ML25350C334
| ML25350C334 | |
| Person / Time | |
|---|---|
| Issue date: | 12/12/2025 |
| From: | James Corson, Joseph Staudenmeier NRC/RES/DSA |
| To: | |
| Shared Package | |
| ML25350C331 | List: |
| References | |
| Download: ML25350C334 (0) | |
Text
IMPACT OF FUEL DISPERSAL FROM HIGH BURNUP PWR CORES Joseph Staudenmeier RES/DSA/CRAB I James Corson RES/DSA/FSCB December 12, 2025 1
INTRODUCTION The NRC staff is currently reviewing the Electric Power Research Institute (EPRI) Alternative Licensing Strategy (ALS) topical reports, which utilize a combination of probabilistic fracture mechanics and leak-before-break (LBB) to assert that large break LOCAs are so unlikely to occur that the effects of Fuel Fragmentation, Relocation, and Dispersal (FFRD) do not need to be considered, consistent with how certain phenomena are treated by the exception within GDC 4 for LBB. The NRC staff is considering an alternative risk-informed regulatory approach which would benefit from a reasonable technical evaluation of the impact of FFRD on the LOCA event progression and outcomes. This information may be able to support regulatory findings that FFRD does not need to be explicitly analyzed up to a certain point based on the fact that the impact would not be significant. This technical basis could be used in similar risk-informed approaches for regulatory findings on other approaches to disposition FFRD for LOCA analyses.
The Office of Nuclear Regulatory Research (RES) is supporting this effort by performing a series of analyses intended to address certain technical issues associated with fuel fragmentation, relocation and dispersal (FFRD) of high burnup/extended increased enrichment (HBU/IE) nuclear fuel due to large-break loss of coolant accidents (LOCAs) per a request from the Office of Nuclear Reactor Regulation (NRR). Task 2 requests that RES perform an evaluation of the impact of postulated fuel dispersal that accumulates on spacer grids using the dispersal estimates from Task 1.
This report is organized as follows:
Section 2 summarizes the fuel dispersal estimates from Task 1 [1].
Section 3 describes how the dispersal estimates are used to calculate the impact of the fuel dispersal.
Section 4 describes the conclusions and recommendations of the work.
2 DISPERSAL ESTIMATES FROM TASK 1 TRACE and FAST calculations performed using the WHAM methodology [2] were used to calculate fuel dispersal estimates in Task 1 using three different fuel dispersal models:
Model C from RIL 2021-13 [3]: All fuel disperses through the burst opening from contiguous nodes with burnup > 55 GWd/MTU and hoop strain > 3%.
Model A from RIL 2021-13: Fuel disperses through the burst opening from contiguous nodes with burnup > 55 GWd/MTU and hoop strain > 3%. For this case, the amount of fuel that disperses from a given node increases linearly from 0% at 55 GWd/MTU to 100% at 80 GWd/MTU.
Model A, but accounting for the impact of spacer grids: This case is the same as the model A case, but the amount of fuel susceptible to dispersal is limited to the grid span containing the burst. Note that the distance between grid spans in the upper half of the assembly is less than 12 inches for the fuel assembly design used in these calculations.
Table 1 summarizes the results of the fuel dispersal calculations for different axial power profiles, offsite power availability, available trains of pumped ECCS and dispersal models.
Figure 1 shows an example of the axial distribution of high burnup rod bursts in the FAST calculations. The majority of the bursts are near peaks in the axial power profiles shown in Figure 2.
The results are heavily influenced by the loading pattern developed by Southern Nuclear Operating Company, which has a significant number of fuel assemblies in the interior of the core that are above the fuel fine fragmentation threshold at the end of life. This will probably increase the percentage of burst rods, and thus the total dispersal, compared to existing core designs.
The TRACE and FAST calculations demonstrate that the assumption of loss of offsite power at coincident with the LOCA has little impact on the peak cladding temperature or dispersed fuel mass. The axial power profile has a pronounced impact and the chopped cosine power profile results in a higher PCT and more fuel dispersal than the top peaked profile calculated by PARCS. In the chopped cosine power profile cases, the bursts tend to occur lower in the bundle, where burnup is slightly higher and the distance between grid spans is greater than higher in the bundle, where most of the bursts occur in the PARCS power profile cases. The number of trains of pumped ECCS that are available has a significant impact on the calculations.
