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

Abstract

Nepheline precipitation in nuclear waste glasses during vitrification can be detrimental due to the negative effect on chemical durability often associated with its formation. Developing models to accurately predict nepheline precipitation from compositions is important for increasing waste loading since existing models can be overly conservative. In this study, an expanded dataset of 955 glasses, including 352 high-level waste glasses, was compiled from literature data. Previously developed submixture models were refitted using the new dataset, where a misclassification rate of 7.8% was achieved. In addition, nine machine learning (ML) algorithms (k-nearest neighbor, Gaussian process regression, artificial neural network, support vector machine, decision tree, etc.) were applied to evaluate their ability to predict nepheline precipitation from glass compositions. Model accuracy, precision, recall/ sensitivity, and F1 scores were systemically compared between different ML algorithms and modeling protocols. Model prediction with an accuracy of ~0.9 (misclassification rate of ~10%) was observed for different algorithms under certain protocols.This study evaluated various ML models to predict nepheline precipitation in waste glasses, highlighting the importance of data preparation and modeling protocol, and their effect on model stability and reproducibility. The results provide insights into applying ML to predict glass properties and suggest areas for future research on modeling nepheline precipitation.

Data-driven machine learning for disposal of high-level nuclear waste: A review

“…Table 1 Main research on the status of the DDML for the disposal of HLW. Unsupervised learning type (Lu et al, 2021); (Sirdesai et al, 2019); (Yoon et al, 2019) Artificial neural network (ANN) Supervised learning type (Solans et al, 2021); (Elodie et al, 2020); (Tsai et al, 2019) Genetic algorithm (GA) Supervised or Unsupervised learning type (Suh et al, 2020); (Xu et al, 2020) Clustering method (CM) Unsupervised learning type (Stanfill et al, 2020); (Suh et al, 2018) Logistic regression Supervised learning type Deep…”

“…6. It is also adopted to predict the nepheline precipitation from compositions (Lu et al, 2021) and the specific heat capacity of bentonite buffer materials (Yoon et al, 2019).…”

“…Gong et al (2021) conducted the leaching behaviors of A2B2O7-structures with the LR and kernel ridge regression (KRR) method. Finally, Lu et al (2021) predicted the nepheline precipitation from compositions with the kNN, GPR, ANN, SVM, etc.…”

“…(1) hybrid DDML algorithms. Firstly, the hybrid DDML algorithms are nice choices as presented (Lu et al, 2021;Benbouras and Lefilef, 2021;Gong et al, 2021;Bang et al, 2020;Lillington et al, 2020;Ebiwonjumi et al, 2021). The accuracy of the DDML is related to many factors including the learning algorithm and model parameters.…”

Glass formulation and composition optimization with property models: A review

“…(B) Accuracy of machine learning models (k-nearest neighbor, KNN; support vector machine, SVM; Gaussian process regression, GPR; decision tree, DT; random forest, RF; artificial neural network, ANN; Gaussian naive Bayes, GNB; quadratic discriminant analysis, QDA; logistic regression) to predict nepheline formation in waste glasses fitted using three model fitting protocols (e.g., using mole fraction or mol% as model inputs). 28 modeling, and developing efficient/optimized formulation approaches are critical to accelerating the design of new glass materials. [18][19][20][21][22] The objective of this paper is to review the recent development of glass property models, including model fitting, feature extraction, model evaluation, and uncertainty quantification (UQ).…”