Using ensemble model to predict isothermal hydration heat of fly ash cement paste considering fly ash content, water to binder ratio and curing temperatureopen access
- Authors
- Sun, Y.; Lee, Han Seung
- Issue Date
- Jul-2023
- Publisher
- Elsevier BV
- Keywords
- Fly ash (FA); Cement; Isothermal hydration heat; Ensemble regression model
- Citation
- Case Studies in Construction Materials, v.18, pp 1 - 12
- Pages
- 12
- Indexed
- SCIE
SCOPUS
- Journal Title
- Case Studies in Construction Materials
- Volume
- 18
- Start Page
- 1
- End Page
- 12
- URI
- https://scholarworks.bwise.kr/erica/handle/2021.sw.erica/112549
- DOI
- 10.1016/j.cscm.2023.e01984
- ISSN
- 2214-5095
- Abstract
- It is widely accepted that fly ash (FA) addition in cement impairs the reaction of silicate phase and promotes the reaction of aluminate phase. Meanwhile, FA also retards early-age hydration heat release of cement. No generalized hydration prediction model is available for accurate description of the isothermal heat release of FA cement paste. In this study, we try to adopt an ensemble model, one of the machine learning approaches, to predict the early-age isothermal hydration heat flow of FA cement paste. Effect of curing temperature, water to binder ratio (w/b), FA content and curing time on heat release is thoroughly considered in the ensemble model, and the hyperparameters of the model are tuned by Bayesian optimization with 5-fold cross validation. The results indicate that whether for the training data or the testing data, the root-mean-square error (RMSE) and mean-absolute-error (MAE) have been well controlled at approximately 0.1 and 0.06 mW/g paste, respectively. The optimized model is also applied to predicting the time -dependent isothermal hydration heat flows of three different FA cement systems cured under different temperatures. The coefficient of determination (R2) values for both the predicted heat flow rates and the calculated cumulative heat have attained 0.98, proving the applicability of the trained model on isothermal heat prediction for FA cement paste. Sensitivity analysis is also conducted for the optimized model. Curing temperature and time are revealed to be the two most important input variables for isothermal heat prediction using machine learning, followed by FA content, and the least important one is w/b. Neglecting any of them reduces the prediction ac-curacy of the model.
- Files in This Item
-
Go to Link
- Appears in
Collections - COLLEGE OF ENGINEERING SCIENCES > MAJOR IN ARCHITECTURAL ENGINEERING > 1. Journal Articles
Items in ScholarWorks are protected by copyright, with all rights reserved, unless otherwise indicated.