ML19209B945
| ML19209B945 | |
| Person / Time | |
|---|---|
| Site: | Rancho Seco, Crane |
| Issue date: | 10/08/1979 |
| From: | Mattimoe J SACRAMENTO MUNICIPAL UTILITY DISTRICT |
| To: | Ross D NRC - TMI-2 BULLETINS & ORDERS TASK FORCE |
| References | |
| NUDOCS 7910110257 | |
| Download: ML19209B945 (29) | |
Text
..
i SMUD V SACRAMENTO MUNICIPAL UTILITY DISTRICT C 6201 s street, Box 15830, sacramento, California 95813; (916) 452-3211 October 8, 1979 I
{
Mr. D. F. Ross Jr., Director Bulletins and Order Task Force Office of Nuclear Reactor Regulation U.S. Nuclear Regulatory Comission Washington, D.C.
20555 Docket No. 50-312 Rancho Seco Nuclear Generation Station, Unit No. 1
Dear Mr. Ross:
The Sacramento Municipal Utility District herein provides additional information on items required by your letter of August 21, 1979. The District's letter of August 31, 1979, provided initial information which in addition satis-fied the September 15, 1979, information committment made in the District's July 26, 1979, letter. At the B&W Owner Group /NRC meeting on September 13, 1979, in Bethesda, Maryland, and in subsequent phone conversations between B&W and the NRC staff, the schedule was modified.
The fcllowing items required in Attachment A to Enclosure 1 of your August 21, 1979, letter are attached:
Item TB Safety Valve Model Justification Item 2A OTSG Model Justification Item 4 CRAFT-2 Simulation of TMI-2 Accident The remaining items required by Attachment A to Enclosure 1 of your T
letter dated August 21, 1979, will be submitted per this schedule:
Item 1A CRAFT Analyses with Sequential 12/11/79 AFW Flow Item 2B Repressurization to PROV Setpoint with PORV then Sticking Open Qualitative Assessment 10/25/79 Detailed Analysis 2/11/80 ry Items 3A-3D Noncondensable Gasses in the 11/5/79 (d
- 6 N
y Primary System p
4 11.1 07 ef s g
7910110 P
Mr. D. F. Ross Jr. October 8,1979 Item 5, Semiscale Experimental Prediction and Item 6, LOFT pre-test predictions are being handled by NRC staff /B&W direct comunication.
If you have additional questions concerning this submission, please contact us.
Sincerely, hta$,,%
C
/John. Mattimoe Assistant General Manager and Chief Engineer Enclosures o
I i
ATTACHMENT 1 7
RESPONSE TO QUESTION 13 0F ATTACILME. T A TO D.F. ROSS (NRC) I.ETTER DATED S/21/79 Ouestion 1B.
Provide justification of relief and safety valve flow models used in the CRAFT 2 code.
RESPONSE
~
1 The CRAFT 2 code, which is documented in topical report BAW-10092, Rev. 2,
does not have any special models for prediction of the fluid discharge through the relief and safety valves.
Rathe, they are modeled. as leak paths from'the pressurizer control volume to the cont'ainment. Thus, the Bernoulli (orifice) equation is used for subcooled discharge, while the Moody correlation is used for saturated steam or two-phase discharge.
These models are the same as those used in B&W's ECCS Evaluation Model.2 Since little information exists on the flowrate through pressurizer valves fo'r cubcooled or two-phase fluid conditions, it is impossible to ascertain the accuracy of this modeling technique.
Since pressurizer leaks are in-herently less severe than the breaks in the cold leg pump discharge piping analyzed to demonstrate compliance to 10 CFR 50.46, a truly realistic model for the discharge rates is not necessary.
However, the modeling technique utilized is expected to reasonably approximate the' discharge rates and'their subsequent effect on the RCS.
System response to relief valve actuation have been analyzed and submitted to the Staff in Section 6 of the May 7,'1979,' report. The cases speci-3 fically analyzed were:
~
1.
A loss of main feedwater accident which results in actuation and a subsequent sticking open of the pressurizer relief valve was addressed.
Offsite power was assuced to remain available and only one HPI train was used for emergency core cooling. This analysis is similar to the TMI-2 event that. occurred on March 28, 1979, and demonstrated that, if one HPI pump remained availabic, no core uncovery would have oc-a curred.
4
~
2.
A stuck open PORV assuming a loss of offsite power and only one HPI train availabic was analyzed. Results of this evaluation demonstrated that core uncovery would al:;o,not occur.
1131 030
An additional analysis of the effect of a pressurizer break which supple-mented those presented in reference 3, was provided to the Staff in a Ictter from J.l!. Taylor (B&W) to R.J. Mattson (NRC) dated May 12, 1979.
That analysis c>amined the effect of the stuck open PORV case, Case 2 above, except the auxiliary feedwater system was assumed inoperable.
The results of that evaluation showed that, even without auxiliary feedwater, one HPI pump can handle the accident provided that realistic decay heat valves are utilized.
In all of these evaluations, the PORY was modeled via a leak Path representation in the CRAFT 2 code. The orifice area of the PORV was modeled as the leak area (1.05 in.2) and a discharge coefficient of 1.0 was utilized.
