ML062640349

Vermont Yankee Hearing - Entergy Exhibit 27, GE Topical Report, Qualification of the One-Dimensional Core Transient Model for Boiling Water Reactors, Report No. NEDO-24154-A, Vol. 2

ML062640349
Person / Time
Site:	Vermont Yankee
Issue date:	08/31/1986
From:	Fischer D, Jeffery Wood General Electric Co
To:	NRC/SECY
Byrdsong A T
References
50-271-OLA, Entergy-Licensee-27, RAS 12278 78NED290R1, NEDO-24154-A
Download: ML062640349 (259)

Text

DOCKETED USNRC NEDO.24154.A 78NED290R1 DOCKET NUMOMi 4 PM SEP 19 PM 3: 37 CLASS I PROD. &UTIL FACO...Ul._h6.L1 AUGUST 1986

.... , -, --SECQFETlkRy AD JU0iCAIONS STAFF LICENSING TOPICAL REPORT QUALIFICATION OF THE ONE-DIMENSIONAL CORE TRANSIENT MODEL FOR BOILING- WATER REACTORS VOLUME 2 Dat m Offia wcialEmdub kR(~gýafC fthrm_____ _____

GENERAL 0 ELECTRIC

-j~ep ILe-- - s, 1 - -Y'- N99 S../C 69 - 6;t

NEDO-24154-A 78NED290RL Class I August 1986 LICENSING TOPICAL REPORT QUALIFICATION OF THE ONE-DIMENSIONAL CORE TRANSIENT MODEL FOR BOILING WATER REACTORS VOLUME 2 Approved :C 2c."Jars-t JkE. Wood, Manager Approved:__________________

D. L. Fischer, Manager C re & Fuel Technology Core Nuclear Design NUCLEAR ENERGY BUSINESS OPERATIONS o GENERAL ELECTRIC COMPANY SAN JOSE, CALIFORNIA 95125 GENERAL 4 ELECTRIC

NEDO-24154-A X&5CLADIIMER OF P'ZSnSI*rBlI Nei'ther -.he General E'Zectric Company nor any of the contr-rbutore to this document mrakes any,war rantyU or representation (ezprese or implied) with respect to the accuracy, corleteress, or use fislness of the inforitzion contained in this docanent or that the me of such information may not inrirnge on privately owrned ighte; nor do they as*ume wz responsibility for Liability or damage of any kind which ry result the use of any of the infaomation contained in this document.

UNITED STATES NUCLEAR REGULATORY COMMISSION It WASHINGTON. 0. C. 2=5 FEB 2 Dr. G. G. Sherwood, Manager Safety and Licensing General Electric Company 175 Curtner Avenue San Jose, California 95114

SUBJECT:

ACCEPTANCE FOR REFERENCING GENERAL ELECTRIC LICENSI"NG TOPICAL REPORT NEDO-241 54/NEDE-241 54P The Nuclear Regulatory Commission has completed its review of the General Electric Company Licensing Topical Report REDO-24154 Volumes I and I1 and NEDE-24154 Volume III entitled "Qualification of the One-Dimensional Core Transient Model for Boiling Water Reactors" and the supplemental infor mation submitted by l H. Buchholz, letter (MFN 155-80) dated September 5, 1980. This report describes the General Electric transient analysis code.

ODYN. This code is to be used for-transient analyses of the following eight transients:

A. For Thermal Limit Evaluation Thermally Limiting or Event Near Limiting (Typie'ally)

1. Feedwater Controller Failure- X Maximum Demand

2. Pressure Regulator Failure - Closed

3. Generator Load Rejection X

4. Turbine Trip X

5. Main Steamline Isolation Valve Closures

6. Loss of Condenser Vacuum

7. Loss of Auxiliary Power - All X Grid Connections iii

FEE 4 1981 Dr. G. G. Sherwood -2 B. For ASME Vessel Overpressure Protection Pressure Limiting

1. MSIV Closure with Position Switch Scram Failure (i.e., 14SIV Flux Scram) X Other transients and anticipated operational occurrences may be analyzed with the REDY Code described in NEDO-10802.

For all operating BWRs and those currently under operating license review, we propose that the following pressurization events be reanalyzed using ODYN:

generator load rejection/turbine trip without bypass (whichever Is limiting),

feedwater controller failure-maximum demand and main steam isolation valve closure-flux scram (to satisfy ASME code pressure requirements) , These are the same pressurization events presently included in reload submittals. As discussed in the September 5, 1980 letter from R. H. Buchholz (GE) to USNRC subject: "Response to NRC Request for Information on ODYN Computer Model,,

the events not included are considered to be bounded by these three.

This letter also provides the information required for application of ODYN and justification for continued plant operation during transition from transient analyses based on ODYN. The safety evaluation and a supplementary safety evaluation are enclosed.

As a result of our review, we find the Licensing Topical Report NEDO-24154 Volumes I and II/NEDE-24154P Volume III as augmented by GE/Buchholz letter No. MFN4 155-80 dated September 5, 1980 acceptable for referencing in applications for operating license within the conditions specified in the topical report, the supplemental letter and the attached safety evaluation and supplement.

We do not intend to repeat the review of the safety features described in the topical report and found acceptable in the attachment. Our acceptance applies only to the features described in the topical report and under the conditions discussed in the attachment.

In accordance with established procedure, it is requested that General Electric Company publish an approved version of these reports, proprietary and non-proprietary, within three months of receipt of this letter. The revisions are to appropriately incorporate the supplementary Information into the body of the report and incorporate this letter and the attached safety evaluation and supplement following the title page and thus Just in front of the abstract. The report Identifications of the approved reports are to have a -A suffix.

iv

FEE6 4 Dr.' G. G. Sherwood Should Nuclear Regulatory Commission criteria or regulations change such that our conclusions as to the acceptability of the report are invalidated, General Electric and/or the applicants referencing the topical report will be expected to revise and resubmit their respective documentation or submit justification for the continued effective applicability of the topical report without revision of their respective documentation.

Sincerely, Robert L. Tedesco. Assistant Director for Licensing Division of Licensing

Enclosure:

As Stated vivi

SAFETY EVALUATION FOR THE GENERAL ELECTRIC TOPICAL REPORT QUALIFICATION OF THE ONE-OIMENSIONAL CORE TRANSIENT MODEL FOR BOILING WATER REACTORS NEDO-24154 and NEDE-24*54-P Volumes 1, 11 and III Prepared By Reactor Systems Brunch, DS0 Reviewers F. Odar, Laud (RES)

L. Beltracchil

0. B. Fleno G. M. Holahan M. M. Mendonca J. Voglewede June 2980 vli/viil

TABLE OF CONTENTS Pace I.

SUMMARY

OF THE TOPICAL REPORT .................................... I-1 A. INTRODUCTION . . I-1 B. SCOPE . 1-2 C. SUIMARY OF ANALYTICAL MODELS ............................... 1-2 II. STAFF EVALUATION ........................................... -1 A. REVIEW OF ANALYTICAL MODELS ............................. 11-2

1. Recirculation and Control System ...................... 11-2

a. Recirculation Loop Model ........................ 11-2

b. Contral System Model 11-66.................

c. Stem Separator Model .......................... .I-S

d. Upper Plenum, Vessel Dome and Bulkwater Model ....................................... 11-22

2. Steam Line Model ....................................... 11-12

3. Core Thermal Hydraulics Model .......................... 11-14

a. Drift Flux Model .............................. Il-15

b. Mechanistic Boiling Model ....................... 11-19

4. Core Physics Model .......................... ........... 11-20

a. Assumptions in the Neutronics Model ............... 11-20

b. Derivation of Equations for the One-Group, One-Olmenslonal, Time-Dependent Neutronic Nadel .......................................... 11-21

c. Calculation of Input Parameters ................... 11-23 S. Fuel Heat Transfer Model .......................... 11-25

6. Summary of Code Uncertainties ........................ 11-3

a. Margin in ACPR Calculations ....................... 11-31

b. Margin in Pressure Calculation .................... 11-33 B. QUALIFICATION OF THE ODYN CODE .............................. 11-35

1. Qualification of Neutronics Model - Comparison of ODYN with BWR Core Simulator ....... -................. 11-35

2. Qualification of Thermal Hydraulic Model ............... .11-38 ix

TABLE OF CONTENTS (Continued)

Pac_..e

3. Qualificat*on Using Integral Tests ..................... 11-40 a,

b.

Peach Bottom Tests ................................

KXM Test.......................

11-40 11-54

4. Qualification Using Another Computer Code o Audit Calculations ................................. 11-59

a. Development of Calculatlonal Method ............... 11-5s

b. Peach Bottom Tests and Audit Calculations ......... 11-61 S. Summary of Code Qualification ........................... 1-75 C. EVALUATION OF THE MARGIN................................... 11-76 III. STAFF POSITION ........................................... 111-1 IV. REFERENCES .............................................. IV-1

• ** ee * *i o, oeo e aoeel e o o ~oi. ee o oee ee*o K

I. SUMfARY OF TOPICAL REPORT A. INTRODUCTON Between April 9, 1977 and April 27, 1977, three turbine trip tests were performed at the Peach Bottom, Unit 2, to examine the validity of the General Electric transient analysis methods and verify the computer codes. The first scram signal which normally would have been initiated on the position of the turbine stop valve, was bypassed in order to provide a transient comparable in severity to the worst transients analyzed in FSARs. Using the transient analysis method and the REDY computer code used in the licensing applications at that time, General Electric made pre-test predictions of pressure, neutron flux and &CPR on a best estimate basis. The neutron flux and ACPR predictions were signifi cantly nonconservative and the pressure predictions ware somewhat nonconser vative.

After the tests General Electric performed post-test predictions of pressure, neutron flux and ACPR using the actual or measured plant parameters with best estimate modeling assumptions as well as the licensing model asswuptions. The ACPR and the neutron flux predictions were again nonconservative for both sets of calculations. The pressure peaks were predicted conservatively. It should be noted that General Electric showed that the predictions of pressure end ACPR were conservative with licensing basis inputs when the first scram signal initiated on turbine stop valve position was not bypassed, I.e., under normal conditions.

The comparisons of the test results vn the REDY code, the licensing basis model, confirmed the existence of a steam line pressure wave propagation phenom enon in a turbine trip transiant ad time varying nature of the axial core I-1 xi

power distribution. General Electric accelerated its model development program to include steam line dynamics and representation of the core physics and thermal hydraulics in space-time domain and produced the ODYN computer code which is the subject of this review.

.8. SCOPE The scope of this review is the evaluation of the 0DYN code for use in the analysis of certain transients in Chapter 15 of the FSA~s.

C. SUTA.ARY OF ANALYTICAL MODELS The overall system modal in the 00YN coda consists of a one-dimensional represen tation of the reactor core, and the recirculation and control system model.

These two models are coupled to each other. A steady state initialization is made initially, and then the parameters for the transient are calculated.

First, the recirculation and control systems are solved for the steady state conditions. Soes of the initial conditions are input and they may be plant unique. Other initial hydraulic values such as core pressure drop and bypass flow fraction, which are also input to the steady state recirculation end control model, are calculated elsewhere. These parameters are calculated in the staadj state multi-channel core code (Reference 1). Using all these input values, the steady state recirculation and control model calculates the remaining hydraulic parameters in the plant. The steady state initialization in the recirculation and control model provides the loop pressure drop, core exit pressure, core inlet flow and enthalpy Uo the one-dimensional reactor core model. These values are used In the reactor core model to calculate the neutron kinetics, thermal hydraulics and fuel parameters for the steady state conditions.

1-2 xii

The steady state axial power distribution is calculated by the neutronics model. The model uses cross section fits obtained from an analysis about cross sections for different relative coolant densities and control states and that art radially averaged for each axial plane. The fits are such that the axial power in the oae dimensional model is required to yield the same axial behavior as in the three-dimensional BWR Core Simulator solution. The Steady state thermal hydraulic solution permits the calculation of the steady state fuel temperature distribution.

During the transient, the recirculation and control system model calculates the time derivatives. At the end of the time step, the recirculation and control system model supplies the new external boundary conditions to the reactor core model. The reactor core model calculates the new neutron flux, thermal hydraulic parameters and fuel temperatures. It also provides reactor core exit quality, flow and pressure as input to the recirculation and control system model. The recirculation and control system model calculates the loop pressure drop and the reactor core model calculates the core pressure drop. These pressure drops are compared. If they are not equal within a certain limit, the recirculation and control system model derivatives are modified and the time step calculations are repeated.

The recirculation system is modeled by solving the mass, energy and moment.um conservation equations for the steam line, reactor vessel and recirculation loop components; which included jet pumps, recirculation pawps and associated piping. The centrol system is modeled as a series of connected gains, filters, xiii

integrators, and nonlinearities (limiters and function generators). The control system outp ut is-valve position and thus flow control. The one-dimensional core model comprises equations describing the neutron kinetics, thermal-hydrautics and heat transfer behavior of the core.

Major assumptions used in the modeling of the recirculation system art as follows:

1. Pressure variations in the system are described with tan nodes. One node is used for the reactor Inlet; another node is used for the reactor vessel dome, and the reaining eight nodes are used to describe the behavior of the steam line.

2. Liquid and vapor mass volume balances are used to predict the reactor vessel water level changes.

3. The recirculation loop model can simulate any combination of multi-loop systems. The entire recirculation loop is assumed to be subcooled and incompressible.

4. Steam in the steam linte is treated as single phase flow. Condensation of steam in te stnm linte Is precluded during the transient.

Major assumptions used in the reactor core model are as follows:

1. A one-dimensional neutron kinetics model is assumed. The neutron flux varies axially with time. One energy group diffusion theory and six delayed 1-4 xiv

neutron groups are used. Decay beat is modeled using a simple exponential decay heat model. The one dimensional neutron diffusion parameters are obtained by collapsing the parameters obtained from the GE three-dimensional BWR Core Simulator (Reference 2).

2. A single active heated channel represents the core average conditions and another single channel represents the core bypass. A five equation model representing mass and enerV conservation for the liquid and vapor, and the mixture momecm conservation are used to calculate core thermal-hydraulic behavior.

3. Heat transfer to the moderator and fuel temperatures are calculated using an average fuel and cladding model at each axial location of the core. The gMp conductance is an input parameter which may vary axially in time. The cnductian parameters are temperature dependent. A radially uniform (flat) power distribution is assumed in the fuel rods.

1-5 x'r/vi

11. STAFF EVALUATION The staff evaluation was performed in three parts:

A. Review of the analytical models in the ODYN code and determination of uncertainities in the code modeling.

a. Review of the qualification of the code. This part of the review is accomplished in three areas:

L Comparison of specific bodels in the code with separate effects test data.

1. Comparison of integral response of the code with the integral test data.

3. Comparison of the code predictions with the predictions of an independent code; i.e., audit calculations.

C. Review of the safety margin; i.e., evaluation of the margin when the code is used with the uncertainties assigned In the licmnsing basis transient.

The uncertainties of the calculations were evaluated as part of the calculatlonal model review.

The measure of all code uncertainties is made in terms of ACPR/ICPR ratio. The

• CPR* Is an acronym for critical power ratio. It is the ratio of.the critical power of the limiting bundle in the core to the power of the same bundle at the operating power of inerest. The critical power is an artificial bundle power obtained by increasing the power analytically until the critical quality is reached. The analysis is performed using the GEXL correlation. Since the hydraulic and neutronic parametars change during the transient, CPR also changes during the transient. The minimu, value of the CPR is called MCPR and UZ-i xvii

the difference between the Initial critical power ratio, ICPR, and MCPR is the ACPR. Hence, the ratio of ACPR/ICPR is a measure of the relative severity of the transient.

The uncertainties In the code are determined by making sensitivity studies. An independent parameter in the code is perturbed and the resulting change in ACPR/ICPR is calculated for a turbine trip without bypass transient, which is generally Jimiting. These independent parameters pertain to the various models such as the parameter of Cc in the Zuber drift flux model or frictional loss coefficients in the steamline. They do not pertain to system parameters which determine the actuation of the valves since licensing basis analysis require limiting settings for these systems parameters.

A. REVIEW OF ANALYTICAL MODELS

1. Recirculation and Control System

a. Recirculation Looo Model The recirculation loop system consist.s of the upper plenum, steam separators, vessel dome, jet pump antd reciculration loop. Mass, energy and momentum conservation equations are used to describe thermal and hydraulic behavior of the components. These equations are solved using an explicit finite differencing method which it presented in Reference 3.

During the steady-state initialization, the time derivatives are set equal to zero. A multi-channel steady state hydraulics code provides the steady state core pressure drop and the bypass flow fraction to the recirculation system model. This code is presented in 11-2 xviii

Reference I and has been reviewed and approved by NRC (Reference 4).

The other inputs used by the recirculation syste model are plant specIfic such s dimensions related to plant geometry, pressure loss coefficients,, separator carryunder fraction and Jet pump and redir culation pump characteristics.

In the initial steady state conditions the Jet pump drive and suction flows can be determined from the equation of continuity and the jet pump O" ratio. This ratio is defined as the ratio of the suction to the drive flow. It is valid for the rated conditions which are selected to correspond to steady state initial operating conditions.

Using the moment*m equation and the um" ratio, the suction flow and the suction flow loss coefficient are determined. During the transient the ratio changes. The jet pump suction and drive flows (consequently recirculation loop and core inlet flows) are calculated using the momentum equations keeping the suction flow loss coef ficient constant. The sum of the suction and drive flows provide the recirculation loop flow and the sum of all recirculation loop flows provide the core inlet flow.

The recirculation system models used in the ODYN and REDY codas are the same. The REDY codd (Reference S) has been reviewed for AT'S analyses and the recirculation system model has been found acceptable with some limitations (Reference 6). The following discusses and evaluates the recirculation system model. This evaluatiton, exce~pt for the uncertainties, is the same both for OYh and REDY codes. The I1-3 xix

limitations found in the REDY code are equally applicable in the ODYN code.

During the transient, momentum equations are used t calculate the jet pump suction and drive flows. Hence, the foar loss coefficients in the recirculation system affect the care flow and consequently the calculated ArPR. A sensitivity study performed by General Electric using the ODYN code by decreasing the diffuser farm loss coefficient by 1= showed an incriue of 0.001 in ACPR/ICPR. General Electric estimated uncertainties in the jet pump loss coefficients about 20%.

These uncertainties are Inferred from the uncertainity in the jet pump I' ratio. General Electric noted that the decrease In the jet pump pressure drop loss on the order of 2=Z changed ACPR/1CPR by 0.01. This is the biggest uncertainty estimated by General Electric in the recirculation system. According to General Electric reason able variations In other parameters such as drive flow L/A, jet pump areas or lengths (which are manufactured to close engineering tolerances) and loss coefficients at the nozle, plenu and bulkwater did not change ACPR/ICPR ratios significantly. Based on these sensitivity studies.the impact of these uncertainties an the values of ACPR/XCPR in the generally limiting transient is small.

During the transient., the transient terms of the momentum equation representing Inertia may become important in determining the core flow.

Recirculation pump trip tests were performed at 50%, 75%, and 12D power levels in the Oyster Creak plant and reported In Reference S.

Good agreement exists between the measured and REDY calculated I1-4 xx

core flows for the transient. This shows that the momentum equations were solved correctly to predict flow transients. Recirculation pump trip tests were also performed In Dresden-2 and they were reported in Reference S. However, in these tests measured Core flows were higher than those calculated because the actual pump inertia was higher than the value used in the analysis.

One of the Jet pump modeling assumptions is that the region from the nozzle to the throat is considered to have no inertia. In order to validate the transient modeling of the jet pump, transient jet pump tests were conducted at the Moss Landing Generating facility, Reference S. In these tests the jet pump drive flows were oscillated at several frequencies and measurements were made of the gain and phase relationship of the drive flow. CWmparison of the measurements and model predictions showed good agreement up to 5 Hz. The model did not predict a resonance condition In the cold test data at 6.5 Hz; consequently, the use of the model is limited to 5 Hz. This limitation means that the code will have errors if recirculation loop flow variations are sudden. The harmonic components of the flow variation should be less than S Hz.

Another assumption which has been validated by tests is the assumption of complete nixing at the core inlet. Tests were performed in Monticello to verify this assumption. Care flow distributions for three core flow rates, at 29%, 5% and S5M of rated flow rates, were measured for symmetric operation of the recircu lation pumps, Reference 7. Tests results indicate that the bundle flow rate does not vary more than Z.1M from that in the average 11-6 xxi

bundle with 95% confidence level. This indicates that the assumption of uniform pressure distributtion at the inlet of the care and colete mixing is a valid assumption for the recirculation system modeling.

The review of the analytical models and the comparison of the predictions with the tests above lndicate that the recirculation loop model and the impact of associated uncertainties on ACPR/ZCPR as presented by General Electric are acceptable. The harmonic coo ponents of the flow variation should be less than S iz and the model should be valid for the analysis of transients where the fluid in the recirculation loop, downcomer and core inlet remains subcooled (incompressible). In the transients to be analyzed by the ODYN code, it is expected that these limits will not be exceeded.

b. Control System Model The control system models were evaluated for structure as well as the methodology for evaluating plant specific properties. Plant specific properties consist of response functions, gains, and time constants for the control system.

The system models are composed of transfer functions, limiters and function generators. The transfer functions are based on typical filters and proportional, integral, derivative control laws.

Limiters and function generators are used in the modeling of flow valves as a means of linearizing the gain within control loops. We have reviewed the model structure for the motor generator flow control 11-6 Xxii

model, the feedwater control model and the pressure regulation with the Meca*nical Hydraulic Control. We find the structure of these models acceptable and typical of the type of modeling conducted with classical control system theory.

With respect to the description of the contr*ol models, the following models were evaluatedi (a) Valve Flow Control System (b) Motor-Generator Flow Control (c) Feedwater Flow (d) Pressure Regulator and Turbine Controls (e) Reactor Safety System For input signals, the Valve Flow Conti.ol model receives a turbine governor signal, a sensed steauflow signal, a filtered neutron flux signal, a recirculation drive flow signal and a manual setpoint signal. The control system is modeled as a series of connected gains, filters, Integrators, and nonlinearities (limiters and function generators). The contraol system output is valve position and thus flow central.

For input signals, the Motor-Generator Flow Control model receives a load demand error, a master manual or automatic signal as well as a loop manual or automatic signal. The control system is modeled as a series of gains, integrators, function generators, and with actuators of a drive motor, variable speed coupler, generator, and motor pump.

The controlled variable is recirculation drive flow.

11-7 xxiii

For input signals, the Feedwatar Control System receives feedwater flow disturbances, vessel pressure corrections, a level setpoat.

signal, a mixture level signal, and a steam flow signal. These signals are operated on by a contral modeled as a series of connected gains, integrators, filters, and non-linearlties (limiters and function generators). The controlled variable is feedwater flow.

For input signals, the Pressure Regulator receives a turbine inlet pressure signal, a pressure setpoint, a turbine speed setpoint and a turbine load setpolnt. These signals are operated on by a control modeled as a series of gains, filters, control laws, control valve servos and non-linearities. The controlled variable is turbine inlet pressure.

The staff review finds that these models are conditionally acceptable.

Technica1lly, the models are composed of transfer function, gains, filters, and synthesized nonlinearities such as deadbands and saturation limits. The technical form of the control system models is acceptable to the staff.

However, the model is uied to establish initial control system settings such as gains, time constants, and control functions. Since the selection of these settings is made on a plant specific basis, the staff requires that each applicant's Safety Analysis Report reference a clearly defined basis for making these selections. The design criteria must be provided for each control system of the plant. The initial control system characteristics shall be verified as conforming to the design critaria for each control system of the plant.

11-8 xxiv

C. Steam Separator Model The separator is modeled using a one dimensional momentum conservation equation whereas the flow in a separator is rotational and clearly multl-dimensional. However, using separator test results (Reference 8), it was possible for General Electric to develop an e*pirical one dimensional momentum equation describing the flow behavior. Tests indicated that the thickness and configuration of the layer of swirling water along separator walls is Independent of the inlet flow (for 200,000 lb/hr < Flow c 800,000 lb/hr) but dependent on the inlet quality. The water layer primarily affects the effective L/A in the momentum equation of the separator. Due to differences between the densities of steam and water, the primary inertial effects are due to the liquid. The tests of Reference B provided a relationship between the effective L/A and the inlet quality, and an empirical separator pressure drop coefficient.

General Electric states that the value of pressure drop coefficient has a conservative bias in it. The higher the pressure drop or the pressure drop coefficient, the higher is the value of ACPR/ICPR.

However., General Electric did not quantify the conservatism in this model in terms of &CPR/ICPR relative to actual plant conditions.

Therefore, no credit is given to this conservatism.

General Electric performed sensitivity studies decreasing the value of L/A by 30. This resulted in an increase of 0.002 in ACPR/ICPR:

In order to assess if the scatter of 30% in the separator L/A is sufficient, the staff reviewed the separator data In Reference S.

II-9 X2v

The data indicates that the scatter in the separator L/A values can be high. The thickness af water layer can be used to make a fairly good estimate for L/A. Tests indicate that the thicknesses of water layer for the same conditions can vary from each other by a factor of four. Reference 8 describes the reason for these variations as an instability.

Discussions with General Electric indicate that the value of L/A used In the ODYN code included the value of L/A for the standpipe and therefore, the scatter was not a factor of 4 but It was judged to be 30=. The staff has no information how these separate L/A effects (one due to the separator and the other due to the standpipe) can be assessed.

Reviewing the analytical model we find that the separator model is acceptable; however, based on available information we judge that a factor of 2 in separator L/A variation (rather than 30%) would be more appropriate in assessing the uncertainty. Nence, we estimate the component of that ACPR/ICPR uncertainty for L/A will increase from +/- 0.002 to t 0.015.

d. Utoer Plenum. Vessel Dome and Bulkwater Model These components are modeled using mass, energy and momentum conservation equations. Peach Bottom tests indicate that dome pressures calculated to predict the data are higher than the experimental values. In the opinion of General Electric, the reason for the overprediction is that the energy equation for the dome 11-10

region predicts that the bulk water mass very quickly becomes subcooled,' the system becomes stiff, and therefore, the pressure rises very quickly. Since the rapid pressure rise leads to a rapid void co)lapse the staff concludes that the model is conservative.

However, the Peach Botto tests also indicate that ACPR predictions are not conservative. This implies that the conservatism of the bulk water model is offset by the nonconservatism somewhere else.

General Electric did net quantify the conservatism in this particular model. In view of Peach Bottom tests where a trade off has occirred, no credit for conservatism can be given. .

We find that the analytical methods used in these models are acceptale; however, as stated, no credit for conservatism will be given.

2. Steam Line Model The steam line Is modeled assuming single phase mass and energy conservation equations which are solved using an explicit finite dif ferencing method. The steam is assumed to behave isentropically. The steam line is nodalized Into six segments while the bypass line is modeled using two nodes. Safety and relief valve flow rates are treated as separate flow branches.

Sensitivity studies were performed by General Electric for various numbers of nodes for a sample test problem wherein the inlet pressure is kept constant and at the outlet turbine stop valve closure is simulated. These sensitivity studies were performed using nodal arrangements of 3,4,5,6,7, I1-11 Xxvii

8,20, and 40 nodes and compared with the analytical model predictions using the method of characteristics. The analyses indicate that a minimum of 7 nodes Is required to predict frequencies to a reasonable degree. The comparison of amplitudes of pressure oscillations between the 8 node model and the analytical model is also reasonable. The conservatism of a model is dependent upon the integral of the pressure oscillations over a relatively short period of time since it is the integral of the pressure that is imposed on the core. The void collapse and the subsequent power increase is dependent upon the rate of change of this integral pressure.

Judging from the pressure oscillations calculated from the 8 node model and the analytical model based on the method of characteristics the staff concludes that the Integrated pressures are approximately the same for both models and perhaps there is a very slight conservatism in the 8 node model. Consequently, we find that the finite differencing scheme and the solution method employed in the steamiline model are acceptable.

Other uncertainties In the model are in the form of friction loss coefficients and in the value of the average specific heat ratio. General Electric conducted sensitivity studies by varying the specific heat ratio and form loss coefficients. The Peach Bottom tests Indicated an average specific heat ratio of 1.15. The change of this ratio U 1.25 caused an increase of 0.01 in the ACPR/1CPR ratio.

We reviewed the values of the average specific heat ratios for steam at 1000 psi&. The value of 1.15 is valid for saturated steam with very little amount of drplets in It. The value of 1.25 is valid for a slightly superheated steam. Since there is a pressure drop along the 11*12 XXviii

steam line, we do not expect steam to be superheated. Hence, the value of 1.15 is acceptable. We also find the calculation of uncertainty of 0.01 in ACPR/ICPR ratio acceptable.