The parameter that has the greatest impact on the dispersed fuel mass is the fuel dispersal model itself. The most conservative model from RIL 2021-13 (i.e., model C) results in over 2%
(> 1,900 kg) of dispersed fuel in these cases, whereas the least conservative model used here (i.e., model A limited to a single grid span) results in a much smaller dispersed fuel mass (as low as ~0.5%, or ~500 kg of UO2) for the PARCS power profile cases. Adding a second ECCS train further reduces the dispersed fuel mass even more ( ~0.4% or ~400 kg of UO2) for the PARCS power profile case and the least conservative dispersal model used here.
Table 1. Summary of Fuel Dispersal Estimates from FAST Calculations PARCS
- Power, Offsite Power Available, 2 ECCS Trains PARCS
- Power, Offsite Power Available PARCS
- Power, Offsite Power Unavailable Chopped
- Cosine, Offsite Power Available Chopped
- Cosine, Offsite Power Unavailable FAST Peak Cladding Temperature (oC) 774 816 834 852 863 Burst Rods
(% of total) 25 49 55 44 49 Second Cycle Burst Rods (% of total) 24 50 58 38 41 Dispersed Mass (RIL Model C) (kg UO2) 1300 2000 2100 3400 3700 Dispersed Mass (RIL Model A) (kg UO2) 700 940 980 2300 2500 Dispersed Mass (RIL Model A, single grid span) (kg UO2) 380 530 540 1400 1500
Figure 1. Number of high burnup rod bursts (burst node burnup > 55 GWd/MTU) as a function of elevation for the PARCS (top) and chopped cosine (bottom) power profiles; offsite power is available in both cases
Figure 2. Power shapes used in the FFRD calculations.
The size distribution of the particles in the dispersed fuel can affect the impact of the dispersed fuel. The particle size distribution was measured in the NRC-sponsored tests at Studsvik [4],
and more recent tests in the SATS facility at ORNL [5]. The particle size distributions measured in the Studsvik and ORNL tests are shown in Figures 3 and 4.
Figure 3. Measured particle size distribution from the Studsvik tests.
Figure 4. Measured particle size distribution from the Studsvik and ORNL tests.
3 POSSIBLE IMPACT OF FUEL DISPERSAL The amount of fuel dispersed from the rupture and the size distribution of the dispersed particles was used to estimate the possible impact of the dispersed fuel.
The estimated particle drag was used to determine the size of particles that can be carried with the steam flow in the bundles during reflood for a range of estimated steam velocities. Water droplets in the size range of 1.5-2 mm are carried with the steam during reflood but the density of UO2 is more than 10 times that of water so the maximum fuel particle size that can be carried with the steam will be smaller. The drag force on a particle was estimated by assuming the particles were spheres. The drag force on a particle is given by
=
ll 2
(1) where A is the projected area of a sphere. The Clift-Gavin model was used to approximate the drag coefficient for a spherical particle. The model is accurate to within 6% of experimental data for the Reynolds number range less than where the drag crisis occurs at a Reynolds number of
~2.0e5. This covers the range needed for the particle sizes and steam velocities in this report.
The Clift-Gavin model for the drag is given by:
= [
24 (1 + 0.150.687)] +
0.42 1+ 42500 1.16 (2)
The drag force needed to balance the gravitational for was used to determine the maximum size of a particle that can be suspended by upward steam flow. Table 2 shows the steam velocity needed to suspend a given particle size. The steam velocity during reflood is in the range of ~5 to 10 m/s so it is likely that only particles in the 0.5 mm or less size range could be carried by the steam flow.
Table 2. Steam velocity needed to suspend a given particle size for saturated vapor at 250 kPa.
