Title:
Application of Data Science Techniques to Estimate Soluble Alkali Contribution from Fly Ashes for Determination of Concrete Pore Solution Chemistry
Author(s):
Saraswatula
Publication:
Web Session
Volume:
ws_F23_Saraswatula.pdf
Issue:
Appears on pages(s):
Keywords:
DOI:
Date:
10/29/2023
Abstract:
Although the available alkali (AA) test based on ASTM C311 is the only test currently available to measure soluble alkali contribution from fly ashes (FAs), the test is currently being discontinued due to criticisms stemming from its lack of correlation with the ASR expansions and high operator/laboratory variability of test measurements. Therefore, a Bayesian machine learning model (ML) is developed to estimate the AA from FA based on their bulk chemical composition. However, ML models only explain correlation but not the cause-effect. Subsequently, advanced analytical approaches coupled with thermodynamic modeling were used to evaluate the influence of composition, mineralogy & reactivity on AA contribution from fly ashes to pore solution. Over 400 experimental data points of FA - AA measurements, QXRD-based phase quantification, degree of reaction, water-soluble alkali measurements, pore solution extraction measurements, etc., from published literature and our labs were compiled. The data were analyzed to understand alkali dissolution from FA, evaluate their overall soluble alkali (SA) contribution, and further, refine the ML model predictions. The research findings support the judicious use of ASTM C311-based AA measurement in estimating pore solution alkalinity. Furthermore, predictive equations based on the ML model are incorporated in a new pore solution model (currently under development) to directly estimate the SA contribution from FA based on their bulk oxide composition.