General Electric also performed a sensitivity study by decreasing the loss coefficient by 20. This was based an the upper licit of st1amline-loss coefficient uncertainty. Decreasing the loss coefficient by 20 increases the ratio of ACPR/ZCPR by 0.01. Decreasing the loss coefficient by 20M is a reasonable assuption and we find the calculation of uncertainty of 0.01 in ACPR/CPR due to pressure loss coefficients acctable.

In conclusion, our review indicates that the analytical methods used in steam line modeling and associated uncertainties are acceptable.

3. Core Thermal-Hydraulics. Model Two-phase mass, energy and momentum conservation equations were used to predict the behavior of the thermal-hydraulics of the core. Two mass and two energy conservation equations representing each phase separately and one momentum equation representing the mixture comprised the five equation model. In addition to these equations, correlations for 1) interfacial heat flux, 2) Zuber drift flux model (Reference 9), 3) two-phase pressure drop, and 4) heat transfer, are used.

The interfacial heat transfer correlation i's based on the "mechanistic nodel" presented in Reference 9. The selection of the heat transfer correlations is based on the flow regimes. In the single-phase liquid region, the Dittus-Goelter correlation is used. Zn the subcooled ant bulk 11-13 xxix

bolling regions, the Jens-Lottes and Chen correlations are used. respec tively. Two-phase pressure drop correlations are based on the Martinelli-Nelson correlation. The five equation model together with the correlations are solved using a fully implicit finite diflerencing method in the space-time domain. The space domain is ant dimensional In the axial direction and the care is represented using 24 axial nodes.

To improve the accuracy of predictions within a node, a boiling boundary concept is dafined. This concept defines a location In the axial direction for which the mixture enthalpy is equal to the enthalpy at which point subcooled boiling begins. This location establishes the boundary between the liquid and two-phase regions within an axial node at each time stop and the program selects the appropriate correlation for the appro priate region. The variables solved for each node are volumetric flux, vapor fraction, pressure, vapor enthalpy and liquid enthalpy.

Two models are particularly significant in the assessment of unertainties in the five equation model; they are Zuber drift flux and the subcooled boiling models. These are discussed in the following sections.

a. Drift Flux Model The choice of two parameters, C0 aand V is important in this model.

The first coefficient (C ) Is the concentration parameter which describes the slip due to cross sectional averaging of a nonuniform void fraction profile. The second em (V ) is the drift velocity which describes the local slip between the phases. The value of C is strongly dependent on the flow regimes and geometry. This 11-14

dependence has been shown in many tests (Reference 9). The drift velocity is dependent on the density differences between the phases as well as on the flow regimes.

In the model used by General Electric, these parameters are empirically determined in the form of correlations based on the test data. *The data were obtained both from tubes and channels, and are reported in References 10 through 14. When the vapor fractions obtained from these parameters were used to calculate power shapes observed in Sris, some discrepancies were observed. Consequently, General Electric introduced another correlation for Co, and a concept of neutron effective void fraction, to provide a better fit with measured power shapes. Based on physical considerations it is conceivable why C0 used in thermal hydraulic calculations is different from Ca for niutron power calculations. The thermal hydraulic C0 is based on tube geometry while neutron effective C0 is obtained from actual core geometry. The value of C should be different for tubes and rod bundles because of different vapor fraction profiles and flow roomes. However, in a telecon General Electric stated that C0 valid for thermal hydraulics gave good agreement with Atlas data and C valid for neutron effective void fraction gave good agreement with the core data. Hence, the differences cannot be explained based on geometrical considerations alone and there is an artificial fix In the model. According to Reference 34, this fix is necessary to compensate for deficiencies in lattice physics methods.

11-35

General Electric estimates that the uncertainty in the concentration parameter, Co. is about

• 3a at a void fraction of .70 for neutron effective vapor fraction calculations. This corresponds to a t 5 uncertainty in void reactivity coefficient which leads to an uncertainity of t 0.008 in the value of &CPR/ICPR. However, General Electric uses t 1= uncertainty in the value of Cc for thermal hydraulic calculations. We find no reason that the uncertainty in C0 for neutron power calculations should be different because the correlation Is used to calculate voids the same way as in the thermal hydraulics. General Electric does not state any uncertainty in Vgi for neutron power calculations but states an uncertainty of t 2OX for thermal hydraulic calculations.

Based an Reference 15, we assessed the uncertainties for thermal hydraulic C t Z2a and Vgj

• 30 respectively. Reference 25 has a different data base from the references that the General Electric used. Extrapolating the General Electric results, we estimate that t 2C uncertainty in C0 would result in +/- M3.uncertainty in the void fraction or void reactivity coefficient.

We also reviewed the void fraction data taken In the FRIGG loop, Reference 16. The FRIGG tests were performed using rod bundles. The review of the data indicated that the scatter of a 3= in void fraction was reasonable In the low quality region. This finding also substantiated the estimate of a 2= uncertainty in the value of C0 .

11-16 xxxii

Assum ing the same uncertainty for the neutron effective Ca and extrapolating the General Electric results, we have estimated that the uncertainty of +/- 2% in Ca resulting in 1 33% uncertainty in the void fra.ct*on or in the void reactivity Coefficient would produce an uncertainty of t 0.053 in ACPR/ICPR. We presented these findings in the ACRS hearing, Reference 30.

In response to the above staff assessment, General Electric submitted additional information, Reference 31, requesting the reduction of the uncertainty in &CPR/ICPR. The primary argument was that the uncertainty of +/- 2 in the value of Ca (seven times the uncertainty i

of

• U which had been proposed by General Electric) leading to an uncertainty of approximately

• 3= in void fraction was applicable for a low quality "nd a low vapor fraction region. The uncertainty becomes smaller at higher qualities. In addition, General Electric submitted another sensitivity study using. neutron effective Co 1.0 and noted that this would be the bounding value for ACPR calculations.

General Electric also noted that the transient results waere weakly dependent on void fractions at low qualities in the subcooled region, Reference 32.

We reviewed the new information submitted in Reference 31, and agree with General Electric that uncertainties in vapor fraction can be reduced at higher qualities and that C0 = 1.0 is a bounding value for bulk boiling. General Electric stated an uncertainty of +/- E% in void reactivity coefficient at a void fraction of 70%. This corresponds approximately to an uncertainty of +/- 5% in void fraction. Further 11-17 xxxiii

review of the void fraction data in the FRIGG loop shows a scatter of t IZ in void fraction at qualities of 5% and M*. These qualities are considered relatively high and they correspond to vapor fractions of 40 arnd 60 respectively. It appears that the FRIGG loop data

,show a larger-scatter of void fraction than that assumed by General Electric. At high qualities, the uncertainty of :tI In void fraction corresponds to approximately t 1= uncertaintn in Cc.

However, the theoretical limit for C0 in the bulk loading region is 1.00. It can be higher but not lower in this region. The scatter of 1CM on neutron effective Cw would bring the value of C0 below 1.00.

Some of the scatter in the FRIGG data was due to measurement errors which could be as high as +/- 1= Hence, we accept the limit of C

• 1.00 as bounding for the uncertainty studies. Further, sensitivity studies performed in Reference 32 Indicate that the transient is weakly dependent an changes of neutron effective Ca in low quality region. Hence, we accept the calculation of uncertainty of +/- O.011 in ACPR/ICPR as suggested in Reference 31. This is approximately 3C% larger than that originally proposed by General Electric.

b. Subcooled Boiling Model The phenomenon of subcoole*d boiling is modeled.by the *Mechanistico subcooled boiling model developed by R. T. Lahey in Reference 9. The model provides a relationship for interfacial heat flux between the bubbles and surrounding liquid. It consists of two terms: one tem shows the effect of the temperature difference between the phases and the other shows the effect of the wall heat flux.

Il"1B xxiv

The model has been verified using the data obtained by S. Z. Rouhani (References 17 and 18). These data were obtained fro, a vertical annular channel. In determining the uncertainty of the correlation, General Electric provided a sensitivity study using a coefficient un" in the correlation. The nominal value of In* is 1.0. For n - 1.25, a change of 0.009 in ACPR/ICPR is obtained. If 1.50 is assumed, the change in ACPR/ICPR is 0.014. GE states that the value of 1.25 provides a reasonable uncertainty for the model but does not provide any supporting evidence or data.

We raviewed the void fraction vs. axial he t curves drawn for various In" values and find that the void fraction difference between the t*o curves drawn for n a L.0 and n a L.5 is about 3.S In absolute or 2C relative t the average measured value of the void fraction in the subcooled region. Some of the rod bundle experiments performed in the Frigg loop (Reference 16) show 100% (relative) scatter of the data. In general, the scatter is 15 - 3= relative to the average void fraction. We believe t 30. scatter is a reasonable estimate of uncertainty. Therefore, we increased the uncertainty in ACPR by a factor of 1.67 (30/U8) which results in

• 0.023 In the uncertainty value of ACPR/ICPR for the subcooled boiling model. We estimate the corresponding minimum and maximum values of In' to be 0.5 and 2.0 respectively. General Elective is required to make sensitivity studies to verify that these values correspond to i 0.023 uncertainty in ACPR/ICPR.

• 11-19

The review of the analytical models describing the thermal-hydraulic behavior of the core indicates that these models art acceptable provided the uncertainties of various components are Increased Uo the values recommended by the staff.

4. C*re Physics Model

a. Assumptions in the Neutronics Model The neutroics model of ODYN is based on time-dependent, one dimensional, one-group, diffusion theory. The model includes the effect of delayed neutrons and the calculation is performed in the axial dimension of a VWR. Radial effects due primarily ti Doppler, moderator, and control state are taken into account in collapsing a three-dimensional model to the one-dimensional axial model. Ofthe three effects., the control state variation due to scram during a transient is the east important. Some care must, therefore, be taken in choosing the Initial weighting functions to account for these effects.

We have reviewed the assumptions in this neutronic model with the current state-of-the-art for performing space-time coupled neutronic and thermal-hydraulic calculations. We conclude, based on our review, that the assumptions on which the neutronic model of ODYN are based are acceptable.

b. Derivation of Ecuations for the One-Graut. One-Dimensional, Time-Devencent Neutronic mocel We have followee in a step-by-step manner the derivation of the one-group, space-time neutronics model presented primarily in U1-20 x:xvi

Appendix A of Volume I of the report. This derivation proceeds from the time-dleendent form of the three-dimensional neutron diffusion equation for the fast flux as used by the General Electric three dimensional reactor simulator (Reference 2) along with appropriate equations for delayed neutrons. The three-dimensional time-dependent

,neutron flux is represented as a product of radial and axial time dependent components. Weighting functions are next introduced to make this factorization unique and to minimize errors in the procedure in some sense. The weighting functions are taken, according to the adiabatic approximation, as the solution to a steady-state eigenvalue problem to be solved at. various points in time. In practice, the weighting functions are calculated only at time zero for as many BWR operating states as is necessary. This procedure results in the final form used for the one-group, one dimensional, time-dependent equations along with defining equations for the nuclear parameters that are used. The derivation also includes discussion of the average axial power distributions, initial normalization procedures, and boundary conditions.

Section 5 of Volume I of the report discusses the Integration of the spat*al and time variables to obtain the discrete form of the one group, one-dimensional, time-dependent equations. The procedures used for this are straight forward. This section also discusses the radial weighting function and the treatment of the control state.

Cross section related parameters are functions of axial core heighi, control state, and relative water density. These parameters are fit to quadratics in the relative water density.

11-21 x*cvii

Out review of the derivation of the equations for the one-group, one-dimensional, time-dependent neutronics model has been performed by deriving and verifying each of the equations presented in Volume I of the report. We conclude, based on our step-by-step review of all the neu-trnic equations, that the derivation of nuclear parameters and equations for the one-group, one-dimensional, time-dependent model is acceptable.

c. Calculation of Neutronic Inout Parameters General Electric uses Its Lattice Physics Model and its Three Dimensional BiWR Core Simulator to process nuclear data for the ODYT code. The Lattice Physics Model Is described in Reference 19. The Three-Oimensional BWR Care Simulator is described in Reference 2.

Both of these codes have been reviewed and approved by the NRC for use in BWR applications.

The Lattice Physics Model, as its name implies, is used to generate nuclear parameters for use as input to the BWR Coar Simulator. This dataIs generated as a function of fuel type, control, temperature, void fraction, void history, and exposure. Before being used by the BWR Core Simulator, the data is transformed from the Lattice Physics Model void fraction to the Neutron Effective Void (NEY) model void fraction. This empirical procedure was developed by GE to remove a discrepancy between 94 Core Simulator results and operating reactor data. The BWR Core Simulator Is used to perform the three-dimen sional analyses that are required for obtaining the data for processing Into. parameters and cross sections for the OYT code.

N-22 xxxviii

Our review of the calculation 9f neutranic input parameters is based on the use of NRC reviewed and approved codes and on comparisons of three-dimensional and O0Th steady-state neutronic analyses. The approved codes are (1) the Lattice Physics Model (NEDE-20913-P, "Lattice Physics Methods," C. L. Martin, June 1976 and NEDO-2093S, "Lattice Physics Methods Verification," C. L Martin, June 2576) and (2) the BWR Core Simulator (NEDO-205.3, "Three-Olmensional BWR Care Simulator." J. A. Woolley. May 1976 and NEDO-20946, "BWR Simulator Methods Verification," G. R. Parkos, May 1S76). The steady-state calculations compared th BeWR Core Simulator and ODYN results for scram reactivity and core averaged axial power distributions, among other things, for a number of different reactors and operating states.

Some of the uncertainty values used by General Electric in response to our Question 12 need to be revised in our judgemmnt. We believe that the Doppler reactivfty coefficient uncertainty should be increased from t 6 percent to about

• 10 percent. This increase is based on the uncertainties inherent in the calculation of Uranium-238 resonance absorption, the calculation of the Dancoff factor in the complex !WR lattice, the calculation of spatial weighting factors, and the computation of effective fuel temperatures. This change in the Doppler uncertainty will have very little effect on the calcu lated ACPR/ICPR ratio. We estimate that this will increase the uncertainty in ACPR/ZCPR from a 0.0015 to t 0.002. We believe that the scram reactivity uncertainty should be increased from

• 4 percent to about 1 10 percent. This increase is based on the uncerainties 11-23 xxxix

inherent in calculating the initial scram reactivity rate and total control rod worths. We estimate that this will increase the uncertainty in ACPR/ICPR from 1 0.01 to +/- 0.02. The General Electric values for the Lattice Physics Model and BWR Coar Simulator uncer tainties in the void reactivity coefficient calculation are acceptable as given In response to our Question 12.

Since the uncertainties in the neutron effective void fraction are assessed in the thermal hydraulic section, we did not consider it as part of void reactivity uncertainty. Hence, the uncertainty in ACPR/1CPR value Is reduced from t 0.020 to t 0.018.

We conclude, based on our review, that the procedures and calculations performed to provide the neutronic parameters for input to the 0DYh code art acceptable.

5. Fuel Heat Transfer Model Heat transfer to the coolant end temperatures within the fuel are calculated assuming a single cylindrical fuel element for each axial location. The fuel heat transfer model used in the OYT code calculates fuel temperatures as a function of time in the transient as input to the Doppler reactivity calculation. The cladding wall temperatures are also calculated as input to the transient cladding-to-coolant heat transfer model. The ODYh code allows for axial variation of the neutron flux, as well as of coolant flow, density, and pressure. This results in an axially varying set of input conditions for the fuel heat transfer model.

The resulting temerature calculations are then solved for a series of discrete axial elevations in the core.

11-24 xl

The fuel and cladding conductivity and heat capacity are assumed to be temperature dependent. A gap thickness is specified between the fuel and the cladding and an input gap conductance is used. Axial and time variations in the gap conductance may be given, but a constant value is used for safety analyses. The external heat transfer coefficient and coolant temperature are obtained from the thermal-hytraulic portion of the code. The heat generation rate in the fuel pellet is obtained from the axial power distribution which is determined by the neutronics segment of ODYN. The radial heat distribution in the fuel rod is assumed Uo be independent of axial position and independent of time.

General Electric derived the fuel heat transfer model from the general heat flow equation. The equation is expressed with axisymetry and zero axial conduction assumed. The resulting, one-dimensional, transient heat conduction equation is solved by the Crank-Nicholson finite-diffterence technique. The solution is approximate, but the procedure is widely, practiced and is well documented in the open literature. General Electric has limited its description of the fuel heat transfer model to the formulation of this final equation.

The resulting heat conduction equation is applied to a single rod with a radially averaged heat generation rate. This rod is used to represent all of the fuel rods in the reactor core. Because axial conduction is assumed to be negligible, the equation can be salved independen+tly for each discrete axial position in the core. The finite-difference technique also requires a radial nodalization of the fuel rod. The nodes may be of arbitrary size. General Electric has assumed that the fuel pellet is 11-25 xli

divided into seven radial nodes and the cladding into two nodes. The coefficients for both the steady-state and transient forms of the resulting finite-difference equations are given in Tables 7-1 and 7-2 of the model description.

A number of limiting assumptions have been considered in our review of the fuel heat transfer model.

1. The ODYK care transient model is designed to handle short-term events which occur on a time scale of seconds. This makes it possible to ignore the effects of long-term fuel behavior phenomena, such as creep and swelling. Pre-transient conditions, such as the average fuel-to-cladding gap size, are calculated with more detailed fuel performance codes, such as GEGAP-I11 (Reference 20) and subsequently used as input t OOYM.

2. The ODYN core transient. model is designed to handle average, rather than extreme, fuel conditions. The fuel rod input parameters represent an average of all fuel rods at a given axial location.

Boundidn input parameters are, in general, more difficult to establish and thus are more critical in the overall analysis.

However, the ODYN coda, as a whole-cort analysis, requires only the average conditions. In this respect, we note that the ODYN transient model does not have a hot channel capability, where extreme fuel conditions would be required as input.

11-26

%ii

Both of these limiting assumptions were considered in our review of the gap conductance values used by the ODY0code. We have reviewed (Reference 21) the selection of the axial and time variation of gap conductance to determine whether the selected values are appropriate for different transients. General Electric stated that the core average gap conductance values are calculated by GEGAP-II (Reference 20) which is approved by NRC. The calculated conductance is input for all axial nodes and is kept constant during the transient.

A sensitivity study was also performed for the most limiting pressurization event in which the ACPR decreases when axial varying gap conductance Is us"d. It was shown that most of the high power axial nodes have higher than core average gap conductance. During the transiant, higher gap conductance will lead to faster heat transfer from the fuel to the moderater/coolant which generates more steam voids. This results in lower stored heat In the higher power nodes. In addition, the faster conversion of fuel stored energy to steam voids in the care helps to mitigate the transient due to negative void reactivity feedback Therefore, the transient with axial varying gap conductance is less severe than that with constant gap conductance.

During limiting pressurization transients, It is expected that the fuel gap conductance will be higher than its initial stead-state value due to the increase in the thermal expansion of the fuel pellet. As discussed above, higher gap conductance leads to a less severe transient. General Electric has not taken credit for this fact, but has stated that the use of constant conductance throughout the transient compensates for uncer 11-27 xliii

taintles in thermal Conductivity and specific heat of the fuel and cladding. We have examined these properties and find that they are appropriate ever the temperature range specified by GE (300-1500 1K for fuel thermal conductivity). Therefore, it is concluded that the use of a constant, Care average gap conductance in the proposed ODYN licensing calculations is approprIate.

We have also questioned the use of a specific core average gap conductance value of 1000 Rtu/hr-tzt-OF for the analysis of the Peach Bottom Unit-2 tur~ine trip event. General Electric has shown (Q-11, Volume II) the calculated peak neutron f.lux as a function of time for gap conductance' values of 500, 1000, end 1500 Stu/h-ftO-*F. S*all differences in neutron flux are observed for the 500 and 1000 Btu/hr-ftt-F values. This is because the entire flux pulse is only a few tenths of a second wide and a fast fuel time constant is needed to produce a moderate density feedback through the rod heat flux. The peak neutron flux is minimum for the 1500 Btu/hr-ft2-OF value, showing that large values of gap conductance will mitigate the calculated flux response. This conclusion is in agreement with that found for axial and time varying conductance values.

It also shows that a core average gap conductance value of 1000 Btu/

hr-ftz-*F is not, in itself, an adequately qualified Conductance value for core transient analyses. We conclude that conductance values should be based on an approved fuel peWformance code.

We have also reviewed the use of a radially averaged heat generation rate rather than a radially-dependent heat generation rate. We questioned the conservatism of this assumption because flux depressions, and therefore a u2-z x1iv

radially-dependent heat generation rate is expected in BwR fuels. General Electric has acknowledged that the radial power distribution within the fuel red is not uniform. This is because the plutonium build-up and self-shieldIng of the fuel results in a radial power shape peaked sharply at the outside of the fuel pellet. Heat transfer from the inside of the pellet to the cladding occurs by diffusion through the fuel material.

When the power Is peaked at the outside of the pellet, the average distance from the area of maximum heat generation to the edge of the pellet is less. This results In a shorter time constant than in the uniform power production case. A reduction in the thermal time constant results In faster feedback of heat flux to the moderator/coolant and reuces the consequences of the pressurization transient in the same manner that higther gap conductance does. Hence, a uniform power dis tribution assumption inside the fuel pellet is conservative from the moderator/cool ant standpoit.

Although the use of a uniform radial pin power distribution and small gap conductance values lead to conservative moderator/coolant conditions, these assumptions also lead to higher fuel temperatures. The higher fuel toweratures, In turn, lead to increased Doppler broadening in the fuel pin which is non-conservative for transient analysis. The ODY0code assumes that all fuel at the same axial location in the core has the same temerature profile. Analyses have shown that this approach may tend to underestimate the Doppler reactivity effects because the fuel pins which have the greatest resonance capture rates are near the bundle periphery and operate at higher average temperature than that calculated by the code. This assumption is valid only for fuel assemblies with uniform 11-29 xlv-

enrichmnt. However, the Doppler reactivity Eontrtbutton to BWR transient analysis appears to be of lesser importance than the scram and moderator void reactivity contributions. The use of a uniform radial pin power distribution is therefore appropriate in the analysis of events where Doppler reactivity effects are small.

We have also questioned the application of the Crank-Nicholson method to the fuel heat transfer equation. This method suffers complications when heat generation varies with position and time, when thermal.properties vary and when non-linear boundary conditions are used. General Electric has stated that the method of solution suffers complications only when the time steps are too large relative to the fuel thermal time constant or when the fuel properties change more rapidly than the time step of the solution. It was further stated that the BWR fuel thermal time constant is in the range of 5-8 seconds compared to 0.01 second time steps taken by the ODYN code. Such extensive time stepping Is required for the hydraulic analysis and will accomnodate all non-linearity problems of the fuel behavior. It was also noted that the gap conductance is conservatively held constant in the transient calculation. We therefore conclude that the method of solution is appropriate for safety analyses.

In summary, we find that the 00Yh fuel heat transfer model is appropriate for whole-core analysis of short-term events. We note that the code is used for whole-core analysis and is not proposed for hot channel calcu lations. We have also examined the list of events selected (Volume III, table 2-1) for analysis with 0DYT and find that these events are of short duration or are limited in expected fuel temerature inc,ease. We 11-30 xlvi

conclude, therefore, that the ODYN fuel heat transfer model is appropriate for the safety analysis of these events.

6. Summary of Code Uncertainties

a. Mar*in In ACPR Calculations In sumery, the staff agrees with some of the coda uncertainties calculated by General Electric. However, some of the code uncertainties are low and the staff recommends higher values. A comparison of the code uncertainties and the corresponding bounding values as reco=mended by General Electric and the staff is presented in Table 1.

General Electric claims an expected conservative bias of 0.02 (Table 3-3, Volume II) in the calculation of the value of ACPR/ICPR due to the modeling of the gap conductance. However, the sensitivity studies performed using different values of gap conductance (Q-11, Volume 11) as well as the comparison of the Peach Bdtto test data with the ODYN predictions do not indicate that such a conservatism in ACPR calculations exists. Consequently, we do not believe that the predictions have a conservative bias.

Our review shows that the ODYN code is a best estimate code and there is no inherent conservatism in predictions of ACPR/ZCPR when best estimate input values are used. Consequently, we do not give credit for this claimed conservatism of 0.02.

General Electric estimated the total code uncertainty (Table 3-3, Volume :11) using the method of linearization. This method can 11-31 xlvii

TABLE-I COMPARISON OF CODE UNCERTAINTIES AND CORRESPONDING

- BOUNDING VALUES AS ESTIMATED BY GENERAL ELECTRIC AND THE STAFF STAFF Bounding Boundi ng

.Values of 2ACPR Values of t-.%CPR Parameters =CPR Parameters -CP I*. Reactor Core Model (1) Nuclear Model (a) Void Coefficient 0.018 O0.020 OZO at (b) Doppler Coefficient Sd1 6% 0.002 0.002 (c) Scram Reactivity '24 a5 t4% 0.010 a" t 0.020 (d) Proupt Neutron Heating 0.006 Co +/- 1 0Z 0.006 (2) Thermal Hydraulic Model n .S (a) Drift Flux Parameters V t-= O.008 V .at30 0. 01 (b) Subcooled Void Model n 125 L 0.009 a a 0.5 0.023 2.0 (3) Fuel Heat Transfer Model (a) Pellet Heat Distribution (Conservative)

(b) Pellet Heat Transfer Parameters (Conservative)

II. Recirculation System model (1) System Inertia (W/A) + 20= 0.002 L/A

• 200% 0.002 (2) Jet Pump losses K o 20 0.010 K - 20% 0.010 (3) Coae Pressure Drop A 1.5 psi 0.005 AD + L 5 psi 0.005 (4) Separator.(L/A) o30% 0.002 -200 0.015 (5) Separator aP (Conservative) 111. Steam Line Model (1) Pressure Loss Coefficients K 20 0.010 K -Z (2) Specific HeMat Ratio r* .10 0.01o_ .10 0.010 Total: 0.031 0.044 11-32 xlviii

estimate the output distribution only approximataly. The mthod also assumes the independence of the parmeters. The appropriateness of the linear method should be verified by response surface and Monte Carlo analyses. However, as will be shown subsequently, the results of the statistical analyses performed in Volume-III are not acceptable.

R~ew statistical analyses, If performed by General Electric, should be based on code uncertainties based on comparison of code predictions with the test data. Consequently, we use the value of total code uncertainty calculated from model sensitivity studies and method of linearization in detarmining the margin of ACPR/ICPR in Option A (to be presented in Staff Position) where statistical analysis is not required. The total code uncertainty in Table 3-3 of Volume III as per General Electric is i 0.031. Based on our review we Increase this value to

• 0.044.

b. Margin in Pressure Calculations General Electric has not performed analyses to determine the uncertainties in the calculation of pressure. Hence, it will be necessary for General Electric to perform these calculations using staff recommended values of the parameters listed In Table I for the Main Staam Isolation Valve closure event. We believe that there is sufficient conservatism in the ASME vessel overpressure limit to permit General Electric to use approximate linear methods to determine the uncertainty in the output. This uncertainty (2a) should be added to the ODYN calculated pressure. If General Electric demonstrates that this uncertainty is very small (e.g., by a factor of 10 or more) relative to the uncertainty in determining ASIE vessel overpressure limit, no addition of uncertainty to the calculations of pressure is needed.

11-33 xlix

B.* QUALIFICATION OF THE ODYN CODE

1. quallflcation of Neutronics Model - Comnarison of ODYN with BWR Core Simulator One of the ways in which the ODYN code may be qualified is by comparison of ODYN results with those obtained by using other codes and analytical methods. These comparisons should include both steady-state and dynamic calculations. A calculation of a BWR turbine trip without bypass licensing basis transient is compared in a later section to a calculation performed by our consultants at Brookhaven National Laboratory (BNL). This section will discuss some steady-state comparisons made by General Electric of ODYN and the BWR Core Simulator.

The EWR Core Simulator Code (NEDO-20953, Thrte-Oimensional BWR Core Simulator,' J. A. Woolley, May 1976 and NEDO-2094, "BWR Simulator Methods Verification," G. R. Parkos, May 1576) has been reviewed and approved by the NRC. This code, as used by General Electric, predicts measured power distribution peak to average ratios as follows:

(a) Axial power distribution - 5%for uncontrolled assemblies

- 10 for controlled assemblies (b) Radial power distribution - E9 underestimate relative to the process computer (c) Nodal power distribution - 4%, for gama scan data

- 7 to 8W for process computer data 11-34 1

The 8WR Core Simulator calculation Of, the criticality of first cycle and' reload BWRs results in a small bias which is taken into account for reactivity determinations of cold, xenon-free and hot operating condi tions. The standard deviation of these criticality calculations is about 0.002 in units of reactivity.

