ML23115A008

SER for TR-0116-21012, Revision 1, NuScale Power Critical Heat Flux Correlations Public Version

ML23115A008
Person / Time
Site:	99902078
Issue date:	04/25/2023
From:	Bavol B NRC/NRR/DNRL/NRLB
To:
References
TR-107522, Rev 1, EPID L-2021-TOP-0033
Download: ML23115A008 (36)

Text

SAFETY EVALUATION BY THE OFFICE OF NUCLEAR REACTOR REGULATION SUPPLEMENT 1 TO TR-0116-21012-P-A TOPICAL REPORT 107522, REVISION 1 APPLICABILITY RANGE EXTENSION OF NSP4 CRITICAL HEAT FLUX CORRELATION NUSCALE POWER, LLC EPID NO. [EPID NO. L-2021-TOP-0033]

1.0 INTRODUCTION

By letter dated November 5, 2021 (Reference 1), NuScale Power, LLC (NuScale) submitted a request for review and approval of Topical Report (TR)-107522, Revision 0, Applicability Range Extension of NSP4 Critical Heat Flux Correlation: Supplement 1 to TR-0116-21012-P-A, Revision 1, to the U.S. Nuclear Regulatory Commission (NRC). By letter dated October 27, 2022 (Reference 16), NuScale submitted TR-107522, Revision 1, Applicability Range Extension of NSP4 Critical Heat Flux Correlation, to NRC that incorporates the requested updates from staff. The purpose of this report was to provide the bases for an extension to the range of applicability for the NSP4 critical heat flux (CHF) model1 to be used for the safety analysis of the NuScale Power Module (NPM) with NuFuel-HTP2TM fuel. The range of applicability is expanded to ensure the NSP4 model encompasses the operating domain of the NPM at higher rated power levels.

The complete list of correspondence between the NRC staff and NuScale is provided in Table 1 below which contains the correspondence relevant to this review.

Table 1: List of Key Correspondence Sender Document Document Date Reference NuScale Topical Report - Supplement 1 November 5, 2021 1

NRC staff Request for Supplemental Information December 1, 2021 3

NuScale Supplementary Information to Topical Report January 14, 2022 4

NuScale CHF Notes and Slides February 18, 2022 5

NRC staff Request for Additional Information (eRAI 9899)

March 30, 2022 6

NuScale Response to eRAI 9899 July 20, 2022 7

NRC staff Request for Additional Supplemental Information (eRAI 9899)

September 8, 2022 8

NuScale Supplemental Response to eRAI 9899 September 30, 2022 9

NuScale Revision 1 to TR-107522 October 27, 2022 16 In performing this review, the NRC staff applied a credibility assessment framework which focused on critical boiling transition (CBT)2 models. The framework is fully described throughout the safety evaluation (SE).

1 The terms model and correlation are synonymous. While this SE primarily uses the word model, there is no difference between a CHF correlation and a CHF model.

2 CBT is the name given to the phenomena which occur when a flow regime that has a higher heat transfer rate transitions to a flow regime that has a significantly lower heat transfer rate. Historically, terms such as CHF, departure from nucleate boiling, and critical power have been used. However, the NRC staff needed a way to separate the general phenomena occurring (i.e., CBT) from a specific type of phenomena which may occur (e.g., departure from nucleate boiling, dryout) and from the specific values of certain parameters which are often used to signify that such a transition has occurred (e.g., CHF, critical power).

2.0 REGULATORY EVALUATION

General Design Criterion (GDC) 10 of Title 10 Code of Federal Regulations (10 CFR) Part 50, states that The reactor core and associated coolant, control, and protection systems shall be designed with appropriate margin to assure that specified acceptable fuel design limits are not exceeded during any condition of normal operation, including the effects of anticipated operational occurrences. GDC 12 states, The reactor core and associated coolant, control, and protection systems shall be designed to assure that power oscillations which can result in conditions exceeding specified acceptable fuel design limits are not possible or can be reliably and readily detected and suppressed. Thus, GDC 10 and 12 introduce the concept of specified acceptable fuel design limits (SAFDLs).

In essence, SAFDLs are those limits placed on certain variables to ensure that the fuel does not fail. One such SAFDL is associated with CBT. CBT is defined as a transition from a boiling flow regime that has a higher heat transfer rate to a flow regime that has a significantly lower heat transfer rate. If the reduction in the heat transfer rate and resulting increase in surface temperature is large enough, the surface may weaken or melt. In a nuclear power plant, this condition could result in fuel damage.

In order to ensure that such a CBT does not occur, SAFDLs have been developed, as described in Standard Review Plan, Section 4.4, Thermal and Hydraulic Design (Reference 10). For NuScale, one SAFDL has been proposed as an acceptable means for satisfying GDC 10 and 12 as documented in Section 4.4 of its Design-Specific Review Standard (Reference 15).

(A) For CHF correlations, there should be a 95-percent probability at the 95-percent confidence level that the hot rod in the core does not experience a boiling crisis during normal operation or anticipated operational occurrences (AOOs).

Therefore, the main objective of the NRC Staffs review was to determine if the NSP4 model could result in accurate predictions, such that there would be a 95-percent probability at the 95-percent confidence level that the hot rod in the core does not experience CBT during normal operation or AOOs.

3.0 TECHNICAL EVALUATION

The purpose of TR-107522, Applicability Range Extension of NSP4 Critical Heat Flux Correlation: Supplement 1 to TR-0116-21012-P-A, Revision 1 (Reference 1 and 16), is to provide the bases for an extension to the range of applicability for the NSP4 model to be used for the safety analysis of the NPM with NuFuel-HTP2TM fuel. The NRC staffs technical evaluation focused on determining if the model is acceptable for use in reactor safety license calculations (i.e., that the model can be trusted) for the extended range.

