Public Abstract

DE-SC0023988: Machine Learning Methods for Early CIPS Detection

Award Status: Active

Institution: Veracity Nuclear, LLC, Lenoir City, TN
UEI: U8C8MCVHS5B4
DUNS: 118479089

Most Recent Award Date: 09/23/2024
Number of Support Periods: 2
PM: Goetz, Sujata

Current Budget Period: 09/10/2024 - 09/09/2025
Current Project Period: 09/10/2024 - 09/09/2026
PI: Godfrey, Andrew

Supplement Budget Period: N/A

Public Abstract

Machine Learning for Early CIPS Detection-Veracity Nuclear, LLC, 1488 Blackberry Ridge Dr., Lenoir City, TN 37772-4028 Godfrey, Andrew, Principal Investigator, andrew.godfrey@veracitynuclear.com Collins, Benjamin, Business Official, ben.collins@veracitynuclear.com Amount: $1,150,000 The build-up of crud deposits on fuel in operating nuclear reactors has been a problem of concern for the past three decades which has resulted in fuel failures, lost electrical generation, and unplanned outages due to the phenomena of Crud-Induced Power Shift and Crud-Induced Localized Corrosion. Crud risk is currently managed using conservative fuel management and reactor operation strategies, resulting in higher fuel costs, due to the purchase of additional fuel, and higher maintenance costs due to periodic ultrasonic fuel cleaning. These costs are estimated to be $80 million per year across the national operating reactor fleet. The efficient management of crud risk is therefore a significant opportunity to reduce nuclear energy production costs. In addition, crud is a critical barrier that must be overcome for reactors considering power uprates, longer operating cycles, advanced fuel products, and flexible operations. This proposal will deliver a novel reactor core monitoring technology to assess crud risk during reactor operation based on a first-of-a-kind, crud machine learning engine. The key advantage of our product is that it enables the real-time assessment of crud risk by integrating historical reactor operating data, in-core reactor flux measurements, and high-fidelity reactor simulation in the forward prediction of crud risk. By monitoring the onset of crud risk, sufficient forewarning is provided to allow for mitigating actions and the avoidance of disruption to reactor operation. The differentiator for the technology being developed is the coupling of high-fidelity reactor core simulation, which provides detailed three-dimensional crud modeling on a fuel rod-by-rod basis, with reactor measurements obtained during the operating cycle. The resultant crud machine learning engine would be reactor-specific and be integrated within existing utility risk assessment workflows improving overall plant economic performance. Immediate benefits of the crud machine learning engine include, beyond de-risking future plant improvements, the identification of crud risk margin, reduced fuel costs, and the elimination of fuel cleaning. Phase 1 of this work demonstrated the methodology for crud machine learning based on the use of generated synthetic data for the calculation of Crud-Induced Power Shift with application to crud monitoring using fixed in-core detectors. Phase 2 will focus on extending the methodology to Crud-Induced Localized Corrosion for moveable and fixed in-core detectors. In addition, validation of the complete crud methodology will be performed using historical reactor operating data for several operating reactors. This will entail working closely with utility partners to benchmark the reactor models and establish crud thresholds based on cycles with and without Crud-Induced Power Shift and Crud-Induced Localized Corrosion.

Machine Learning for Early CIPS Detection-Veracity Nuclear, LLC, 1488 Blackberry Ridge Dr., Lenoir City, TN 37772-4028

Godfrey, Andrew, Principal Investigator, andrew.godfrey@veracitynuclear.com

Collins, Benjamin, Business Official, ben.collins@veracitynuclear.com

Amount: $1,150,000

The build-up of crud deposits on fuel in operating nuclear reactors has been a problem of concern for the past three decades which has resulted in fuel failures, lost electrical generation, and unplanned outages due to the phenomena of Crud-Induced Power Shift and Crud-Induced Localized Corrosion. Crud risk is currently managed using conservative fuel management and reactor operation strategies, resulting in higher fuel costs, due to the purchase of additional fuel, and higher maintenance costs due to periodic ultrasonic fuel cleaning. These costs are estimated to be $80 million per year across the national operating reactor fleet. The efficient management of crud risk is therefore a significant opportunity to reduce nuclear energy production costs. In addition, crud is a critical barrier that must be overcome for reactors considering power uprates, longer operating cycles, advanced fuel products, and flexible operations. This proposal will deliver a novel reactor core monitoring technology to assess crud risk during reactor operation based on a first-of-a-kind, crud machine learning engine. The key advantage of our product is that it enables the real-time assessment of crud risk by integrating historical reactor operating data, in-core reactor flux measurements, and high-fidelity reactor simulation in the forward prediction of crud risk. By monitoring the onset of crud risk, sufficient forewarning is provided to allow for mitigating actions and the avoidance of disruption to reactor operation. The differentiator for the technology being developed is the coupling of high-fidelity reactor core simulation, which provides detailed three-dimensional crud modeling on a fuel rod-by-rod basis, with reactor measurements obtained during the operating cycle. The resultant crud machine learning engine would be reactor-specific and be integrated within existing utility risk assessment workflows improving overall plant economic performance. Immediate benefits of the crud machine learning engine include, beyond de-risking future plant improvements, the identification of crud risk margin, reduced fuel costs, and the elimination of fuel cleaning. Phase 1 of this work demonstrated the methodology for crud machine learning based on the use of generated synthetic data for the calculation of Crud-Induced Power Shift with application to crud monitoring using fixed in-core detectors. Phase 2 will focus on extending the methodology to Crud-Induced Localized Corrosion for moveable and fixed in-core detectors. In addition, validation of the complete crud methodology will be performed using historical reactor operating data for several operating reactors. This will entail working closely with utility partners to benchmark the reactor models and establish crud thresholds based on cycles with and without Crud-Induced Power Shift and Crud-Induced Localized Corrosion.