The quantities U be compared are the core averaged axial power shape, the scram reactivity, and toe void reactivity coefficient. These neutronic parameters were selected for comparison because of their importance in the turbine trip without bypass licensing basis transient. In addition, it is the space time evaluation of these -quantities that. distinguishes the ODYN calculation from a point kinetics evaluation of pressurization type transients.

The comparison of the c*re averaged axial power distribution, as computed by the BWR Care Simulator and ODYN, is given by the response by GE to our Question 36. This response states that the collapsing scheme employed in the generation of nuclear parameters ensures that the steady-state cort averaged axial power distribution and criticality computed by OYh are identical to the BW Core Simulator results. The response also Indicates that, for a nmber of plants and operating states, The ODYN core averaged axial power distribution agreed to within 0.5 percent of the results obtained with the thre-dimensional BWR Core Simulator.

The scram reactivity was compared for three BWR-4 reactor operating states. The initial scram rate (ISR), defined as the scram reactivity insertion rate during the first second from the time scram is initiated, 11-35 ii

is the quantity chosen for comparison rather than the total scram worth.

The ISR has been shown to be a critical quantity for short, duration power burst transients of the load rejection type.

One operating state was for the beginning of cycle 2 but with all control rods out. OYN results for the neutron flux and screm reactivity as a function of control rod insertion (or time) compared well with results obtained with the BWR Core Simulator. The ZSR for ODYN was about 0.93 of the value obtained with the BVR Core Situlator. The second comparison was for the same BWR-4 but with some centrol rods inserted to achieve a critical condition. The ISR for OYh was about 0.86 of the value obtained with the BWR Care Simulator. The third comparison was for another BWR-4 at 50 percent of full rated power and 100 percent of full rated core flow.

This reactor had a considerable control rod inventory and, therefore, provides a severe test for the ODYN code. The ISR for ODYN was about 0.94 of the value obtained with the BWR Core Simulator. For all three comparisons the 0TYN result for ISR was smaller than that obtained with the DWR Core Simulator and was therefore conservative.

The void reactivity coefficient derived from ODYN calculations was compared to the void reactivity coefficient derived from the BWR Core Simulator. This coefficient is obtained by knowing the core averaged void fraction and reactivity at two different reactor operating states. The different reactor operating states were obtained by changing either the reactor pressure or flow. For the variety of cases examined, GE states that the ODYh and BD Core Simulator void reactivity coeffficients agree to within S;.

11-36 Ilii

Our review of the comparison of steaay-state BURs calculated using the one-dimensional 0DYN code with comparable calculations using the three dimensional BWR Core Simulator code has been performed (1) by reviewing GE results for, the scram reactivity and void reactivity coefficients and (2) by reviewing the GE response to our request for additional information an steady-state comparisons between the two codes.

We conclude, based on our review, that steady-state ODYN code calculation of core averaged axial power distributions, scram reactivity, and void reactivity coefficients are either in-good agremnt with or conserve tively calculated with respect to comparable steady-state results obtained with the BWR Core Simulator code.

2. Qualification of the Thermal Hvdraulic Model Several comparisons of the CDYN thermal hydraulic model to standard GE design models were performed. The standard GE design model was submitted in Reference 1 and was approved by NRC In Reference 4. Both steady state and transient cornditions were analyzed.

The steady state analysis first compared the thermal hydraulic character istics (void fraction vs. axial location) of two typical BWR fuel channels (lhigh and low power channel). The results of this comparison show good agreement betwen the models. This was expected since both models are very similar. The max imum void fraction variation between these models was approximately 5% for the high power channel and about V% for the low power channel. These variations are for the axial locations where the void reactivity change is expected to be most significant for 11-37 liii

the transient calculations, i.e., A4 ft axial height. The seady state analysis also compared the change in void fraction vs. axial location for a 10 psi pressure change. The maximum void fraction variation between the models for this comparison was approximately 0.5% for the high power channel and about 5% for the low power channel. These variations are within the range of uncertainty for these type of thermal hydraulic calculations. See also discussions in Section II.A.3.a and b.

For the transient analysis comparison, the ODYN channel thermal hydraulic model result was compared to an analytical solution for exponential flow decay. The comparison riquired tht 00MY be modified to include a constant axial heat flux distribution, and steam and drift flux proper ties. This was done because the tests were run with a uniform axial heat flux and calculational convenience required the choice of constant steam and drift flux properties. The tests were performed using a single heated tube containing Freon-.14 at relatively low pressures and temperatures; about 125 psla and 160F respectively. There is t 10% uncertainty in the measurements of the void fraction. The calculational uncertainty seems to be on the order of

• 5%. These tests verified that the analytical modeling technique Including the drift flux model is acceptable and can be used to predict the vapor fraction. Judging the comparisons beteen the predic tions and the test data and the special nature of these tests, the staff estimates that an uncertainty of t 3= in transient void fraction in low qualities and +/- 5% in high qualities for a red bundle geometry during reactor transients is reasonable and consistent with the findings in the analytical model review in the previous section.

11-38 lt'.

3. Qualification Using Interral Tests In the past several years General Electric has undertaken a test program to-verify the analytical methods for reactor pressurization transients.

The tests of major interest for the current discussion consist of four turbine trip experiments. Three of these tests were performed at Peach Boatt Unit 2 (PB-2) in April 1977 and the remaining test was performed at a foreign reactor (KM) in June 1977. These tests provide the experi mental data base for verification of the W0YK code. The test results will be sumarized In this section. A detailed description of the PB-2 test is presented in Reference 22.

General Electric stated that 0DYN has been developed from first principles and independent of these results. The staff notes that in the ODYN code the only artificial fix is the neutron effective void correlation. The comparisons with integral plant tests provide an independent check of the 00Yh code. The evaluation concentrates on the differences beteen test results and corresponding 00YN predictions. The parameters which are considered in these comparisons are steamline pressure, reactor vessel dome pressure, core exit pressure, and transient neutron flux distri bution. These parameters are of primary importance In simulation of the pressurization transient. An accurate, 00Yh simulation of these param eters would provide some verification of the assumptions for the transient model s.

a. Peach Bottom Tests The inputs used for this comparison were best estimate or measured values for the current (April 1977) Peach Bottom Unit 2 EOC2 condi tions. The three Peach Bottom Unit 2 (P9-2) tests were conducted at 11-39 lv

power levels of 47.4, 6L6, and 69.1 percent of full rated power.

The tests were intended to be conducted at 10O percent of the rated flow. However, the second test was conducted at 82.1 percent of rated flow due to xenon. These three tests had different control rod distributions and fractions. For these three tests the first scram signal on the position of the turbine stop valve was disabled so that the scram would occur an high neutron flux. Disabling of the primary scram signal was necessary tU obtain'a significant power increase as a function of time for tests. Control rod insertion was assumed to vary linearly with time and was based on measured data. A constant value of 1000 ITU/hr-ftz-*F was used for the fuel rod gap conduc tance. Sensitivity studies performed by General Electric showed that the neutron flux as a function of time was insensitive to large changes in the gap conductance for these tests (See Q-f, Volume 1I).

The GE DMR Core Simulator was used to generate three-dimensional power distributions and to collapse the nuclear parameters according to the ODYN procedures. The initial core averaged axial power distributieon calculated with ODYN can then be compared with power distributions obtained with the PB-2 process computer. Comparison shows that OYN agrees quite well with the process computer core average axial power distributions for all three tests. This means that the GE neutronic procedures for generating .nuclear parameters are internally consistent and provides the proper initial conditions for the start of the transient calculations.

11-40 lvi

A comparison of the total core power as a function of time provides an integral test of the important reactivity fetdback due to scram and moderator density changes. This comparison would also be indicative of the adequacy of the core pressure and inlet flow calculations. The comparison shows that ODYI predicts the initial andfall-off part of the turbine trip. transients correctly but overpredicts the peak total core power response for, all three test.

It should be noted that the calculated consequences of the turbine trip tasts are sensitive to scram delay time and the power fraction for prompt moderator heating. It should also be noted that small changes in reactor operating state conditions such as, for example, core pressure, cause relatively large changes in the flux transient because of the large net reactivity of the transients.

The reactivity components displayed for these 0OYN calculations show that when scram occurs the power burst is quickly quenched. This is due to the control rod distribution and fraction for each test. The Doppler reactivity component plays only a secondary role. The reactivity components again demonstrate the necessity for their accurate assessment in any calculations of these type of transients.

A further indication of the adequacy of the ODYN calculation can be ascertained by comparing the core power as a function of time at the Local Power Range Monitor (LPRM) detector positions. The miniature fission detectors that comprise this LPRM system are distributed both radially and axially within the reactor core. Analysis of the PS-2 data shows that the radial variation of the neutron power with 11-41 lvii

time is similar for each detector on an axial level. This means that a one-dimensional axial calculation such as O0Yh should be an adequate representation for these tests. The neutron power as a function of time does vary, however, with axial position as shown by the experimental data for-the A, B, C, and 0 level LPRMs which are located at 1-1/2, 4-1/2, 7-1/2, and 10-1/2 feet from the bottom of the core. Comparison of the ODYh results for the neutron power as a function of time for these four detector levels with test data shows similar trends as observed for the total core power. The OYN results and test data agreement for each LPRM level is similar to that for the total core power for each of the tests. This indicates that the ODYN calculation correctly models the axial neutron flux variations as a function of time for these PB-2 turbine trip tests.

Figures 1 through 3 (reproduced from NEDO-24154) present comparisons of axial neutron flux variations as measured (calculated by the process computer) and calculated by the OYh code before the initiation of the tests. Figures 4 through 6 present comparisons of prompt neutron power as measured and calculated by the ODYN code during the tests.

The 0TYN calculated steamline pressure compares well with the PB-2 data for predicted wave travel time and frequency of pressure oscillations. However, the calculated pressure curves are more spread out and the amplitudes are smaller than the measured steamline pressure. General Electric has attributed this difference to the coarseness of the spatial mesh.in the steamline modeling. As 11-42 lviii

MO-24L54 4 In IA ammaool ompIJT3 AXiALAVER41l L.2f-.

1. ý-

I 2 LfVIL u I-I ma a.S "l-A&IVUL

@A ,-

.2 1-

! I 0  !

8iS I I I I I 4 a I I0 i2 AXIAL "wowN ffm Yngue I AxalZ Pawl hPuf.ta !ub1aa Trip I 11-43 lix

NEO-24LS/A 5

2 I

AXIAL ,JMaw tftl firm-a 2 A+/-lia love? Profl~,a Tubisa Trtp 1 11-44 Ix

la /

ASIVILIf AXAL

£:v oa -c:aT bmTi 11-4 lxi

I s-IB~. j see 1.3 va"119 Firjaet 4 leath lattou-2 Yuyblne Tvfp I Prompt Neutron Power

sons a-is 0.1.

91.4

'-I Agi It 4 afg aSO aSV PQ5 Ie. 6.25 M~ I." 61j Sao $AS &to0 In 240 Two "m f3lutte 5 r'each Noiioms-i Timbine Tirlp 2 Froqt "o~utran rowear

son BIG I 30 4% se W a .6 ci .s 1.419 Ime tIo flailre 6 PemAc lo~tto-2 Tutbine TrIp ) trompt Weitweao Povet

evidance of this hypothesis, General Electric showed that the steam line pressure calculation for the KiM test, which had a finer spatial mesh, was quite accurate. General Electric has also pointed out that the steamline pressure response shape is not as important to the transient behavior as Is the integrated value of the steamine pressure response.

We do not agree entirely with General Electric. In answer to Q-19 in Volume I, General Electric performed a sensitivity study showing the effect of nodalization (different mesh sizes) and comaring the results with the analytical model which uses method of character istics. The difference in amplitudes in this comparison is on the order of 10M while the difference in amplitudes in Peach Bottom tests and ODYN predictions Is 5. In addition, expected differences have opposite trends. The accuracy claimed by General Electric in the rJK test can be due to the adjustment of the valve opening time. This adjustment was made by General Electric Uo obtain a better agreement with the measured pressure data. It appears that the steamline model does not predict the amplitudes of oscillations accurately. This is also substantiated by the staff audit calculations. However, we agree with General Electric that the integrated steamline pressure response is more important in determining the transient behavior than the amplitud of individual oscillations occurring at these frequencies. The Peach Bottom tests indicate that the dome pressures do not oscillate and this is the pressure U which the reactor is subjected. Comparison of the dome pressures indicate that the dome pressure calculations performed by the ODYN code are conservative 11-49 1xv

relative to data; i.e., the overall rate of pressure rise as well as the tagnitzde of the calculated pressure were higher than the data indicate.

The initial doae pressure rise for the P9-2 tests was predicted accurately by ODYN. The calculation overpredicts the pressure rise near the peak of the first pressure oscillation, thus conservatively modeling void collapse for reactivity feedback. ODYN appears to overpredict the peak vessel pressure rise which demonstrates a conservative basis for overpressure protection analyses. General Electric states (Reference 23) that the overprediction Is due to the assumption made in the energy equation for the dome region. The overprediction of dome pressure is considered a desirable conser vatism. Within the period of time that the neutron flux pulse occurs, the dome pressure overprediction is approximately 16 to 28%

higher than the data.

Reviewing the steamline and dome pressure transients and based on the sensitivity studies performed by General Electric, we require that the steamline be modeled by at least 8 nodes with maximua size of 100 ft for a node.

The core exit pressure is one of the most Important parameters for the prediction of the pressurization transient neutron flux response.

As was the case in dome pressure comparisons, the initial rise In core exit pressure was followed wall by the model for all three tests. From the comparisons of ODYN to the PB-2 test results,

'I-so lXVi

General Electric has concluded that the steamineI dome and vessel thermal hydraulic models simulate the overall core pressure rise rather well In all three experiments. This lends confidence to the code predictions through the full range of power levels. Measure ments indicate sone oscillations in core exit pressure. These oscillations have been attributed to instrument line effects by General Electric. This Is corroborated by the lack of associated oscillations In neutron flux measurements.

The neutron flux predictions by the OMYN code were conservative relative to data. We estimate that the peak neutron flux is higher by 54 to 86% than the data and the integral of the nuclear power (which is a measure of the amount of energy generated) is also higher by approximately 16 to 4ZZ than the data. Hence, the neutron fluxes were predicted conservatively In all three tests.

As a final step, General Electric has presented a calculation of ACPR/ICPR for test and model. We reviewed the calculational procedure and consider it appropriate. The results show that the ACPR/1CPR foa" ODYh predicted transient conditions is within 0.01 of the values which would be predicted from test conditions; i.e., the ACPR/ICPR values calculated using the measured flow from jet pump Ap measurements, the measured pressure and the measured power during the tests. The ODYN transient conditions predicted two out of three ArPR/,CPR values conservatively. The differences are between -5.2X and 6.&% relative to values calculated using the data (minus means nonconservative). The differences in these three test results in 11-s1 lxVii

terms of ACPR/ICPR give p w L14 whtich represents a very slight conservatism for the mean and o 9:6.39% for the standard deviation.

Since the data (thre, points) are very limited, the results do not have a high degree of confidence. Table 11 presents these values of ACPR/ICPR.

TABLE 11 COMPARZSON OF MAX*MUM ACPR/ICPR VALUES FOR PEACH BOTTOM AND KKJM' TESTS ACPR/ICPR ACPR/ICPR Test Initial CPR (Data) (ODYN)

Peach Bottom Turbtine trip 1 2.536 0.170 0.173 Turbine trip 2 2.U5 0.136 0.12.

Turbine trip 3 2.048 0.132 0.141 KKM Turbine trip 1.279 0.077 0.084 Review of the test results indicate that all model conservatisms claimed by General Electric such as conservatism in calculation of the steam dome pressures and neutron flux, conservatism in collapsing of 3-0 core neutronics and thermal hydraulics, conservatism in the gap conductance input parameters and any other conservatism claimed in the computer model are either so small that it did not make any difference in calculating &CPR for these three tests or all of these claimed conservatisms are offset by an unidentified nonconservatism somewhere else, perhaps in calculation of flow. It is evident that the calculations of ACPR are not conservative for all of the tests.

1I-52 lXViii

They can only be regarded as best estimate or accurate predictions.

Hence, based on the Peach Bottom tests we do not give any credit for the conservatism in the models used in the OYh code. The code will be regarded as best estimate for ACPR calculations and any discrepancy between the test results and the code will be treated as an uncertainty or an error. Further tests would be needed to reduce these uncertai nti es.

b. KKM, Test Czmari son A brief summary of the test conditions Is contained in Volume II.

KEN plant has an unusual configuration, in that, It has two turbines and two sets of. stsamlines with a reheater line in each steamlIne.

It presents some special model considerations for .00Yh simulation. A special version of ODYN was developed to simulate this configuration.

Also uniqu to this test comparison as opposed to the PB-2 comparison is the modeling of turbine stop valve and bypass valve actuations.

Measured turbine stop valve end bypass valve positions between initial and end of actuation were not available for this transient.

The stop valve behavior can be reasonably estimated from the opening to closing time. However, the transient response is quite sensitive to the bypass valve behavior. The bypass valve opening speed of the ODYN model Lwas adjusted until the calculated transient turbine inlet pressure agreed with measurament. This adjustment was made for only the initial bypass valve opening speed and, thereafter bypass valve position was controlled based on the plant control parameters. The remainder of the test modeling Is similar to that of the PS-2 test 11-63 Lxix

I comparison. The fuel rod gap heat transfer coefficient was selected to be 600 Btu/hrft -F.

The turbine trip test conducted at KKM provided a reactor and operating state that was quite different from PB-2. The test at KJM corresponded to an end-of-cycle condition vith all control rods fully withdrawn and with the reactor at 77 percent of full rated power and 86.5 percent of full rated core flow. The reactor itself is considerably smaller in size than PB-2 and has a somewhat different system including two turbo-generating units. The turbine trip test at WI( resulted in a milder transient than the tests at PE-2. The ODYN results compared to the test data showed the same general agreement as was observed for PE-2 fo r the response of the core power as a function of time. The calculated steamline pressure response for the KM turbine trip appears to be in good agreement with measurement. The WIM comparisons appear to be in slightly better agreement with measurement than do the P9-2 comparisons. As previously discussed, the characteristics of the bypass valve were adjusted to give a good agreement with the measured steamline pressure.

The measurements of dome pressure showed some oscillations. Since these oscillations did not manifest themselves in the neutron flux measurements, they were attributed to instrument line disturbances and were not considered to be actual pressure oscillations. There were also oscillations in care exit pressure measurements. Similar oscillations were not observed'in the PB-2 tests. These oscillations 11-54 lxx

were also attributed to the instrument line disturbances since no oscillations were observed in neutron flux.

The calculated pressure responses pass through the data up to 1.8 seconds of the transient time. After 1.8 seconds the calculated pressure are higher raa those measured. The calculated care exit pressure had a 40 millisecond delay behind the data. This was attributed to the modeling of steam separator inertia. We agree with General Electric that the overall shape of the core exit pressure response is duplicated well by the ODYN coda. The agreement between the calculated and measured pressures in the dome and the, steamlin. is also reasonably good. There is no conservatism in calculation of pressures up to 1L8 second of transient time.

The measurement of neutron flux indicates a double peak behavior.

This double peak was attribdt*d to an oscillation in core pressure which was thought to be enhanced by KJX4 bypass characteristics. The ODYN code overpredicted the initial neutron flux peak by approximately 53U and underpredicted the second peak. We estimated that the integral of the calculated nuclear power was higher by approximately 2CS than the data. Figures 7 and 8 present the comarisons of measured and calculated axial neutron and prompt neutron fluxes respectively.

The calculated value of ACPR/ZCPR was about 9.2X conservative relative to the value calculated using measured quantities (see Table I1).

11-$5 lxxi

~ .. *

-WOML AXIAL AVE WAGE I-a I

.4 .4 bq S

U' a'

Ilk 0

3-4 Is 3m 4.0 4.0 .S '0 Ma 126 AXIAL UIGV 5IM ritRerm 7 AveraRe Avial Pwmm 0"N Test Coudltiona

'fee

'-ION a

-ft

-Ito TIoIt 1 fliure a KKih Turbine Trip rroupt Neutron iroer

A, Although there is conservatism of 9.2% in ACPR/ICPR, the difference In absolute values is small; i.e., 0.007 in terms of ACPR/ICPR.

Since the value of ICPR was L.279, the value of ACPR is app.roxImately 0.01. According to General Electric the maximum practical accuracy In ACMR is 0.02 (Velma III). In addition, rKJM transient Is a relatively mild transient. Hence, we do not give any credit for conservatism in ACPR/ICPR prediction.

4. Qualification Using Another Computer Code - Audit Calculations Another important means of qualifying a code Is to c"part the results' of calculations with the results obtained from another code. The two codes should be as independent as possible including the neutronic Input parameters. The BNL-TVIGL (Reference 24) and RELAP-38 (Reference 2S) codes are fully capable of analyzing these BWR turbine trip tests and satisfy the requirement of independence. The nuclear data base for deriving the Input for the BNL-lIGL/RELAP-38 codes also satisifies the requiresent of independence.

a. Develo=ment of Calculational Method A calculational method for the analysis of the turbine trip transients was developed at BNL using the RELAP38 and SNL-1WIGL computer codes.

This method was developed under two NRC Technical Assistance Programs supplementing each other, and uses the codes in an iterative manner.

The details of the meftod are presented in Reference 26. The RELAP3B code is used to perform the system transient analysis for the audit calculations. The BNL-TNGL code is used to calculate the reactivity feedbacks and core power transient. The DNL-TWIGL code performs a U-"8 lxxiv

space-time analysis of core neutron4cs and thermal hydraulics with feedback in two dimensions (reference 21).

The BNL-TWIGL code has a number of advantages over the 00YT code.

The calculation can be performed with two neutron energy groups in two-dimensional (r,z) cylindrical geometry. It has the capability of allowing for five radial scram zones. Any important radial effects will, therefore, be calculated by BNL-TWZGL The BML-TWIGL code also.

has two disadvantages relative to the ODY0 coda. These disadvantages art: (1) the lack of a bypass flow channel, and (2) the independence of the Doppler reactivity with void fraction. Weighing these advantages and disadvantages of SNL-TWNGL relative to the ODYN Coda, it is our Judgment that they will not adversely affect the comparison of the two codes for the turbine trip transient discussed herein.

The calculational method was developed using the Peach Bottom tests as a bench mark. Assuming the measured power history (power vs.

time) in the core as Input, RELAP38 calculates the system thermal hydraulic parameters and provides the BNL-TWIGL code with the time dependent core inlet boundary conditions, i.e., pressure, flow and temperature variations with time. Then, the BNL-TWGL code performs the space-time analysis of the core neutronics and thermal-hydraulics.

The calculated power history is then compared with the measured power which was input to the RELAP3B code. If the differences are large the calculated power history Is used in the RELAP3B code and the calculations are repeated until the power history calculated by the BNL-TWIGL code is in good agreement with the power history input to 11-59 1xxv

the RELAP38 coda. This method was used for both the Peach Bottom tests and the licensing audit calculations (turbine trip without bypass transient).

b. Peach Batto Tests and Audit Calculations The RELAP38/BNL-TTIGL calculational method described above was employed by BSL to analyze the Peach Bottom transient tests. The calculated power history agreed well with the measured power history.

There Is also good agreement between the other calculated and measured parameters. The core physics results that were obtained by BNL are presented In Reference 26. Reference 26, also discusses the geometric modeling, the neutron cross sections, and the initial ization of the transients. These calculations confirmed the adequacy of the BNL-TWICL/RELAP-3B modeling for computing MWR turbine trip transienti. Figures 9 through 14 show samples of these agreements.

"Revised BNL" curves in Figures 12 through 14 refer to a more detailed SUL model which will be explained subsequently.

Calculations performed with the ODYN code also agread very well with the experimental results although the neutron flux predictions were slightly conservative. We believe that this can be attributed to the slightly higher 07YN pressure predictions during the transient. The power history calculated by BNL provides better agreement with the experimental data on a best estimate basis than the 0DYN code predictions.

11-60 lxXvi

p~ni otm2rom oTrpTs .nta oe 1.0 1.4 12 0 1.0 toI 0.0

,J.r W4- 0.0 0.4 02 0.0 1.0 F'ractionail Cor-e hleiglit

Ilench Ilottoovi 2 I'm-blivo Trip Test 2, Itilthil flower 1.0 1.4 0

I-'

0 IA)

I~4 I

I.'

I-..

'Ii fr

.e-. 0.0 4-,

~i1 0) 0.6 0.4 0.2 0.3 0.4 0..3 0.0 0.7 I'ractional Core Ileight

Iflem.:Is Iottomn 2 I'lri-ille TipmIj Tesit 3. Initial Power' 1.4 1.0 0

0.0 D4 Cd 0-4 4) m 0.0 0.4 03 0.4 0.5 Oil 0.7 Fractional Corc.- Ifeight

Peach Bottom 2, Turbine Trip, Test 1 300.0 25,0

-W t) 200,0 4J 9:

'a CL v B-.

-a 150.0 1oo.0 CL 50.0 0.0 DAK Ti me(s)

2 Peach Bottom 2, Turbine Trip, Test 2 300.0 250.0 TI 4-.

200.0 150.0 C,

-a

~8 5

Eu

£1 L.

p 3:

0 IL 50.0 0.0 0.0 TI me(s')

Peach Bottom 2, Turbine Trip. Test 3 400.0 350.0 300.0

.01 a

250.0

• 12 200.0 150.0 I

4' 100.0 60.0 0.0 Time(s)

At a meeting an July 14, 1978 attended by GE and our consultants from BHL, a turbine trip without bypass transient (TTOB) was defined. for calculation by GE with the O0YN code and by BNL with the INL-T1WGL/

RELAP-3B codes. This TTNOB transient was for PB-2 at end-of-cycle 2 with the reactor at an all rods out condition and with a Haling core power distribution. The reactor trip was assumed to occur from the primary trip signal for this transient, i.e., the position of the turbine stop or conti-l valve. All of the system input paramters were discussed and values were asslined. The reactor was assumed to be operating at a 104.5 percent of full rated power and at 100 percent of rated core flow.

The Initial calculations by GE and SNL differed considerably. The total core power as a function of time calculated by ONL ýas about 60 percent greater in energy output although the initial rise and falloff of the power was about the same. The BNL calculation predicted a peak power of over 7 times the initial core power at about 0.9 seconds. The GE calculation resulted in a peak power of about 4 times the Initial core power at about 1.0 second.

A GE evaluation of Its calculation resulted In finding two significant errors that led to a new GE calculation. Ohe of the errors was the steamline length. It had originally been input as 460 feet whereas the value should have been 400 feet. GE also found that one of Its processing codes had improperly accounted for the Doppler reactivity feedback variation with void fraction. This new GE calculation resulted in a more severe transient than the earlier 11-67 1ymiti

calculation. The new GE prediction of peak power was about 5.3 times the Initial core power at about 0.9 seconds. This GE calculation had an earlier ritse nd an earlier fall-off of the total core power than the BNL calculation. The IBL calculation still predicted a greater ener, output for the transient by about 20-25 percent. A comparison of the reactivity components showed that the void, scram, Doppler, and not reactivities as a function of time differed significantly between the BNL and new GE calculation. As an examle, the BNL calculatiorl resulted in a prompt critical calculation with a maximum net reactivity of over one dollar. The GE calculation resulted In a maxitmum net reactivity somewhat less than 0.8 dollars.

Since it was our expection that the 8NL and GE calculations would be In better agreement, a meeattng with General Electric at BNL was held to resolve differences between the calculations. This meting was held from September V7 through September 29, 1978 with GE, BNL, and NRC in attendance, Reference 27. The main differences noted between the two calculations are listed below.

Im REMARK L. Relief valves (a) Set Points GE values too large (b) Delay time BNL did not include (c) Bank capacities BNL did not use GE values (d) Time constant for full flow BNL did not include

2. Turbine inlet pressure GE value too large

3. Steam separator modeling (a) Separator 1./A SNL value low (b) Separator mass BNL water inventory too low 11-*8

I'EM (Cant.) REMARK (Cant.)