To perform this evaluation, the NRC staff used a framework similar to the framework used in the NRC staffs SE of the original NuScale Power CHF model (Reference 2). More details about the framework applied in this review can be found in NUREG/KM-0013 (Reference 12). Note that many of the findings are based on the initial review performed on the original submittal, TR-0116-21012-P-A, Critical Heat Flux Correlations, (Reference 2), which TR-107522 supplements.

The review framework is generated from a single main goal; then that main goal is logically decomposed into subgoals. Logical decomposition is the process of generating a set of subgoals which are logically equivalent (i.e., necessary and sufficient) to the main goal. This decomposition is expressed using Goal Structure Notation. Each subgoal can either be further logically decomposed into other subgoals or if no further decomposition is deemed useful, the subgoal is considered a base goal and evidence must be provided to demonstrate that the base goal is true.

For CBT models, the top goal is: The CBT model can be trusted in reactor safety analyses.

Based on the engineering judgement and experience from multiple NRC technical staff members and a study of previous SEs, this goal is decomposed into various subgoals as given in the figures below, starting with the decomposition of the main goal into the three subgoals given in Figure 1.

Figure 1: Decomposition of G - Main Goal The NSP4 model has already been approved by the NRC, and therefore the NRC staff has previously considered these three goals to have been met. The expansion of the range of applicability for the NSP4 model would not impact the NRC staffs findings on G1 and G2, as

those are independent of the application domain in which the model is applied. For the NSP4 model, the only exception within these subgoals is the subgoal related to equivalent grid spacers and this is addressed below. The expansion of the range of applicability would impact G3 so the NRC staff focused its review in this supplement on ensuring that the validation of the NSP4 model did not change with the extended range of applicability.

3.1 Experimental Data Experimental data is the cornerstone of a CBT model. Not only is the data used to generate the coefficients of the model and validate the model, but previous data are often used to generate the models form. Therefore, it is essential that the experimental data are appropriate.

Demonstrating that the experimental data are appropriate is accomplished using the three subgoals given in Figure 2 below.

Figure 2: Decomposing G1 - Experimental Data As stated above, the NSP4 model has already been approved, and therefore the NRC staff has previously considered these three goals to have been met. The expansion of the range of applicability for the NSP4 model would not impact the NRC staffs findings on G1.1 and G1.2, as both the Stern and Kathy experimental facilities have been determined to be credible test facilities and it has previously been determined that this data has been accurately measured.

Therefore, the subgoals G1.1 and G1.2 are considered satisfied through the staffs review documented in Reference 2, and only G1.3 was further investigated in the staffs review in this supplement.

Reproduced Local Conditions The next subgoal in demonstrating that the experimental data are appropriate is to demonstrate that the local conditions in the reactor have been reproduced in the experiment. This is typically demonstrated using the five subgoals as given in Figure 3 below.

Figure 3: Decomposing G1.3 - Reproduced Local Conditions The NSP4 model has already been approved in Reference 2, and therefore the NRC staff has previously considered all of these goals to have been met. This expansion of the range of applicability for the NSP4 model would not impact the NRC staffs findings on G1.3.1, G1.3.3, G1.3.4, and G1.3.5 and these subgoals are considered to have been met in the prior review of the NSP4 model. However, the NRC staff decided to re-evaluate subgoal G1.3.2 again for this review as there is a difference in the grid spacers used in the Stern data (which supports the range extension) and the grid spacer used in NuFuel-HTP2TM fuel. Therefore, only this goal is evaluated below.

Equivalent Grid Spacers Equivalent Grid Spacers The grid spacers used in the test bundle should be prototypical of the grid spacers used in the reactor assembly.

G1.3.2, Review Framework for CBT Models The primary source of data for the range extension is the high mass flux data from Stern. As described in the original CHF TR (Reference 2), NuFuel-HTP2TM contains five grid spacers, the bottom of which is an HMP spacer and the top four being HTP spacers. However, the data from Stern are based on simple grids, that is grid spacers which were primarily designed to ensure the fuel rods maintain their distance from other fuel rods and not designed to increase flow mixing. All grid spacers induced some flow mixing downstream of the spacer which evens out

the qualities and enthalpies in the subchannels of the fuel assembly resulting in more margin to CHF. That is, if the grid spacers were removed, CHF would occur much sooner. While even simple grids provide some benefit, additional benefit can be gained if the grid spacer is specifically designed to induce mixing. Depending on the amount of mixing and the grid spacer design, the increase in CHF margin varies, but a 10 percent - 20 percent increase in CHF margin would be expected.

It is common to use CHF data from simple grids and mixing grids to generate and validate a CHF model. Moreover, using a model developed on grid spacers with simple grids has been previously considered by the NRC staff to be conservative for predicting the performance of fuel with mixing vanes (Reference 11). However, this is different from what NuScale is requesting in this supplement.

To that point, instead of using the data from simple grids as a conservative prediction of CHF performance, NuScale used the data from simple grids to validate the NSP4 model. The main challenge with this approach is that the NSP4 model was developed for mixing grids, not simple grids. Thus, the staff expected the model to over-predict the CHF performance of the simple grids in the Stern tests as the NSP4 model is based on mixing grid data. The staff determined that the amount of the over-prediction varies, based on the magnitude of mixing in the grid spacers which were used to generate the data for the NSP4 model.

