Vol.:(0123456789) 1 3 MRS Advances (2021) 6:131–137 https://doi.org/10.1557/s43580-021-00031-2 ORIGINAL PAPER Machine learning to predict refractory corrosion during nuclear waste vitrifcation Natalie J. Smith‑Gray 1  · Irmak Sargin 2  · Scott Beckman 1,2  · John McCloy 1,2 Received: 1 January 2021 / Accepted: 9 February 2021 / Published online: 22 February 2021 © The Author(s), under exclusive licence to The Materials Research Society 2021 Abstract The goal of this study was to determine the efects of model nuclear waste glass composition on the corrosion of Monofrax ® K-3 refractory, using machine learning (ML) methods for data investigation and modeling of published borosilicate glass composition data and refractory corrosion performance. First, statistical methods were used for exploration of the data, and the list of features (model terms) was determined. Several model types were explored, and the Bayesian Ridge type was the most promising due to low mean average error and mean standard error as well as high R 2 value. Parameters and model results using previously identifed model features and those from this study are compared. ML methods appear to give results at least as good as previously available models for describing the efects of glass composition on refractory corrosion. Keywords Glass · Machine learning · Waste management · Corrosion · Nuclear materials Introduction At the Hanford Site, there are approximately 56 million gallons of nuclear waste stored temporarily in underground tanks [1]. The U. S. Department of Energy has contracted to build the Waste Treatment and Immobilization Plant to vitrify much of this waste, beginning with low-activity waste vitrification by December 2023. There are many parameters related to both glass properties and glass pro- cessing to consider when developing the glass composi- tions for nuclear waste vitrifcation. For the current study, the processing parameter of interest is the impact of the glass melt on Monofrax ®1 K-3 refractory, which lines the joule-heated ceramic-lined melters (JHCM) at Hanford [2]. The potential for refractory attack by glass in low-activity waste (LAW) melters is more concerning than for analogous high-level waste (HLW) melters. The LAW waste contains a much higher alkali fraction, on average, which is known to degrade typical refractories like the spinel-alumina/chromia Monofrax ® K-3 [containing (Mg, Fe)O·(Al, Cr) 2 O 3 spinel and (Al, Cr) 2 O 3 as the primary crystalline phases] [2, 3]. This refractory is typically (wt%): 58 Al 2 O 3 –27 Cr 2 O 3 –6 MgO–6 Fe 2 O 3 –1.6 SiO 2 –0.3 Na 2 O–0.4 other [2]. Several factors act simultaneously to destroy refractory during nuclear waste vitrifcation, such as the arrangement of refractories within the melter, its operational characteristics, and the glass composition [4]. Physical (erosive) degrada- tion occurs when molten glass circulates within the melter, agitating the refractory lining [4]. Chemical (corrosive) destruction can occur when elements from the glass perme- ate into the refractory causing chemical reactions [2, 4]. For radioactive glass JHCMs, it has typically been assessed that when the refractory is degraded, the entire melter must be replaced and decommissioned due to the cost and technical challenge with relining [4]. Similarly, refractory lined com- mercial glass melters typically have a ‘campaign lifetime,’ which is strongly afected by the refractory lifetime, typically John McCloy was an editor of this journal during the review and decision stage. For the MRS Advances policy on review and publication of manuscripts authored by editors, please refer to mrs. org/editor-manuscripts. Supplementary Information The online version contains supplementary material available at https://doi.org/10.1557/s4358 0-021-00031-2. * John McCloy john.mccloy@wsu.edu 1 Material Science and Engineering Program, Washington State University, Pullman, WA 99164, USA 2 School of Mechanical and Materials Engineering, Washington State University, Pullman, WA 991654, USA 1 Monofrax ® is a trademark of the GmbH & Co. KG of Vienna, Aus- tria.