4. Doppler reactivity feedbackk IML does not Include variation with void fraction

5. Bypass heating effects BNL does not include S. Fuel gap conductance BNL used variable value ( 400-500) whereas GE-used a constant value of 1000 In addition, some of the BNL neutronic data were also Compared to corresponding GE data. The beginning-of-cycle infinite eultip1i1 cation factor (K.) was compared for a number of fuel types both with and without a control blade and as a function of void fraction. Only small differences were noted between the various sets of data. The Initial axial K. distribution was also compared and again only small differeces were noted. It was also noted that the BNL control worth was about 15 percent larger than that of GE. These neutron cross section and 1 data as a function of fuel exposure an void fraction were later provided to the staff by GE for the two dominant fuel types in the PS-2 core (References 28 and 29).

Void reactivity coefficients extracted from the BNL and GE TTWOB calculations indicate that the BNL value at the start of the transient is larger in magnitude by about 14 percent than the GE value. This larger BNL void coefficient is consistent with the lattice physics data that is used. However, the neutron effective void correlation used by General Electric compensates for a large part of this difference (Reference 34).

A reanalysis of the turbine trip without bypass transient was performed by BNL. Both analyses were presented in Reference 33.

11-69 XzzzV

Some sensitivity studies were also performed. Some of the differences noted In the meting did not change the ONL results substantially.

However the difference In the separator inventories and the required renodalizatin around the separators to accommodate proper Inven tories Unside and outside of the separators and inclusion of a separate flow path between the steam dome and outside of the separators (bullwatar), reduced the neutron power and the total energy output in the licensing basis transient CTTWB). The total anergyoutp*u, the integral of the neutron power, predicted by BNL during the transient was less than that In the second calculation performed by General Electric. However, the General Electric calculation still Indicated an earlier rise and an earlier fall-off of the total core power than the ONL calculation. Figure 15 presents these two BHL calculations as well as the General Electric calculation.

The primary reason for the change In the energy outputs as well as*

disagreement in the shape of the neutron power transient was the new inlet core fl*aw calculation by the HL. The core inlet flow in the second SNL calculation was In closer agreement with the second GE calculation In that it exhibited similar oscillatory behavior.

However, there were still some differences in amplitudes. It should be noted that the differences In flow variation during the transient between the two BHL calculations were within 35% of each other.

Judging from the 8NL studies, Reference 33, we conclude that the modeling of separators is significant in predicting the core inlet flow. We also note that the BNL modeling of the separators is still deficient in that the inertia term, L/A, does not depend on the 11-70

BWR Llcensing-basis p=

Transient 000.0 700.0 600.0 500.0

.4-a a1 I.' 400.0 L

300.0 0~

(L 200.0 100,0 0.0 0.0 TI me(s)

quality at the entrance of the separateors. The 0DY0 code contains the modeling of the L/A term which derived itU basis from experimental dat obtained from separators. Hence, we judge that the-ODYN model should predict the core Inlet nlow more accurately than does the BNL model, and that the neutron power transient should also be moar accurately predicted by the 0DYh code.

Reference 33 indicates that previous and revised 8ML models predict almost similar core inlet flow vafiations at the beginning of the Peach Bottom turbine trip tsts. However, there are large differences betwen predictions after 0.8 sec. Since the power peaks occur before 0.8 sec In the Peach Bottom tests, these large differences-in inlet flow predictions do not alter the predictions of neutron powers as illustrated In Figures 22 through 14. However, as shown in the analysis of licensing basis transient, the separator inertia term and its modeling is important in predicting the transient behavior (or amplitude of oscillations) of the core inlet flow.

Reference 33 also indicated that the heat flux predictions In the second BNL calculation were lower for the portion of the transient where highest ACPRs were expected to occur than those calculated by General Electric. This was expected since the integral of neutron flux in the second MHL calculation was smaller than that calculated by General Electric. BNL did not perform ACPR calculations. However, the predictions of heat flux would suggest, and we would expect that the second 8NL calculations would produce less ACPR than that calculated by General Electric.

11-7Z lXXXViii

BNL did nbt report beat fluxes in their first calculations. However, the integral of the neutron flux was almost the same as that calculated by General Electric. Hence, we expect similar severity for ACPR values If ACPR calculations were &ade using the first calculations performed by BNL.

A third analysis of the TTOB transient was performed by BNL using GE calculated values of the core exit pressure and core Inlet flow. The BNL-*WIGL calculation now predicted a transient with similar initial power rise end fall-off characteristics as the GE ODYN calculation although the peak power was higher in the BNL calculation. A sensitivity calculation with a 5 percent change in the void mac tivity feedback resulted in a DNL-TWIGL power transient that comared very well with the corresponding GE results. The change of 5% In void reactivity or the coefficient Is well within the calculated uncertainty of 12% as presented In Section UI.A.6.

These audit calculations established the fact that core inlet flow Is a very sensitive parameter. The core inlet flow measureaents (i.e.,

the jet. pump Ap measurements) In the Peach Bottom tests contained some errors. In Section 11.2.3.a, where qualification of the OYN code using Peach Bottom tests was evaluated, many comparisons between various-parameters (such as pressure and neutron flux) were made and these parameters were always found to be conservative relative to data. However, despite these conservative featus, the calculated values of ACPR cannot be considered as conservative. This was also pointed out In Section I1.1.3.a. Based on the information submitted, 11-73 lxrzxX

it is or judgment that there is some nonconservatism in core inlet flow calculation in the ODYN code over.oaIng all other conservatisms.

We conclude that, although a precise audit of the GE ODYh code was not.,obtained by the BNL-TWIGL/RELAP-38 codes for this licensing basis TTWOB transient, the analyses performed by SHL provided us with valuable insights concerning this transient. We conclude that the primary reason for the disagreement of the BL-TWGL/RELAP-3B TTWOB results with the GE results is due to core inlet flow differences.

Judging from the audit calculations as well as Peach Bottom test results, we also conclude that differences in predictions on the order of Z= in prediction of peak neutron flux can be expected using different computer codes which represent the state of the art.

5. Summary of Code Qualification In summary, we find that the ODYh Is a best estimate code containing models developed from first principles and provides good predictions of existing experimental data. The experimental datea wer obtained from separate effects and integral plant tests. The separate effects tests include core power measurements from various plants and heated tube tests to verify the void fraction model. Integral plant tests were performed at Peach Bottom Unit 2 and KI. Comparison of the test data and calculations indicates that the agreement is within the uncertainties calculated In Section A. We find that the ACPR predictions from the OOYN and SCAT codes are neither conservative nor nonconservative. They predict the available data well.

11-74 Xc

The ODYN-SCAT prediction of the three Peach Bottom transient tests and one WI0 transient test demonstrated a 2a uncertainty of approximately 37% of ACPR/1CPR at a 95% confidence level. We have determined this using x distribution. No credit was given for measurement errors. This results in a 2a ACPR/ICPR uncertainty of 0.068 for a transient which degrades the CPR from an Initial value of 1.30 to the limit of 1.06. Since these tests represent a very limited data base, it is likely that the 2o uncertainty can be reduced significantly by the acquisition of additional test data for comparison to code predictions. Hence, we recommend that additional integral plant tests be performed tU qualify the code with a higher confi dence.

C. EVALUATION OF THE MARGIN The ODYN statistical analysis was performed b~y General Electric at our request in order to provide a quantitive basis for determining if the ODYN licensing basis contains an acceptable level of conservatism. Two quantities were calculated in this analysis: the probability of the expected ACPR exceeding the licensing basis ACPR; and the probability of exceeding the thermal hydraulic design basis (i.e., probability of exceeding 0.2% of fuel in Boiling Transition).

The 00YT Code is intnded to be used to calculate the change in Critical Power Ratio (CPR) during rapid pressurization transienU such as the loss of load and feedw*ter controller failure transients. This information Is used in com bination with the General Electric Thermal Analysis Basis (GETAB) CPR safety limit to establish the operating limit CPR. GETAB is a statistical analysis 11-75 xci

which determines the value of CPR which Corresponds to 0.2% of the fuel rods in Boiling Transition. The GETAB analysis considers the effects of uncertainties on input parameters such as power, coolant temperature and flow as well as uncertainties inthe GEXL correlation.

Uncertainties In the ODYK Code need to be considered since these will affect the probability of exceeding the thermal-hydraulic design basis. One method of accounting for the effects of ODYh code uncertainties is to Include uncer tainties direcly into the GETAB statistical analysis. A second method is to assure that the ODYN licensing calculation gives a sufficiently conservative value of ACPR to assure that the thermal-hydraulic design basis is not exceeded.

GE has chosen to use the second method In demonstrating the acceptability of the ODYN licensing basis.

We have determined that this approach is acceptable in principle. In addition, we have determined that an acceptable level of conservatism for the ODY0 licensing basis corresponds to a S% probability of exceeding the thermal hydraulic design basis.

General Electric has provided statistical analyses of the loss of load (Turbine Trip) and feedwater controller failure transients. These analyses use Monte Carlo calculations to predict ACPR with a second order response surface which simulates ODYN calculations. The input parameters In the response surface are:

initial power; control rod drive (CAD) speed; exposure index (a measure of axial power shape effects): ODYN code uncertainty; and response surface fitting uncertainties.

11-76 Xcii

The response surface was generated through a regression analysis of ACPR calculations performed using input from the ODYN Code. The accuracy of the response surface was tasted by General Electric by comparing the results of ODYN calculations to the results of response surface calculations. The accuracy checks were done for 15 load rejection transients for a BWR/3 EOC-6 and for 15 load rejection transients for a BWR/4 EOC-4. These comparisons showed good agreement between the two methods. In addition, a regressional fitting error was developed from these comparisons and this fitting error was added to the response surface. This error was found to have a range of one standard deviation values of 0.0076 ACPR/ICPR to 0.0126 ACPR/ICPR depending on the plant type, the time in cycle, and the transient of interest. This range of errors is three to four times .smaller than General Electric's estimate of the 0TYN code uncertainty (0.031 ACPR/ICPR) or our estimate of (0.044 ACR/ICPR) see Table 1. This indicates that the response surface is a faithful repro duction of the ODY0calculational results and that the response surface can be used to establish the effect of ODYN code uncertainties on the probability of exceeding the thermal-hydraulic design basis.

The distribution functions of each of the Input variables (initial power, CRD speed, exposure index, and code uncertainty) were reviewed. The uncertainties of the ODYN code are discussed extensively In the code review section of this report and will not be repeated here. The uncertainty on initial power level used by GE was + 2. We requested additional information to substantiate this value and were given extensive information on the various elements in the plant energy balance and the uncertainties associated with each of these elements.*

The elements of the energy balance were checked against the ASHE standard for determining energy output from a nuclear plant, RASME Performance Test Codes, 11-77 Xciii

I Test Code for Nuclear Steam Supply Systems (PTC 32.1-1969).0 In addition, the uncertainty values for each element ware reviewed and found to be reasonable.

We have concluded that the 2 uncertainty (at one standard deviation) is an acceptable value for power measurement uncertainty.

In support of the assumed distribution of CRO speeds, General Electric has provided the results of tests from 13 operating BWRs. The total data base Includes 3,985 individual CRD sc*rs. The information was presented in considerable detail, including the mean values and standard deviations of the times to 5%, 20%, SC% and 91M insertion for various plant types and for full core and partial care scram tests. An extensive and convincing statistical analysis of the data was also presented. Each data set was tested to determine if it could be tested as part of a larger data set; and only those data set which were found to be statistically alike, at a high confidence level, were triated together. Statistical tests were perform"e by General Electric to determine the significance of: variations among BWR designs; variations betowen full core tests and partial core tests; variations among operating plants; variations among scram tests; and variations among individual drives within scram tests. We conclude that these CC scram tests are indicative of past operating wxperience and that the man values and standard deviations of CRD speed can be chosen for the statistical analysis of the OCYN code. However, we cannot conclude that these CRD scram tests will be indicative of future reactor scram speed performance. Zn addition, it is also necessary to demonstrate that the scram characteristics of an individual reactor to be licensed can be rtpresented by the distribution used in the analysis. The scram characteristics of an Individual reactor should belong to the same population. General Electric should provide an assurance or appropriate 11-78 xCiv

modifications to Technical Specifications to demonstrate that the scram characteristics indeed belong to the same population or can be represented by the same distribution. General Electric should also assess the impact of the use of best estimate distributions an providing this assurance.

The transient response to rapid overpressure events is dependent on the core average axial power distribution and axial exposure distribution since these strongly influence both the void and control rod reactivity feedback. General Electric has defined Exposure Index as a measure of the axial exposure distri bution. Exposure Index indicates the extent to which an actual axial exposure distribution differs from the ideal, design axial exposure distribution (Haling distribution). ODYN licensing calcqulations use the Haling distribution as inpu. General Electric proposed to show the conservatism associated with this assumption by establishing that the axial exposure distributions actually encountered during operation are more favorable than the Haling distribution.

This conservatism was quantified as part of the overall ODYN statistical analysis by including Exposure Index as one of the input variables in the response surfaca.

To establish a basis for the expected distribution of Exposure Indices, General Electric presented data from 11 operating reactors at end of cycle conditions and 15 data points for S operating reactors at mid-cycle conditions. Zn response to a request for additional data on observed Exposure Index, General Electric provided 8 additional data points. Because of the limited namber of data points and the large scatter in the data we were led to question the assumption that the'data was normally distributed. The individual data points obtained from General Electric were subjected to the V-tast for normality by 11-79 XcV

the NRC, Applied Statistics Branch. This test indicated that there was not a sufficient reason to reject the assumption that the data were normally distributed. Based on this information, the inclusion of Exposure Index in the statistical analysis as a normally distributed variable is acceptable. As in the case with the'use of measured CRD speeds, the implications of using best estimate values of Exposure Index based on past operating experience and the associated need far assurance and modifications to Technical Specifications to demonstrate continued acceptable performance were not addressed. Since we cannot determine appropriate modifications necessary for demonstration of the conservatism due to inability to operate at Haling power shape for each reactor to be licensed, we find that the use of the variation of power shape from that of a Haling shape in the statistical analysis is not appropriate.

General Electric has performed the statistical analysis using several different sets of assumptions relative to the response surface input parameters. The probability of exceeding the ODYN licensing basis ACPR was calculated for each case.. This corresponds to the probability of exceeding the GETAB CPR safety limit. The probability of exceeding the criteria of O.2% of fuel rods in Boiling Transition was also calculated for some of these cases. Since General Electric has proposed that the safety limit for BWRs be based on the GETAB CPR safety limit, it is appropriate to use the probability of exceeding this value as the basis for accepting the proposed licensing method. As stated previously, we have determined that a IM probability of exceeding the GETAE CPR limit is acceptable. Unless the safety limit for BSRs is redefined by General Electric and reevaluated by the staff, the use of the probability of exceeding O.3% 6f fuel In boiling transition is not an appropriate basis for judging the acceptability of the ODYN licensing basis.

11-80 XCVi

In conclusion, we recomend that General Electric reperform the statistical analysis to demonstrate the appropriateness of the margin to the GETAS limit.

This statistical analysis should not take credit for conservatisa in the Haling powr distribution. It my take cridit for distribution in scram speeds if General Electric demonstrates that the distribution used In the analysis is applicable to the plant specific case. The analysis should also be performed using the code uncertainties as revised by the staff (t 0.068 ACPR/ICPR) which was based on the plant test data. General Electric may wish to convolute additional variables in the statistical analysis If assurance for conservatism

- for eachspecific application is provided.

1I-81 xcvii/xCviii

I11. STAFF,POSITION We stated our position on the ODYN code and Its application in Reference 35. The following Is a statmsnt of that position.

I. ACPR Calculations The analysis for ACPR must be performed in accordance with either approach A or approach B.

A. ACPR Calculations with Margin Penalty This approach is comprised of the three step calculation which follows:

1. Perform ACPR calculations using the ODYN abd the improved SCAT (Reference 36) codes for the transients In Table III and using the input parameters in the manner proposed in pages 3-1 through 3-4 of NEDE-24154-P. The sensitive input parameters are listed in Table IV.

2. Determine ICPR (operating initial critical power ratio) by adding ACPR calculated in step 1 above to the GETAB safety limit. Calculate ACPR/ICPR.

3. Determine the new value of ICPR by adding 0.044 to the value of ACPR/ICPR Calculated In step 2 above. Apply this margin to Chapter 15 analysis of the FSARs submitted for OLs, and CPs and to reloads.

II-i xcix

The Margin of 0.044 is obtained from consideration of uncertainties in components listed in Table 1.

A sample calculation is presented below:

Assume that ACPR calculations using the OYN licensing basis have been performed and the result Is W7PR 0 - .14 where the subscript c refers to calculations.

Calculate .CPR based on the calculations.

1CPRca L06 + .14 z 1.20 where the GETAB limit is 1.06.

A"PR c .14 s-_a r-.. ..

Stec 3 AcPR~

=rP.117 + 0.044 a .161 MC~nW ACPRnew ACPRnew = A ACPRe c . 6 14 .192 1C 1c ICPRnew = 1.06 + .I a 1.25 111-2 C

8. Statistical Approach for Reduction of Marain Penalty General Electric usessed the probability of the ACPR during a limiting transient exceeding the ACPR calculated for the proposed licensing basis translen*t'(NEDE-25254-P response to question 4). The General Electric study demonstrated that this probability, based on operating data over several fuel cycles from a group of plants, is very low. The key parameters In the study ire scram speed, power level, power distribution, and an estimate of ODYN uncertainties. The proposed approach utilizes the conservatism inherent In the statistical deviation of the actual operating conditions from the limiting conditions assumed for the first three parameters in licensing basis calculations to compensate for potential non-conservatisms from the ODYh uncertainties.

The staff has concluded that the use of end-of-cycle power distributions from multi-cycles for several reactors to obtain credit for margin conservatisms relative to Haling power distribution is not appropriate.

There is no usurance that the end-of-cycle power distribution conservatisms obtained from operating reactor history are representative of the eand-of cycle conditions .which will exist for the specific core. We have also concluded that scram speed data used In the GE statistical assessment must be proved applicable to specific license and reload.applications. In order to take credit for conservatisa in the scram speed performance for reloads, it must be demonstrated that there is Insufficient reason to reject the plant-specific sc.m speed as being within the distribution assumed in the statistical analysis. For CP and OL, the scram speed 111-3 ci

distribution for the specific plant must be demonstrated consistent with those used in the statistical approach. Similar design and prototypic performance characteristics coupled with appropriate technical specifications on scram speed performance could provide acceptable evidence of the applicability of the data base.

Statistical convolution of the power masurement uncertainties to take credit for full power operation at a power level value below that used in licensing calculations is acceptable to the staff. However, plant specific procedures to operate within the licensing limit must be taken Into account in these calculations.

The code uncertainty penalty (0.044 in ACPR/ICPR) applied to the licensing calculations described in (A) does not account for unknown contributors.

Past experience has shown that additional margin in safety calculations is often needed to compensate for unknown non-conservatism in licensing calculations due to code errors or other factors. The ODYN prediction of three Peach Bottom transient test.s and one KKt transient test demonstrated a Zn uncertainty of approximately 37' of ACPR./ICPR at a 9S% confidence level. This was determined using x% distribution. No credit was given for measurement errors. This results in a Za ACPR/1CPR uncertainty of 0.068 for a transient which degrades the CPR from an initial value of L.30 to the limit of 1.06. Since these tests represent a very limited data base, it is likely that the 2c uncertainty can be reduced significantly by the acquisition of additional test data for comparison to code predic tions. Therefore, the magnitude of the code uncertainty used in the statistical convolution may be reduced to a value consistent with the 2n 111-4 Cii

value of ACPR/ICPR uncertainty at a, 95% confidence level when such a redazction can be Justified by additional transient test data.

In s*mmary, the staff has concluded that the statistical approach to compensata for potential non-conversatisms from the ODYh uncertainties is acceptable vith t.he following limitations.

1. Power distribution conservatisas should be excluded.

• . Scram speed conservatisms must be demonstrated to be applicable to plant specific cases.

3. Calculations should be performed using a code uncertainty value which is 37% of the ACPR/ICPR for a limiting transient to account for code uncertainties, including unknown contributors (e.g., code errors),

based on the approved transient test data base. This results in a value of t 0.068 In ACPR/ICPR uncertainty for a transient extending over a CPR range of 1.30 to 1. 06.

4. The transient test data base must be expanded and submitted for staff review to justify any reduction in the value of ODYh Code uncertainty (2= value of hCPR/ICPR at a 951 confidence level).

S. A now statistical analysis conforming with these limitations must be provided.

• -'u-.

ciii

An acceptable licensing basis using the Option 8 statistical approach is a 95/95 ACPR/ICPR for the limiting event. This can be established in one of two ways:

a. Option B can be applied on a plant-specific basis - i.e., statistical analyses performed on a particular plant to determine its 95/95 ACRR/ICPR. The statistical analysis procedures to be used art those defined in the ODYN Licensing Topical Report (LTR), Volume 3, except for the modifications required by the NRC in Reference 35.

b. Option B can be applied on a generic basis. This Involves the establish ment of generic ACPR/ICPR adjustment factors for groupings of similar-type plants (the groupings used in the OMYN LTR are considered to be an acceptable matrix) which can then be applied to the plant-specific ACPR/ICPR calculations from the ODYN LTR deterministic approach to derive the estimated 95/95 values. Each plant group and transient type correction factor is based on an analysis of a typical plant in that group (e.g., BWR 2/3, 4/5, and 6), in which the differences between the 95/95 ACPR/iCPR calculated per the ODYN LTR deterministic approach is detaerined for a specific transient (e.g., load rejection without bypass). The difference, which may be positive or negative, is designated the plant group adjustment factor for that transient.

The generic ACPR/ICPR adjustment factors established for the various plant groupings must be submitted to the NRC for review.

1ll-6 Civ

11. PRESSURE CALCULATIONS Calculations should be performed for the Main Steam Isolation Value closure event with position switch scram failure using the values listed in Table I* as per staff evaluation to arrive at the overall coda uncertainty in pressure calculation. Add this uncertainty to the ODYN calculated pressure for this event in OL, CP and reload applications. If General Electric can demonstrate that this uncertainty is.very small (e.g., by a factor of 10 or more) relative to the bias in determining ASME Vessel Overpressure limit, no addition of uncertainty to the calculations of pressure Is needed.

We note that there is an error in Enclosure 2 of Reference 35. The bounding values of the drift flux parameters should have been in conformance with Table I as per staff evaluation.

111-7 cv

TABLE III TRANSIENTS TO BE ANALYZED USING THE ODYN CODE A. For Thermal Limit Evaluation Thermally Limiting or Near Limiting Event (Typical ly)

1. Feedwater Controller Failure o x MaI*mu Demand

2. Pressure Regulator Failure - Closed

3. Generator Load Rejection x

4. Turbine Trip x

5. Main Steamline Isolation. Valve Closures

6. Loss of Condenser Vacuum

7. Loss of Auxiliary Power x All Grid Connections B. For ASME Vessel Overaressure Protection Pressure Limitina

1. MSIV Closure vithi Position Svwtct Scram Failure (i.e.. MSIV Flux Scram) x "I1-8 cvi

TABLE IV INPUT PARAMETERS SENSITIVE FOR THE ANALYSES

1. CRD scram speed - at tahenical specification limit.

2. Scram setpoints. at technical specification limits.

3. Protection-system logic delays - at equipment specification limits.

4. Relief valve capacities - minimum specified.

S. Relief valve setpoints and response - all valves at specified upper limits of setpoints and slowest specified response.

6. Pressure drop fro, vessel to relief valves - maximum value.

.7. Stemuline and vessel geometry - plant-unique values.

8. Initial power and steam flow - maximum plant capability.

9. Initial pressure and core flow - design values at maximum plant capability.

10. Core exposure/power distribution - consistent with Haling mode of operation.

11. Feedwater conditions - axium temperature (maximum core average void content).

111-9 cvii

III. Oter Limitations

1. Listing of important input variables such as listed in Table IV and initial plant parameters Including but not limited to control system characteristics as depicted in Figures 4-13 through 4-16 of NEDO-24154, Vol. 1, but with numerical values provided should be provided with each submittal. The initial control system characteristics, including the modal used In the selection of initial settings, shall be defined and substantiated in terms of the design basis for each control system of the plant. We .understand that neutronic parameters which were originally obtained from the GE 3-0 Core Simulator and collapsed to provide input to the OOYN code, are best estimate. If there is a significant change In this calculatlonal method altering the input palrmeters, General Electric should submit the new procedure to NRC for its evaluation. The code uncertainty value of t 0.068 ACPR/ICPR based on the Peach Bottom and KKM test data includes uncertainties in this calculatlonal method since this method was used in comparison of test data with code predictions. Hence, any significant change in this procedure will change the code uncertainty.

2. A minimum of eight nodes should be used to represent the steam line.

However, the maximum length of any node should not be more than 100 ft.

3. The code cannot predict accurately core inlet flow oscillations with frequencies above 5 Hz. Although we do not expect any inlet flow oscilla tion above frequency of 5 Hz for the transients listed in Table 111, General Electric should verify that the harmonic coaponents above 5 Hz are indeed very small if very rapid variations of flow in these transients are predicted.

111-l0 cviii

4. The transients listed in Table III are short term licensing transients.

If the code is intended to be used for long tarm transients or different types of overpressurization transients such as ATVS, appropriate modifications should be made.

111-32 cix

References

1. NEDE-24011-P, "Generic Reload Fuel Application," Chapter 4, Hydraulic Model Description, May 1977.

2. NEDO-20953, "Three-Dimensional SWR Core Simulator," J. A. Wooley, May 1976.

3. NEDO-20566, "General Electric Company Analytical Model for Loss-of-Coolant Analysis In Accordance with 10 CFR 50, Appendix K, Volume Z," January 1976.

4. NRC letter from D. 9. Elsenhut to R. Gridley, dated May 12, 1978.

5. NEDO-10802, Amendments 10802-01 and 02, "Analytical Methods of Plant Transient Evaluations of the General Electric Boiling Water Reictor,'

R. 8. Linford, February 1973.

6. NUREG-0460, V. 11, Anticipated Transients Without Scram for Light Water Reactors, April 1978.

7. NEDO-10299, MCore Flow Distribution in Modern Boiling Water Reactor as Measured In Monticello," January 1971.

S. APED-4762, uPerformance Tests of Axial Flow Primary Steam Separators,"

C. H. Robbins, January 1965.

9. NEDD013388, "2-Phase Flow in Boiling Water Nuclear Reactors," R. T. Lahey, July 1974.

10. Isbin, H. S., Sher, N. C., Eddy, K. C., 'Void Fractions In Two-Phase Steam-Water Flow," AlChE Journal, Vol. 3, No. 1, 136-142, March, 1957.

1I. Isbin, H. S., Rodriquez, H. A., Larson, H. C., and Patti.e, B. 0., "Void Fractions in Two-Phase Flow," AIChE Journal, Vol. 5, No. 4, 427-432, Dec.

1959.

12. Marchaterre, J1 F., ,The Effect of Pressure on Boiling Density in Multiple Rectangular Channels,N AL-o522, Feb., 1956.

13. Janssen, E., Kesvinen, J. A., "Two-Phase Pressure Drop in Straight Pipes and Channels; Water-Steam Mixtures at 600 to 1400 psia," GEAP-4636, May 1964.

14. Cook, V. H., NBoiling Density in Vertical Rectangular Multichannel Sections with Natural Circulationu AHLo5621, Nov. 1956.

15. Ishii, N., "One-Dimensional Drift-Flux Model and Constitutive Equations for Relative Motion Between Phases in Various Two-Phase Flow Regimes,"

AHL-77-47, October 1977.

111-12 CX

16. Frigg Loop Project, Frigg-2, AB Atominerg1, Stockholm, Sweden, 1968.

17. Rouhani, S. Z.8 "Void Measureents in the Region of Subcooled and Low Quality Boiling," Symposium on Two-Phase Flow, University of Exeter, Devon, England, June 1965.

1. buhani, S. . , "Void Measurements in the Region of Subcooled and Low-Quality Boiling," Park 11, AE-RTL-788, Aktiebolaget Atomenergi, Studsvik, Sweden, April, 1966.

19. MEDO-20913-P, "Lattice Physics Methods,' C. L. Martin, June 1976.

20. "GEGAP-III: A Model for the Prediction of Pellet-Cladding Thermal Conductance in BWR Fuel Rods,* General Electric Report NEDC-20181, November, 1973.

21 0. F. Ross (NRC) letter to E. D. Fuller (GE) dated June 2, 2978.

22. EPRI NP-564, "Transient and Stability Tests at Peach Bottom Atomic Power Station Unit 2 at End of Cycle 2," L A. Carmichael and R. 0. NMiemi, June 1978.