The NSP4 model is primarily based on fuel with HTP grids. HTP grids are a unique design in that they do not contain mixing vanes. Instead, the grid spacers contain flow channels built into the grid spacer whose purpose is to mix the flow. While all mixing vane designs are proprietary and the mixing performance is difficult to quantify, in the NRC staffs experience, the HTP grids were not primarily designed to increase CHF performance. While the grids do increase CHF margin, this margin increase is not as great as other grids which were designed primarily to increase that margin. Thus, it is the staff view that the NSP4 models over-prediction of the CHF performance of simple grids (such as those used in the Stern test) would not be as great as the over-prediction produced by a typical mixing vane CHF models analysis of those same simple grids.

Additionally, the distance between the grid spacers on NuFuel-HTP2TM is approximately (([

], while the grid spacers of the Stern assemblies was (( }}. Generally, in the staffs view, applying CHF data from a longer span between grid spacers (e.g., the Stern data) to a shorter span (e.g., the NuFuel-HTP2TM fuel) is considered conservative as CHF performance of the assembly with the longer span would be expected to be worse than of the assembly with the shorter span. Even though there is a difference between the grid spacers used in the development of the range extension and those of the NuFuel-HTP2TM assembly, the NRC staff finds that the grid spacers used in the test bundle are appropriate even though they are not prototypical as there is reasonable assurance that the CHF data obtained from Stern can be conservatively applied to the NuFuel-HTP2TM. That is, the staff finds there is reasonable assurance that the NuFuel-HTP2TM assembly will have better CHF performance than that measured from the bundle tested at the Stern facility. The NRC staff concludes that this goal (G1.3.2) has been met. (([ ]}} - Information Considered Proprietary to Framatome

3.2 Model Validation Validation is the accumulation of evidence which is used to assess the claim that a model can predict a real physical quantity (Reference 13). Thus, validation is a never-ending process as more evidence can always be obtained to bolster this claim. However, at some point, when the accumulation of evidence is considered sufficient to make a judgment that the model can be trusted for its given purpose, the model is said to be validated. Demonstrating the model validation is appropriate is accomplished using the five subgoals given in Figure 4 below. Figure 4: Decomposing G3 - Model Validation Validation Error Validation Error The correct validation error has been calculated. G3.1, Review Framework for CBT Models The validation error is obtained from a ratio of the predicted CHF value and the measured CHF value, which is consistent with the method used to determine the validation error in the original TR (Reference 2). Because NuScale is using the same validation error in this supplement, the NRC staff finds that the correct error has been calculated. The NRC staff concludes that this goal has been met. Data Distribution The second subgoal in demonstrating that the models validation was appropriate is to demonstrate that the data is appropriately distributed throughout the application domain. This is typically demonstrated using the six subgoals as given in Figure 5 below.

Figure 5: Decomposing G3.2 - Data Distribution The evidence the staff considered in determining whether the goals were met is provided below. Validation Data Validation Data The validation data (i.e., the data used to quantify the models error) should be identified. G3.2.1, Review Framework for CBT Models NuScale identified the validation data for the extension to the NSP4 model as the data taken from the Stern facility. Therefore, the NRC staff concludes that this goal has been met. Application Domain Application Domain The application domain of the model should be mathematically defined. G3.2.2, Review Framework for CBT Models NuScale identified the application domain of the NSP4 model in Table 8-4 of the original topical (Reference 2). In its initial request (Reference 1 and 16), NuScale requested an extension of the upper mass flux limit from 0.635 (Mlbm/hr-ft2) to 0.7000 in Table 5-1. In a later RAI response (Reference 7), NuScale increased the value in this Table 5-1 to 0.7500. This safety evaluation is focusing on the this increase in the application domain from 0.6350 to 0.7500 (Mlbm/hr-ft2).

Because this applicability domain is defined in Table 5-1, the NRC staff concludes that this goal has been met. Expected Domain Expected Domain The expected domain of the model should be understood. G3.2.3, Review Framework for CBT Models The expected domain of the NSP4 model has not been further defined from the application domain. The expected domain is a useful construct which enables reviewers to better focus on specific areas of the application domain where the use of the given CHF model is expected. However, given the small increase in the application domain due to the addition of the extended mass flux range, the benefit of defining a separate expected domain is limited. Therefore, the entire application domain will be used as the expected domain. Because the expected domain is not defined separately from that application domain and is only used to further focus on the review on regions of the application domain in which the use of NSP4 model would be expected, the NRC staff has concluded that this criterion does not apply. Data Density Data Density There should be an appropriate data density throughout the expected domain. G3.2.4, Review Framework for CBT Models To understand the data density, the NRC staff created plots (Figures 6-11) demonstrating the data density of the initial NSP4 model (Reference 2) along with data in the extended domain. The primary data supporting the validation of the NSP4 model is from tests K8500 (Kathy data) and tests U1 and U2 (Stern data). These tests were not used in the initial approval of the NSP4 model. Additionally, the NRC staff did consider predictions of test C1, but did not believe it was reasonable to include this data in the validation analysis for reasons discussed in Section 3.2.3.1 of this SE. To determine the data density, the staff created 2D plots of the ((

}} These plots were used by the staff to confirm that the density of the validation was sufficient in the application domain.

First, the staff plotted the original data for the approved NSP4 model (o - black circle). This data is an example of a reasonable data density over the application domain that the staff previously found to be acceptable in Reference 2. Second, the staff plotted the NSP4 data from tests K8500, U1, and U2 in the currently approved mass flux range ( - blue square). Finally, the

staff plotted the NSP4 data from tests K8500, U1, and U2 in the extended mass flux range (x - red x), that is data above a mass flux of 0.635 (Mlbm/hr-ft2). A summary of these data is provided in Table 2. Table 2: Legend for Data Density Plots Original NSP4 data o Current Mass Flux Range Extended Mass Flux Range x Because the staff was evaluating a four-dimensional application domain (pressure, mass flux, local quality, and inlet subcooling), the staff created six plots to compare each dimension with each other. The data density plots are given in Figure 6 through Figure 11 below.

