A data‐driven approach for predicting nepheline crystallization in high‐level waste glasses

Abstract

This is the author manuscript accepted for publication and has undergone full peer review but has not been through the copyediting, typesetting, pagination and proofreading process, which may lead to differences between this version and the Version of Record. Please cite this article as

Up-Cycling of Iron-Rich Inorganic Waste in Functional Glass-Ceramics

“…The interpretation of the durability of a glass-ceramic is not straightforward, since stable crystals may be embedded in a poorly durable matrix or vice versa. Nepheline crystallization (stimulated by Fe), as an example, is generally seen as negative in nuclear glasses, not directly, but as the condition for the formation of a residual glass phase with poor durability [24]. In the present case, in our opinion, it had a positive impact in immobilizing alkali ions.…”

Predicting nepheline precipitation in waste glasses using ternary submixture model and machine learning

“…This approach provides a visualization of data and modeling results that otherwise would be impossible; however, either the false positive or the false negative rate would be higher than acceptable values according to the chosen threshold when modeling nepheline precipitation. Most recently, a new descriptor, "difference based on correlation", was introduced by Sargin et al, 8 which has a higher accuracy compared to previous (single) descriptors, and also has more balanced false positive and false negative rates. In a study by McClane et al, 9 influence of compositional shifts in the residual glass on the measured durability was studied and a distinct deviation in leaching behavior as a function of a structural feature (Q unit) was observed.…”

“…In a study by McClane et al, 9 influence of compositional shifts in the residual glass on the measured durability was studied and a distinct deviation in leaching behavior as a function of a structural feature (Q unit) was observed. In addition to these efforts, machine learning (ML) approaches such as artificial neural network, 2 support vector machines, random forests, and decision trees 8 were also investigated to test their ability to predict nepheline precipitation. However, the accuracy of the models in that study was found to be limited by an "overlapping" region (i.e., a region of similar compositions that exhibited different crystallization behavior).…”

“…However, the accuracy of the models in that study was found to be limited by an "overlapping" region (i.e., a region of similar compositions that exhibited different crystallization behavior). 8 ML has been recently used in materials science, [10][11][12] and for predicting glass properties such as glass transition temperature, 13 density, 14 elastic modulus, 14,15 and dissolution. [16][17][18][19] However, misleading information or conclusions can be obtained when using ML methods without robust protocols.…”

Derivation of the Structural Integrity of Residual (SIR) glass model for the enhancement of waste loading

“…Some of these methods include: the Optical Basicity (OB) model which focuses on calculating the basicity of the cations, the Neural Network (NN) model which was developed to incorporate complex nonlinear interactions, the Sub-Mixture model (SM) which applies a polynomial fit to a pseudo-ternary diagram, and the Difference based on Correlation (DC) descriptor which combined the differences between the mass fractions of oxides observed to have positive and negative association with nepheline formation. [9][10][11][12] While these techniques have generally resulted in incremental advances needed to successfully increase the concentration (i.e., waste loading) of Al 2 O 3 in the resulting waste glass, several shortcomings still exist. These techniques are largely computationally based, rely on absolute glass composition inputs (wt%), do not necessarily consider how changes between component ratios influence the glass structure, and focus solely on circumventing crystallization instead of maintaining glass durability.…”