Inorganic Materials Group
Exploratory Project (2018)
Machine Learning Approaches to Predicting Concrete Performance
Objective: Develop and validate a materials science driven Machine Learning (ML) approach for the discovery of structure-performance relationships in concrete binders. The project will evaluate the capability of neural networks and complementary supervised algorithms to be trained with large datasets, taken from industry and from the technical literature, to robustly predict cementitious material properties, such as compressive strength and heat of hydration, as a function of their composition, mixture proportions, and curing conditions. If successful, such algorithms could streamline concrete binder mixture design by partially supplanting costly trial-and-error physical testing with rapid calculations on a computer notebook application.
Standard Reference Materials (SRMs) and Standards
Standard Reference Materials (SRMs®) are paramount for instrument calibration and verification to measure the properties of cementitious materials. In addition, they are used in the development of test methods and for performance verification of laboratories using standardized test methods.
The Inorganic Materials group is very active in developing and maintaining SRMs. A list of SRMs resulting from their efforts are provided here, with detailed information found at the Standard Reference website :
- SRM 114q and SRM 46h: Fineness of cement powder by ASTM and AASHTO methods
- SRMs 2492, 2493 and 2497: series of Bingham fluids used to measure calibrate rotational rheometers designated for paste, mortar and concrete. (2492, and 2493 are available, 2497 will be available soon)
- SRMs 2686a, 2687, and 2688, portland cement clinker for phase abundance.
In addition, IMG is strongly involved in ASTM and AASHTO to champion standard test methods. A list of the test methods is provided here.