23. Letter from K. V. Cook to R. L. Tedesco, MFN 3137-78, "Transmittal of Responses to Round 2 Questions on the OYN Transient Model," datead Dec. 13, 1978.

24. BNL-NUREG-2.925, "BNL-TWIGL, A Program for Calculating Rapid LWR Core Transi*ents," 0. J. Diamond, Ed., October 1976.

25. BHL-NURE;-22011, "User's Manual for RELAP-3B-MOD 120: A Reactor System Transient Code," 1977.

26. BNL-NUREG-24903, *Core Analysis of Peach B.ttom-2 Turbine Trip Tests,"

H. S. Cheng, 0. J. Diamond, September 1978.

27. Memo from F. Odar to Z. R. Rosztoczy, "Meeting with General Electric at Brookhaven National Laboratory,$ October 17, 1978.

28. Letter from K. W. Cook of GE to Frank Schroeder of NRC, October 10, 1978 on Poetntial Differences between GE and BNL Models.

29. Letter from K. W. Cook of GE to Frank Schroeder of NRC, October 26, 1978, on Transmittal of Exposure Dependent Data.

30. Transcript of ACRS hearings held on March 19-20, 1979, Los Angeles, California.

31. Letter from K. W. Cook of GE to R. L. Tedesco, Clarification of ODYN Model Uncertainties, MFN 323-79, dated April 30, 1979.

32. Letter from K. W. Cook of GE to R. P. Denise, Additional Void Fraction Information Recquested for ODYN Review, MFN-2ZL-79, August 27, 1979.

111-13 Cxi

33. BNL-NUREG-26684, "Analysus of Licensing Basis Transients for a BWR/4,"

M. S. Lu, H. S. Cheng, W. G. Shier, 0. J. Dimaond, M. H. Levine, September 1979.

34. BUL-HUREG-23501, *A Space-Time Analysis of Void Reactivity Feedback in Boiling Water Reactors," M. S. Cheng, M. S. Lu, D. J. Diamond, October 1977.

35. Latter from R. P. Denise (NRC) to G. G. Sherwood (General Electric), dated January 23, 1980.

36. Letter from K. W. Cook (General Electric to F. Schroeder and 0. G. Elsenhut (NRC), NFN 171-79, "Implementation of a Revised Procedure for Calculating Hot Channel Transient ACPR," July 20, 1979.

111-14 cxii

Supplemental Safety Evaluation For The CencralElectric appical Mepct Qualification Of The One-Dimensional Core Transient Hodel For Boiling Water Reactors NEDO-241 S4 and NEDE-241 S4F Volumes I, 11 and III Prepared By Reactor Systems Branch, MSI cxiMi/cxiv

mThe Safety Evaluation Report on the ODT code (Reference 2) is primarily an evaluation of the calculational model with little discussion of implementation requirements. Reference 3 provides the information required to bridge the gap between evaluation and implementation. Specifically, there are eight Items covered In Reference 3; these are:

1. ODYN Option B statistical adjustment factors.

2. Control rod drive scram insertion time conformance procedure for plants licensed under ODY0Option B,

3. Uncertainty in ODYN pressure calculations,

4. ODYN model temperature limits, S. Uncertainty in subcooled boiling model,

6. Description of electronic hydraulic control model,

7. Listing of ODT input variables, S. Comparison of minimum critical power ratio operating limits established by REDY and ODY0.

Each of these items is discussed below.

Item 1. Statistical Adjustment Factors Page 111-6 of Reference 2 allows tw statistical approaches; one Is a plant specific statistical analysis and the other is a generic analysis for plant roups (e.g. BWR/2, 3, 4, 5, 6) and transients. The second approach involves the establishment of generic aCPR/1CPR adjustment factors for groupings of similar-type plants which can be applied to plant-specific ACPR/I*CPR calculation from the ODYN licensing topical report (LTR) deterministic approach.

Reference 3 provides the statistical adjustment factors for the three transients which are normally limiting transients (load rejection or turbine trip without CXV

'bypass, feedwater controller failure to maximum demand and pressure regulator downscale failure). These generic statistical adjustment factorsare shown in Table 1; we find them to be acceptable.

Item 2. CRD Scram Insertion Time Conformance Procedure Page I11-3 of Reference 2 states *In order to take credit for c=nservatism in the scram speed performance for reloads,-it must be demonstrated that there is insufficient reason to reject the plant-specific scram speed as being within the distribution assumed in the statistical analysis. For CP and OL, the scram speed distribution for the specific plant must be demonstrated consistent with those used In the statistical approach."

GtGera&2 rlectrlc presents the fol©vizng procedure as one vhich astisfles te Staff's objectives for scram coforausae. It. shobld be noted that some utilities uhing 0MT Option $ W7 desire to erabliash their ova conformance procedures.

The procedure consists of teasing, at the S significance lefte, the scra surveillance data a: the 20: Insartioz position which is generated several tines each cycle as required In the taactivit7 Control System TecbhIcsl Specification (20 insertion is tePresentastI. of that portion of the OcTa&

most affecting the pressurisa:tion transient). The uique rod notch position closest to 20: (and the appropriately adjusted tine of Insertion) is expectalt to be utilized In actual plant appillcasion of this generic Concept. l©o" m*os plants, the surveillauce requirements are as follws:

(1) al Control rods are easured at beg*nnftg of c*cle (Wc), and (2) Z of cotrol rod are measured every 20 days during cycle MX io planstdepandent and ranges from 10 to 50),

c*Xvi

the completlon of each surveillance test lerforfed In compliance with the At technical epeciflcateot surveillance requirmemts, the average value of an uurveillance data at the 201 Insertion position generated to the cycle to data Ls to be tested at the 5Z silgiflcance level against the dlstribu:ion assmed In the CM analyses. The surveillance Inforaation vh1C*ocb Phlazt using this procedure vwil 1ave to retain throutiout the fuel cy*a Lasthe Viabor of active control rods ueaesurd for each ourveillance test (the first tust LI at tba D=C ad Ls dezoted %: the M test Is dteoteod I) d the average cram time to the 20: Insertion position for the active rods measured In test I( 1).T he equatloc used to calculate the overall average of all the scrae data generated to date In the cycle is:

C2-1)

Sol

'tav where a ai*ber of surveillance tests perforsed to date Sz the cycle;

- total number of active rods ueesared to date In the cycle; and 1

N" , out of the Poras tines to the 2V&iseartion position of l lei active rods measured to date in the cycle to cmply vith the Technical Specification survaeliancae requirements.

CxVii

The average scra** t.e, tis, is tested against the analysis msa. using the following equzation:

,e. 3 (2-2) where S*4. C.2-3)

The paraeters ivande are the uan and standard deviation of the distribution for average scram insertion time to the 2= position used In the CM option I

.... , e cycl. averaSe scran im satisfies the Equation 2-2 crterbib, onntinuee plant. operation. under the OMT Option 3 operating limit ziniz= critical pove:

ratlo (OLC*OR) for pressurization events is permitted. If not, the OLUCR for pressuriza:ion events oust be re-established, based an a Linear LInerpolation between the Option I and Option A CIMs. The equation to establish the new operating imit for pressurization events ia given beWow:

what*

Tmand T3are defined In Lquations 2-1 and 2-3, respectively; IA *the present technical specification limit on core average scram time to the 20: Insertion position; and 6MV the difference bietween the MLAOR calculated using Option A and that =sIng Option I for prussurizat ion events.

cxViii

goer that *quactio 2-3, which establishes the W-10 a vable cram *1ar tion tim for operation urder Option 3, say also be epressed It the fullovin;

• Its AV (2-5)

A 1.65 51 1/ (2-6)

The relaticuonbip Ueten the coefficient, A, and the aunt of eurveillance data Senerated during the cycle is illus:rated In fCrte 2-1. As sore data beqoe available through the performance of in-cycle surveillanace-:sts, the coe.fflcent decreases, as does the acceptance criterion, is" Thus, the ecram steed criter-lo Is %eing tightened as the cycle progresses, based on the assuuptlo* that, as acre scraz data hbeco*e available during the cycle, the mcertainty in the meat value calculatiot should decrease.

We find the scram Insertion time conformance procedure to be acceptable.

Ituem 3. Uncertainty in ODYR Pressure Calculations Page 111-7 of Reference 2 states that if GE can demonstrate that the uncertainty in calculated pressure is small (e.g. bi a factor of 10 or mare) relative to the bkias in determining AM vessel averPreasure limit, no addition of uncertainty to the calculation's of pressure Is needed. A sensitivity study varying ODYN Input parameters over the range of Table 1 of Reference 2 shows the RMS uncertainty in the peak vessel pressure to be 11 psi. GE estimates the bias In the AS1E code to account for the material uncertinty to be approximately 310 psi. Therefore, there is no need to account for pressure uncertainty in cxix

.. the ODYN calculations.

Item 4. ODYN Model Temperature Limit An early draft of the ODYN SER limited the code calculation to fuel temperatures less than lSOO5K (ipproximately 2240 0 F). This was because Figure 8-2 of Reference 1 limited the thermal conductivity of U02 to 1SO0K. The actual equation used in 0OYN, 38.24 13 3 K' 57.T÷T + 6.07123 x 10 (T+273) where T a fuel temperature (CC) and K a thermal conductivity (watts/CmOC) is based on data which extended to the U02 melting temperature (3080 0 K).

Therefore, the fuel temperature limit for ODYN0analyses is the U02 melting temperature (30800K or SlOO5F).

Item S. Uncertainty in Subcooled Boiling Model Page 11-19 of Reference 2 states "We estimate the corresponding minimum and maximum values of 'n' to be 0.5 and 2.0 respectively. General Electric is required to make sensitivity studies to verify that these values correspond to 0.023 uncertainty In ACPR/ICPR." GE analyzed the turbine trip without bypass transient for n a 2.0, as requested for a 251 BWR/4. The peak core average beat flux (S rated) increased from 121.6% (for n a 1.0) to 124.0 (for n a 2.0). This leads to a ACPRIICPR sensitivity of about 0.024. It is concluded that the Staff's estimate of 40.023 for n values between O.S and 2.0 is valid.

CXX

Item 6. Description of Electronic Hydraulic Control Model An early draft of Reference 2 stated 'Wherein electronic hydraulic controls are used in the design, the model used in selection of initial control setting shall be submitted for staff review." This statement was made because Reference I provided information only for the mechanical hydraulic control.

GE claims that there is no functional difference between the two types of control. However, they provided a description of the model in Reference 3.

We agree with the GE claim that there Is no functional difference between the two types of contral.

Item 7. Listing of ODYN Input Variables Page IZI-1O of Reference 2 states "Listing of important input variables such as listed In Table IV and initial plant parameters Including but not limited to control system characteristics as depicted In Figures 4-13 through 4-16 of NEDO-24154, vol. 1, but with numerical values provided should be provided with each submittal. The initial control system characteristics, Including the model used in the selection of initial settings, shall be defined and substantiated in terms of the design basis for each centrol system of the plant.' Item 7 of Reference 3 lists typical values of these initial parameters which may be Included by reference into individual plant submittals provided the values are appropriate to the individual submittals.

item 8. Comoarison of MCPR Overatina Limits Established by REDY and ODYN The Staff requested GE to provide a comparison of CPR operating limits based on REDY and ODYN prediction. The purpose of such a comparison was to cxxi

-evaluate the appropriateness of continued plant operation under the current REDY based operating limits during the transition period In which 00Th Is imple mented for rapid pressurization events.

In addition, the Staff Indicated that the initial ODYN analysis for each BWR operating plant must include all the pressurization events identified in Table 2-1, Volume 3 of Reference 1, unless justification could be provided that fewer events (such as the limiting events presently analyzed for reload sub mittals) would be sufficient.

Table 2 shows ODYN and REDY-based CPR operating limits for the limiting pressurization events (load rejection without bypass and feedwater controller failure-maximum demand) for plants in which both ODYN and REDY calculations are available. Two sets of ODY0numbers are provided: ODY0deterministic calcu lations per the GE letter (Reference 1), labelled "ODYN-GE LTR" in Table 2; and ODYN Option B statistical calculations, labelled "ODYN Option BI in Table

2. Also Included in the table are the plant minimum operating limits, based on all the abnormal events, when using REDY, ODYN GE LTR, and ODYN Option B to calculate the rapid pressurization events.

Because the overall plant operating limits are, in all cases, either uneffected or improved. £Econcludes that implementation of ODYN will not represent a significant change to the operating limits for BWR plants. For those plants. which use ODYN Option B, It is generally expected to either produce no change t the limit or else to Improve It slightly. We agree with this conclusion.

cXxii

The events for which CM has ben quallfied and approved are lLs:ed In Reference 1, Tolre 3, and Isclude the faelovin: (1) fee"dater controller falluresax-iu= demand; (2) pressure regulator fallure-closed directlio; (3) gen erator load rejection with and without bypass operation; (4) main steauline Isolation valve closure (trIp.scran and flux scrai); ($) lost of condenser vacv=; (6) tvrbine trIp vith and without bypass; and (7) loss of auzil .ry pover - all grid connections. CZ proposes that Only the following three events be reported for reload Submittals or safey mnal7sis report revtilsos:

generator load reject:ion/turbie trip without bypass (whichever Is lLim:ing).

fe&dvater controller fallure-zaxiui demand, and main stealine Isolation ialve closure-flux scr-a (to satisfy ADM code pressure requirements). 'hese are the a=* pressur+/-:ation events presently Included In 1*eald subuittals, end reflct?

the ¢onistency in the 0 and 3=T retults. The events sot included In the submittal are am& ls" &tre, Lto tM naams diwcuased Utlm.

1) Turbine/Generator Trips With Bypass These events are aonsiderstl7 less severe thar the transien.s In which the bypass sys:te is assumed to fall. Typical turbine bypass capacities range froz 23-4C: of rate steamflow. This bypass capacit7 results tn a consid era&bl7 mlder therula and overpressurization event.

2 Pressure Regulator Failure - Closed ,[rection The standard event evaluated in M analysis to one In vihic the controlling pressure regulator Is assamed to fail In the closed direction. Under these failure conditions, the backup regulator takes over control Of the turbine admission valves, preventing any serious transient. The disturbance is mild and similar to A pressure set point cha&gt Vith nO Significant reductions of fuel therual margins occurring. As shown It the SAjts, this event Is considerably less Severe than the generator and turbine trips without bypass.

cxxiii

-10 -

. Ioss of Condenser VEacuum L)

Various syst:: malfuctiazs can cause a less of coiden*se vacuum due to some single equipment failurs. The reduction or less of vacuum In the ain turbine condenser vii.,sequiotially trip the saen and feedvater turbines and bypass sys te- and, for scue plants, close the sain stema.nlne isolation valves. While these are the major events occurr*ig, othe: resultant actions Viii Include scram (frbm stop valve closure) mid bypass opening vith the salt turbine tri;.

5ecause the protective *cti*ns are actuated at various levels of condenser vscuum, the severity of the resultin transient Is directly dependent upon the rate at vhich the vacvu pressure Is lost. lormal loss of vacuum due to loss of cooling vSter pums or steam et air e*jectar problem products a very slov rate of loss of vacuum (imnutes. not seconds). If corrective actions by the reactor operators are act successful, thea simultaneous trips of the uui an feedvater tVrbines, and ultimately eoMlete isolation by closing the bypass ralves (opened with the main turbine trip)and the HlVs. vwill occur. This event is bounded by the turbine trip without bypass event.

4) Loss of Auxiliary Power . ATlGrid Connections thle'event Is Initiated by a generator load Tejection. Since the turbine bypass aysten Is assimed to operate during the initial portion of this eve..:.

It Is cow*araMe to the losa rejection with bypass and Is Considerably less severe than the withort bypass events.

5) MIV Closure - Trip Scram This event has a slower shutoff of ateaz flow than the turbine tri; vtthav:

byP&ss event. Thertfore, the transient Is not as severs. This bas been con ftrmed b OIr calculiatims.

cxxiv

- 11 -

The staff agrees with the SE assessment of the relative severity of the transients listed. Therefore, the following events should be reanalyzed with ODYN for plants which have analyses of record using REDY:

1) generator load rejection/turbine trip without bypass

2) feedwater controller failure maximum demand

3) main steam line isolation valve closure-flux scram.

If for a particular plant another event should be more limiting than those just listed, then the other event should also be reanalyzed with ODY0. For the new plants with transient analyses supplied by GE, all of the events listed in Table 3 of Reference 1 should be analyzed with ODYN.

CxXv

Re ferences

1. NEDO-24154 and KEDE-24154P, Volumes 1, 1I and III. qualification of the One-Dimensional Core Transient Model for Boiling Water Reactors,"

October, 1978.

2. NPerandum for T. Novak and R. Tedesco from P. S. Check, "Safety Evaluation for 4uali ficatton- of-the--One-Dimensilnal Core -Transient Model for Boiling Water Reactors," REDO-24154 and NEDE-24154P, Volumes 1.11 and I11,0 October 22. 1980.

3. Letter to P. S. Check (NRC) from . H. Buchholz (GE), 1FN-155-80, OResponse to NRC Request for Information on 0DYN Computer odel ,"

September 5, 1980.

Table 1.

SZOfrA. 07CWDL C SZA7ZS74AL ADJDISr%&*7 ACOWS (CAPL/ZCPR)

Groupicts LR/TTWOSPO'T I SU-1 2/3 - ZOC 40.006 -0.016 ma'./5 ,/o iaRnl - zOC -0.039 -0.009

]R,'A/5 v/o In - - 0.111 -..0009 su 415 vwi - zoC -0.02,4 0.016 513Z 1/5 v/ITM - M -0.001 +0.026 U11_6 -_ 'C -0.022 40.003 4-0.017

• With the exception of FW or PRDF events, this set of adjusttent factors will be applied to all pressuriuation events analyzed vith the orn code to establish the CPR operating lizLt, since they tgypically involve generator or turbine trips.

ecxxvii

Table2 CORWARl!1O or R"ITIOMNW OPSINTS!

Operating CPR Limits

.. L"10o BP - vaP Plant IInfutu OM 01151 (CE (C.F (Ortion" ( fion I(op ion Plant .urOT Lt r) K IITr itt) Or"P I.tr) R)

M/h3-205 rOC 1.4S 1.36 1.21 1.41 1.36 .146 1.36 A.3S h.k MM14-1896 Ioc 1.42 1332 1.27 1.31 1.25 it) 1.42 1.33 1.27 M14-2801 RC4 1.1a 1.20 1.11 1.,1 LIZ 11202) 1.21 1.21(2)

M'IG-S231 forC 1.16 1110 1,13 1,20(l) 1.08 1.11 1,20) 1,20()

PateIe (3) Limited by Rod UItldrvaI Frror (Z) Limited by Lane of IOO' Feedveter fleeting (3) Plant with RPT

m L

mcclc AM6TCPCAM MOIRTION "TA Of POVIi INCV=L 1+/-S~a 2-1. 5: Significance Coefficlent (A)vs. Survueillance* Data in Cycle cxxix

GENERAL ELECTRIC NUCLEAR POWER SYSTEMS DIVISION GENERAL ELECTRIC COMPANY. 175 CURTNER AVE.. SAN JOSE. CALIFCRNIA 95125 1 . MC 682, (408) 925-5722 MFN-013-81 January 19, 1981 U.S. Nuclear Regulatory Commission Division of Systems Integration Office of Nuclear Reactor Regulation Washington, DC 20555 Attention: Paul S. Check, Assistant Director of Plant Systems Gentlemen:

SUBJECT:

ODYN ADJUSTMENT METHODS FOR DETERMINATION OF OPERATING LIMITS

References:

1) Telecon, H. C. Pfefferlen, R. E. Engel and R. T.

Hill (GE) with M. W. Hodges (NRC), January 5, 1981

2) R. P. Denise to G. G. Sherwood, January 23, 1980 The purpose of this letter is to document the agreement reached in the Reference 1 telecon concerning the application of the Initial Critical Power Ratio (ICPR) adjustment factors in conjuction with ODYN analysis results.

Option A The ICPR adjustment method to be used with the non-statistical approach (Option A) for rapid pressurization events is that given in Step 3 of Reference 2. This method is to be considered a defini tion for use specifically with the NRC derived adjustment factor of 0.044.

The application of the adjustment factor in this case can be defined as:

ICPRnew = 1.044 (ICPR) c where (ICPR)c = calculated value of ICPR Option B The method to be used with the statistical approach (Option B) has been defined to be consistent with the adjustment factors derived by General Electric. This method requires the addition of the Cx.oc

CENERAL 0 ELECTRIC U.S. Nuclear Regulatory Commission Page 2 adjustment factor (AF) to the ratio of the calculated values of ACPR and ICPR (ACPR/ICPR) c:

CPRnew = (CPR + AF TMR~e TMF c This equation can be simplified to:

ICPRne = ^SL C- E CR)

+ AF]

Where SL = Safety Limit MCPR It should be noted that in both the Option A and Option B cases, the ICPR is defined as Safety Limit plus ACPR for the event being analyzed.

If you have any additional questions or comments, please contact me or H. C. Pfefferlen on (408) 925-3392 of my staff.

Very truly yours, R. H. Buchholz, Manager BWR Systems Licensing Safety and Licensing Operation RHB:sem/1158-59 IF cc: L. S. Gifford M. W. Hodges C=ixl/cxxxii

NEDO-24154-A PREFACE This document represents a compilation of the information provided to the Nuclear Regulatory Comission by the General Electric Company for the staff's review of GE's One-Dimensional Core Transient Model.

Volume I of this report contains the One-Dimensional Core Transient Model description, as well as the questions/respouses associated with the staff review of this description document.

Volume II of this report contains the document for Qualification of the One-Dimensional Core Transient Model for Boiling Water Reactors, as well as the associated staff questions/responses.

Volume III of this report contains two parts: Part I provides the General Electric proposal for application of the One-Dimensional Core Transient Model, as well as the staff questions/responses related to the proposal; Part II provides two questions/responses related to the model description which involve information propri etary to the General Electric Company.

The questions referenced within this report were transmitted to General Electric by letter from the NRC staff: untitled letter, D. F. Ross to E. D. Fuller, dated June 2, 1978.

cxxxiii/cxxxiv

NEDO-24154-A TABLE OF CONTENTS Page ABSTRACT cxli

1. INTRODUCTION AND

SUMMARY

1-1

2. MODEL-MODEL COMPARISONS 2-1 2.1 Nuclear Model Comparisons 2-1 2.1.1 Scram Reactivity 2-1 2.1.2 Void Coefficient 2-2 2.2 Thermal-Hydraulic Model Comparisons 2-3

3. ANALYSIS OF TURBINE TRIP EXPERIMENTS 3-1 3.1 Peach Bottom Turbine Trips 3-1 3.1.1 Test Summary 3-1 3.1.2 Model Inputs 3-2 3.1.3 Data Comparisons 3-3 3.1.3.1 Transient Pressure 3-3 3.1.3.2 Neutron Flux 3-4 3.1.3.3 Critical Power Ratio 3-7 3.1.4 Conclusions 3-8 3.2 KIM Turbine Trip 3-8 3.2.1 Test Summary 3-9 3.2.2 Special Model Considerations 3-10 3.2.3 Model Inputs 3-10 3.2.4 Data Comparisons 3-11 3.2.4.1 Transient Pressure 3-li 3.2.4.2 Neutron Flux 3-13 3.2.4.3 Critical Power Ratio 3-15 3.2.5 Conclusions 3-15 REFERENCES 3-74 APPENDICES A RESPONSE TO NUCLEAR REGULATORY COK4ISSION QUESTIONS ON SECTIONS 1 THROUGa 3 OF VOLtThW II A-i cxxxv/cxxxvi

NEDO-24154-A LIST OF ILLUSTRATIONS Figure Title Page 2-1 Plant A (BWR/4) All-Rods-Out Scram 2-6 2-2 Plant A All-Rods-Out Scram Reactivity 2-7 2-3 Plant A Initial Control Rod Pattern in Nodes Withdrawn (24 is Completely Out) 2-8 2-4 Piant A Critical Rod Pattern Scram Reactivity 2-9 2-5 Plant B Rod Pattern in Number of Nodes Withdrawn (24 is Completely Out) 2-10 2-6 Plant B (BWR/4) Scram Reactivity 2-11 2-7 Plant B Transient Axial Power 2-12 2-8 Steady-State Void Fraction Profile High Power Channel 2-13 2-9 Steady-State Void Fraction Profile Low Power Channel 2-14 2-10 Change in Void Fraction from 10 Psi Pressure Change High Power 'Channel 2-15 2-11 Change in Void Fraction from 10 Psi Pressure Change Low Power Channel 2-16 2-12 Void Fraction Versus Time Exponential Flow Decay Problem 2-17 3-1 Axial Power Profile Turbine Trip 1 3-19 3-2 Axial Power Profile Turbine Trip 2 3-20 3-3 Axial Power Profile Turbine Trip 3 3-21 3-4 Peach Bottom-2 Rod Insertion Fraction Versus Time 3-22 3-5 Peach Bottom Eight-Node Steamiine Schematic 3-23 3-6 Peach Bottom-2 Turbine Trip 1 Steamline Pressure 3-24 3-7 Peach Bottom-2 Turbine Trip 2 Steamline Pressure 3-25 3-8 Peach Bottom-2 Turbine Trip 3 Steamline Pressure 3-26 3-9 Peach Bottom-2 Turbine Trip 1 Dome Pressure 3-27 3-10 Peach Bottom-2 Turbine Trip 2 Dome Pressure 3-28 3-11 Peach Bottom-2 Turbine Trip 3 Dome Pressure 3-29 3-12 Peach Bottom-2 Turbine Trip 1 Core Exit Pressure 3-30 3-13 Peach Bottom-2 Turbine Trip 2 Core Exit Pressure 3-31 3-14 Peach Bottom-2 Turbine Trip 3 Core Exit Pressure 3-32 3-15 Peach Bottom-2 Turbine Trip I Prompt Neutron Power 3-33 3-16 Peach Bottom-2 Turbine Trip 2 Prompt Neutron Power 3-34 3-17 Peach Bottom-2 Turbine Trip 3 Prompt Neutron Power 3-35 3-18 Peach Bottom-2 Reactivity Turbine Trip 1 3-36 cxxxvii

NEDO-24154-A LIST OF ILLUSTRATIONS (Continued)

Figure Title Page 3-19 Peach Bottom-2 Reactivity Turbine Trip 2 3-37 3-20 Peach Bottom-2 Reactivity Turbine Trip 3 3-38 3-21 Peach Bottom-2 A Level LPRM's Turbine Trip 1 3-39 3-22 Peach Bottom-2 B Level LPRM's Turbine Trip 1 3-40 3-23 Peach Bottom-2 C Level LPRM's Turbine Trip 1 3-41 3-24 Peach Bottom-2 D Level LPRM's Turbine Trip 1 3-42 3-25 Peach Bottom-2 A Level ?lux Turbine Trip 1 3-43 3-26 Peach Bottom-2 B Level Flux Turbine Trip 1 3-44 3-27 Peach Bottom-2 C Level Slux Turbine Trip 1 3-45 3-28 Peach Bottom-2 D Level Flux Turbine Trip 1 3-46 3-29 Peach Bottom-2 A Level Flux Turbine Trip 2 3-47 3-30 Peach Bottom-2 B Level Flux Turbine Trip 2 3-46 3-31 Peach Bottom-2 C Level Flux Turbine Trip 2 3-49

.3-32 Peach Bottom-2 D Level Flux Turbine Trip 2 3-50 3-33 Peach Bottom-2 A Level Flux Turbine Trip 3 3-51 3-34 Peach Bottom-2 B Level Flux Turbine Trip 3 3-52 3-35 Peach Bottom-2 C Level Flux Turbine Trip 3 3-53 3-36 Peach Bottom-2 D Level Flux Turbine Trip 3 3-54 3-37 Schematic of KM Stenmline Model 3-55 3-38 Average Axial Power MKH Test Conditions 3-56 3-39 Control Rod Motion Pattern KMH Turbine Trip Test 3-57 3-40 Turbine Pressure Turbine A KKQ Turbine Trip 3-58 3-41 Turbine Pressure Turbine B M1 Turbine Trip 3-59 3-42 KDM Turbine Trip Steamline Pressure, Turbine A 3-60 3-43 KXM Turbine Trip Steamline Pressure, Turbine B 3-61 3-44 E( Turbine Trip Dome Pressure 3-62 3-45 KD1 Turbine Trip Core Exit Pressure 3-63 3-46 K1M Turbine Trip Prompt Neutron Power 3-64 3-47 Reactivity Components MI Turbine Trip 3-65 3-48 A Level LPRM Flux rKM Turbine Trip 3-66 3-49 B Level LPRM Flux M Turbine Trip 3-67 3-50 C Level LP.RM Flux KLM Turbine Trip 3-68 cxxxviii

KEDO-24154-A LIST OF ILLUSTRATIONS (Continued)

Figure Title Page 3-5-1 D Level LPRM Flux KKK Turbine Trip 3-69 3-52 A Level Flux KKK Turbine Trip 3-70 3-53 B Level Flux KMM Turbine Trip 3-71 3-54 C Level Flux KM Turbine Trip 3-72 3-55 D 'Level Flux KWM Turbine Trip 3-73 i1-1 Peach Bottom-2 Turbine Trip 3 (69% Power) Neutron Flux A11-2 15-1 Peach Bottom-2 Turbine Trip 1 Core Exit Pressure (35% Bypass) A15-3 15-2 Peach Bottom-2 Turbine Trip 2 Core Ex*i Pressure (35% Bypass) A15-4 15-3 Peach Bottom-2 Turbine Trip 3 Core Exit Pressure (35% Bypass) A15-5 cxxxix/ cxl

NEDO-24154-A LIST OF TABLES Table Title Page 2-1 Comparison of Void Coefficients Obtained with Three Dimensional and One-Dimensional Core Models 2-5 2-2 Summary of Hydraulic Channel Conditions 2-5 3-1 Peach Bottom-2 Turbine Trip Test Conditions 3-16 3-2 Peach Bottom-2 Dynamic Test Signals 3-16 3-3 Peak Vessel Pressure 3-17 3-4 Maximum ACPF Values for Peach Bottom Turbine Trip Tests 3-17 3-5 KM( Turbine Trip Test Conditions 3-17 3-6 KM1 Dynamic Test Signals 3-18 3-7 Maximum &CPR Values for the KM Turbine Trip Test 3-18 5-1 Peak Flux Values as a Function of Scram Delay Time A5-2 6-1 Summary of One Dimensional Model Sensitivity Studies A6-2 15-1 Summary of &CPR/ICPP Results for Peach Bottom Turbine Trips A15-2 cxli/cxlii

NEDO-24154-A ABSTRACT Atranaien: nmodZ for the boilin zxter reactor MBWR) has been deve :oed which contains a one-dimensiona: neutronic and thermaZ hydrauZic simuZation of the reactor core, as weZl as a noda:

representatioitof the pressure variations in the main steaZine.