The method for modeling the PORV described above does result in a pre-dicted steam flowrate, at the valve rated pressure, which is in excess of the design (rated) flowrate. An alternative modeling approach is to use a discharge coefficient (C ) which, at the valve rated pressure, would D
yield the valve rated flowrate.
For the 177-FA plants, this is a CD approximately 0.85.
For the first two cases described above, this model-ling approach would result in a slower system depressurization and a
~
slower discharge of the RCS inventory.
Thuc, the use of a C used
=
D in previous evaluations results is a conservativa assessment of the tran-sient. For the third case, the use of a smaller C v uld result in a D
larger repressurization following the loss of the SG as a heat sink and the chaage in the discharge'from steam to two-phase flow. However, use of a C 0.85 would result in an inventory loss less than that calcu-D lated fa reference 4 and no core uncovery would occur.
Besides the cases involving actuation of the pressurizer relief valves, analyses were perfor=ed for a total loss of SG heat sink and are provided in references 5 and 6.
In those evaluations, the pressurizer safety valves were exercised. To model these valves, the leak path representa-tion was used with the Icak path opening and closing at the opening set-point of.the valve.
The valve area and C was chosen such that the rated j
D flowrate for the valve would be simulated at the valve rated pressure.
I, Because of the large relief capacity of the valve, the system pressure oscillated within a few psi of the valve setpoint and the valve was exercised intermittently., Thus, any discrepancies between the modeled 2
1131 08d
~
and the actual relief capacity of the pressurizer safety valve is not ex-pected to significantly alter the system response.
While there is little information availabic on the discharge rates through the pressurizer valves, it is also important to note the breaks in the pressurizer are bounded by breaks in the cold leg pump discharge piping.
8 Pdmp' discharge breaks are analyzed to show conformance of the ECCS to meet lthe criteria of 10 CFR 50.46. The reason that cold leg breaks bound breaks in the pressurizer was discussed in detail in reference 3.
Therefore, it is not necessat:y to si=ulate the actual relief capacities of the pressuri-zer valves in order to demonstrate the ability of the ECCS to mitigate the consequences of a loss of RCS inventory through the valves within the
~
criteria o'f 10 CFR 50.46.
REFERENCES 1
BAW-10092, Rev. 2, " CRAFT 2 - FORTRAN Program for Digital Simulation of a Multinode Reactor Plant During LOCA," R.A. Hedrink, J.J. Cudlin,
. and R.C. Foltz, April 1975.
i IAW-10104, Rev. 3, "B&W's ECCS Evaluation Model," B.M. Dun, et al.,
August 1977.
3' Letter J.H. Taylor (B&W) to R.J. Mattson, May 7,1979, " Evaluation.
of Transient Behavior and Small Reactor Coolant System Breaks in the 177-FA Plant."
4 Letter J.H. Taylor (B&W) to R.J. Mattson, May 12, 1979.
5 Letter from R.B. Davis to 177 Owner's Group,
Subject:
" Complete, Loss of Feedwater Traasient," September 11, 1979.
6 Letter from R.B. Davis to Mr. C.R. Domeck,
Subject:
" Complete Loss of Feedwater Transient.on Davis-Besse," September 11, 1979.
1131 08k
~
ATTACHMENT 2 RES"0dSE TO QUESTION 2A 0F ATTACHMENT A 0F 8/21/79 D.F. ROSS LETTER Question 2A.
Provide justification that the 3 node steam generator model used in the CRAFT 2 analysis of small breaks is adequate for the prediction of steam gcnerator heat transfer.
RESPONSE
The B&W ECCS Evaluation Modell for small breaks utilizes a three-node repre-sentation, in the CRAFT 2 simulation, for the predicti,on of steam generator heat transfer following a small break. Two of the nodes, stacked verti-cally, are used to model the primary side of the once through steam genera-tor (OTSG). The upper node includes the hot ieg piping, from the center on the 180* U-bend at the top of the vertical section of the hot leg to the SG upper head, the upper head of the SG, and the upper one-half of the tube region. The lower node simulates the lower one-half of the tube region. The thire' node is used to model the secondary side of the OTSG.
To evaluate the suitability of this modeling technique, the unique charac-teristics of the OTSG and its effects on the s=all break transient must be examined. As is shown later, for small breaks evaluated with the auxiliary feedwater system operable, heat removal via the SG is not necessary for the worst case breaks, i.e.,
those that result in core uncovery, in order to successfully nitigate the transient. For the smaller breaks, heat re-moval via the SG is necessary.
The three-node representation utilized appropriately models the heat transfer characteristics of the OTSG. For the smaller breaks, heat removal via the steam generator is cessary and the heat transfer characteristics of the OTSG must be appropriately con-sidered. Althoug5 the 3-node SG model does not rigorously account for the heat transfer process that will occur, it does provide a reasonable representation of the effects of these heat transfer prrcesses in the OTSG. Since these smaller breaks exhibit large margin to cure uncovery, i
the CRAFT 2 SG =odel is adequate for demonstrating compliance to 10 CFR f
50.46.
