113 Computer and information sciences
N. Ranjbar would like to acknowledge the research grant (42098) from Villum Fonden. H. Nguyen and P. Kinnunen would like to acknowledge The University of Oulu & The Academy of Finland Profi5 (#326291). Part of this work was carried out with the support of the Centre for Material Analysis, University of Oulu, Finland.
© 2023 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).