ML23262A984
| ML23262A984 | |
| Person / Time | |
|---|---|
| Issue date: | 09/19/2023 |
| From: | Alexander Chereskin, Raj Iyengar, Pillai R, Pint B, Wendy Reed, Savara A, Sulejmanovic D NRC/RES/DE, Oak Ridge |
| To: | |
| Wendy Reed 301-415-7213 | |
| References | |
| Download: ML23262A984 (13) | |
Text
Evaluation of Molten Salt Compatibility with Structural Alloys Wendy Reed(1)*; Bruce Pint(2), Dino Sulejmanovic(2), Ashi Savara(2), Rishi Pillai(2),
Alexander Chereskin(1), Raj Iyengar(1)
(1) United States Nuclear Regulatory Commission (NRC), Washington D.C., USA (2) Oak Ridge National Laboratory (ORNL), Oak Ridge, USA 1
MPA Seminar, October 10-11, 2023
Outline
- NRC's Advanced Reactor Program
- Molten salt reactors and materials compatibility
- Evaluation of molten salt on structural alloys
- Assessment of models for long term corrosion predictions
- Conclusions and areas of potential future research 2
NRCs Advanced Reactor Program https://www.nrc.gov/reactors/new-reactors/advanced.html 3
Reactor Engineering Branch Advanced Reactor Technical Support Molten Salt Chemistry Reliability of Components &
Risk Reduction Fuel Cycle 4
MSRs: Structural Materials Behavior ALLOY THIN, PROTECTIVE OXIDE Dry/Wet Air or Steam Environment ALLOY Molten Salt Environment UNPROTECTIVE OXIDE MSRs: Reduce the corrosion through salt purification, redox control, and special alloys.
5
Compatibility Experiments
- 3 sets of experiments were carried out to look at the effect of:
- changing the ratio of molten salt volume (V) to alloy specimen surface area (SA)
- changing the capsule material; and
- changing the test methodology including welded capsules, sealed crucibles and open crucibles in a glove box
Salts and Sample Materials FLiNaK and NaCl-MgCl 7
Analysis SEM Electrochemistry ICP-OES 8
Molten Salt Compatibility: Insights
- Need understanding of:
- Specimen Mass Change
- Post-exposure microstructure characterization
- Pre-and post-exposure salt chemistry
- Electrochemistry
- Important Considerations:
- Have impurities been analyzed?
- Test environment
- Temperature
- Experimental materials
- Duration of experiment 9
Preliminary Assessment of Models for Generating Long-term Corrosion Predictions
- Predicting materials corrosion resistance (over years) is a technical gap to commercial realization of MSRs
- Long-term predictions must be made with sparse experimental data and limited physicochemical simulations
- Can AI/ML/DS help to fill these data gaps?
- There are two approaches: AI/ML/DS can augment physics-based simulations, or AI/ML/DS can attempt to directly predict outcomes
- The prediction must be made as a function of material and conditions (alloy composition, concentration, temperature) 10
Augmentation Example
- In molten salt-based alloy corrosion, individual element thermodynamic activity is the most important factor. Fe alloys with 10 elements of varying composition is one type of alloy of significant interest (Fe, Cr, Mn, Si, C, Ti, Mo, Al, Nb, Ni)
- A piecewise GP AI/ML model was trained on 50,000 compositions and 3 temperatures. It was accurate at predicting activity values and its own errors for 95% of the points. A sampling of Cr activity is shown on the left, the full parity plot on the right 0.55 0.6 0.65 0.7 0.75 13.95 14.45 14.95 15.45 15.95 16.45 16.95 Cr Activity Concentration of Cr (Atomic % )
Predicted Actual Cr Activity Parity Plot Predicted Actual 11
AI/ML/DS Future Expectations
- AI/ML has potential to help close these data gaps and to help increase our understanding of which factors matter most
- Multi-scale modeling augmented by AI/ML/DS will likely be encountered
- Understanding of propagation of uncertainties during AI/ML/DS augmentation, including uncertainties from experiment, will be needed 12
Summary
- Molten salt reactors pose unique considerations with regards to compatibility of materials with molten salt coolants during operations
- Insights on key experimental parameters used to assess molten salt compatibility of structural materials in static halide salts that could aid development of test standards were developed
- AI/ML techniques have the potential to address long-term testing data gaps 13