Inperformingsmallbreakevaluations,theCRAhT22 code is used to pre-dict the hydrodynamic response of the primary system _ including the effect,
. s N 31 08)
O
i of GG heat transfer during the ' transient. The option 2 SG model, which is s.plained in detail in Section 2.6 of topical report BAW-10092, Rev. 2,
? utilized to predict heat flow in the SC.
The calculation progresses' basically as follows:
1.
Based upon the initial steady-state heat transfer characteristics of the OTSG and the initial primary and secondary fluid temperatures, an overall UA for each region of the SG is calculated.-
2.
The calculated steady-state UA can be modified by user-specified in-put options.
These include an input multiplier table versus time, multiplied based on the primary side control volame mixture height during the transient, and a multiplier for reverse heat transfer, i.e.,
^
heat flow from the secondary to the primary side of the SG.
3.
Using the modified UA and the calculated primary and secondary side control volume temparatures, the amount of heat transferred is calcu-lated.
In performing the small leak calculations for de=onstrating c0=pliance to 10 CFR 50.46 for the operating B&W plants, no input multiplier versus time is utilized, nor is the modification based on prfmary side mixture level used. However, a multiplier for reverse heat transfer of 0.1 is utilized. This cultiplier and its basis is explained in the ECCS evalu-ation model topical reportl and is utilized to reflect the change in heat transfer regime on the secondary side of the SG for reverse heat flow.
The OTSG design of the B&W designed operating NSSs allows use of a simplis-tic model for calculation of SG performance during a small LOCA transient.
With the loss-of'-offsite power, assumed in design calculations for small breaks, and t'he subsequent loss of main feedwater, the auxiliary feeJ-water system is actuated and will become operable in t.nproximately 40 sec9nds and control the secondary side level.
The auxilf ary feedwat'r
~
entsrs the SG very high, approximately.2 feet below the upper SG tube i
sheet, and is sprayed onto the tube bundles.
Thus, heat transfer will k
occur in the upper portion of the SG inctpendent of the actual 1cvel in the SG.
The introduction of auxiliary feedwater to th_ SG has two ef-fccts on the small LOCA transient.
First, it raises the thermal center in the SG during the natural circu'1ation phase of the accident which 113i 08 >
e
.m
results in a centinuation of circulation through the RCS, for some period of time, even while inventory is lost from the primary system.
Later in the transient, after sufficient inventory has been lost from the system, circulation will be interrupted and the auxiliary feedwater, for a certain range of small breaks, will condense steam on the primary side of the SG; thereby maintaining the primary system pressure near the secondary side pressure. The analytical approach utilized for the small break evaluation is consistent with this performance of the auxiliary feedwater system.
It should be noted that between the time that circulation through the loops is lost and the time that' the primary side SG level has dropped to the point where condensation heat transfer will occur, system repressuri-zation can occur as heat removal via the SG will be lost.
This phenomena occurs only for the very small sized small breaks in which the SG heat removal is necessary.
If simulation of this repressurizatica phenomena of the very small breaks is desired, an additiondl node would be needed in the small break model in order to separate the hot leg and SG upper plenum volumes'from the tube region. This will allow steam to accumulate in the upper regions of the RCS without being affected by heat removal that occurs in the steam generator.
In the analyses presented in reference 5 for these smaller sized breaks, a model which included the additional node was utilized and showed that the repressurization phenomena does not result in core uncovery..
It is also important to note the role of the SG on the small break tran-sient in order t6 evaluate the appropriateness of the SG model utilized in small break evaluationa. Licensing calculations for the operating B&W units have previously been submitted to the Staff in references 3 and 4.
These evaluations have shown that the worst case small breaks, i.e., breaks which result in core uncovery, occur for breaks in excess of 0.05 ft2 As demonstrated in the May 7, 1979 reports, SG heat re= oval is not necessary for breaks of this size. For smaller breaks, SG heat remeval is necessary as the break alone is not sufficient to rc=ove enough fluid volume and I
cnergy to depressurize the RCS. However, as demonstrated in reference 5, these breaks are of no consequence as.the SG heat removal and the slower discharge rate for these breaks easily prevents core uncovery.
/
'l 1131 0842
-. - -. ~...
As demonstrated, the SC model utilized in tha small break evaluations for the operating plants appropriately accounts for the effcet of the spatial heat removal processes that will occur in the OTSG during a small break.
It was also shosn that the SG performance is not important for the worso case small breaks.
Thus, the CRAFT 2 SG model is adequate for demonstrating compliance of the ECCS to 10 CFR 50.46.
REFERENCES I
"B&W's ECCS Evaluation Model," BAW-10104, Rev. 3, Babcock & Wilcox,
. August 1977.
2 R.A. Hedrick, J.J. Cudlin, and R.C. Foltz, " CRAFT 2 - Fortran Program for Digital Simulation of a Multinode Reactor Plant During Loss-of-Coolant," BAW-10092, Rev. 2, Babcock & Wilcox, April 1975.
3 Letter, J.H. Taylor (B&W) to S. A. Varga (NRC), July 18, 1979.
4 "Multinode Analysis of Small Breaks for B&W's 177-Fuel' Assembly Nuclear
, Plants With Raised Loop Arrangement and Internals Vent Valves,"
BAW-10075A, Rev. 1, Babcock & Wilcoi, March 1976.