Figures 6, 7, 8, 9, 10, and 11 (([ ]}}

Based on the data density displayed Figure 6 - Figure 11, the NRC staff considers that there is no significant difference between the data density of the original NSP4 model and the data density in the extended mass flux domain. Because there is no significant difference, the NRC staff concludes that this goal has been met. Sparse Regions Sparse Regions Sparse regions (i.e., regions of low data density) in the expected domain should be identified and justified to be appropriate. G3.2.5, Review Framework for CBT Models Based on the review Figure 6 - Figure 11, the NRC staff was not able to identify any sparse regions. The NRC staff therefore concludes that this goal has been met. Restricted Domain Restricted Domain The model should be restricted to its application domain. G3.2.6, Review Framework for CBT Models The staff already concluded in the original NSP4 SE (Reference 2) that NuScale appropriately restricted the NSP4 model to its application domain. Because this TR supplement would only modify that domain, the NRC staff finds that the change in the upper mass flux limit would not necessitate a new review of this goal. The NRC staff therefore concludes that the restricted domain goal has been met. Consistent Model Error The third subgoal in demonstrating that the models validation was appropriate is to demonstrate that the model error is consistent over the application domain. This is typically demonstrated using the three subgoals as given in Figure 12 below.

Figure 12: Decomposing G3.3 - Consistent Model Error The evidence demonstrating that the following goals were met is provided below. Poolability Poolability The validation error should be investigated to determine if it contains any subgroups which are obviously not from the same population (i.e., not poolable). G3.3.1, Review Framework for CBT Models Figure 4-3 of the TR supplement (Reference 1 and 16) provides the predicted to measured (P/M) CHF values for three different tests as a function of mass flux. Tests U1 and U2 have a uniform power shape while test C1 has a cosine power shape. In the ((

}}, the cosine test is predicted very conservatively3, while the uniform tests have some conservative and some non-conservative predictions. For the uniform tests, of the (( 
}} in the high mass flux region, (( }} exceeded the requested departure from nucleate boiling ratio (DNBR) limit of 1.21 while (( }} of the cosine tests exceeded the 95/95 (or even a P/M value of 1.0).

In response to RAI 9899, NTR-02 (Reference 7) NuScale provided further justification for the few non-conservative predictions. First, it reiterated that the tests of U1 and U2 contained simple grid spacers, while the NSP4 model is based on mixing vane grids. Based on this explanation, the staff would expect that a model which has been trained on mixing vane data 3 For P/M plots, a prediction is said to be conservative when the P/M value is less than 1.0. This means that that the CHF has been measured at higher heat flux than the model predicts. Likewise, a prediction is said to be non-conservative when the P/M value is greater than 1.0. This means that that the CHF has been measured at lower heat flux than the model predicts. Generally, only non-conservative predictions above the DNBR limit are a concern.

would be non-conservative in predicting simple grid tests as that model would predict better CHF performance than would be expected for the simple grids. As a demonstration of the consistency of the NSP4 models prediction of simple grid data, NuScale pointed to Region 2 of Figure 4-1. The NSP4 model has been approved in this region with a DNBR limit of 1.21 (Reference 2). However, the NSP4 models predictions of U1 and U2 in (( }}. Thus, in response to RAI 9899, NTR-02 NuScale stated that it is reasonable to assume that the non-conservative behavior is due entirely to using the NSP4 model to predict the simple grid data, and ((

}}. NuScale also discussed the conservative trend in mass flux is likely due to (( 
}}.

In general, the NRC staff would expect a model such as NSP4 to non-conservatively predict simple grid data. Thus, the trends in Figure 4-3 are not surprising. In further assessing the validity of the model, the NRC staff investigated both (( }} to determine if there was any reason to believe that the same DNBR limit of 1.21 which was approved for ((

}}. This included an investigation into the very conservative prediction of the cosine data from test C1.

One of the first items the staff noted was the inability of the NSP4 model to correctly predict the elevation of CHF. The elevation error for tests C1, U1 and U2 (% difference between the location predicted and location measured) is given in Figure 13, Figure 14, and Figure 15 below. (( }} Figure 13: Elevation Error vs. Local Mass Flux for Test C1 For tests C1, the NSP4 model predicts CHF to occur at a much different elevation than where CHF was measured. Because test C1 has a cosine power shape, this incorrect prediction not only impacts the predicted value from the correlation, but also the measured value from the test. The measured and predicted location is determined from the subchannel which has the minimum DNBR (i.e., the subchannel which is believed to be closest to experiencing CHF). For uniform power shapes, the heat flux of all channels is the same. Thus the measured heat flux is the same irrespective of which subchannel has the minimum DNBR. However, for cosine

power shapes, the heat flux peaks at the center and falls off at the edges, thus the measured CHF will vary depending on the subchannel which is predicted to the have the minimum DNBR. One concern as demonstrated in Figure 13 is that while the NSP4 model does predict conservatively, ((

}} The NRC staff was concerned about if this trend (( 
}}. Further, the staff was concerned that if the trend (( 
}}.