Comparisons dre made between various parts of the core model and other nc* lear and therma hydraulic models. MvodeZ results are also compared to data acquired from turbine trip ex-periment8 carried out at the Peach Bottom and M Atomic Power Stations.

Cxliii/Cxliv

NEDO-24154-A

1. INTRODUCTION AND SMMRY The transient behavior of a boiling water reactor (MR) power plant depends not only on the mechanical response of the recirculation and control systems but also on the power behavior of the reactor core. Previous computer modelsI for the BWR system have used a point model to describe the thermal hydraulic and neutronic response of the reactor core. The dynamic steamline response has been described by a lumped compressible steamline node. An improved BWR transient model, which includes a more detailed desciiption of the reactor core and steamline, has been constructed and is described in detail in Reference 2. This report describes qualification studies support ing the one-dimensional core transient model. These studies include comparisons with other computer models, as well as comparison with data from actual BWR pressurization transients.

The improved transient model consists of an integrated one-dimensional reactor core model which is coupled to recirculation and major system control models.

The recirculation and control system models are, for the most part, identical to the point reactor model described in Reference 1. The reactor core is hydraulically coupled to the recirculation loop through the core exit pressure and core inlet flow. The core exit quality and pressure drop are computed by the core model, which, in turn, interacts with the loop parameters.

The improved model also contains a more detailed description of the pressure variations in the main steamline. The model includes a nodalized description of the mass and momentum balances in the steamline and is capable of predicting the wave phenomenon present in the steamline during transients such as turbine trips.

The reactor core model allows for spacial variation of the neutron flux, fuel temperature, coolant flow, coolant density and pressure in the axial direction.

The neutron kinetics behavior is described by one energy group diffusion theory with six delayed neutron groups. The thermal-hydraulic model contains separate conservation and energy equations for the vapor and liquid phases as well as a mixture momentum equation. Fuel rod heat conduction equations are solved for each axial elevation in the core.

1-1

NEDO-24154-A Two types of studies will be presented in this report; model-model comparisons and model comparisons with experimental data.- The model-model comparisons are designed to show that, the one-dimensional nuclear and thermal-hydraulic models of the reactor core are consistent with core models currently used in BWR design. Also, the collapsed one-dimensional nuclear mcdel is compared with transient results pbtained from more detailed three-dimensional simulations.

The model compari;ons with experimental data consist of calculations of four instrumented turbine trips conducted with a delayed direct scram by bypassing of the direci trip scram, such that the scram occurred following a high flux trip. The experiments were performed at the Peach Bottom-2 and IM BWR power plants. The model data comparisons are given for a large number of transient variables, including system pressure, reactor flow, and neutron flux. The model-data comparisons form the basis for the qualification of the one dimensional transient model.

The nuclear model-model comparisons show that the one-dimensional model agrees well with three-dimensional results for scram calculations with no thermal feedback. Also, good agreement with three-dimensional simulator results are obtained for static void reactivity coefficients. The thermal-hydraulic model steady-state results agree well with results obtained with the standard BEW design models. The comparison of transient thermal-hydraulic results is carried out on a simplified flow decay problem for which an analytic solution exists.

Turbine trip data were acquired at the Peach Bottom Atomic Power Station Unit 2 (Peach Bottom-2) for three power levels: 48%, 622 and 69% of rated power.

The comparisons show good agreement between data and calculation for all three turbine trips. The experimental data show a steamline wave phenomenon which is duplicated by the model calculation. The model also shows good agreement in total neutron flux. The axial shift of the flux during the transient is apparent in the data and is duplicated by the calculation. Thermal margin calculations carried out using both the experimental and calculated transient heat generation rates showed that the calculated transient change in critical power ratio (LCR) divided by the initial CPR (ICPR) is within 0.01 of the change inferred from the experimental data for all three experiments.

1-2

NEDO-24 154-L An additional turbine trip experiment was carried out at the I*M plant in Muehleburg, Switzerland. This turbine trip was initiated from 77% of rated power and approximated end-of-cycle conditions. The KEM reactor is a two turbine plant and contains a more complicated steamline configuration. There fore, modifications were made to the model described in Reference 2 to adequately describe this complexity. The results of this comparison again show good agreement between data and calculation. The calculated change in critical power ratio agrees with the change inferred from the data to within 0.Ol ACPR/ICPR.

The four turbine trip experiments represent the most significant part of the one-dimensional transient model qualification data base. The comparisons shown in this document show that the one-dimensional model is successful in calculating all four turbine trips, which were initiated from different power levels and core conditions.

1-3/1-4

NEDO-24154-A

2. MODEL-MODEL COMPARISONS The comparison of computer model results to experimental data obtained from actual reactor plant transients is the most desirable qualification basis.

However, comparisons with other models are also useful in showing that a specific assumption yields accurate results or is consistent with other approved design methods.

2.1 NUCLEAR MODEL COMPARISONS The solution techniques employed in the one-dimensional nuclear model are described in Section 5 of Reference 2. The one-dimensional diffusion param eters are derived by collapsing three-dimensional steady-state solutions obtained from the General Electric three-dimensional BWR simulator 3 . This collapsing process is described in detail in Appendix A of Reference 2. The ability of the one-dimensional core model to calculate scram reactivity and void coefficients will be examined in this section.

2.1.1 Scram Reactivity In order to test the validity of the one-dimensional collapsing procedures described in Section 5 of Reference 2, a series of scram calculations has been compared with the one-dimensional solutions with three-dimensional solutions obtained with the three-dimensional BWR simulator. These scram calculations have no void feedback and represent a test of the collapsed one-dimensional model in properly estimating scram reactivity worth. The basic nuclear data used in the two calculations are identical and the one-dimensional parameters have been obtained from the steady-state three-dimensional solution. Three cases are considered here. The first case examined is a BWR/4 core at the beginning of Cycle 2 (Plant A). To simulate an end-of-cycle scram situation, all of the control rods were assumed to be withdrawn. Figures 2-1 and 2-2 show a three-dimensional, one-dimensional comparison of total neutron flux and scram reactivity versus time. The agreement is quite good over the entire scram range. The second example, shown in Figure 2-4, shows the same core in a critical rod pattern shown in Figure 2-3. In this case, the flux and scram reactivity agreement is still good, although not quite as good as the all-rods out case. This second case is more difficult to calculate because of the rapid 2-1

NEDO-24154-A radial flux changes caused by the motion of the inserted rods. The third example shows a BWR/4 reactor at 50% power, 100% flow (Plant B). In this case, there is a considerable number of control rods in the core, as shown in Figure 2-5. Also, the radial power distribution is complicated because of the lower power level and the complicated rod pattern. Here the agreement is still good, especially in the early part of the scram, which is the important region in the analysis of turbine trip accidents. Finally, a comparison of axial power shapes for the third example is shown in Figure 2-7 for a number of times during the scram.QT The change in average axial shape is followed quite well by the one-dimensional model. In summary, the collapsing procedures used in generating the one-dimensional neutron kinetics model yield results which compare well with three-dimensional scram calculations.

2.1.2 Void Coefficient The one-dimensional nuclear model must accurately predict the void reactivity response of the core in order to follow the neutron flux response during .3'A transients such as turbine trips. Table 2-1 contains a comparison of void coefficient values obtained with the one-dimensional nuclear model and the three-dimensional BEWSimulator. In each case, the void coefficient is cal culated as the following eigenvalue difference:

1 (a1 ) - K (a 2 )

Void Coefficient - -a 8

where a and a2 are core averaged void fractions at two different reactor states.Q The two reactor states were generated by varying the reactor pressure and flow, which undergo the greatest change during a turbine trip transient. In all cases, the one-dimensional and three-dimensional coefficients agree to wirhin 5%. A wide variety of power levels and control rod configurations has been covered in the cores analyzed. Thus, the one-dimensional model is expected to favorably reproduce the three-dimensional transient void reactivity response.

• Q7 - Replies to -RC questions on the text are documented in Appendix A. The symbcl Q7 denotes tLhat this topic is discussed further in the reply to NRC Questi~on 7.

2.-'

NEDO-24154-A 2.2 THEERMAL-HYDRAULIC MODEL COMPARISONS Steady-state comparisons between the thermal-hydraulic model-described in Section 6 of Reference 2 and the standard GE BWR channel hydraulic model have been carried out. Figures 2-8 and 2-9 show axial void fraction profiles for a high power and low power channel. Some differences exist in the subcooled void region because of different numerical treatments of the subcooled boiling process. Also, small differences exist at the top of the channel because the current model allows pressure and saturation enthalpy to vary along the channel, while the standard GE model assumes constant pressure and saturation enthalpy.

Considering the above model differences, the overall agreement is good, to within 0.003 in void fraction. These same two channels were also run at 10 psi higher pressure and the change in void fraction calculated as a function of axial height. The results obtained by the two models are compared in Figures 2-10 and 2-11 for the low and high power channels, respectively. Note that excellent agreement is obtained for the change in void profile generated by a pressure change. Q9 These calculations show that the channel hydraulic model gives steady-state results consistent with standard GE design tools, which have been verified against void fraction measurements in test assemblies.

In the transient mode, model-model comparisons are more difficult because the five-equation model is more detailed than currently employed design models.

However, a closed form solution exists for the one-dimensional vapor continuity equation with constant steam and drift flux properties.5 The initial conditions for the transient considered are listed in Table 2-2. During the transient, the axial heat flux distribution is constant and uniform. The time-dependent boundary condition is:

Inlet volumetric flux - 0.0002 (1 + 2 9 9 8 6 . 4 9 5 e-5t), ft/sec The one-dimensional hydraulic model was modified to include constant steam properties and drift flux correlations and then used to obtain a solution to the above problem. Spacial nodes were placed every 0.5 ft up the channel.

2-3

INEDO-24154-A This solution is compared with the closed form solution in Figure 2-12. The cransient void fraction at 0.5 and 12 ft from the channel entrance is plotted.

The maximum deviation at 0.5 ft is 0.006 in void fraction. The maximum devia tion at 12 ft is 0.0003. This numerical test shows that the overall numerical procedures employed in the thermal-hydraulic model are adequate.

2-4

NEDO-24154-A Table 2-1 COMPARISON OF VOID COEFFICIENTS OBTAINED WITH TAREE-DIMENSIONAL AND ONE-DIMENSIONAL CORE MODELS Percent Percent

. of of Rated 'Rated Average 3-D Void 1-D Void Power Flow Void Perturbation Coefficient Coefficient Plant (z) Fraction Mechanism (C/AV) (e/Av)

Plant B (BWR/4) 48 100 0.232 Pressure change -33.9 -33.2 Plant B (BWRI4) 62 80 0.289 Pressure change -29.9 -31.5 Plant B 69 100 0.301 Pressure change -28.5 -29.2 (BWR/4)

Plant C (BWR/5) 104 100 0.413 Pressure change -21.3 -22.0 Plant C 104 100 0.413 Flow change -24.2 -24.0 (BWR/5)

Plant D 77 93 0.321 Pressure change -26.3 -27.6 Table 2-2

SUMMARY

OF HYDRAULIC CHANNEL CONDITIONS Inlet Volumetric Bundle Sub Flow Hydraulic Channel Flux Power cooling Area Diameter Length Pressure Label (f t/see) (Btu/1b) (f0 2 ) (f t) (ft) (psi)

High Power 6.45 6.55 23.4 0.1065 0.0433 12.5 1055 Channel Low Power Channel 4.305 1.64 23.4 O0.1065 0.0433 12.5 1055 Transient Calcula 5.997 1.00 0 0.1115 0.0484 12 1000 tion 2-5

NEDO-24154-A 100 2

z z

z 2.U T"ME irni sgure 2-1. Plant A (3WR/4) AIll-Rods-Out Scram 2-6

S~. A. - ~

K:.

0.3 0.2 I

-,J I 0.1 U'

~b.

0 a

TIME find Figure 2-2. Plant A All-Rods-Out Scram Reactivity

NEDO-24154-A INITIAL CONTROL ROD ARRAY J/1 I 3 5 7 9 11 13 1

3 20.42 5 12.43 0.45 7 14.43 18.42 9 0.45 0.45 20.42 18.42 13 0.45 0.45 0.45 Figure 2-3. Plant A Initial Control Rod Pattern in Nodes Withdrawn (24 is Completely Out) 2-8

0.3 SOLUITION SLUTION 0.2 494 IxI U

1. Aa TIME IcIl Figure 2-4. Plant A Critical Rod Pattern Scram Reactivity

• ~ ~.

141T7 IAL CONTROL 500 ARRAY ji 3 5 1 9 11 13 16 17 19 21 23 25 27 29 3 19.17 14.17 19.17 6 0.17 0.17 0.17 017 20.17 10.17 8.17 10.17 20.17 9 0.17 0.17 0.17 0.17 0.17 0.17 I1 10.17 10.17 20.17 20.17 10.17 19.17 13 0.17 0.17 0.17 0.17 0.17 0.17 IS 14.17 8.17 20.17 i-i 8.17 14.17 0*

17 0.17 0.17 0.17 0.17 0.17 0.17 19 19.17 10.17 20.17 20.17 10.17 19.17 21 0.17 0.17 0.17 0.17 0.17 23 20.17 10.17 8.17 10.17 20.17 25 0.17 0.17 0.17 0.17 27 19.17 14.17 19.17 29 1;1gure 2-5. Planc B Rod Patterfl in Number of Nodes Withdrawn (24 is Completely Out)

NEDO-24154-A w/

I 0 12 TIME (lce Figure 2-6. Plant B (BWR/4) Scram Reactivity 2-11

NEDO-24154-A I

(A 3

w a

Ill 2

w AXIAL HEIG.T Ift)

Figure 2-7. Plant B Transient Axial Power 2-i

NEDO-24154-A

/ mmmmmONE.DIMENSlONAL MODELI 20 0.2 oAL 24 10 AXIAL. HEIGWT Iftl Figure 2-8. Steady-State Void Fraction Profile High Power Channel 2-13

NEDO-24154-A 0.6 O.5 U

0 0.2 0.1 0.1 6

AXIAL HEIGHT Oftt Figure 2-9. Steady-State Void Fractiot Profile Low Power Channel 2-14

NEDO-24154-A z

oI-*

U 0

z z

0 AXIAL HEIGHT Ift)

Figure 2-10. Change in Void Fraction from 10 Psi Pressure Change High Power Channel 2-15

NEDO-24154-A 0.010 O-M U0.012 0

Fiuew-1 AXIAL HEIGHT (0t)

Change in Void Fraction from 10 Psi Pressure Change Low Power Channel 2-16

NEDO-24154-A U.

0.1

-I w

2 2

4 2

U U.

a UA I

4 C

I' U.

0.4

.4 C

'.1

--&I a 2 TIME Inmc Figure 2-12. Void Fraction Versus Time Exponential Flow Decay Problem 2-17/2-18

NEDO-24154-A

3. ANALYSIS OF TURBINE TRIP EXPERIMENTS Four special turbine trip experiments were carried out on BWR reactors during 1977. An extensive amount of data was taken at each event, providing an outstanding set of benchmarks for transient model qualification. These tests were specifically designed to obtain data which would be useful in verifying assumptions for trinsient models. The first three turbine trips were carried out on the Peach Bottom-2 reactor and the fourth on the M reactor in Switzerland.

3.1 PEACE BOTTOM TURBINE TRIPS Three instrumented turbine trips were carried out at the Peach Bottom-2 reactor during April 1977. These tests were conducted with the direct scram on stopvalve position bypassed such that a trip on high flux was obtained.

This departure from the normal reactor condition was required to obtain a sufficiently large flux response to allow a model-test comparison. A detailed description of the test conditions and measurement process can be found in Reference 4.

3.1.1 Test Su=tary The initial power and flow conditions for each test are shown in Table 3-1.

These test conditions were selected in order of increasing power along a line of constant reactor flow. Prior to the second turbine trip test, it was necessary to reduce core flow to hold the power to within 1 of planned test power level due to the xenon level in the core at the time of the test. In each of the three tests, the trip scram was disabled and the flux scram setpoint adjusted. The scram setpoints are also listed in Table 3-1.

A total of 153 signals were recorded by a digital data acquisition system. A summary of the measured quantities appears in Table 3-2. The comparisons presented here will concentrate on important aspects of both the one-dimensional reactor core model and distributed steamline model.

3-1

NEDO-24154-A 3.1.2 Model Inputs As pointed out in Reference 2, most of the mathematical modeling for the recirculation and control system part of the BWR is identical to the model currently used for transient design analysis. 1 The vast majority of the external loop description for the Peach Bottom plant is identical to that used in design and licensing analysis, and is taken from a description of the plant conditions found in Reference 7. A few quantities have been changed to be compatible with the one-dimensional core and eight-node steamline models. These changes are all discussed in this section.

The initial conditions for all three tests were used by the GE BWR Simulator to generate three-dimensional power distributions. These nuclear data were then collapsed into one-dimensional parameters according to the procedures outlined in Reference 2. A comparison of process computer output and the resulting cue-dimensional axial power profiles appears in Figures 3-1, 3-2, and 3-3.

The control rod motion pattern was taken from measured data. A linear rod motion pattern as shown in Figure 3-4 was assumed. Figure 3-4 also shows the measured rod positions as a function of time after the trip signal. The same motion pattern was assumed for all three turbine trips.

The steamline representation is shown in Figure 3-5. The lengths and areas associated with each node are also shown. The coefficients for pressure losses were calculated by considering the geometry of the steamline &nd using pressure drop correlations6 to compute the appropriate losses over each section of pipe. These coefficients are also shown in Figure 3-5. The resistance quantity Xi is defined in Reference 2. The core thermal-hydraulic geometry used a weighted average of the dimensions from the 576 7x7 bundles and the 188 8x8 bundles.Q1 0 Temperature-dependent conduction parameters were used for the fuel pellet and cladding. & value of 1000 Btu/hr-ft -.F was used for gap conductance.

For the three turbine trip test conditions, the transient flux response was found to be a very weak function of gap conductance. That is, variations of 1

100% in gap conductance produced only 5% changes in peak neutron flux.Ql 3-2

NEDO-24154-A This low sensitivity results from the fast nature of the Peach Bottom flux

'transients, where thermal feedback contributes only a small part to the reactivity shutdown.

Turbine stopvalve and bypass valve positions were also recorded during each transient4 and were then used to describe valve motion in the calculation.

3.1.3 Data Comparisons 3.1.3.1 Transient Pressure Comparisons Dynamic pressure measurements were recorded at the turbine inlet, in the steamline 90 ft downstream from the vessel, the vessel dome, and near the core exit plenum. In all of the pressure comparisons listed in this section, the data shown are the unfiltered data as recorded by the pressure sensors.

The sensors are connected to the appropriate measurement locations by water filled sensor lines. These sensor lines have their own second-order response which can often give rise to oscillations in the recorded data. Further discussion of the sensor line effects is contained in Reference 4.

Figures 3-6 through 3-8 show comparisons of calculation and data for the pressure measured in the steamline near the safety/relief valves. Note that the model accurately predicts the wave travel time down the steamline and the frequency of pressure oscillations. However, the calculated waves are more spread out and the amplitudes are smaller than the measured waves. The spreading out of the calculated wave is due to the coarseness of the spatial mesh. However, this change in shape does not appreciably affect the dome pressure profile because the vessel dome pressure is influenced in large part by the total steamline flow, which is proportional to the integral of the steamline pressure. Hence, the area under each oscillation is of more interest than the detailed shape.

Plots of dome pressure over the first 1.2 sec of the transient are shown in Figures 3-9 through 3-11. Note that the calculation simulates the initial pressure rise rate quite well in all three tests. Note also that the calcula tion overpredicts the pressure rise near the peak of the first pressure oscilla tion. The peak dome pressure occurs about 2.5 sec into the transient. Values for the peak dome pressure are listed in Table 3-3. The transient model 3-3

NEDO-24154-A overpredicts the peak pressure by approximately.30 psi in all three cases.

The overprediction of peak pressure is thought to be due to the bypass flow model used in the transient model. The rated bypass capacity of 26% of total steam flow was assumed. It is felt that the real bypass capacity is somewhat larger than 26%. Alsot the model calculations assume choked flow through the bypass valve for all valve positions. It is possible that, immediately after the valve opening, the bypass flow will overshoot the steady-state value. If this is true,, the peak pressure could be o"ercalculated by the model.

The core exit pressure is compared in Figures 3-12 through 3-14. Accurate prediction of the core exit pressure is most important in determining the core neutron flux response. Note that, again, the initial rise in pressure is followed well by the model for all three test conditions. The oscillations in the data between 0.4 and 0.7 sec are thought to be caused by the second order behavior of the instrument line. Note that after 0.7 sac, large oscillations occur.Q4 These oscillations begin just about the time that control rod motion starts and could be connected with a disturbance caused by the initial motion of the control rods. No oscillations in neutron flux are observed during this period, so that it is most likely that these oscillations are connected with an instrument line disturbance.

In evaluating the model against the measured data, the steamline and dome model simulate the overall core pressure rise rather well in all three experiments. As the power level is changed, the model tracks the experimental behavior and that about the same differences are observed in all three experiments. There are, however, localized portions of the transient where the model overpredicts the pressure rise. In some cases, these differences can be important, such as the portion of each curve between 0.7 and 0.8 sec where the peak neutron flux is observed. In this region, small changes in pressure can have noticeable effects on neutron flux. This behavior will be discussed further in Section 3.1.3.2.

3.1.3.2 Neutron Flux Comparisons The most important and demanding test of this transient model is the neutron flux response. The severity of the flux transient determines the transient critical power ratio and its prediction requires not only good core pressure 3-4

IEDO-24154-A calculations, but an accurate scram and void reactivity feedback model. The total flux results are shorn in Figures 3-15 through 3-17 for the three turbine trips. Note that in all three cases excellent agreement is obtained over the initial rise and falloff portions of the flux spike. The calculation always tends to overpredict the peak flux value.

This overprediction near the peak flux value is primarily due to the fact that the core exit pressure rise is overcalculated during this time period.

Figures 3-18 through 3-20 show the reactivity components as a function of time during the transient as calculated by the one-dimensional model. The defini tions of the reactivity components are given in Appendix A of this report.

Note that near the peak of the total reactivity curve, the net reactivity is about $0.80. When the overall reactivity level is this large, small changes in reactivity can cause large changes in peak flux levels because the prompt change in flux is proportional to 1/(l-p/I): where p is the net reactivity and B is the delayed neutron fraction. For the Peach Bottom transients a i psi change in core pressure will generate a change of about 0.02 in the ratio

/0/, which, in turn, will cause about a lOZ change in peak neutron flux.. Note also that this sensitivity exists over a short time span and that 0.04 sec after the scram motion begins, the calculated flux agrees again with the data.

A second contributor to the peak flux differences is the assumed scram motion model. The model assumes that all control rods start at the sam time and move with the same speed,, whereas there is actually a spread of starting times and speeds. This model approximation tends to overestimate the peak flux but also will overestimate the initial fall off rate in the neutron flux. A change in scram initiation time of 0.02 sec results in a 5Z change in peak neutron flux, so that this effect is not as large as the pressure effect noted above.Q5 The reactivity curves also show that for the Peach Bottom test conditions, the scram reactivity is very strong and almost immediately turns the flux transient around. This is due to the fact that many control rods are inserted in the core, giving rise to a strong scram reactivity.

3-5

.MD0--243.54-A The Doppler feedback is small for the Peach Bottom Transients and the final answer is not sensitive to the Doppler coefficient. Turbine trips initiated from higher power will have larger Doppler sensitivities, but even in those cases, changes in Doppler coefficients of 30 to 40% are required to materially influence peak heat flux results.

In addition to the total APRM response, LPRM signals were recorded at 80 locations in the core. The LPRMs are located at four axial levels, which are 1.5, 4.5, 7.5 and 10.5 ft above the bottom of the fuel. These locations will be referred to as the A, B, C and D levels, respectively. Figures 3-21 through 3-24 show plots of all the Turbine Trip No. 1 LPRM signals for each level plotted as a percent of initial signal. All the LPRM signals for a given level have the same time behavior, which indicates that there is very little radial flux change during the transient. The transient behavior is not the same between the four levels, indicating that there are axial flux shifts during the transient.

The magnitude of the flux shift is considerable. The A level fluxes increase to 350%, whereas the D level increases to 550%. These trends are observed in all three turbine trip experiments (Reference 4). Based on this information, a one-dimensional model appears to be appropriate to describe the transient nuclear behavior of the core. Figures 3-25 through 3-36 show comparisons between calculation and data for the A, B, C and D level flux response for the three turbine trips, The largest flux rise occurs near the top of the reactor core, which has the largest void fraction and the largest void coefficient. Detector A is below the boiling boundary for these experiments and experiences the smallest flux rise. Figure 2-10 shows that the largest void changes occur near the boiling boundary in the low quality regions. In the reactor, the boiling boundary is different in each channel and the variation in void fraction with axial posi tion is smaller than the one-dimensional estimate, which contains a single thermal-hydraulic channel. For this reason, the one-dLmensional model will overestimate flux changes near the boiling boundary. This trend can be noticed in the Peach Bottom results in comparing the B level flux response with the other levels.qO However, the magnitude of this overestimate near the boiling boundary is relatively small, and the A, B, C and D level comparisons are almost identical to those observed for the total flux for all three transients.

3-6

NEDO-24154-A This indicates that the one-dimensional model is accurately predicting the axial flux shifts which take place during the transient.

3.1.3.3 Critical Power Ratio The critical power ratio (CPR) is a calculated quantity which is a measure of how close conditions in a particular channel are to transition boiling. AL good measure of the relative severity of a particular reactor transient is the maximum change in CPR, divided by the initial or steady-state CPR (ICPR).

For the Peach Bottom turbine trips, the CPR comparisons have been made by driving a hot channel transient thermal-hydraulic calculation with both experimentally determined and calculated inlet flow, pressure, and fuel heat generation rate. The pressure input was taken from the core pressure signal, which was filtered with a 5Hz low pass filter. The transient fuel heat gen eration rate was taken to be proportional to the total APEM response. Core

.flow was obtained from pressure drop measurements taken across four of the jet pumps throughout the three turbine trips. Changes in core flow can be detected by assuming the jet pump pressure drop to be proportional to the square of the flow. In practice, however, this is not an accurate measure of core flow because of the large amount of noise in the jet pump pressure drop signal. In this case, a 5 Hz filter was applied to the four jet pump signals to reduce the noise component and then averaged to obtain a pressure drop. The steady-state flow was normalized to the recorded flow at the begin ning of each transient.