5
" Evaluation of Transient Behavior and Small Reactor Coolant System Breaks in the 177-Fuel Assembly Plant," Babcock & Wilcox, transmitted via letter from J.H. Taylor co R.J. Mattson, dated Iby 7,1979.
O e
5 i
~.
.i 113i 08h
ATTACHMENT 3 CRAFT 2 SIMULATION OF THE MARCH 28, 1978 THI-2 TRANSIENT I.
INTRG'7CTION In the May 7, 1979 " Blue Book" reports, a CRAFT 2 simulation of the first hour of the THI-2 transient was presented. That analysis has since been modified and updated to include more recent estimates of the net makeup to the RCS during the event. This report presents the.results of the latest B&W CRAI"I2 simulation of the TMI-2 event and covers approximately the first 2 hours2.314815e-5 days <br />5.555556e-4 hours <br />3.306878e-6 weeks <br />7.61e-7 months <br /> and 20 minutes of the transient.
1 The small break ECCS evaluation model, which is described in topical report BAW-10104 and the July 18, 1978 letter report, was used,with some "best estimate" modifications, for the simulation. Actual TMI-2 data were combined with available information about the operator actions to determine estimates of the HPI and AW injection times and flow rates.
The simulator results (described in detail in the "Results" section) show all the trends and very good comparisons to the actual plant data of system pressure, temperature and pressurizer level.
The analysis also predicts the time for the start of core uncovery which is in reasonable agreement with the NSAC-1 report. Thus, the CRAFI2 code is shown to benchmark very well versus the TMI-2 data and is suitable for the performances of small break evaluations.
2, METHOD OF ANALYSIS The CRAFT 2 code which is documented in topical report BAW-10092, was used to simulate the TMI-2 reactor coolant system hydrodynamics.
The model uses one node for the reactor building, two nodes for the secondary system, and 23 nodes to simulate the reactor coolant system, including four nodes for the pressurizer. A schematic diagram of the model it shown in Figure 1.
i The analytical model used for this simulation is basically the same as B&W's I
ECCS evaluation model. However, certain input assumptions which differ from the evaluation model approach, were =ade in order to obtain a "best estimate" simulation. These assumptions are described below:
a.
The initial core power level used in the model was 102% of 2772.
However, following reactor trip, the fission product decay hea.t was adjusted to 98% power op,cration. The decay heat curve utilized 113108f 3
is 100%, instead of 120% required by Appendix K to 10 CFR 50, of the ANS 5.1 decay heat curve.
b.
A loss of the main feedwater pumps, which is the initiating transient, was assumed at ti=e zero.
In order to account for potential draining of secondary side fluid from the steam generator downcomer into the tube regton, a main feedwater coast-down of 10 seconds was utilized.
c.
A turbine trip coincident with the loss of main feedwater was assumed. This results in the steam generator pressure being controlled by a combination of the turbine bypass valves, the atmospheric dump valves and the main s~ team safety valves.
For the first 90 minutes, the turbine bypass valves control the secondary side pressure.
In the simulation, these valves were set at 1025 psig.
d.
The CRAFT 2 input was adjusted to open the pilot-operated relief valve (PORV) at 8 seconds. This opening time had to be input, and the open valve simulated, since CRAFT 2 code does not have models for the pressure relief systems of the RCS.
Preliminary TMI-2 data was used'to determine the FORV opening time.
Present TMI-2 scenarios indi.cate that the PORV actually opened at 3 seconds.
As will be shown in the results section, if the CRAFT 2 code had an explicit PORV model, it would have predicted the opening at 3 seconds, e
e.
The reactor scram was chosen to occur at 10 seconds based on preliminary TMI-2 data. Since the CRAFT 2 code does not have provisions for a reactor trip on high pressure, this had to be simulated based on time.
f.
The Icak area utilized for the PORV is 1.05 in.
and represents 3
the orifice area of the valve. The Moody critical discharge correlation was utilized to predict the fluid lost through the PORV.
For the first 4 minutes of the simulation, a discharge coefficient (C
f was used. For the remainder of the evaluation a C D
D 1.0 was employed.
9 3
113i08%
I
g.
Actuation of the High Pressure Injection System (HPI) was based on g
ESFAS signal of 1615 psia. This resulted in the actuation of the 2 HPI systems at 1 minute and 45 seconds into the transient, as opposed to 2 minutes an' 2 seconds which was the make-up flow initiatfon time at TMI-2.
Between 275 and 6100 seconds, the HPI flow vas assumed to be throttled by the operator to an average flow of = 34 gpm. This value is based on preliminary assessment on the n_e_t, makeup flow to the RCS. No explicit modeling of t
letdown was used, only net flow was simulated. After 6100 seconda, an average net makeup (HPI) of 42 spm was utilized.
h.
A, four-node pressurizermodel was used in the evaluation in order to reduce instantaneous artificial condensation in the pressurizer.
This phenomenon, which occurs when the subcooled reactor coolant fluid mixes with two-phase pressurizer fluid, results from the equilibrium model limitations of the code. This model is necessary only to predict the response of the RCS.during the initial phase of the loss of main feedwater event. Also, the pressurizer surge line resistance was updated to reflect more realistically the TMI-2 surge line.