Because of the large differences between the measured CHF values and local conditions used in the analysis of the C1 test data compared to the actual CHF values and local conditions at which CHF occurred in the test, the NRC staff does not believe that the C1 tests demonstrate that the NSP4 model would accurately or conservatively predict CHF. While the analysis does show that the NSP4 model conservatively predicted test C1, the NRC staff does not believe it is reasonable to expect the model to have the same performance in the reactor and cannot determine if the reactor performance would be more or less conservative. However, the NRC staff does not believe that the prediction of the C1 invalidates the NSP4 model, as it does not provide any evidence that the model would behave in a non-conservative manner. Therefore, as discussed in Section 3.2.4.3, the NRC staff relied on other data to determine the acceptability of the NSP4 model in the extended mass flux region. (( }} Figure 14: Elevation Error vs. Local Mass Flux for Test U1

(( }} Figure 15: Elevation Error vs. Local Mass Flux for Test U2 For tests U1 and U2, the NRC staff found that the NSP4 model does a good job of predicting the elevation of CHF at lower mass flux conditions. While predicting the exact elevation is not a requirement, being able to predict the elevation is often used as further evidence that the model is behaving as expected. For tests with a uniform axial power shape, CHF should always occur at the end of the heated length. This is because the end of the heated length will have the highest quality, highest enthalpy, and highest void at the exit. In one sense, looking axially down a test section could be considered looking back in time, as the local conditions at lower elevations should have been experienced at the very end of the heated length first. This analogy breaks down just above the spacer grids where there is significant turbulent mixing, but it holds for just below the spacer grid spacers where CHF generally occurs. The staff noted that, in the high mass flux region, the NSP4 model predicts CHF to occur ((

}} based on previous staff experience. (( 
}} To better understand this behavior, the NRC staff performed an analysis similar to that of NuScale and considered the contribution of each term in the NSP4 model to the final predicted value of CHF. The model is given in Equation 7-1 of the initial TR (Reference 2) and is restated below for convenience.

((

}}

In order to investigate the behavior of this model, the NRC staff considered the value of each term and how much that term contributed to the final sum of all terms. ((

4.}} The staff evaluated the model5 by examining the contribution of each term in Eq. 7-1. The contribution of each term for the U1 and U2 tests is given in Figure 16. (( }} Figure 16: Contribution of each term in NSP4 Figure 16 displays the percent contribution of each of the ((

}}. To better understand this plot, consider the first group of data at low mass fluxes (around 0.1). The red triangle ()

represents ((

}}. The yellow star () represents (( 
}}.

Figure 16 provides insight into the mechanics of the NSP4 model, as it visually displays which terms are important in the prediction of CHF and if those terms are acting to increase or decrease the predicted CHF value. For the extended mass flux range currently under review (i.e., mass fluxes greater than 0.635), (( ((4

}}

5 In order to perform this evaluation, the NRC staff recreated the NSP4 correlation in MATLAB. While the correlation form is easy to program and NuScale provided all of the data for the necessary inputs, the staffs version of NSP4 correlation varied slightly (usually within 5%) from the predicted CHF value reported by NuScale. The staff considers that this variation is reasonable for the given analysis.

}}. To demonstrate this better, a zoomed in view of the high mass flux region is given in Figure 17.

(( }} Figure 17: Contribution of each term in NSP4 (High Hass Flux Region) A comparison of Figure 16 and Figure 17 demonstrates that the behavior of the NSP4 model is ((

}}. For example, (( 
}}

From this analysis, the NRC staff concludes that while the same NSP4 model would be used at higher mass fluxes, ((

}} as demonstrated in Figure 16. Thus, this is not a simple extension, (( 
}}. Similar to the NSP4 models inability to predict the correct elevation of CHF in test C1, this analysis does not provide any evidence that the model would behave in a non-conservative manner.

In response to RAI 9899, NTR-02, NuScale stated that ((

}}. However, the NRC staff disagrees with this conclusion.

The NRC staff found that the results from the above analysis demonstrate that there is a ((

}} and the staff therefore concluded that these two datasets should not be pooled together.

In summary, the NRC staff has concluded that that there was not a sufficient justification to pool the data from the ((

}} However, the NRC staff has concluded that similar behavior of the (( 
}} inside the extended mass flux range to determine the appropriate DNBR limit in that range. Based on this analysis of the poolability of the different data sets, the NRC staff has determined that this goal has been met.

Non-Conservative Subregions Non-Conservative Subregions The expected domain should be investigated to determine if contains any non-conservative subregions which would impact the predictive capability of the model. G3.3.2, Review Framework for CBT Models The staff investigated the domain for non-conservative subregions and ((

}}. For this analysis, (( 
}} The non-conservative behavior as a function of (( 
}} can be seen in Figure 18 which shows the P/M values as a function of (( }}. 

(([ ]}} Figure 18: (( }} for U1, U2, and K8500 (view 1) Because it is often hard to interpret a 3-D plot on paper, the staff prepared Figure 19 using the same data but with a rotated view.

(([ ]}} Figure 19: (( }} for U1, U2, and K8500 (view 2) Finally, this 3-D data is collapsed by ignoring the ((

}} value in Figure 20. 

(([ ]}} Figure 20: (( }} for U1, U2, and K8500 Figure 20 illustrates that there is a substantial non-conservative increase in the P/M values for ((

}}. That increase is not observed in the K8500 test. Since K8500 is also a simple grid test, the staff considers it unlikely that the increase is due to the over-prediction of

(( }}. If this were the primary reason, the same non-conservative prediction would be seen in K8500, and while K8500 is ((

}} the DNBR limit of 1.21), the (( 
}}.

From Figure 18 and Figure 19, the staff observed that the ((

}}. This can be seen when comparing the (( 
}} value as given in Figure 21. 

(([ ]}} Figure 21: (( }} for U1, U2, and K8500 As illustrated in Figure 21, the ((

}}. From this analysis, the NRC staff concludes that while there seems to be a non-conservative subregion in the extended mass flux domain, that subregion is (( }}. That non-conservative subregion is addressed in Section 3.2.4.3, Appropriate Bias for Model Uncertainty, below. With the exception of the non-conservative subregion (( }}, the NRC staff did not identify any other non-conservative subregions. Because the expected domain has been investigated for non-conservative subregions, the NRC staff concludes that this goal has been met.

