Fly Ash Dissolution Data Collection for Machine Learning Models

Abstract

Bio-inspired silicate dissolution refers to the use of naturally occurring organic acids (e.g., citric, oxalic, acetic) and their chelating agents in enhancing the breakdown of aluminosilicate structures during natural weathering. Specifically, this research focuses on fly ash, whose chemical reactivity relies on silicate dissolution principles, contributing to the development of sustainable environmental solutions in concrete and other building materials. Fly ash is a by-product of coal combustion and is composed of amorphous (glassy) and crystalline aluminosilicates; this research highlights the descriptors and dissolution processes of Class F, Class C, and CFB fly ashes. To conduct this research, numerical data was extracted from over 30 papers and databases dedicated to the properties of fly ash dissolution. Currently, approximately 110 data records are organized into a machine learning template containing various input variables, including fly ash descriptors, organic acids involved in the dissolution process, and the reaction monitoring metrics during dissolution. However, many papers on fly ash dissolution report these parameters inconsistently or without quantitative detail. This lack of standardized data makes it difficult to develop predictive AI models. Nonetheless, the numerical data was synthesized into the machine learning template. For instance, fly ash descriptor data was divided based on the oxides, trace metals, and crystallinity present in each class. The following data represents Class F oxide presence: 37.0-62.1% of SiO2, 16.6-35.6% of Al2O3, 2.6-21.2% of Fe2O3, 0.50-14.0% of CaO, and 0.3-5.2% of MgO. As for the reaction parameters, it was found that various organic acids are involved in the dissolution process, specifically citric, oxalic, and acetic acid; the template highlights their properties, such as the pKa values of citric acid—ranging from 2.96-3.14 (pKa1), 4.38-4.76 (pKa2), and 5.68-6.40 (pKa3)—and its initial pH values—ranging from 2.08-3.24 depending on the solution. By merging fragmented experimental literature into a cohesive framework, as shown in the examples above, the study establishes a dataset that promotes subsequent AI-driven predictions of fly ash dissolution behavior. Further, this development also supports the more sustainable use of industrial by-products, such as fly ash, in building materials.