- i. Steam Generator Modeling - The steam generator model was modified to account for the following phenomena:
1.
The overall heat transfer coefficient (between - ' mary and secondary) was assu=ed to ramp to zero in one minute to account for the delayed auxiliary feedwater injection.
2.
Full heat transfer coefficient was reinstated at 500 seconds to account for the auxiliary feedwater injection af ter 8 minutes.
3.
Auxiliary feedwater was initiated at 500 seconds with half of the design AW flow capacity and with the SG level controlled to 3.fect.
With the reactor coolant pumps on, AW is controlled
{
by the-ICS to 3 feet.
4 Steam Generator B was assumed to be isolated at 1 hour1.157407e-5 days <br />2.777778e-4 hours <br />1.653439e-6 weeks <br />3.805e-7 months <br /> and 41 minutes based upon preliminary TMI-2 data. This was simulated bysettingtheheatthhnsfercoefficientacrosstheBsteam generator to zero.
f
5.
The auxiliary feedwater control level was manually raised to 50% on the operating range at I hour and 45 minutes into the transient due to the loss of thd RC pumps.
6.
The main steam safety valves were modeled to open at e
$400 seconds, and the feedwater flow was increased at 6100 seconds. This was done to simulate the steam generator A depressurization following the operator's attempt to increase feedwater flow to steam generator A at about 1 hour1.157407e-5 days <br />2.777778e-4 hours <br />1.653439e-6 weeks <br />3.805e-7 months <br /> and 34 minutes,
- j. The RC pumps in the B loop were tripped at 4400 seconds; the A loop RC pumps were tripped at 6060 seconds.
These values are consistent
^
with the THI-2 data.
Table 1 provides a comparison of the assumed times for various system actuations and operator actions to the NSAC scenario. As shown, the values utilized are reasonable compared to the actual performance during the THI-2 transient.
4 e
e O
v e
e i 1.5 i CGP
3.
RESULTS 3.1 System Pressure Figures 2 and 3 compare the zaactor coolant systen pressure calculated by CRAFT 2 to the TMI-2 data. Following the loss of main feedwater, the pressure in the RCS rose sharply due to the decreased heat removn1 across the SG.
As shown by Figure 2, the CRAFT 2 prediction overpredicts the pressure during this phase of the accident due to the delayed opening of the PORV 3 seconds in the transient versus 8 seconds for the CRAFT 2 simulation, and the delayed reactor trip, 8 seconds for the transient versus 10 seconds for the simulation.
If the CRAFT code had an explicit model for the PORV, opening of the valve rould have been consistent with the data and a better comparison would have bcea obtained. After the re-The cal'ulated pressure drops below the actor tripped, the RC pressure decreased.
e actual data after 20 seconds. This is apparently caused by the 10 second main feedwater coastdown employed in the simulation overprediating the drainage of secondary downcomer fluid to the SG.
After the SG drics out, approximately one minute, the difference between the prediction and the data decreases.
Approximately 5 minutes into the transient, the fluid in the hot l'g flashed-e due to the depressurization of the RCS and the system pressure increased. As indicated on Figure 3, the CRAFT 2 code properly predicts the system repressuri-zation time, but overpredicts the actual pressure. The overprediction of sys-tem pressure is probably caused by the assumed net makeup to the RCS during this time period. Although the HPI was throttled ta a net makeup of 34 gpm during this time interval in the simulation, between 4:58 and 6:58, the NSAC scenario of events indicate that the letdown flow was in excess of 160 gpm.
Thus, it is quite probable that there was a decrease in inventory in the RCS due to the high letdown over this time period.
At 8 minutes and 18 seconds, auxiliary feedwater flow was readmitted to the SG and primary system pressure decreased (Figure 3) to approximately 1100 psig and was maintained at that valut up to approximately one hour and 20 minates.
1 As shown by Figure 3,-the CRAFT 2 prediction is greater over this period by
{
about 100 psi. The coolant pressure was measured in the hot leg during the accident; the predicted system pressure shown is the core pressure. The ac-tual predicted hot leg pressure is about 60 psi lower than the predicted core pressure. Also, the pressure in the s'econdary side was held in the CRAFT 2 s
9
simulation at 1025 psig, while the measured value was 1000 psig, resulting in an additional deviation. Thus, the CLAFT2 prediction reasonably follows the transient behav~1or over this period when the deviations are considered.
It should also be noted that the primary syrtem pressure during this phase of the transient is basically controlled by the SG.
The CRAFT 2 prediction did not demonstrate fluctuations in system pressure during this period as the sec-ondary pressure of the SG is assumed to be regulated at 1025 psig.
The plant data shows that the secondary side SG pressure was not held constant over this period, but fluctuated.
At one hour and 34 minutes, the RCS pressure dropped due to an apparent atte=pt
^
by the operator to increase feedweter to the A SG.
The analysis attempted to simulate the depressurization effect of the increased auxiliary feedwater flow by opening the relief valves at 5400 seconds and increasing the auxiliary feed-water flow at 6100 seconds.
This modeling technique was utilized as little information is available on the actual auziliary feedwater flow delivered to the SG during this period. As shown by Figure 3, this resulted in an underpredic-tion of the primary system pressure until 7500 seconds and an overprediction for the remainder of the transient analyzed.
