Understanding and Modeling the Stochastic Nature of Corrosion Reactions in Atmospheric Conditions

Abstract

Corrosion is a detrimental materials degradation mechanism to many in-service alloys and alloy systems. The stochastic nature of corrosion processes makes prediction and identification of degradation a monumental task. Additionally, the wide range of materials utilized in complex systems, coupled with variable environments, further complicates this process. Here, we utilize Finite Element Method (FEM) models coupled with statistical software (Dakota), and machine learning techniques to build a generalized framework to evaluate the probability of corrosion occurring. The ability to input defined, statistically distributed model parameters allows for the identification of corrosion probability and provides for the discovery of important parameters for future experimental efforts. In addition to the statistical predictions of corrosion, machine learning (ML) techniques allow for accelerated predictions from models, potentially reducing both time and computation effort for corrosion predictions (i.e., rate and extent of corrosion). The framework for corrosion modeling will be discussed and applied to various models, including reactive transport models. SNL is managed and operated by NTESS under DOE NNSA contract DE-NA0003525. This document is SAND2023-02444C.