Engineering Applications of Artificial Intelligence and Machine Learning for Mechanical Systems and Component Performance

Homiack, Matthew; Matrachisia, John; Villareal, Tristan; Savara, Aditya; Verzi, Stephen; Iyengar, Raj

Permalink

https://hdl.handle.net/2142/121793 Copy

Description

Title

Engineering Applications of Artificial Intelligence and Machine Learning for Mechanical Systems and Component Performance

Author(s)

• Homiack, Matthew
• Matrachisia, John
• Villareal, Tristan
• Savara, Aditya
• Verzi, Stephen
• Iyengar, Raj

Contributor(s)

McKirgan, John

Issue Date

2023

Keyword(s)

• Artificial intelligence (AI)
• Nuclear
• Regulation
• Research

Abstract

Artificial intelligence (AI) is one of the fastest-growing technologies globally and has the potential to enhance decisionmaking processes for the nuclear industry. This paper explores several recent use cases of AI in support of U.S. Nuclear Regulatory Commission research efforts. The first use case explored machine learning (ML) to monitor the performance of a system. The study used a boiling water reactor full scope simulator for synthetic data generation. A scenario was created to induce a recirculation pump malfunction that may go undetected by operators but would lead to adverse operating conditions. The goal was to detect the malfunction early using AI/ML and thereby provide operators with more time to react. Long short-term memory forecasting was used as the ML method to predict future values of this trend. The study demonstrated the potential for using ML to monitor equipment performance. The scope of this research has been expanded to explore the application of multivariate time-series classification modeling to identify and classify the type of malfunction introduced to the simulated system. The second use case explored ML for sensitivity analysis and time-series prediction to augment probabilistic fracture mechanics simulations. For this use case, an interface was developed between the Extremely Low Probability of Rupture (xLPR) code and open-source ML models. The xLPR code was then used to analyze leak-before-break behavior in a pressurized water reactor piping system. Supervised ML, in the form of random forest regression, was applied to the sample input and output data from the xLPR code to determine the input variables that are most important with respect to the selected quantities of interest, both individually (univariate analysis) and across all the quantities of interest (multivariate analysis). ML was used further to explore time series prediction as a surrogate for the xLPR simulation. The third use case explored ML to overcome sparse data and enable long-term predictions of materials compatibility of nuclear reactor components in molten salt environments. This use case involves the application of ML for lifetime assessment of materials compatibility with a specific focus on the aspects of (a) loss of sound metal due to corrosion attack (e.g., depth of attack), (b) critical depletion of strengthening elements (e.g., chromium as solid solution strengthening element), and (c) corrosion-induced dissolution of strengthening phases (e.g., carbides).

Type of Resource

text

Language

eng

Handle URL

https://hdl.handle.net/2142/121793

Owning Collections