3.2 Pressurizer Level A comparison of the CRAFT 2 predicted pressurizer level to the TMI-2 data is provided in Figure 4.
As shown, there are two pressurizer level predictions given La the figure. The first, entitled mixture level - CRAFT, is the calcu-lated mixture level within the pressurizer. The sec ond, entitled instrumenta-tion reading - CRAFT, is the calculated liquid level that would be "seen" with-in the pressurizer icvel tqps and is directly comparable to the TNE-2 data.
The initial pressurizer response and comparison to the loss of main feedwater event (first 4 minutes of the transient) is not easily discernable in Fig wa 4.
It was, however,
._ ussed in the May 7, 1979, report.
During this phare of the accident, the pressurizer level responded in a similar manner as the system pressure. Also, the comparison of the predicted to the actual response f
of the pressurizer level is similar. That is, the rise in pressurizer lev _.
I during the first 10 seconds is overpredicted and the pressurizer level af ter reactor trip is underpredicted. The reasons for tnis are the same as those discussed previously in section 3.1.
1 i 3 i 091V W
e-
The significant aspect of this comparison is the predicted mixture Icvel re-sponse to the predicted instrument reading response'during the transient. As shown by Figure 4, tne predicted instrument response.and the measured response are in good agreenent throughout the simulation.
However, as shown by the fig-ure, although the instrument reading is on sca'le for portions of the first 101 minutes of the transient, the actual predicted mixture level af ter 6 minutes is at the top of the pressurizer. Thus, a two-phase mixture exited through the valve during this entire period. After 101 minutes, only steam was entering the pressurizer through the surge line (note that the RC pumps have been tripped'-
and the pressurizer mixture had reached a sufficient void fraction to allow for phase separation at the top of the mixture and only steam started to flow out.
E 3.3 System Flow
_ Figure 5 shows a comparison of the predicted an d transient loop flows. As shown, the predicted flow rates do not match well with the actual data.
This disagreement is caused by two factors. First, loop flow was measured by Gentillis tubes, which are calibrated based on single phase flow.
Their actual performance during two-phase flow is unknown.
Secondly, performance of the RC pumps with two-phase flow is not well understood.
In performing the evaluation, a two-phase pump degradation multiplier based on the semiscale pump tests was utilized.
This multiplier results in a sharp decrease in pu=p head once any significant voiding is calculated at the pump inlet. As shown, at 55 minutes, the loop flow sharply decreased due to this effect. Although the agreement is not exec 11cnt, the pump flow calculation does not appear to have significantly affected the simulation.
3.4 Hot and Cold Leg Temocratures
~ Figures 6 and 7 show a comparison of the predicted versus actual response of the -
hot and cold Icg temperature measurements during the transient. After 5 minutes and up to the time the core started touncover, the RCS was saturated, and the fluid temperature comparison has the same deviations previously discussed in section 3.1.
.[:
After the core starts to uncover, which occurs at approximately 110 minutes, the hot leg temperature measurement indicated superheated steam (Figure 6).
How-ever, the CRAFT 2 prediction does not exhibit this behavior. This is due to the one-node representation of the cer6.and the equilibrium assumption of the
CRAFT code. As long as fluid is predicted to rc=ain within the cora node, re-gardless of the actual amount of core unenvery, the one-node representation calculates the exiting steam temperature to be sarurated. However, the actual physical process results in saturated steam at the top of the core mixture level. This steam superheats as it rcccives energy from the uncovered portion of the fuel pins. A multinode representation of the core would be necessary to predict the hot leg temperature response durint,this period.
3.5 System void Traction The average system void fraction evolution for the primary, system, excluding the pressurizer, is shown on Figure 9.
Due to the continued loss of RCS in-ventory through the PORV and the inadequate net makeup to the RCS, the system void fraction increases almost linearly from 10 until 101 minutes into the transient. At 101 minutes all the RC pumps have been tripped.
At this time, the RCS liquid inventory is distributed as follows; the RV is filled to slightly above the top of the core; the loop seal in the B loop is full; the A loop has very little inventory. During the subsequent 30 minutes, the RV inventory is boiled-off and the steam is condensed by the A loop steam generator. Because of the lowered loop design, this inventory remains trapped within the A loop pump suction piping and the steam generator. During this period of time, the core becomes uncovered. Thus, since thc process is a redistribution of water within the RCS with the only fluid loss being steam vented through the PORV, the system average void fraction does not change significantly.
3.6 Core Mixture Level The calculated core misture level for the transient is given in Figure 8.
As shown, no core uncovery was calculated while the RC pumps were operating. how-ever, closely following the termination of the RC pump flow, the level in the core decreased. Core uncovery was calculated to start occurring at 105 minutes into the transient.
This compares reasonably well with the NSAC prediction of approximately 103 minutes. Thus, the calculated loss rate through the FORV and the net makeup to the RCS must be in reasonable agree =ent with the actua:
l behavior during the TMI-2 incident.
As shown by Figure 8, the core was predicted to totally uncover. However, this result occurs due to the insufficient spatial detail in the core region. The simulation assumes that all core heat 'is removed and deposited in the fluid.
