Physically Informed Data-Driven Methods for Greatly Enhancing the Use of Heterogeneous Supplementary Cementitious Materials in Transportation Infrastructure
For reasons of abundance and availability, fly ash is a desirable supplementary cementitious material (SCM) that can be used to partially replace cement in the binder fraction in concrete. However, variabilities in composition and seasonal fluctuations in the availability of fly ashes from familiar sources result in alterations and uncertainty in concrete performance when fly ash is used as an SCM. This has limited fly ash use, for example, in terms of prevailing cement replacement levels. To enhance fly ash use, this research will:
- Reveal a reduced order descriptor of fly ashes that describes their performance as an SCM.
- Develop physically informed machine learning (ML) methods for predicting and optimizing the performance of fly ash-containing binders and concrete.
- Use the insights gained to reduce the cost, extent of overdesign, and environmental impact of high-volume fly ash binders.
These goals will be achieved within a big-data framework, wherein carefully curated reduced-order variables guide, inform, and constrain the ML methods. This transformational strategy allows unprecedented predictions of the properties of original Portland cement plus fly ash binders, thereby enhancing the use of fly ash in concrete due to the resulting certainty and assurance in material properties and performance that is achieved.