In both cases, the initial conditions and channel properties are identical.

Only the transient pressure, flow and flux responses are different so that a good measure of bias in CPR is obtained.

For the transient CPR calculations driven by the experimental data, uncer tainties in the input quantities will contribute to an uncertainty in the ratio CPR/ICPR. Reference 4 quotes a !2 psi uncertainty in core pressure.

This pressure uncertainty, coupled with a t3Z uncertainty in flow, results in a !0.01 uncertainty in the ratio CPR/ICPU.qlS'ql 8 This CPR uncertainty is obtained from sensitivity calculations carried out on pressurization type transients.

3-7

NEDO-24154-A The results of the transient CPR calculations are sumoarized in Table 3-4.

In each case, the model driven CPR/ICFR is within 0.01 of the data driven value. This excellent agreement is an indication of the overall quality of the calculated results and shows that, while differences exist in detail between model calculation and experiment, their overall impact on CPR is small.

3.1.4 Conclusions A great deal of information has been assembled for the three turbine trips carried out at Peach Bottom. The one-dimensional BWR transient model compares well with the measured pressure responses in the vessel dome and core exit.

The steamline pressure comparisons show the model underpredicting the magni tube of the pressure oscillations but this does not affect the dome pressure response. The peak steamline and dome pressures are overcalculated by the model, indicating the possibility of larger bypass flow than assumed by the model.

The flux response agreement is good for all three experiments. The model overpredicts the peak neutron flux due to a slight overcalculation of pressure near the time of maximum neutron flux. The initial and final portions of the flux show excellent agreement. The axial response is also followed well, as evidenced by comparisons with A, B, C, and D level LPRK signals.

Finally, the good qualitative agreement obtained in the pressure, flow and flux responses results in excellent transient CPR comparisons. Overall confidence in the model is strengthened by the fact that about the same degree of agree ment was obtained for all three turbine trip tests. Model input assumptions were identical for all three tests except for the core power level, core flow and known changes in valve actions and scram setpoints. This agreement indi cates that the model can also follow changes in transient behavior due to changes in power level and flow and can be used to predict full power turbine trip transients.

3.2 KM TURBINE TRIP An instrumented turbine trip was carried out at the M power plant in Muehleburg, Switzerland on June 30, 1977. The I Plant is unique for BWR's, having two cur bines and a bypass capacity of 110% of full rated steam flow. At the time of the 3-8

NEDO-24154-A test, the reactor core was near the end of a fuel cycle and all the control rods Vere out of the core at rated conditions. These test conditions provide an addi tional, valuable transient benchmark because they realistically simulate end-of cycle conditions. However, the two-turbine configuration and large bypass capacity provided additional challenges for the steamline and valve flow portions of the one-dimensional transeint model. As in the case of the tests conducted at the Peach Bottom-2 plant, the direct trip scram on stopvalve position was bypassed to obtain a significant flux response during the test.

3.2.1 Test Summary The initial conditions for the M turbine trip test are summarized in Table 3-5.

The KKK plant was designedwitha large bypass capacity in order to avoid a full reactor scram following a turbine trip. Therefore, only selected rods are inserted following the turbine trip signal. For this test, the select rod insertion was disabled. The flux scram setpoint was not changed and remained at 120% of rated for the test. The normal bypass capacity at KM is 110% of rated steam flow. Each turbine has two bypass valves which are located in the same valve chest with the turbine stcpvalves. One of these bypass valves was disabled on each turbine, thereby reducing the total bypass capacity to 55%

of rated steam flow for this test.

As in Peach Bottom, a digital data acquisition system was used to record a total of 118 signals. The signals (summarized in Table 3-6) are essentially the same as those recorded at Peach Bottom. Unlike Peach Bottom, however, the signals recording the turbine stopvalve and bypass valve positions were not available. The KKM plant has a mechanical control system to regulate bypass flow and a continuous signal recording valve position was not available.

Therefore, additional position switches were installed on these valves, but their signals were lost due to equipment problems. The only signals available are the valve initial opening time and the time at which the valve is in the full open position. The comparisons presented here will concentrate on the pressure and flux measurements which form the qualification base for the dis tributed steamline and reactor core models.

3-9

NEDO-24154-A, 3.2.2 Special Model Considerations The one-dimensi6nai transient model used for the description of the Peach Bottom turbine trips is described in Reference 2. Separate models exist for valve flow control and motor generator plants, so that with one exception all BWR/3 through 6 plants can be described. This exception is KIM, which has two turbines and .two sets of steamlines. Also, each steamline has a reheater line, a feature which is not present on any domestic BWE's. Therefore, a special version of the one-dimensional transeint model has been constructed which contains two steamlines and also correctly simulates the reheater line.

The physical dimensions of both steamlines are identical, so the same constants are used for both lines. A schematic of the steamline model appears in Figure 3-37, along with the dimensions and pressure drop constants.

The pressure control system in use at M was also simulated, along with a description of the other unique control features of the KKK plant. These other WM unique model approximations had little effect on the calculated turbine trip transienf response.

3.2.3 Model Inputs With the exception of those items mentioned in Section 3.2.2, the mathematical modeling for the KM transient model is identical to that described in Refer ence 2. The recirculation system parameters are based on the input procedures used for other GE transient analyses.

As in the leach Bottom tests, the GE BWE Simulator was used to generate a steady-state three-dimensional power distribution for the test configuration.

These nuclear data were then collapsed into one-dimensional parameters according to the procedures outlined in Reference 2. A comparison of the resultant one dimensional power profile and the process computer output is shown in Fig ure 3-38. A total of 32 control rod drift relay signals were recorded, wiich were then used to determine the control rod bank position with time. From these data, a linear rod motion pattern was established. This pattern is shown in Figure 3-39 along with the measured position of a few control rods. The time zero is established as that point when the flux reaches the trip setting, which in this case is 120% of full rated power.

3-10

NEDO-24154-A A special steamline model was constructed for this analysis and is discussed in Section 3.2.2. The model dimensions and loss coefficients are listed in Figure 3-37.

The gap conductance used in the KWH analysis is 600 Btu/hr-ft 2 -_F. As in the Peach Bottom analysis, the severity of the test transient is not very large, and the calculated flux response is not sensitive to gap conductance. The core thermal-hydraulic geometry was determined from a weighted average of the dimen sions from the 12 7x7 bundles and the 228 8x8 bundles in the core at the time of the experiment.

As pointed out before, measured turbine stopvalve and bypass valve positions were not available for this transient. Because of the large bypass capacity, the calculated dome pressure and core flux response is sensitive to the bypass valve opening characteristics. In order to arrive at a realistic transient value response, the bypass valve opening time was set at 0.1 sec, which was the observed time, and the bypass valve opening speed was adjusted until the calculated transient turbine inlet pressure response agreed with the measured data. This adjustment affected only the initial valve opening speed. After the initial bypass valve opening, the KM pressure controller system was allowed to change the valve position based on the plant control parameters and the calculated pressure controller inputs. It should be pointed out that the adjusted bypass valve opening speed was more rapid than the value currently used in the KK design calculations.

3.2.4 Data Comparisons 3.2.4.1 Pressure Calculations Figures 3-40 and 3-41 show the transient pressure comparisons at the two turbine inlets. Upon the manual trip signal, the B turbine stopvalve motion was initiated first, followed by the A turbine stopvalve motion 0.024 sec later. The stopvalve motion closing time was 0.042 sec for both turbines.

All of the plots in this section place time equal to zero at the beginning of turbine B stopvalve motion. Initially, the pressure rises and then begins to decrease about 0.2 sec after the start of stopvalve motion. This decrease is due to the opening of the bypass valve, which begins its motion at 0.1 sec.

3-11

NE)O-24154-A Other oscillationi are due to the wave nature of the steamline transient pressure response. The turbine pressure response from 0.2 to 0.6 sec is sensi tive to the bypass valve motion. The best calculated response was obtained with a very rapid valve opening time (.0.08 sec), which is more rapid than that used in normal design calculations. The opening time of 0.08 sec agrees quite well with opening times of 0.07 to 0.09 sec observed in the M startup test results obtiined from the turbine vendor. The overall agreement is about the same in the A and B steamlides, the only difference being that the A results are dispaced about 0.024 sec in time and the initial pressure rise is somewhat slower in the B turbine.

Figures 3-42 and 3-43 show a comparison of transient steamline pressure for steamlines A and B, respectively. The first pressure peak is narrower than those observed in Peach Bottom; the magnitude of the peak is reduced by the opening of the bypass valve. The model predicts a slightly slower wave travel time. The 20 to 30 msec delay in the first calculated pressure peak relative to the data still represents good agreement between experiment and model but will contribute to a slight delay in the initial core pressure rise. After the first pressure peak, the calculations show a 3 Rz oscillation in the steam line pressure. This same frequency is also present in the data. Note that the detailed wave shape is somewhat different in the data and calculation.

The KF1 model does not tndercalculate the magnitude of the initial steamline pressure peak, as observed in the Peach Bottom results. This is due to the finer node structure in the KKK steamline simulation. The KKM used 7 nodes to simulate a 330 ft steamline, whereas the Peach Bottom model used 6 nodes to simulate a 400 ft steamline. The impact of the steamline pressure differ ences is reduced because the reactor vessel pressure will depend in large part on the change in total steamline flow, which is proportional to the integral of the transient steamline pressure.

The dome pressure comparison is shown in Figure 3-44. The data show a large number of oscillations. These oscillations are due to a disturbance in the instrument line. This can be verified in the test data because similar oscillations are not detected in the core pressure or neutron flux response.

Because of this behavior, detailed comparison between data and calculations is difficult. The calculated response does pass through the data, but it is difficult to determine if the timing of the pressure wave is being calculated 3-12

NEDO-24154-A accurately. The dome pressure compares quite weil out to 1.8 seconds, when the data starts to decrease and the calculated pressure continues to increase. No relief valve opening was observed during the test and the cause of the decrease in the observed pressure is not known nor important in the determination of the core flux response.

Figure 3-45 shows a comparison of the measured and calculated core exit pressure. Although there is some oscillation in the data after 0.42 sec,.an overall pattern can be determined. Again, it is felt that these oscillations are due to instrument line response because no oscillations are observed in the neutron flux. The calculated pressure response lags about 40 seac behind the data, reaching an initial peak at about 0.5 sec. Better agreement between calculation and experiment could be obtained with an improved inertia model for the steam separators. The current model assumes the steam to be mixed with the water in the separators, which tends to overestimate its inertia.

The model used here still yields reasonable agreement and also provides a conservative estimate of the neutron flux.

In evaluating the total steam~line and vessel model, it is important to note that all of the significant aspects of the transient pressure behavior have been duplicated. The plotted data are the raw instrument data, and the instrument line effects are still present so that precise comparisons cannot be made at this time. However, it is apparent that the overall shape of the core exit pressure response is duplicated quite well by the model.

3.2.4.2 Neutron Flux Comparisons The neutron flux response from the M Turbine Trip is compared in Figure 3-46.

The KMN transient flux response is interesting because of its double peak behavior. This double peak behavior is caused by an oscillation in the core exit pressure. The pressure rises initially to a peak around 0.4 sec, falls off slightly, then reaches a second peak at 0.72 sec. The control rod scram motion starts at 0.6 see and reduces the magnitude of the second flux peak.

Careful examination of the calculated core exit pressure shows that the model delays the initial pressure rise somewhat, then overestimates the first pres sure rise and underestimates the second pressure peak. This results in a calculated flux which does not have the double peak behavior, but which 3-13

NEDO-24154-A overestimates the initial flux peak and underestimates the second fluk peak.

Note, however, that these are shape differences and that the overall magnitude of the transieni is slightly overpredicted by the model. Also note that dif ferences between calculated and observed flux response correlate quite well with differences between calculated and observed core pressure. This indicates that the one-dimensional reactor core model responds quite accurately to pres sure and scram changes.

The calculated reactivity components are plotted in Figure 3-47. The initial reactivity rise is due to void collapse. However, the scram does not begin until the transient has turned around. Note that the scram reactivity curve has a smaller slope than observed in Peach Bottom. This is due to the all rods-out configuration.

As a further test of the reactor core model, the A, B, C and D level LPRM signals were also compared. Figures 3-48 through 3-51 show the A, B, C, and D level LPRMs all plotted as a percent of initial power. Note that the overall behavior is one-dimensional, similar to the observed behavior at Peach Bottom.

A comparison between calculated A, B, C and D level response and the average observed response at each level is shown in Figures 3-52 through 3-55. In each case, the calculated curve differs from the measured response. However, the deviation in each case is about the same as observed in the total flux. Note that the A level flux decreases much more rapidly after the scram begins because of its close proximity to the control rods, whereas the D level flux has only decreased to ZOZ of initial after 1.2 sec. This same pattern Is also evident in the calculated values.

In attempting to evaluate the one-dimensional model's ability to calculate neutron flux response, it should be pointed out that the UZ transient is a mild one compared to the Peach Bottom transients , In Peach Bottom the neutron flux peak value was as high as 550Z of initial, whereas the maximum WM increase was 220Z. Because of its larger bypass capacity and subsequent reduced pres sure rise, the detailed flux response was more sensitive to small oscillations in the pressure, which, in turn, was influenced by the turbine and bypass valve behavior. Even with the differences in flux shape, the total energy release is approximated quite well by the one-dimensional model. The impact on the cransient critical power ratio Is discussed in Subsection 3.2.'4.4.

3-14

NEDO-24154-A 3.2.4.3 Critical Power Ratio Critical power ratio calculations were also carried out for the KK transient using both the model transient and the data transient as inputs to the hot channel calculation. The CPR results are summarized in Table 3-7. Note that good agreement is achieved between the data driven and model driven maximum

&CPR values. The uncertainty on the data driven &CPR/ICfl values is +/-0.01.

3.2.5 Conclusions Analysis of the M turbine trip presented a few unique challenges for the one-dimensional transient model. The two-turbine configuration and the large bypass capacity required careful consideration of the steamline model and the turbine stopvalve and bypass valve characteristics. The final dome pres sure and core pressure response was followed quite well by the model, although the exact shape of the transient pressure response was not duplicated. The overall magnitude of the flux response was duplicated quite well, even though the double peak shape was not duplicated by the model. Rowever the success of the model in duplicating the transient is demonstrated by the excellent agreement obtained in the transient CPR values.

The M transient was in many ways quite different from those at Peach Bottom, yet the transient model was successful in duplicating its behavior as veil, lending additional confidence to its ability to calculate pressurization transients.

3-1.5

NEDO-24154-A Table 3-1 PEACE BOTTOM-2 TURBINE T*IF TEST CONDITIONS Core Power Core Flow Flux Scram Setting (MWT) (I Rated) (106lb/hr) (%Rtd (ZRated)

Turbine Trip 1 .,1562 47.4 101.3 100.3 85 Turbine Trip 2 2030 61.6 82.9 82.1 95 Turbine Trip 3 2275 69.1 101.9 100.9 77 Table 3-2 PEACE BOTTOM-2 DYNAMIC TEST SIGNALS No. of Signals Recorded Test Signal Description 8o LPRM Signals 4 APIM Signals 2 TIP Signals 4 Jet Pump (Calib) dPs 1 CORE dP 2 Steamfiow Nozzle dPs 2 Steamflow Nozzle Upstream Pressures 2 Turbine Inlet Pressures 1 Reactor Vessel Pressure 1 Core Exit Pressure 2 Reactor Feedpump Flows

1. Reactor Feedvater Temperature 2 Recirculation Loop Flows 2 Recirculation Pump Inlet Temperatures 2 Reactor Water Level 31 Control Rod Drift Relays 1 Scram Solenoid Relay 4 Turbine Stop Valve Position Switches (2 at 102 and 2 ac 90%)

4 Turbine Bypass Valve Posi:ton Switches (2 at 10% and 2 at 90%)

1 Turbine Bypass Position 4 Turbine Stop Valve Position 3-16

NEDO-24154-A Table 3-3 PEAK VESSEL PEESSURE Data* Model Calculation Turbine Trip 1 1042 1070 Turbine Trip 2 1052 1072 Turbine Trip 3 1069 1100

• Data value is biased to the same initial value as calculation.

Table 3-4 MAXIMUM ACPR VALUES FOR PEACE BOTTOM TURBINE TRIP TESTS ACPR/ICPR. &CPR/ICPR Initial CPR (Model)

-(Data)

Turbine Trip 1 2.536 0.170 0.173 Turbine Trip 2 2.115 0.136 0.129 Turbine Trip 3 2.048 0.132 0.141 Table 3-5 KKH TURBINE TRIP TEST CONDITIONS Core Power 769 MWt 772 Rated Core Flow 25.69 x 10 6 lb/hr 86.5% Rated Flux Scram Settig 120Z Rated Vessel Pressure 1023 psi 3-17

NEDO-24154-A Table 3-6 KWM DYNAMIC TEST SIGNALS No. of Signals Recorded Test Signal Description 52 LPRM Signals 4 APRM Signals 2 TIP Signals 4 Jet Pump (Calib) dPs 1 Core dP 2 Steamflow Nozzle dPs 2 Steamflow Nozzle Upstream Pressures 2 Steamline Header Pressures 2 Turbine Inlet Pressures 1 Reactor Vessel Pressure 1 Core Exit Pressure 32 Control Rod Drift Relays 1 Scram Solenoid Relay 2 Relief Valve Positions 2 Relief Valve Downstream Pressures 4 Turbine Stop Valve Position Switches (2 at 10% and 2 at 90%)

4 Turbine Bypass Valve Position Switches (2 at 10% and 2 at 90%)

Table 3-7 MAXIMUM tCPR VALUES FOR THE WKM TURBINE TRIP TEST Initial CPR 1.279 SCPR/ICPR (Data) 0.077 4CPR/ICPR (Model) 0.084 3-18

NEDQ-24154-L IA - AXIAL AVERAGE 1ONE4IMENSONAL MODEL 1.2 8C LEVEL cc0 w

0.6 o A LEVEL 0.4 0.2 a 2 4 6 1 to 12 AXIAL HEIGHT (ft)

Figure 3-1. Axial Power Profile Turbine Trip1 3-19

NEDo-24154-A C

0.8 AX6AL HEIGI41 th';

Figure 3-2. Axial Power Profile Turbine Trip 2 3-20

NEDO-24154-A m I 1.4 mimm PROCESS COMPUTER m

AXIAL AVERAGE ONE-C MENSIONAL MODEL I

t.2 B LEVEL C LEVEL 1.01-us E

4c

0. -- 0 us wj W-A LEVEL 0.6 0.4 -1 0.2 I I I I I 0

2 4 I a to 12 AXIAL HEIGH4T 1111 Figure 3-3. Axial Power Profile Turbine Trip 3 3-21

1.0 5% OF ALL RODS WITHIN THIS RANGE s,, ,,_,

FASTEST ROO*

0.4 SL/OWEST Ron Os - g

&4,

-jp -" "-" ORSEVIVED ROD MOTION IROID MOTION ASSUMED DELAY OF 0.17.. I IN CALCULATION ASSUMED IN CALC 0.2 '

0 0 0.5 1.0 1.5 2.0 2.6 3.0 3.6 4.0 TIME AFTE.R FLUX REACHES TRIP SETTING foal Fi~gure 3-4. Peach Bottom-2 Rod Insertion Fraction Versus Time

~~..

• . ? i: ., :, ., . * : ..... ,,. . . ..;.:*:**: - i**  :*....

AS - 2.84 t12 Ks

• 6.66 X 10-4 IIN STEADY41TATE AMI - Ki (12, A? 2.04 h 2 91 a 1.37KX 10 Tj 32ht OEPGOap C109TAINMIINTI w

N Il is OTSV Figure 3-5. Peach Bottom Eight-Node Steamline Schematic

J".

1M 4-Sloi I

N

~I.

0 0.3 0.6 0. 1.2 1.5 S.A 2.1 2.4 2.7 3.0 3.3 3.6 3.9 TIME Iml Figure 3-6. Pesch Bottoom-2 Turbine Trip I Steamline Pressure

21080 11 0.

-ma Ut J low F TIME Oftl Figure 3-7. Peach Bottom-2 Turbine Trip 2 Steamline Pressure

11300 1080 Z) 1040 I.-

0.600 0.9m 1.200 1 .670 1.600 .2100 2.400 2.7m0 3.000 TIME ("d rigure 3-8. Peach Bottom-2 Turbine Trip 3 Steamline Pressure

1090 1060 1040 1020 lo TIME foa Figure 3-9. Peach Bottom-2 Turbine Trip 1 Dome Preasure

Ca

'-I I

N TIME fnlcl FiSure 3-10. Peach Bottom-2 Turbine Trip 2 Dome Presaure

I t*J

[

ma Figure 3-11. Peach Nottom-2 Turbine Trip 3 Dome Pressure

1100 to" I~ II

~10"0 t

U' tU 0 tIwo 3.3 TIME teel Figure 3-12. Peach Bottom-2 Tuibine Trip I Core Exit Pressure

1020 t:

I-, f'l U'

two I IS TIME fInd Figure 3-13. Peach Dottow-2 Turbine Trip 2 Core Exit Pressure

1100 tor

.,J w 1020 1.8 TIME fwq)

Figure 3-14. Peach Bottom-2 Turbine Trip 3 Core Exit Pressure

626 460 300 I

w.

I:s I N

7s 0

-76

-i6o 0

TIME ind Figure 3-15. Peach Bottom-2 Turbine Trip 1 Prompt Neutron Power

"° IO 00 0

-71 0 0.16 0.30 0.46 0.60 0.15 0.90 l.0 1.20 1.35 1.60 1.66 1.30 1.35 2.10 2.25 2.40 TIME Ivmc Figure 3-16. Peach Bottom-2 Turbine Trip 2 Prompt Neutron Power

521 460

_ SMODEL 375 -- n . DTA I

-II 0

-71

-Iso ~~ ~- to.m.&e....t. in.~n..tm&as 0..fte... t*t iin1sej.wsI 0 0.11 OL30 0.41 8.80 0L75 0.30 l05 1.20 1.35 1.10 1.81 1.00 1.35 Zia10 .25 140 TIME fowl~

Figure 3-17. Peach Bottom-2 Turbine Trip 3 Prompt Neutron Power

NEDO-24154-A 1.0 0.8 0.6 0.4 I

0.2

-0.2

-0.4

-0.6 TIME NOW Figute 3-18. Peact' Bottoo-2 Reactivicy Turbine Zrip 1 3-36

NEDO-2414-~A

.04 C0.2

-0.2 0 0..*2@

TIE0w 0.8 1.0 Figure 3-19. Peach Bottom-2 Reactivity Turbine Trip 2 3-37

1, NEDD-24154-A U

MI TIME lad Figure 3-20. Peach Bottom-2 Reactivity Turbine Trip. 3 3-38

I:,!

700 400 I

LAA U

TIME Iawl Figure 3-21. Peach Bottom-2 A. Level LPRH's Turbine Trip I

-l Z

2 0

TIMF~ 1McI Figure 3-22. Peach Iottom-2 B Level LPRH's Turbine Trip 1

NEDO-24154-A W4 0.

.0 Sd S 61 c.4 C4, (IVIIINI zio im3nuadi xnid nudi 3-41

-a 2

U.

0 U

IA C

'U ab

-a U.

a:

-a 0.5 0.9 1.2 1.5 l.8 2.1 TIME 1Eecd Figure 3-24. Peach Bnttom-2 D Level LPRH'a Turbine Trip 1

MEO-24154-A 700

-Sa 300 200 0

0 0.2 0.4 0.5 0.8 1.0 1.2 TIME Ind Figure 3-25. Peach Bottom-2 A Level Flux Turbine Trip I 3-43

NEDO-24154-A goo 4w

=oa 10a a

0 0.2 0.4 0.6 C 1.0 1.2 TIME ("c)

Figure 3-26. Peach Bctto=-Z B Level Flux Turbine Trip 1 3-44

WMD-24154-A 4w 300 10o 0

OA8 TIME Isec)

Figure 3-27. Peach Boctom-2 C Level Flux Turbine Trip 1 3-45

NEDO-24154-A 700 4w 3

-.J 300 0 0.2 0.4 0.6 0.8 1.0 1.2 TIME fmr.

Figure 3-28. Peach Bottom-2 D Level Flux Turbine Trip 1 3-46

NEDO-24154-A Soo S400 3w 2W 200 /

1' 0 0.2 0.4 0.I 0.8 1.0 TIME bied)

Figure 3-29. Peach Bottom-2 A Level Flux Turbine Trip 2 3-47

IEDO-24154-A 70w S00 300 200 0

TIME Imc)

Figure 3-30. Peach Bottom-2 B Level Flunx Turbine Trip 2 3-48

NEDO-24154-A 700 am IR

,~40 3w 100 0

0 0.2 0. 0.6 1.0 1.2 TIME Ind Figure 3-31. Peach Bottom-2 C Level Flux Turbine Trip 2 3-49

NEDO-24154-A

-IU

-II 200 log 0

00.2 U. O.6 M.8 1.0 TIME (wJ Figure 3-32. Peach Bottom-2 D Level Flux Turbine Trip 2 3-50

NEDO-24154-&

DATA MOD61-Soo 300 0

0 TIME hIm)

Figure 3-33. Peach Bottom-2 A Level Flux Turbine Trip 3 3-51

NEDO-24154-A DATA MODEL

. I 7w0 6oo x 40

,=I 3*0 200 100 0

0 0.2 0.4 0.6 061.0 TIME tudc Figure 3-34. Peach Bottom-2 B Level Flux Turbine Trip 3 3-C2

NEDO-243.54-A 700 o00

_a 2

SJ x 400 AIM 0

TIME (sad Figure 3-35. Peach Boutom-2 C Level Flux Turbine Trip 3 3-53

NEDO-24154-A 700 500 4w 200 100 0

0 0.2 0.4 0.6 0.U 1.0 1.2 TIME IndJ Fi-urez 3-36. Peach Bottcm-Z D Levell Fl.ux Turbine Trip 3 3-54

IN STEADY STATE. API - KI 0*2i w.

La tLn I.

N

~p.

Al lARRA-h 2 l Ki ILOS$

COEFFICIENT Pd...

Figure 3-37. Schematic of KKN Steamline Model

1.8 1.4 I 1.2 tI Lfl 5: 1A 0.8

-01%

1 0.6 A 0

0.4 0.2 I I I 0 1.6 3.0 4.6 5. 7.6 '.0 10.5 12.0 AXIAL HEIGHT Ifl Figure 3-38. Average Axial Power KKI4 Tent Conditions

1.2 ASSUIMED ROD MOTION o6% 3 OF ALL RODS IN THIS FLANGE 0.

Oo $LOWEST ROo

//

04- I ao

/ /

/ /

0.O ESAY-.- /

0 0.8 1.0 1.6 2.0 2.5 3.0 3.6 4.0 TIME AFTER FLUX REACHES TRIP BETTING hal Figure 3-39. Control Rod Motion Pattern KKM Turbine Trip Test

NEDO-24154-A U

E 3

E 0.3 TWTiM(d, Fikare 3-40. Turbine Pressure Turbine A L( Turbine Trip 3-58

NEDO-24154-A Imu

~ 1030 0

0 TIME (mWJ Figure 3-41. Turbine Pressure Turbine B KMX Turbine Trip 3-59

(ha AN I.'

Up TIME (sKl Figure 3-42. KKH Turbine Trip Stenaline Pressure. Turbine A

1120 100 1000--,.J V 1040 I-. ____ ODEL

~~ ~DATA IE.

100_____

TIME bat Figure 3-43. KlH Turbine Trip Steamline Pressure, Turbine B

1000 N

~1000 _ ______MODEL O DATA I P IreI o r . . . e m e m c m e m c m e c m c e m e m e m e m 1.5 1.0 TIME (WuI Figure 3-44. KKM Turbine Trip Dome Pressure

2li-An 1I20 11001 10 l - /

m^

loeo-t:

.040

'C I-a w

U' 10201-r~~~r-~

• 0 0n 0 A S e n i . e 0J0 0.3 0.6 0.9 1.2 1.5 1.8 2.1 2.4 2.7 3.0 33 TIME hIc Figure 3-45. KKN Turbine Trip Core Exit Pressure

4l 160 IL too 60o 01

-60 06Woo* .e

-%sedoo in@0 OM-feof as a."* .in.~ OE* OMMOIn..MI 0 ~ .*Of*

WoMNa~ O.W o 0O* 0 in0in** Of

--Coo 1 1 -L- -L---- I--- -A -~ -- - - L -A L S 0.lb . 0.30 0.46 0.60 0.75 0.30 1.05 1.20 1.3 1.60 1.65 1.80 1.96 2.10 2.26 TIME had Pigure 3-46. KlH Turbine Trip Prompt Neutron Power

4l NEDO-24154-A 1.0 ...