Model Trends Model Trends The model is trending as expected in each of the various model parameters. G3.3.3, Review Framework for CBT Models Due to the limited nature of the review for the mass flux extension from 0.635 to a mass flux of 0.75, the NRC staff did not specifically review all model trends, but instead focused on those trends related to the P/M values as discussed elsewhere in this safety evaluation. The NRC

staff concludes that this criterion has been addressed elsewhere in this SE in the analysis of poolability (3.2.3.1 Poolability), the analysis of non-conservative subregions (3.2.3.2 Non-Conservative Subregions), and the determination of the DNBR limit (3.2.4.3 Appropriate Bias for Model Uncertainty). Quantified Model Error The fourth subgoal in demonstrating that the models validation was appropriate is to demonstrate that the model error has been appropriately quantified over the application domain. This is typically demonstrated using the three subgoals as given in Figure 22 below. Figure 22: Decomposing G3.4 - Quantified Model Error The evidence demonstrating the following goals were met is provided below. Error Data Base Error Data Base The validation error statistics should be calculated from an appropriate database. G3.4.1, Review Framework for CBT Models The applicant determined the validation error by comparing the predictions of the NSP4 model to data from specific experiments. However, during the review NuScale informed the NRC staff (References 7 and 8) that it had removed specific data points from the analysis because those data points were outside the requested range of application. Further, while some of the removed data points were discussed in the TR and its supplement, others were not. In general, the NRC staff considers that data driven models such as CHF models, should have all reasonably available data provided, and if such data is not provided or not used in the analysis, then it should be made clear to the reviewer what data is not being used and why. In the staffs experience, it is common for applicants to disposition specific test runs from various tests and not use the data for a variety of reasons. However, deciding that the data can be ignored is a decision that often relies on engineering judgment, and such judgments are reviewed by the NRC staff and are necessary for making an assessment on the validation of the

model. This is because the data ignored may provide evidence that the models predictive capability is much lower than anticipated. This has been captured in a recommendation in the staffs SE for future reviewers. The NRC staff has determined that these data were appropriate for validation as the data were not used to train the NSP4 model, therefore, the NRC staff considers that this goal has been met. Statistical Method Statistical Method The validation error statistics should be calculated using an appropriate method. G3.4.2, Review Framework for CBT Models Due to the complexities of the validation data (i.e., because the NSP4 model is a mixing vane model, it will non-conservatively predict the CHF performance of a simple grid), the staff could not perform a statistical comparison which demonstrates that the models prediction in the extended mass flux region is bounded by the DNBR limit of 1.21. Because a statistical comparison could not be performed, the NRC staff concludes that this criterion does not apply in this review and therefore engineering judgment must be utilized to ensure that there is reasonable assurance that the DNBR limit used in the high mass flux region will satisfy the 95/95 departure from nucleate boiling (DNB) criterion. The justification for the DNBR limit in the extended mass flux region is described in Section 3.2.4.3 Appropriate Bias for Model Uncertainty. Appropriate Bias for Model Uncertainty Appropriate Bias The models error should be appropriately biased in generating the model uncertainty. G3.4.3, Review Framework for CBT Models In response to RAI 9899, NTR-02 (Reference 7), NuScale provided additional justification for applying the DNBR limit of 1.21 to the extended mass flux range. NuScale also provided additional justification in section 4.1 and Table 4-2 in Rev 1 of the TR. In general, the NRC staff agrees with much of this analysis, as confirmed by the staffs own analysis. ((

}} Both of these are demonstrated in Figure 21 of this SE. However, the NRC staff disagrees with the conclusions drawn by NuScale that the DNBR limit of 1.21 would satisfy the 95/95 criterion in the high mass flux domain, for the reasons described below. 

While the staff does expect to see non-conservative predictions of the U1, U2, and K8500 tests because those tests had simple grids, while the NSP4 model used mixing grids, the magnitude of the non-conservative predictions of the U1 and U2 tests seems to be too high. From Figure 20 of this SE, the staff notes the impact of using the NSP4 model to predict simple grids in the ((

}} to be fully attributed to the fact that the data is from simple grids.

Figure 20 of this SE illustrates that this non-conservatism is (( }}. Therefore, the staff determined that the bias should be separated into a ((

}}. The (( }} was chosen based on the staffs conservative engineering judgment, as the data below this pressure (including K8500 test data and U1 and U2 data) demonstrated similar predictive capability. The validation of the NSP4 model for (( }} is given in Figure 23. 

(([ ]}} Figure 23: Mass Flux vs P/M for U1, U2, and K8500 (( }} The NRC staff determined that there was not enough data from tests U1, U2, and K8500 to determine the 95/95 in the high mass flux region. The data point which had the most non-conservative prediction has a P/M value of (( }}, and while this value is above the DNBR limit of 1.21, it is within the ((

}}. However, the NRC staff had the following concerns for the extended mass flux region:

(a) The staff was concerned about the epistemic uncertainty associated with the lack of sufficient data to demonstrate compliance with the DNBR limit of 1.21 in the extended mass flux domain. (( }}, this value is not a reasonable estimate of the 95/95 given the limited number of data points in the region. The NRC staff would expect the 95/95 value to ((

}}.

(b) The staff was concerned about the epistemic uncertainty associated with the unknown impact of the NSP4 models consistent prediction of CHF at much lower elevations than measured. While the NSP4 models predictions of the U1 and U2 tests in the extended mass flux range were reasonable, the staff was concerned because the models predictions (including those predictions of the C1 tests) ((

}}

(c) The staff was concerned about the epistemic uncertainty associated with unknown impact in predictive capability resulting from the shift in the ((

}} in the extended mass flux domain, the NRC staff believed that it was reasonable to assume that the NSP4 models predictive capability would be changed in that domain.