Particle size (mm)
Steam velocity (m/s)
Reynolds number Drag coefficient 0.25 5.2 140 0.94 0.5 9.4 506 0.57 1
15.6 1680 0.41 2
22.9 4900 0.39 4
30.6 13200 0.43
The Ergun equation [6] was developed to calculate the pressure drop across a bed of spheres of uniform size. It was used to estimate loss coefficients from a bed of dispersed fuel particles that accumulates at a spacer grid or other places such as the debris filter. The Ergun equation is shown in equation 3
=
150 2
(1)2 3
+
1.75 (1) 3 ll (3) where is the pressure drop across the bed, is the fluid viscosity, is the depth of the bed, is the diameter of the particles, is the porosity of the bed, is the fluid density, and is the approach velocity of the fluid based on the unobstructed area before the bed. The first term in the equation is the viscous loss term and the second term is the inertial loss term. The terms in the equation were rearranged to an equation of the form
= (+ ) ll 2
(4) which is the standard form for a loss coefficient in TRACE. The Ergun equation applies to a bed of particles of a fixed size and porosity. The situation here would have a distribution of particle sizes and experimental data would be needed to determine more realistic losses. Also note that the modelling the bed as a loss coefficient does not account for what happens to the water droplets that would be entrained in the steam flow. The droplets would likely be de-entrained by the bed and cause steam to be generated in the bed since the particles have decay heat associated with them. The water in the bed could also possibly cause additional flow resistance through the bed. These two effects could have a significant and unmodelled impact on the heat transfer above the blockages. To show the sensitivity of the results different values for the particle porosity and particle size were used to calculate loss coefficients from the bed of particles. The porosity of the bed is expected to be in the range of 0.2 to 0.4 although there has not been representative data for the actual dispersal and conditions that would occur in the fuel bundle. The particle size distribution was shown in Figures 3 and 4 and it is not clear how many would be trapped and form a bed on the grid spacer. The loss coefficients were calculated for a range of release fractions with a particle size of 2 mm and a bed porosity of 0.3 and are shown in Table 3.
Table 3. Loss coefficients for a particle size of 2 mm and a porosity of 0.3 Percent Released Mass Released (kg)
Depth of Particle bed (m)
Ergun Viscous k-loss Ergun Inertial k-loss Total Ergun k-loss 0.5 500 0.015 37 675 713 1.0 1000 0.03 74 1350 1430 2.0 2000 0.06 148 2702 2850 3.0 3000 0.09 223 4052 4280
The loss coefficients are sensitive to the particle size and the porosity of the bed. The loss coefficients were calculated for a range of release fractions with a particle size of 4 mm and a bed porosity of 0.4 and are shown in Table 4. Note the large variation in loss coefficients between tables 3 and 4 obtained by varying the bed parameters within the possible range of the parameters.
Table 4. Loss coefficients for a particle size of 4 mm and a porosity of 0.4 Percent Released Mass Released (kg)
Depth of Particle bed (m)
Ergun Viscous k-loss Ergun Inertial k-loss Total Ergun k-loss 0.5 500 0.017 3.4 142 146 1.0 1000 0.035 6.7 285 292 2.0 2000 0.069 13.4 570 583 3.0 3000 0.104 20.1 855 875 The loss coefficients were used in TRACE calculations to simulate the effect of the bursts on flow and cladding temperatures during a LBLOCA. A TRACE model that has a less detailed vessel nodalization than was used in the fuel burst calculations was used to examine the impact of the pressure loss coefficients from the released fuel assuming that it collects in a fixed axial location and has a uniform depth across the core. The added losses were implemented by using valves that go from full open to a flow area fraction corresponding to the desired value of the loss coefficient in 1 second starting at an arbitrary time of 100 seconds. A plot of the PCT for a range of loss coefficients at a grid spacer in the upper part of the core is shown in Figure 5. The loss coefficients ranged from the base grid loss coefficient up to 1600. The calculations should be examined from the point of view of whether the losses have a qualitative impact on the evolution and not from an absolute PCT point of view. The calculations show that the postulated added losses that could occur from the assumed bed of fuel particles consistent with the size of the fuel dispersal can have a significant impact on the evolution of the LBLOCA.
It is unlikely that the dispersed fuel would form a uniform bed across the core and calculations were performed using a loss coefficient to model a blockage in ring 1 of the core and the base loss coefficient (no blockage) in ring 2 of the core to look at the impact of an uneven distribution of the bed. Figures 6 and 7 show that the fuel heatup is worse for the non-uniform blockage.
This is because the flow seeks out the path of least resistance and flow is diverted from ring 1 to ring 2 because of the blockage which starves ring 1 of cooling flow. Figure 7 is interesting in that the uniform blockage had a relatively small effect on the PCT, but the non-uniform core blockage had a large impact. These calculations had a coarse nodalization of only two radial rings for the core. Using the large core model from Task 1 with bundle-by-bundle nodalization will not show as big an impact since the lateral flow resistance between adjacent bundles is smaller than the resistance between the two radial core segments in these calculations as will be shown in the following paragraphs. The calculations to this point do not account for the power generation in the debris bed as a heat source to the fluid. Sensitivity calculations
described later were performed to estimate the impact of power generation in the debris bed on the results.
Figure 5. PCT dependence on k-losses.