This results in an overprediction of the core boil-off once the core is~un-covered. A more detailed core model is necessary in order to predict how much core heat is deposited in the liquid region for subseqt:cnt boil-off of the 113 09'I
core _lquid and how much energy is used to superheat the secam.
However, since the simulation was basically made to determine how the core uncovery occurred, the refined core model was deemed unnecessary.
4.
CONCLUSIONS As demonstrated, the CRAFT 2 code simulation predicts reasonably well the system behavior during the first 2 hours2.314815e-5 days <br />5.555556e-4 hours <br />3.306878e-6 weeks <br />7.61e-7 months <br /> and 20 minutes of the TMI-2 transient.
In
~
fact, the co _ uncovery time is predicted within a few minutes of the inferred core level response given in the NSAC report.
Therefore, it is apparent that the net makeup to tie RCS was very low (approximately 34 gpm) over this period which resulted in uncovery of the core and subsequent core damage.
Also, it is shown that the CRAFT 2 code is able to predict the system hydrodynamics dur-
~
ing a small LOCA and is suitable for licensing calculations.
113i 09 O
g
REFERENCES 1
" Evaluation of Transient Behavior and Small Reactor Coolant System Breaks in the 177-Fuel Assembly Plant," Babcock & Wilcox, May 7, 1979.
2 "B&W ECCS Evaluation Model," BAW-10104, Rev. 3, Babcock & Wilcox, August 1977.
~
3 Letter J.H. Taylor (B&W) to S.A. Varga (NRC), July 18, 1978.
4
" Analysis of Three Mile Island, Unit 2 Accident," NSAC-1, July, 1979.
5 R.A. Hedrick, J.J. Cudlin, and R.C. Foltz, "CRAF*2 - FORTRAN Program for Digital Simulation of a Multinode Reactor Plant During Loss-of-Coolant,"
BAW-10092, Rev. 2, Babcock & Wilcox, April 1975.
)l e
e
/
e e
e 9
t g
i
Table 1.
Comparison of CRAFT 2 Assumption to NSAC Scenario Time, hrs: min: see Event NSAC CRAFr2 Loss of feedwater flow / turbine trip 0:0 0:0 PORV opens 0:03 0:08 Reactor trip 0:08 0:10 HPIs actuated 2:02 1:45 HPI throttled 4:38 4:35 i
Auxiliary feedwater block valves opened 8:18 8:20 Reactor coolant pump 2B stopped 1:13:29 1:13:33 Reactor coolant pump 1B stopped 1:13:42 1:13:33 Steam generator B isolated 1:42:00 1:41:40 SG A level raised to 50% on operate range 1:40:00 1:41:40 Reactor coolant pump 2A stopped 1:40:37 1:41:00 Reactor coolant pump 1A stopped 1:40:45 1:41:00 1131 09 O
O 1
T.
s 4
e f
i n
~
n o
.t n
n ho e
t t 6
8 w
s o
i r a
2 (0
h 1
+
o r.
rt rt n
S Te k e t_
~
t a n t
hn en o
lf t o
.l i
aC a e
,n nt 4 r
a t
o r n u o W.
kt t,
a a
t C D :
e
- c. k e
k a
l r C
1 a
R r
a a a e
ue sr a r r n
iB jTB o m
.i a g2 D u R g-e_-
i t r i g 3 X
d a g_
d i 5
a AD l
6 2
ti
. e h
h e
o t
t t
g lo a
a F
P I
6 r>' e g
tne 1
i
,c
~
s n
r
,~
l
~
a L
3 2
T 3
2 A
,e 2
N I
S.
M n
T 1
9g R
I r
P.
o R
f
{
y
)
O 1
3 G
m T'
.4
)-
s<'
F I
a C
R F
r
$g 0
g a
0 i
D
.< re P
g n
i c
d N
o
's g g-)3 '
N
~
T F
A R
C g
i
~
~
O g
y 3
g-Do"
._ g_.
C.s n
O
. g,.
3 2
il
- . i I
2600 FIGURE 2 SYSTEM PRESSURE VS TIME b
2400 ll TEll-2 DATA k
CP. AFT-2 S!!4ULAT10N i
\\
?
I 2200 b
J {s 2000 i
2
\\
\\
J
\\
g 1800
\\.
O E
\\
s m
.N.
1600
- s. N I
l
,b;
-3
+
1400 o
- 1200 1000 i
i 1
i i
1 0
1 2
3 4
5 Time, blinutes 1131 09f 9
=
f
/
I
_y 0
4
.1
/
0 pr 0
0 V
8 N
O I
0 T
2 A.
N 1
E 0
U D
0
, 0 M
I C
I 7
S C
A 2
T 2
6 F
A i
R H
0 0
0 C
T
, 0 1
0 s
6
\\
4 E
R U
S S
s E
e v
0 d 0
t R
0 s u
P n
0 5 n i
T M 8 oc N
A e
s e
L m
O i
e O
T m
C 0
i 0 T RO 0
~
4 T
0 C
6 A
E R
f 0
0 3
0 E
3 R
~
U G
I 0
F 4
0 0
0 2
~
~
~
0 y
2 0
N 0
0 g
\\
1
\\
t 5
A; V
\\
I q b,..