0.8 .

0.6

• 0.4 L/

- .2 TIME I,.c)

Figure 3-47. Reacti~vity Components KKK Turbine Trip 3-65

NEDO-24154-A Ih iz gw M#

XC U 1.2 1. IA 2..I TIME (wo.

Figure 3-48. A Level LPIO4 Flux KP12 Turbine Trip 3-66

NEDO-24 154-A 5:

a TIMIE 6we Figure 3-49. B Level LPRM Flux KKM Turbine Trip 3-67

NEDO-24154-A IM Figure 3-50. C Level LPRH Flux M Turbine Trip 3-68

NEDO-24154-A Z2 Figure 3-51. D Level LPRM Flux KKK Turbine Trip 3-69

NEDO-24154-A 400 3WE 2W

.0 M4 G. 0.5 TIME Wad)

Figure 3-52. A Level Flux WK Turbine Trip 3-70

1=DO-24154-A 2

TIME Ind Figure 3-53. B Level Flux M Turbine Trip 3-71

NEDO-24154-A 300 200

-L I

x

-a U.

0.6 1.2 TIME aacJ Figure 3-54. C Levvl Flux KMC4 Turbine Trip 3-72

NEDO-24154-k

-a 4

K a'

TIME h*,c Figure 3-55. D Level Flux KKK Turbine Trip 3-73

NEDO-24154-A REFERENCES

1. R.. . Linford, AkaZytica* Methods of Plant Transient EvaZuations of the General. Electric Boiling Water Reactor, NEDO-10802, February 1973.

2. One-Dimensionaz Core Transient model, NEDO-24154, Vol. 1, October 1978.

3. J. A. Woolley, Three-DimenvionaZ BW2 Core SimmZator, NEDO-20953-&.

4. L. A. Carmichael and R. 0. Niem+/-, Trasient and Stability Tests at Peac*h Bottom Atomic Power Station Unit 2 at E*d of CycLe 2, EPRI NP-564, June, 1978.

5. K. T. Lahey, B. S. Shiralkiar, J. K. Gonzalez, L. E. Schnebly, The Analysis of Transient Critical Heart FuZ=, GEAP-13249, April 1972.

6. Flow of Fluida Through Valves, Fittings and Pipe, The Crane Co. Technical Paper No. 410, 1969.

7. N. H. Larsen, Core Design ad Operating Data for CýcZes One and Tw of Peach Bottom-2, EPRI-NP563, June 1978.

8. Generic ReZoad .NeZ Application, NEDE 24011-P-3, May 1977.

3-74

NEDO-24154-A APPENDIX A

.RESPONSE TO NUCLEAR REGULAORY COMIaSSION QUESTIONS ON SECTIONS 1 TEROUGH 3 OF VOLUME Ii A-1/A-2

NEDO-24154-A QUESTION 1 (Enclosure 2)

In the case of the steamline pressure predictions, "the coarseness of the spatial mesh," which models the steamline, is said to be the cause of "the spreading out of the calculated wave" as compared to measurement. In order to verify this difference, provide a sensitivity study of the effects of spatial mesh sizes. This study should demonstrate the effect of increased noding in the steamline on the steamline pressure predictions. If this sensitivity study does not verify the postulated effect on steamline pres sure predictions for the three Peach Bottom-2 turbine trip tests, an alternate explanation and verification should be provided.

RESPONSE

The effect of the mesh size on the steamline pressure prediction can be verified by comparing the results obtained with the three Peach Bottom tests and the KKK steamline results. The Peach Bottom analysis shows a spreading out of the initial pressure pulse relative to the measured data, while the agreement is much better for the IM steamline pressure.

The Peach Bottom steamliie is about 400 ft long (100 ft inside the contain ment and 300 ft outside, the containment). Six spatial nodes were chosen to represent this length. The DMH steamline is 327 ft long (93 ft inside and 234 ft outside) and is represented with 7 nodes. The average node length is about 30% smaller in the MM analysis and the improvement to the steamline pressure prediction is clear by comparing Figures 13-14 and 13-15 with Figures 13-1, 13-5, and 13-9 provided in response to Question 13 of Appendix B, Volume I. Despite this noding effect in Peach Bottom, the dome pressure is simulated quite well by the calculations for all three Peach Bottom Turbine Trips.

kl-1/Akl-2

NEDO-24154-A QUESTION 2 (Enclosure 2)

"The fact that heat loss terms are ignored in the transient calculation" was identified as the cause of the discrepancy between the dome pressure predic tions and measurements. Provide a calculation to Justify the basis for this explanation. Provide all assumptions on surface areas, volumes, specific heat, heat transfer coefficient, reactor coolant temperature, and steam and water flow rates. If this calculation does not fully verify the extent of the discrepancy, udditional explanation and verificatiou should be' supplied.

RESPONSE

Studies conducted since the time the draft qualification report was issued have shown that a large energy loss rate is required to explain the observed pressure overprediction. (See response to Question 39, Appendix B, Volume I).

The exact cause of this peak pressure bias is not known. The audit calcula tions carried out by BNL indicate a similar bias if a 26% bypass capacity were used, as in the ODIN calculation.

A2-1/A2-2

NEDO-24154-A QUESTION 3 (Enclosure 2)

Provide calculational verification that the core exit pressure oscillations between 0.4 and 0.7 sec are due to "ringing" in the instrument line. This verification should show that the magnitude and frequency of the oscillations correspond to those for the natural frequency of the instrument line.

Provide a full explanation and verification of the phenomenon.

RESPONSE

A discussion of sensor line effects on the transient pressure response measurements in the Peach Bottom tests is given in Section B.2.1.1 of the Peach Bottom turbine trip test report (Reference 4).

A3-11k3-2

NEDO-24154-A QUESTION 4 (Enclosure 2)

For the larger core exit pressure oscillations at times greater than 0.7 sec, the control rod movement has been postulated as the cause. Provide the core exit pressure response to other scram signals which verify this hypothesis.

These other scrams should be reactor trips vith minimum pressurization transients.

RESPONSE

No other experimental data exist to verify the postulate that the control rod motion was the cause of the sensed core exit pressure oscillations.

It is highly unlikely that these oscillations occurred in the reactor exit pleum. because oscillations in the neutron flux would have been observed.

A4-1/A4-2

NJEDO-24154-A QUESTION~ 5 (Enclosure 2)

In the evaluation of the neutron flux differences, two major contributors were identified (core pressure and scram motion). The variation due to these con tributora should be quantified and the total effect on neutron flux verified.

RESPONSE

The variation in peak neutron flux due to scram initiation time for the Peach Bottom Transient Test conditions is summrized in Table 5-1. The delay time assumed in the model is 0.17 sec for all three tests. A variation of 0.02 sec about the mean was chosen f or this analysis. The peak neutron flux is not very sensitive to the scram initiation time in tests numbers 1 and 2, because the rate of void reactivity increase has decreased by the time the scram has started. M~ore sensitivity is observed in the third turbine trip.

Overall sensitivity of flux to pressure variation is difficult to quantify because of the dynamic nature of the problem. A specific assumption about the transient pressure variation has to be made and the transient reevaluated.

The qualification report quoted atflux change of 10% due to a 1 psi pressure change. This relationship exists over a very short period of time, while the reactivity is high. No generalizations to other transients can be made about this sensitivity because it was only used to explain differences between the calculated and observed data.

A5-1

NEDO-24154-A Table 5-1 PEAK FLUX VALUES AS A FUNCTIOK OF SCRAM DELAY TIE Peak Neutron Flux TT3 Scram Delay. Time TT1 TT2 (sac) (Z) (Z) (2) 0.15. 330 420 340 0.17 338 438 420 0.19 338 "0 560 A5-2

NOEDO-24154-A QUESTION 6 (Enclosure 2)

In order to assure that no unexpected anomalies occur in the usage of ODYN, provide a sensitivity study on the key parameters and explain and verify any observed anomalies. The sensitivity study should be conducted for the Peach Bottom-2 tests, and for turbine trip without bypass and HSIV closure transients for typical BWR licensing conditions. Some of the key parameters which should be varied through the range of potential BWR transient conditions are scram rate, void reactivity coefficient, void distribution, flow rate and distribution, core pressure drop, stopvalve closure time, MSIV closure time (for pressurization transient), core exit pressure, axial power distribution, etc. Refer to the response for other studies requested in these questions as necessary.

RESPONSE

A considerable number of sensitivity studies have been carried out for license basis conditions. Input parameters have been varied in the response to ques tions in Enclosure 3. The peak neutron flux peak heat fluxes and ACP~s have been tabulated in these responses. Model sensitivities have been carried out

.in response to the questions in Appendix B, Volume 1. Table 6-1 contains a sumary of the results of the ODYN model sensitivities.

A6-1

NEDO-24154-A Table 6-1

SUMMARY

OF ONE DIMENSIONAL MQDEL SENSITIVITY STUDIES Peak Core Average Heat Flux

.Model Perturbation (% Rated)

(1) Base Casf 121.6 (2) Void Response Reduced 15% 117.8 (3) Void Response Reduced 5% 120.6 (4) Void Response Increased 4.8% 122.4 (5) Void Response Increased 14% 123.7 (6) Prompt neutron heating - 0.015% 122.6 (7) Prompt neutron heating - 0.022% 121.4 (8) Bypass flow fraction increased 20% 121.2 (9) Jet pump suction loss coefficient decreased 20Z 121.7 (10) Jet pump suction and diffuser losses decreased 20% (increased M ratio 5%) 122.5 (11) Separator L/A decreased 30% 122.0 (12) Separator L/A d-creased 90% 120.4 (13) Separator pressure drop decreased by 50% 118.9 (14) Recirculatiou loop L/A increased by 100% 121.4 (15) Core Pressure drop increased by 0.8% psi 121.4 (nominal - 24 psi)

Core pressure drop decreased by 0.7 psi 121.8 (17) Core pressure drop decreased by 2.4 psi 122.3 (18) Carry under fraction w 0.2% (Nominal - 0.1%) 121.4 (19) Carry under fraction - 0.01% 121.7 (20) Steamline pressure drop losses decreased 20% 122.5 (21) Steamline specific heat ratio y - 1.3 123.8 (22) Steamline specific heat ratio y - 1.2 122.6 (23) Core thermal hydraulic drift flux parameter Co increased 10% 121.5 (24) Co decreased 10% 121.9 (25) Drift flux parameter Vgj increased 20%

121 .6 (26) Vgj decreased 20% 121.5 (27) Subcooled void model R1 decreased 20Z 121.5 (28) Ri increased 20% 121.6 (29) Subcooled void parameter R2n n - 0.5 116.2 (30) R2a n - 1.25 122.4 (31) R2n =1.50 123-.1 A6-2

NEDO-24154-A QUESTION 7 (Enclosure 2)

In Section 2.1.1 the scram reactivity is compared using one-dimensional and three-dimensional procedures. In the three cases that were presented, the scram reactivity insertion calculated using the one-dimensional model is more conservative than that calculated using three-dimensional model except for the initial stages in the third case. At the initial stages of the scram in the third case, the power predicted by the three-dimensional pro cedure is higher than that predicted by the one-dimensional procedure.

Explain the reason for this nonconservatism, and assess its effect on the neutrop flux in the Peach Bottom tests and in ACPR prediction. Discuss how the turbine trip without bypass transient would be affected because of these differences in three-dimensional versus one-dimensional procedures.

RESPONSE

The case analyzed in Figure_2-70_of the qualification report is quite similar to the core conditions for Peach Bottom Turbine trip test number one from 50% power. At around 0.1 sec into the scram, the one-dimensional model under calculates the scram reactivity by about 101. With the large number of con trol rods in the core, the radial flux distribution is quite complicated, and the approximation scheme chosen to represent the time dependence results in a nonconservative bias. This bias in scram reactivity becomes smaller as the rod insertion fraction increases and is -11 at 0.4 sec. In the analysis of the Peach Bottom test conditions, the one-dimensional scram reactivity model would predict fluxes about 6% lower than the three-dimensional scram curve for that portion of the transient after the start of rod motion. If this particular scram curve were to be used in a turbine trip without bypass transient, there would be no appreciable flux transient at all because the start of rod motion would be triggered by the trip scram mechanism, causing the rods to begin moving about 0.5 sec earlier, so that at 0.4 sec, when the pressure wave hits the reactor core, about $0.35 negative reactivity has been inserted.

At 0.8 sec, about $3.0 scram reactivity exists, which is much larger than the positive void reactivity contribution.

At end-of-cycle conditions, the scram reactivity increases much more slowly with time, as can be found by comparing Figures 2-2 and 2-4 of the qualifica tion report. In the end-of-cycle case, the most important portion of the scram curve occurs from 0.2 to 1.2 sec after the initiation of rod motion and during the time of the rapid pressure increase. Here the end-of-cycle compar isons show the oue-dimensional model to be conservative. Also, biases occurring early in the scram motion, when the total reactivity is small, do not have a significant impact on the transient severity.

A7-l/A-7-2

NEDO-24154-A QUESTION 8 (Enclosure 2)

In Section 2.1.2i the void coefficient is defined.

a. Is the void fraction the effective neutron void fraction or is it the void fraction calculated in the thermal-hydraulic analysis?

b. It is stated that one-dimensional and three-dimensional void coefficients agree to within 5Z. Present the details of the wide variety of power levels and control rod configurations to show how this 5% is determined.

c. How is ihis uncertainty of 5% introduced in the ODYN model? What is the sensitivity to ACPR in the licensing basis transient?

RESPONSE

a. All void fractions quoted are void fractions calculated by the thermal hydraulic analysis.

b. Further discussion of the one-dimensional and three-dimensional void response comparisons is presented in the response to Question 2, Appendix B, Volume 1.

c. The difference between one-dimensional and three-dimensional void coefficients is a bias and not an uncertainty. It is shown in the response to Question 2 that this bias is always in the conservative direction for license basis conditions, and, therefore, no model uncertainty is assigned to the one-dimensional collapsing process.

AS-l/AS-2

NEDO-24154-A QUESTION 9 (Enclosure 2)

In Section 2.2 thermal and hydraulic models are compared.

a. Present the details of the standard GE BWE channel hydraulic model referred to in this section.

b. It is stated that some differences in void fraction exist between the standard GE EBWmodel and the ODYN code. These are attributed to dif ferent numerical treatments and assumption of constant pressure in the standard GE model. No difference is attributed due to the thermal nonequilibrium model and drift flux model in the ODYN code, particularly in the subcooled region. Discuss the sensitivity of the void fractious to these models.

C. Present all verification details of the standard GE BWR channel hydraulic model with void fraction measurements.

d. Provide a comparison of the ODYN thermal-hydraulic model with experi mental data.

RESPONSE

a. The details of the standard GE BW*l are presented in Chapter 4 of Reference 8.

b. The model used in the one-dimensional model is a mechanistic subcooled boiling model and is described in more detail in the response to Question 28 in Appendix B, Volume I. This model was used because it is compatible with the separate liquid and vapor continuity and energy equations used in the transient model. The steady-state hydraulics model uses a dif ferent model for subcooled boiling in which the subcooled flow quality is determined from the "profile fit", which is a function of the bulk enthalpy and the quantity hLd, the enthalpy at which subcooled boiling begins. The "profile fit" is an analytical expression for the interfacial heat flux up the channel. The sensitivity of the void fractions and transient results to subcooled void models is discussed in the reply to Question 28 of Appendix B, Volume I. The drift flux correlations are identical to those used in the standard GE hydraulic model.

c. Comparisons with void fraction and pressure drop measurements using the GE correlations are presented in the response to Question 28 of Appendix B, Volume I.

A9-L

NEDO-24154-A

d. The ODYI thermal-hydraulic model has been specifically compared to experimental data. It has been compared to the standard GE design models with close agreement. The standard GE design model or equiva lent equations have been used to test the appropriate void fractions and pressure drop estimates against experimental data.

A9-2

NEDO-24154-A QUESTION 10 (Enclosure 2) in Peach Bottom tests a certain weighted average for 7x7 and 8x8 bundles was used for calculations.

a. Present the details and basis for the weighting procedure.

b. Is the same procedure to be used in licensing basis transient analysis?

C. What is the uncertainty in &CPR associated with this procedure in the licensing basis transient?

RESPONSE

a. In the initial analysis of the Peach Bottom transient, the rod geometry parameters were weighted in proportion to the number of fuel rods of each type in- the core. Since that-time, a design procedure for transient analysis has been formulated which uses the rod dimensions of the dominant fuel type. That is, the fuel dimensions of the rod type having the largest number in the core is used. The Peach Bottom reactor at end-of cycle 2 has 576 707 bundles and 188 8W8 bundles. Use of the 77 param eters rather than the weighted average yeilded the same transient results for the Peach Bottom tests.

b. The dimensions of the dominant rod type will be used in the licence basis analysis.

c. The 707 fuel rods are larger in diameter and therefore have a longer time constant. Therefore, their use will yield a iore severe transient than the 8%8 parameters resulting in a conservative &CPR estimate. For Peach Bottom-2 this estimate is 0.002 ACPR/ICPR greater than the result obtained with average fuel rod parameters. A nonconservative result may be obtained, however, if 8%8 fuel were the dominant type but a significant amount of 7x7 fuel remained in the core. An examination of the current reload configurations shows that when the 8O8 fuel is dominant, there is less than n25% 707 fuel remaining. Based on the sensitivity study quoted above and a 20% loading of 7W7 fuel, the uncertainty .connected with this procedure is estimated at -0.002 &CPR/ICPR.

AI0-1/AI0-2

a.

NEDO-24154-A QUESTION 11 (Enclosure 2)

A value of 1000 Btu/ft 2-hr-*F was used for the gap conductance in Peach Bottom tests. It is stated that a 100% change produced only 5% change in peak flux.

Describe how the gap conductance was changed. What is the estimated change in ACPR?

RESPONSE

Figure 11-1. ihows the Peach Bottom Turbine Trip 3 total neutron flux calculated with gap conductances of 1000 Btu/hr-ft 2 and 500 Btu/hr-ft 2 . Small differences in neutron flux are observed because the entire flux pulse is only about 0.3 sec wide. Therefore, a very fast fuel time constant is needed to produce a moderate density feedback through the rod heat flux. Much larger values of gap con ductance will produce changes in the calculated flux response, as evidenced by the response with a gap conductance of 1500 Btu/hr-ft 2. The estimated difference in ACPR/ICPR between these extreme values is 0.005.

Al1-1

NEDO-24154-A W0O 40O C

f x

320 z

0 z

200 100 a

0 0.2 0.4 0.6 0.3 1.0 1.2 TIME 4,O Figure 11-1. Peach Bottcm-2 Turbine Trip 3 (69% Power) LReuzron Flux A1-2

NEDO-24154-A QUESTION 12 (Enclosure 2)

Present the details of instrument line daiping and equations for the second order response, including the characteristics of the transducer. Assess measurement uncertainty for both Peach Bottom and KIM tests.

RESPONSE

A discussion of the instrument line model appears in Section B.2.1.1 of Reference 19: This discussion references three papers which give the details of the equations used. A physical description of the sensor lines is also provided in Appendix B of Reference 4.

A12-1/A12-2

NEDO-24154-A QUESTION 13 (Enclosure 2)

Extend the pressure plots of Figure 3-6 through 3-14 and 3-40 through 3-45 through a period of 3.0 sec to include the peak pressure region both for Peach Bottom and KM tests.

RESPONSE

Figures 3-6 through 3-14 and 3-40 through 3-45 have been modified to include in excess of 3.0 sac in the qualification documentation.

A13-1/A13-2

NEDO-24154-A QUESTION 14 (Enclosure 2)

Assess the effect of not considering system heat loss in terms of 6CPR.

RESPONSE

The current practice of neglecting system heat loss has a negligible effect on ACPR. See the response to Question 39' of Appendix B, Volume 1.

A14-1/A14-2

NEDO-24154-A QUESTION 15 (Enclosure)

It is stated that a 1 psi change in pressure causes a change of 10% in peak neutron flux.

a. Assess this difference in terms of &CPR.

b. Assess the differences in &CPR for the Peach Bottom and M~ tests assuming the calculated conservative neutron flux inputs and the measured pressure inputs.

RESPONSE

a. It is quite difficult to assess the transient influence of pressure differences on flux calculations without some assumption about how the pressure differences vary in time. The statement about pressure differences in the qualification document was meant to show that small changes in pressure could change the flux near the peak value because the net reactivity was quite high, near $0.80.

b. In order to simulate the measured pressure inputs, the Peach Bottom transients have been calculated with a higher bypass capacity (i.e.., 35%.

instead of 26Z). (This is not meant to infer that the actual bypass capacity is 35%, but is assumed for pressure simulation purposes only).

The core exit pressure traces are shown in Figures 15-1 through 15-3.

Critical quality depends directly on pressure and will be influenced mainly by the pressure near the time of peak heat flux, which for the Peach Bottom tests is around 1.1 to 1.2 sec., The 352 bypass calculations show the flux to be virtually identical to the 26% bypass results shown in Question 13. However, after 1.0 sec, the core pressure has been reduced to agree much better with the experimental data. In the case of MD, the calculated pressure agrees quite weil with the experiment out to about 1.8 sec. The 4CPI./IC PR calculated with these pressure responses are sumarized in Table 15-1 along with the data-given and previously quoted. Hence, there is negligible difference between ACPR's obtained with measured and calculated pressure responses.

A15-1

NEDO-24154-L Table 15-1 SUlHARY OF ACPR/ICPR RESULTS FOR PEACH BOTTO!- TUREINE TRIPS ACPR/ICPR &CPR/ICPR &CPR/ICPR Test (Data) (26% bp) (351 bg)

PB2TTl 0.170 0.173 0.170 PE2TT2 0 .136 0.129 0.135 P!2TT3 O.132 0.141 0.131 0.077 0.0840 0.084*

• KKK results obtained with 50% bypass.

A15-2

31040

[t020 I

TI..

0 TIME (oft Figure 15-1. Peach Bottom-2 Turbine Trip I Core Exit Pressure (35% Bypass)

<N.j I"

'U N

1-0 0.2 0.6 0.9 1.2 1.5 1.8 i1 2.4 2.7 3.0 3J 3A 3.9 Ir MEI,.v Figure J5-2. Peach Bottom-2 Turbine Trip 2 Core Fxit Pressure (35% Bypass)

1040 j Ivv 1020 I

xl~

U cc' 100 Os6ue 0 0 r*dý*re~e IAe TISAE fuel Figure 15-3. Peach Bottom-2 Turbine Trip 3 Core Exit Pressure (35Z Bypass)

NEDO-24154-A QUESTION 16 (Enclosure 2)

Discuss the effect of the uncertainty in the Doppler coefficient on ACPR calculations for the turbine trip without bypass transient in a typical BWR/4.

RESPONSE

Uncertainty in the Doppler coefficient and its impact on the transient analysis is discussed in the responses to Questions 12 and 38 of Appendix B, Volum' 1.

A.16-1/A16-2

NEDO-24154-A QUESTION 17 (Enclosure 2)

It is stated that in the reactor the boiling boundary is different in each channel and the variation in void fraction with axial position is smaller than the one-dimensional estimate. It was also stated that the one-dimensional model represents the average conditions. Present the procedural details for representing the three-dimensional void fraction variation in one-dimensional.

RESPONSE

The procedural details are presented in the One-Dimensional Transient Model report. The effects on the void response and the transient are discussed in the response to Question 2 of Appendix B, Volume I.

A17-1/A17-2

NEDO-24154-A QUESTION 18 (Enclosure 2)

It is stated that measurement uncertainty in pressure is t2 psi and in flow it is -3Z. This gives an uncertainty of 0.01 in ACPR/ICPR.

a. Provide ACPR/ICPR versus time curves for all Peach Bottom and KM tests.

b. Discuss the effects of instrumentation filters on the uncertainty assessment.

c. It is noted that the model and data driven ACPR/ICPR values are also within 0.01, although the differences between the experimental and calculated values of pressures may be on the order of 10 psi. Discuss this apparent discrepancy in view of the statement that was made on pressure measurement uncertainty of +/-2 psi and also an earlier statement that 1 psi difference in pressure may make a difference of I0Z change in peak neutron flux.

RESPONSE

a. The General Electric Company does not normally generate plots of critical power ratio (CPR) versus time. Following General Electric Thermal Anal ysis Basis (GETArB) procedures, the most Limiting pressure and power increase transients are evaluated to determine the largest change in the critical power ratio (ACPR). The addition of the ACPR to the safety limit HCPR provides the minimzn operating CPR required to avoid violating the safety limit should this limiting transient occur. The operating limit and the safety limit MCPR are independent of the path taken by the transient.

Therefore, plots of CPR versus time serve no useful purpose and are not generated.

b. Instrumentation effects on the experimental uncertainty are discussed in Appendix B of Reference 4.

c. The draft qualification report unfortunately gives the impressiou that the +/-0.01 uncertainty was entirely due to the pressure and flow uncer tainty. In reality, the change in CPR due to flow and pressure changes are significantly smaller than inferred from the report. In fact, based on ACPR sensitivities in pressurization transients, a 2 psi pressure uncertainty results in a 0.001 uncertainty in &CPR/ICPR, and a 3% flow uncertainty results in a 0.002 &CPR/ICPR uncertainty. The draft report should have stated that the 10.01 &CPR/ICPR uncertainty was based purely on a judgment of the uncertainties involved in the entire process of A18-1

NEDO-24154-A evaluating &CPR from the experimental data. Based on the pressure sensitivities, the 10 psi pressure difference observed between the calculation and experiment would result in- a 0.005 ACPR/ICPR. In reality, the dependence on pressure is more complex. A more accurate evaluation of the ACPR effects due to pressure is presented in the response to Question 15.

A18-Z

NEDO-24154-A QUESTION 19 (Enclosure 2)

In the meeting. on May 3, 1978 to discuss' the review of the ODYN computer program, it was stated by General Electric that they would provide a letter discussing the subject of scram reactivity and how it would be included in the ODYN documentation. This documentation of the scram reac tivity calculations by General Electric is required to satisfy commitments on, for example,, the GESSAR-251 docket. Therefore, General Electric should provide this information describing how they intend to address the topic of scram reactivity.

RESPONSE

The ODYN Engineering Computer Program contains the analytical models for treating the interaction between neutron physics, core thermal hydraulics, and the fuel heat transfer mechanisms . These coupled sets of differential equations are solved so as to include the effects of time dependent varia tions in the axial direction. The formulation of this model, coupled with the required input from the GE three-dimensional steady-state simulator program, provides a technically sound way of computing scram reactivity.

The required inputs for performing plant transient analyses for reactor scram is reactivity data as a function of control blade insertion. As. stated in our previous communication concerning the implementation of the ODYK code, we plan to continue to use the REDY code for analyses of specific transients, even after the NRC has approved the application of ODYN for a prescribed set of transients. In order to use our best technology available, GE plans to use the ODYN code to calculate scram reactivity as an input for those cases where the RED! code will be used to perform the plant transient analysis.

This will be Implemented by a change to GE's Technical Design Procedures to use the ODYN code to calculate scram reactivity for all cases where it is required as an input for REDY plant transient analysis.

The primary ground rules for this analysis will remain identical to those used in our current Technical Design Procedures. These are:

1. Initial conditions are calculated using the standard steady-state three-dimensional core simulator.

A19-1

NEDO-24154-A

2. Pressure is maintained at a constant value.

3. Recirculation pump speed is maintained at a constant value.

4. The effects of void collapse in the core due :o decreasing fuel surface heat flux are included in the calculation.

Implementation of this improved Technical Design Procedure will assure application of our best technology on a consistent basis.

A19-2

NEDO-24154-A QUESTION 20 (Enclosure 2)

Provide References 4, 5 and 6 in MFN 014-78.

RESPONSE

Reference 4 of MFh 014-78 has been published and the Staff has received copies.

This report is EPRI-NP-564, dated June 1978.

Reference 5 is attached.

Reference 6 is a standard reference for fluid flow and should be readily available to the Staff.

A20-1/A20-2