Figure 6. Non-uniform core blockage effect on PCT with ring 1 k-loss = 400 Figure 7. Effect of non-uniform core blockage with ring 1 k-loss = 200
The results of the simplified model calculations showed that the distribution of the blockages can have a significant impact on the peak temperatures due to the high lateral flow resistance across rows of fuel rods. The temperature increases caused by the blockages should be less if the flow that routes around the blockages has to traverse a smaller number of rows of fuel rods.
Therefore, it was decided to perform calculations with the large bundle resolution model since unblocked bundles are adjacent to bundles with blockages. The increased transverse flow resistance (x-y flow resistance) from the blockages was not modelled. The calculation used two trains of ECCS and created 4 separate failure time and mass groups to model the 45 bundles with fuel releases, based on the fuel dispersal results from Task 1. The blockages were modeled characterized into 4 groups with the maximum release in each group used to calculate the blockage to try to bound the effect of the blockages. The characteristics of the blockage groups are listed in Table 5.
Table 5. Characteristics of Modeled Blockages Number Mass (kg)
Elevation (m)
Time (s)
Effective k-loss 22 14.61 3.04 100 1096 17 5.9 3.04 100 331 4
0.85 1.82 30 48 2
33.18 1.22 26 1871 The location of the blockages is shown in Figure 8.
Figure 8. Location of the 45 core blockages The results of the calculation with and without the blockages are shown in Figure 9. The blockages delay the quench time but do not significantly increase the PCT.
Figure 9. Peak Cladding Temperature with and without blockages.
The region near the blockages was examined to see how localized the delay in quench times was. Figure 10 shows the quench profile in one of the late quenching bundles with a blockage modeled with a loss coefficient in the upper part of the heated core at approximately 2.965 meters. The figure shows that both the lower and upper quench fronts stall near the location of the blockage.
Figure 10. Quench front propagation with and without a core blockage The clad temperatures of the permanent fine nodes near the blockage shown in Figure 11 illustrate how localized the delay in quench time is. Fine node location A60 is immediately above the face with the large loss coefficient that models the blockage. The nodes are ~0.0508 m (2 in.) apart. Only the nodes very close to the modeled blockage have a quench time that is extended significantly beyond the unblocked quench times. Note that there is no data available for this situation to assess how well TRACE can calculate temperatures near a blockage. The heat transfer coefficients at the level of the peak permanent fine node (A60) are shown in Figure
- 12. The figure shows that the vapor heat transfer coefficient is the dominant one at that node until it quenches at approximately 520 seconds. Note that a small heatup occurs after the quench. The vapor heat transfer coefficient before the quench is approximately 50 W/m^2*K or less which is a low heat transfer coefficient and is in the range of natural convection heat transfer coefficients. This is because of the low fluid flow through the large resistance of the modeled blockage because the vapor flow from lower in the core takes a path around the blockage through the adjacent bundles that do not have blockages.
Figure 11. Clad temperatures for permanent fine nodes near the blockage.
Figure 12. Heat transfer coefficients at the elevation of permanent fine node A60.
It is expected that clad temperatures and quench times in a model with subchannel resolution would be less impacted than is the case for the bundle level resolution since the flow has a shorter lateral distance and a smaller lateral flow resistance to traverse to route around the blockages.
Another impact of fuel dispersal is that the particles are a source of heat to the fluid flowing through the bed. Two cases were run using the coarse nodalization two-ring core model to investigate the impact adding power to the fluid in the cell downstream of the blockage modelled with a k-loss of 200. The change to the heat transfer coefficient due to the mixture of particles and fluid was not modelled and the power added to the fluid was not subtracted from the fuel rods. The power was added directly to the fluid since the heat transfer resistance from the particles to the fluid is much smaller than the heat transfer resistance from the power in the fuel rod to the fluid because of the greatly decreased volume to surface area for the particles compared to the fuel rods. Figure 10 shows cases with the decay heat from 0.5% and 1% of the fuel corresponding to a fuel dispersal of 500 kg and 1000 kg of fuel added to the cells above the blockage. The figure shows that there is not a significant impact on the peak cladding temperature in the calculations.
Figure 10. Effect of decay power from debris particles added to fluid.
