Abstract
To reduce the energy-intensity and carbon footprint of Portland cement (PC), the prevailing practice embraced by concrete technologists is to partially replace the PC in concrete with supplementary cementitious materials [SCMs: geological materials (e.g., limestone); industrial by-products (e.g., fly ash); and processed materials (e.g., calcined clay)]. Chemistry and content of the SCM profoundly affect PC hydration kinetics; which, in turn, dictates the evolutions of microstructure and properties of the [PC + SCM] binder. Owing to the substantial diversity in SCMs’ compositions–plus the massive combinatorial spaces, and the highly nonlinear and mutually-interacting processes that arise from SCM-PC interactions–state-of-the-art computational models are unable to produce a priori predictions of hydration kinetics or properties of [PC + SCM] binders. In the past 2 decades, the combination of Big data and machine learning (ML)—commonly referred to as the fourth paradigm of science–has emerged as a promising approach to learn composition-property correlations in materials (e.g., concrete), and capitalize on such learnings to produce a priori predictions of properties of materials with new compositions. Notwithstanding these merits, widespread use of ML models is hindered because they: 1) Require Big data to learn composition-property correlations, and, in general, large databases for concrete are not publicly available; and 2) Function as black-boxes, thus providing little-to-no insights into the materials laws like theory-based analytical models do. This study presents a deep learning (DL) model capable of producing a priori, high-fidelity predictions of composition- and time-dependent hydration kinetics and phase assemblage development in [PC + SCM] pastes. The DL is coupled with: 1) A fast Fourier transformation algorithm that reduces the dimensionality of training datasets (e.g., kinetic datasets), thus allowing the model to learn intrinsic composition-property correlations from a small database; and 2) A thermodynamic model that constrains the model, thus ensuring that predictions do not violate fundamental materials laws. The training and outcomes of the DL are ultimately leveraged to develop a simple, easy-to-use, closed-form analytical model capable of predicting hydration kinetics and phase assemblage development in [PC + SCM] pastes, using their initial composition and mixture design as inputs.
Highlights
Concrete–a mixture of Portland cement (PC); water; sand; and stone–is the principal material used in the construction of all forms of physical infrastructure; and, more generally, the built environment
The Fourier transform (FT)-deep learning (DL) model described in section 2.0 differs from the machine learning (ML) models used in our previous studies (Cook et al, 2021b; Lapeyre et al, 2021; Xu et al, 2021) (i.e., DL model based on random forests) in one key respect: In the Fourier transform-deep learning (FT-DL) model, the database is Fourier transformation (FFT)-transformed, prior to the model’s training, so as to reduce the database’s dimensionality; whereas in the DL model, the database is used in its pristine form
The synthetic database was populated with ∼20,000 data-series (i.e., Y as a function of X), created by randomly assigning an independent set of coefficients; while varying X from 0.2 to 4.0 with a step-size of 0.2.75% of data-series were randomly selected from the database, and used to train the FT-DL and DL models
Summary
Concrete–a mixture of Portland cement (PC); water; sand; and stone–is the principal material used in the construction of all forms of physical infrastructure; and, more generally, the built environment. The combination of isothermal calorimetry and GEMS still cannot produce a priori predictions of time-resolved phase assemblage of a new binder This is because experimental measurement of the new binder’s heat evolution profiles, or PC’s hydration kinetics, would still be required. Recent studies (Cook et al, 2021b; Lapeyre et al, 2021) have shown that machine learning (ML) models–once trained from a sufficiently large calorimetry database–can produce a priori predictions of heat evolution profiles (i.e., time-dependent heat flow rate and cumulative heat release) of PC-based binders, including binary and ternary [PC + SCM] pastes. A deep learning (DL) model–trained from a heterogenous, low-volume database of heat evolution profiles of [PC + SCM] pastes–is implemented to produce a priori, highfidelity predictions of composition- and time-dependent hydration kinetics, and phase assemblage development in MODELING METHODS. As the number of inflection (i.e., non-differentiable) points in the transformed profiles are significantly lower than in the original ones, it is much easier for the FT-DL model to establish input-output correlations from the transformed profiles as compared to the original ones
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