(d) The staff was concerned about the lack of representative data for the NuFuel-HTP2TM fuel and the non-conservative P/M predictions in the extended mass flux domain. While simple grid CHF data has been used previously to demonstrate the conservative nature of a mixing-grid CHF model, the model generally conservatively predicts the data. That is not the case for the NSP4 model, as the NRC staff had to determine the degree of non-conservatism expected due to using the mixing-vane correlation on simple grid fuel. While the NRC staff and NuScale did perform an analysis to determine this non-conservatism, the NRC staff notes that this is not a common analysis and believes there may be uncertainties which have not been addressed. Based on concerns (a) - (d) above, the NRC staffs previous experience with CHF models, and the NRC staffs conservative engineering judgment, the NRC staff concludes that a penalty of ((

}} of the extended mass flux domain. This is reflected in condition and Limitation 1 as stated; For mass fluxes greater than (( 
}}

The validation of the NSP4 model for (( }} is given in Figure 24.

(([ ]}} Figure 24: Mass Flux vs P/M for U1, U2, and K8500 (( }} Like the (( }}, there is not enough data from tests U1, U2, and K8500 to determine the 95/95 in the (( }}. ((

}} Further, as demonstrated in Figure 24 above, the P/M value of (( }} is not an outlier as there are multiple values which are close to this value in the high mass flux domain.

Based on concerns (a) - (d) above, on the few P/M values at (( }} and their non-conservative values which exceed the safety limit of 1.21 and ((

}}, the NRC staffs previous experience with CHF models, and the NRC staffs conservative engineering judgment, (( 
}}. This is reflected in condition and Limitation 2 as stated; For mass fluxes greater than (( 
}}

Based on the staffs analysis and application of the penalties as condition and Limitations 1 and 2 of this SE, the NRC staff concludes that, subject to the satisfaction of these conditions and limitations, the goal of applying an appropriate bias for model uncertainty has been met.

Model Implementation The fifth subgoal in demonstrating that the models validation was appropriate is to demonstrate that the model will be implemented in a manner consistent with its validation. This is typically demonstrated using the two subgoals as given in Figure 25 below. Figure 25: Decomposing G3.5 - Model Implementation The evidence demonstrating the following goals were met is provided below. Same Computer Code Same Computer Code The model has been implemented in the same computer code which was used to generate the validation data. G3.5.1, Review Framework for CBT Models Sections 3.3.1 and 3.3.2 in the original TR-0116-21012 (Reference 2) show that the VIPRE-01 models are used by NuScale to perform the data reduction calculations in accordance with TR-0915-17594, Subchannel Analysis Methodology (Reference 14). To ensure that the NSP4 model is used in a manner consistent with its validation, in the NRC staffs original SE the staff established Limitation 2 on the use of VIPRE-01 calculations using the NSP4 model. Based on the description in Sections 3.3.1 and 3.3.2 of the original TR, and pursuant to Limitation 2, the NRC staff found that the NSP4 model is implemented using the same computer code used to generate the validation data.

Same Methodology Same Methodology The models prediction of the CBT is being applied using the same methodology as it was when predicting the validation data set for determining the validation error. G3.5.2, Review Framework for CBT Models As described in Section 3.1.3.5.1 of this SE, the NRC staff established Limitation 2 in the original SE to ensure that the NSP4 model is used in a manner consistent with its validation. Based on the description in Sections 3.3.1 and 3.3.2 of the original NRC-approved TR-0116-21012-P-A (Reference 2), and pursuant to Limitation 2, the NRC staff found that the NSP4 model is being applied in the same manner as when predicting the validation data set. Transient Prediction Transient Prediction The model results in an accurate or conservative prediction when it is used to predict transient behavior. G3.5.3, Review Framework for CBT Models The focus of this review was to determine the DNBR limit for an extension in the mass flux range. Because this increase in mass flux range would not impact the use of this models ability to predict transient behavior, the NRC staff has determined that this goal does not apply.

4.0 CONCLUSION

Based on the NRC staffs review in Section 3.2 of this SE, the NRC staff concludes that the NSP4 model has sufficient validation in the extended mass flux region up to a mass flux of 0.7500 (Mlbm/hr-ft2) as demonstrated through appropriate quantification of its error. Therefore, the NRC staff concludes that the NSP4 model can be relied upon in reactor safety analyses such as determining whether the SAFDL (as defined in GDC 10 and 12 of 10 CFR Part 50, Appendix A) of DNBR satisfies the criterion for CHF correlations. Further, there should be a 95-percent probability at the 95-percent confidence level that the hot rod in the core does not experience a boiling crisis during normal operation or AOOs, as provided in the NuScale Design-Specific Review Criteria (Reference 15). The staffs conclusion is subject to the conditions and limitations listed below.

4.1 Conditions and Limitations The following conditions and limitations must be met to apply the NSP4 model in the extended mass flux range.

1. For mass fluxes greater than ((

}}

2. For mass fluxes greater than ((

}}

3. The NSP4 model is limited to mass fluxes below 0.7500 (Mlbm/hr-ft2). The full application domain is given in Table 5-1 of the TR.

4. The application of the NSP4 model is limited to type NuFuel-HTP2TM fuel.

5. Any application deviation from the modeling options or deviation from the use of the subchannel code which was used to perform this validation assessment would require re-validation similar to the validation provided in the TR and would require NRC review and approval. Any application to a new fuel type or new mixing vane spacer type, any decrease in the CHF design limits, or any expansion of the application domain would require NRC review and approval.