The dispersed fuel particle bed will also affect the wall to fluid heat transfer coefficient. There is heating of the fluid by the particles and heating of the fluid by the clad. There is also some contact between the wall and the particles. The question in the ideal case is how the heat transfer coefficient from the clad to the fluid is affected by the bed and the local fluid temperature. A heat transfer correlation for wall heat transfer to a packed bed was used to see if the heat transfer coefficient predicted by the correlation value is significantly different than the heat transfer coefficient calculated by TRACE during reflood. Although there are no models that have been developed for this specific problem, heat transfer models have been developed for packed bed heat transfer in chemical engineering applications [7]. The packed bed wall to fluid heat transfer coefficients can be split into the sum of a flow independent part which accounts for the effective wall heat transfer to the debris bed and a flow dependent part that accounts for the wall heat transfer to the fluid. Each part is correlated separately. The flow dependent part can be correlated as a Nusselt number that is a function of the Reynolds and Prandtl numbers as:
= 12 (5) where the characteristic length in the Nusselt and Reynolds number is the diameter of a particle and the properties are the steam properties. The Prandtl number for steam is ~1 over a wide range of conditions so there is little variation due to the exponent on the Prandl number. The exponent on the Reynolds number is in the range of 0.7-0.8 for some correlations and C is in the range of 0.15-0.2. The steam velocity in the cell with the blockage is ~2.5 m/s. For that range of values, the wall to vapor heat transfer coefficient is approximately 400-500 W/m**2-K which is ~8-10 times the wall to vapor heat transfer coefficient that TRACE is calculating in the cell with the blockage. The flow independent part of the heat transfer coefficient would further increase it. This means that the wall to vapor heat transfer in the TRACE calculation is less than what it would be if the effects of the particle bed were accounted for explicitly, if the explicit modelling does not significantly change other aspects of the calculation.
4 CONCLUSIONS The calculations performed here show that only the smallest particles from the dispersed fuel from ruptured fuel rods can be carried with the steam flow. The dispersed fuel has the potential to have a significant effect on the results of LBLOCA calculations if it collects in a bed at the grid spacers. An assumed bed can add significant flow resistance and increase the peak cladding temperature in LBLOCA calculations performed in this task. The results showed that the changes in the spatial distribution of the packed debris bed have a significant impact on the results. The more realistic cases with the bundle level modeling show less impact on the results than the coarsely nodalized 2 radial zone core. The primary impact in the bundle level modeling case is to extend the time until the core fully quenches. The peak cladding temperatures calculated with the bundle level modeling do not increase significantly compared to the base case and they meet the 10CFR50.46 acceptance criteria although no uncertainty analysis was performed for the calculation. Note that only the increased axial flow resistance was modelled.
The increased transverse flow resistance (x-y flow resistance) from the blockages was not modelled and might impact the results. The impact of dispersed fuel on LBLOCAs cannot be quantified with high confidence without more data and analysis to answer questions about the amount, spatial distribution, particle sizes, bed porosity, and the impact on reflood heat transfer of the dispersed fuel that could potentially collect and form a bed above grid spacers. A self-consistent modelling approach that allows for a detailed and rigorous assessment of the impact
of dispersed fuel in LBLOCAs would need an integrated testing, analysis, and code development.
5 REFERENCES
[1] J. Corson and J. Staudenmeier, "Realistic estimates of fuel dispersal from high burnup PWR cores," U.S. Nuclear Regulatory Commission, Washington, DC, 2025 (ADAMS Accession Number ML25350C333).
[2] A. Bielen, J. Corson and J. Staudenmeier, "NRC's methodology to estimate fuel dispersal during a large break loss of coolant accident," Nuclear Engineering and Design, vol. 426, p.
113377, 2024.
[3] M. Bales, A. Chung, J. Corson and L. Kyriazidis, "RIL 2021-13, Interpretation of Research on Fuel Fragmentation, Relocation, and Dispersal at High Burnup," U.S. Nuclear Regulatory Commission, Washington, DC, 2021.
[4] M. E. Flanagan, P. Askeljung and A. Puranen, "NUREG-2160: Post-Test Examination Results from Integral, High-Burnup, Fueled LOCA Tests at Studsvik Nuclear Laboratory,"
U.S. Nuclear Regulatory Commission, Washington, DC, 2013 (ADAMS Accession Number ML13240A256).
[5] N. Capps, Y. Yan, A. Raftery, Z. Burns, T. Smith, K. Terrani, K. Yueh, M. Bales and K. Linton, "Integral LOCA fragmentation test on high-burnup fuel," Nuclear Engineering and Design, vol. 367, p. 110811, 2020.
[6] S. Ergun, "Fluid flow through packed columns," Chemical Engineering Progress, vol. 48, no.
2, p. 504, 1952.
[7] B. Koning, Heat and Mass Transport in Tubular Packed Bed Reactors at Reacting and Non-Reacting Conditions. PhD Thesis, University of Twente,The Netherlands., 2002.