Foundations for a Fission Battery Digital Twin. (3rd July 2022)

• Record Type: Journal Article
• Authors: Browning, Jeren; Slaughter, Andrew; Kunz, Ross; Hansel, Joshua; Rolston, Bri; Wilsdon, Katherine; Pluth, Adam; McCardell, Dillon
• Is Part Of: Nuclear technology. Volume 208:Number 7(2022)
• Page Start: 1089
• Page End: 1101
• Publication Date: 2022-07-03
• Subjects: Digital twin -- fission battery -- microreactor -- machine learning -- multiphysics; Nuclear engineering -- Periodicals; Nuclear engineering; Nuclear Physics; Periodicals; Periodicals; 621.48
• DOI: 10.1080/00295450.2021.2011574
• Languages: English
• ISSNs: 1943-7471

Abstract

Abstract: As the nuclear industry moves toward construction of microreactors and next-generation reactors, these efforts pose new challenges. A digital-twin tool will reduce costs and risk through integration of the disparate systems used in the design, construction, and operation of these reactors. Recent investments at Idaho National Laboratory (INL) in open-source digital engineering and multiphysics framework development provide a foundation from which to create and evaluate a digital twin for nuclear reactors. This digital-twin tool will use the Single Primary Heat Pipe Extraction and Removal Emulator (SPHERE) and Microreactor AGile Non-nuclear Experimental Testbed (MAGNET) as case studies to develop a digital twin with both single and 37 heat pipe test articles. The digital twin will provide the capabilities of remote monitoring and unattended operation (autonomous control) of these systems. A digital twin is a digital replica of an operating asset that can display data received from live sensors, update a physics model for the asset with the received data, compute predictive results of operational status with artificial intelligence (AI) to aid in optimizing asset use, and apply asset control accordingly. This twin will be developed through integration of the open-source technologies Deep Lynx (a data-warehouse technology) and the Multiphysics Object-Oriented Simulation Environment (MOOSE), physical-asset sensors, and physical-asset controls. Specifically, the general AI will successfully predict the events described as MAGNET heat pipe article test cases (such as heat pipe failure) using integrated data from the MAGNET sensors and physics-based models, including developed meta models. The integration of open-source INL software and AI assets with sensor data from a test bed will lead to a repeatable framework and guide for the creation of future digital twins. The team will also perform AI model training and experimentation to determine what models and features are most important to enable intelligent, autonomous control as well as to evaluate and determine best practices for digital-twin cybersecurity.