4.2 Staff Recommendations The following recommendation is made for NRC staff reviews of future supplements to or revision of this TR:

1. The NRC staff believes that data driven models such as CHF models should have all reasonably available data provided, and if such data are not provided or not used in the analysis, then it should be made clear to the reviewer what data are not being used and why. The NRC staff should ensure that there is reasonable justification for ignoring any such data when performing the validation of such a data driven model.

5.0 REFERENCES

1.

NuScale Power, LLC, Applicability Range Extension of NSP4 Critical Heat Flux Correlation: Supplement 1 to TR-0116-21012-P-A, Revision 1, TR-107522, Revision 0, November 2021, ADAMS Accession Nos. ML21309A755 (Proprietary Version) and ML21309A754 (Non-Proprietary).

2.

NuScale Power, LLC, Critical Heat Flux Correlations, TR-0116-21012-P-A, Revision 1, December 19, 2018, ADAMS Accession No. ADAMS Accession No. ML18360A633 and ML18360A634 (Proprietary Version) and ML18360A632 (Nonproprietary Version).

3.

Email from Bruce Bavol (NRC) to Rebecca Norris (NuScale), TR-107522 NuScale Supplement to CHF, dated December 1, 2021. (ADAMS Accession No. ML22020A030).

4.

NuScale Power, LLC, Submittal of Supplementary Information to Topical Report Entitled Applicability Range Extension of NSP4 Critical Heat Flux Correlation: Supplement 1 to TR-0116-21012-P-A, Revision 1, TR-107522, January 14, 2022, ADAMS Accession No. ML22014A249 (Proprietary Version) and ML22014A248 (Nonproprietary Version).

5.

NuScale Power, LLC, Submittal of Supplementary Information to Topical Report Applicability Range Extension of NSP4 Critical Heat Flux Correlation: Supplement 1 to TR-0116-21012-P-A, Revision 1, TR-107522, Revision 0, LO-114135, February 18, 2022, ADAMS Accession No. ML22049A790 (Proprietary Version) and ML22049A789 (Nonproprietary Version).

6.

Final Request for Information eRAI 9899, dated March 30, 2022 ADAMS Accession No. ML22089A024 (Nonproprietary Version). ADAMS Accession No. ML22089A023 (Proprietary Version).

7.

NuScale Power, LLC, NuScale Power, LLC Response to NRC Request for Additional Information (RAI No. 9899) on the NuScale Topical Report, "Applicability Range Extension of NSP4 CHF Correlation," TR-107522, Revision 0, July 20, 2022, ADAMS Accession No. ML22201A533 (Proprietary Version) and ML22201A532 (Nonproprietary Version).

8.

Email from Bruce Bavol (NRC) to Thomas Griffith (NuScale), NuScale TR-107522, Revision 0, Critical Heat Flux Correlation, dated September 8, 2022 ADAMS Accession No. ML22251A391(Nonproprietary Version). ADAMS Accession No. ML22251A380 (Proprietary Version).

9.

NuScale Power, LLC, NuScale Power, LLC Supplemental Response to NRC Request for Additional Information (RAI No. 9899) on the NuScale Topical Report, "Critical Heat Flux," TR-0116-21012, Revision 1-A, September 30, 2022, ADAMS Accession No. ML22273A169 (Proprietary Version) and ML22273A168 (Nonproprietary Version).

10. U.S. Nuclear Regulatory Commission, Thermal and Hydraulic Design, Section 4.4 of NUREG-0800, Standard Review Plan for the Review of Safety Analysis Reports for Nuclear Power Plants: LWR Edition, Revision 2, March 2007, ADAMS Accession No. ML070550060.

11. Morey, D. C., U.S. Nuclear Regulatory Commission, letter to Gary Peters, Framatome Inc., Final Safety Evaluation for Framatome Inc. Topical Report ANP-10341P, The ORFEO-GAIA and ORFEO-NMGRID Critical Heat Flux Correlations (CAC No. MF8400; EPID L-2016-TOP-0008), dated September 24, 2018 (ADAMS Accession No. ML18236A371).

12. Kaizer, J.S., Anzalone, A., Brown, E., Panicker, M., Haider, S., Gilmer, J., Drzewiecki, T., and A. Attard, "Credibility Assessment Framework for Boiling Crisis Transition Models, NUREG/KM-0013-DRAFT, 2018.

13. Oberkampf, W.L., and C.J. Roy, Verification and Validation in Scientific Computing, Cambridge University Press, Cambridge, United Kingdom, 2010.

14. NuScale Power, LLC, Subchannel Analysis Methodology, TR-0915-17564-P-A, Revision 2, March 19, 2019, ADAMS Accession No. ADAMS Accession No. ML19067A257 and ML19067A258 (Proprietary Version) and ML19067A256 (Nonproprietary Version).

15. U.S. Nuclear Regulatory Commission, Thermal and Hydraulic Design, Section 4.4 of Design-Specific Review Standard for NuScale SMR Design, June 17, 2016, ADAMS Accession No. ML15355A468.

16. NuScale Power, LLC, Applicability Range Extension of NSP4 Critical Heat Flux Correlation: Supplement 1 to TR-0116-21012-P-A, Revision 1, TR-107522, Revision 1, October 2022, ADAMS Accession Nos. ML22300A244 (Proprietary Version)) and ML22300A243 (Non-Proprietary).

6.0 LIST OF ACRONYMS AOO anticipated operational occurrence CBT critical boiling transition CFR Code of Federal Regulations CHF critical heat flux DNB departure from nucleate boiling DNBR departure from nucleate boiling ratio NPM NuScale Power Module NRC U.S. Nuclear Regulatory Commission NuScale NuScale Power, LLC SAFDL specified acceptable fuel design limit SE safety evaluation TR topical report Principal Contributor: J.S. Kaizer R. Sugrue}}