L)'
0 o
~o 0
0 0
0 0
0 0
0
. 0 0
0 0
0 5
0 5
0 5
3 2
2 7v 1
1
- 5. =.= y=e
I o
n 1
s I
l
/
o, e
i
(
I l
w a
W CC l
l u
\\,
LB a
.f zw I
b C3 w a
an' I
o f.
o m ce
/
m
.o I
l.
- = ~_.
-- m W >- D m.
I
~ ~ m a
. zm a
1 w m l
w u r m =
N.
- l g
.D t
-><. z r.
8 m -=
l
=
- 3 J
m w>
g, l
\\
=
5 I
m.
l g
3
-g.
a.
d t
o a
i
- 1. I
~
H
-p i
Gi I
..l..
u w*
=
l rD m
l C
~
m b! -
s
~$
I a.
- 1 CD l
a C2
.1 W
8 1
(-
=
c:D t
l e:s o _.
I r.
D I
9 I
l
-k-a i
n
.y g
I' I
\\
I a
l
\\
c<
I s
I b
T y
.5,.
g a.
IN
, '"===
s.
- ~,.. _
I
..
i
.J_**._.-
O C
O O
M.
o c
y ur 'ia.sai e
FIGURE 5.
HOT LEG FLOW 20000 LOOP A HOT LEG
- - - LOOP B HDT LED C '-
45.
- - LOOP A-TMI DATA
~
'..* k.s, 7
....... LOOP. B-TMI DATA g
.Z'm....,
16000 t
'N.A:
3, g
fg,-.
12000
~
i s.
\\,:." s, t
NI N*
\\
q 3
g, t
\\
e 8000 Il
+
\\
1
,a it i.
. i.
{
i:*
4000 i/
\\
ti
\\
t-
\\
\\.
r s
c
\\.
0 g
e 0
20 40 SG 80 100 120 140 Time, Minutes
-4 vs l
I ca
(
I'
I FIGUiiE 6 HOT LEG TEMPERATURE
~
620 1000 2000 3000 4000' 5000 6000 7000 8000 i
i i
i i
i f.l 5 : i 43 Time, Seconds
/
\\
s TH0T LPA NR-TMl-2 DATA ;
- l* i l.'
's TH0T LPB NR-TMI-2 DATA f
600
,ls,
............T H0T LPB CRAFT 2
/
HOT LPA CRAFT 2 I
I l} W'.
l
\\.
1
!/.'.
I l*:
580
\\
D./
is '.
l-
~
i 1
i.:
.t
/
[:
y
?.
lli I
h.
s
/
/
4 :'.\\ '.:
i 560 i.!
~
li /
\\'
ns.,........................gg. n...,,_,,.ge,.d./'j ki I
,~.
I I
y'./
%., q r
'*%w/
- g,,g
{
I tg g
7{:.
540 g pg f
. /
I s x's y.
f
/
N.j g.
%4 520
)
'(...'.f.".2'3'b-v2 500 i
i 0
20 40 60 80 100 120 140 M
Time, Minutes l
f 6
- e
~
r A A T T A A
~
\\
0 D
~.
N 2 2 l
l 0
A a 2 2 I
l
,. g' s
2 T T T T t
0 F F 1
0 A A 0
R R R R 7
W W C C Nw.
A B B A P P P P O O O O O O O O L L L L r
i 0
6 0
D D D D
'."4-0 0
L L L L e
i 0
O O O O 1
N 6
C C C C T T T,
T
\\
E 0
R 0
i U
0
~
0 T
5 8,
I A
REP l
M E
h O
T s
s d
0 e
N t
n 0
i G
o 0
u E
c 4
n' IR i
L e
S M
0 D
6 I
L e
e O
m m
O i
C i
T T
0 R
0 7
0 0
3
~
E 0
RU P
G 0
I 4
I F
0 0
i 0
2
,s 0
+
2 0
t N 0
\\\\
0 1
b'\\
~
!j
- g
~
0 0
0 0
0 0
0 6
2 8
4 0
6 6
6 5
5 5
4 o*
yf:
~
~ [~
i FIGURE 8 CORE MIXTURE LEVEL 12 g
CRAFT 2 - SIMULATION 10
\\
CORE LEVEL PRE 0lCT10N FROM REFERENCE 4
}
t
\\
0
~
y
\\
~
\\
\\
6
=
o n
\\
.m
\\
\\.
4 g
x
\\~
~4 M
t.:
I i
e i
1 e
" i. '
S000 7enfl 8000 9000 10,000 11.000 oq l'me, seconds h*
i.......
T-
~
i e
e s
e m.
(
I l
~."
N O
j o
c g
E M
i g
E w
>=
M
- E N.
sn u
M 3
%c,,,,
a s
9 y
..=
Q
~
2 M
k 5
- =
CD 9
N 5-Ct:".
m g
g 2
o n I
= a W
o
+-.
3
=,
=
o e.e
-a r,
=~
m
... = -
w
~
5 o.
H u,
~
aee
- =
e e
i i
1
=
3 a
a o
~
~
=
5 **DisA saisi! 33
.i bi10b
~
9 O'