ML20033A354

Forwards NRC Evaluation of ORNL Rept on Pressurized Thermal Shock.Rept Does Not Change Previous NRC Conclusion That Probability of Occurrence of Severe Pressurized Overcooling Transients Too Low to Require Immediate Action

ML20033A354
Person / Time
Site:	Diablo Canyon
Issue date:	10/30/1981
From:	Dircks W NRC OFFICE OF THE EXECUTIVE DIRECTOR FOR OPERATIONS (EDO)
To:	Bradford P, Gilinsky V, Palladino N NRC COMMISSION (OCM)
Shared Package
ML16340C101	List: ML16340C100 ML20033A354
References
TASK-2.K.2.13, TASK-TM TAC-45202, NUDOCS 8111250239
Download: ML20033A354 (23)

Text

o*

k e

e e

OCT 3 o Issi "E' ORAP!Cf.T1 FOR: Chaiman ?alladino Croissioner Gilinsky Ccrmissioner Bradford Crmissicner Ahearne Ccrmissicner Roberts '.

RC'? :

'.lilliam J. Dircks Executive Director for Oceratiens STAFF 'tE'!IEf p CR!:L 3EPORT Cf3 PRESSURIZED THEval

. st'BJECT:

SHCCK.

the Ccmission en Octcier 9,1961, I transt'itted a In : v e-crardun ::

ccy of the draf t creliminary report en cressuri:ed themal shock crepared The staff has cc pleted its review of by Oak Ridge '!ational Laboratory.and the staff's evaluatien is enclosed for your inferratien.

Also enclosed is a letter frc9 Duke Power Ccreany cresenting results of the crport, thair evaluation of the ORNL mport.

A Orinci:a1 conclus'icn frcm the ORNL analysis is that an evercooling transient similar to the ecst severs. transient that has occurred (the Dancno Seco event of arch 20,1978) will not oose a threat to the ORfit uamined over-Ocenee-1 aressure vessel for several ecre years.

ccoling transients r re severe than the Pancho Seco event Nith lower ec:urrence croba flities) and, using fracture cachanics calculations thougnt by tha stiff to be c0nservative, credicted failurt of the Oconee-1 vessel..if the cost severe transients were to occur today.

The staff has previously concluded and discussed with the Cr9issien tht t.'le pr0cability of occurrence of tere pressuri:ed overceo11rc transients is sufficiently Icw that irrediate cerr=ctive action is not

.iarranted but that corrective acticns ay be recuired for scre clants within a year.

(

f Centact;

7. E. ?'u rl ey, :RR 49-27517 O

%e

\\

\\e 1

l 8111250239 811116..

PDR ADOCK 05000275 A

PDR i

W M M hme

.-e e

eu.em ag,.

og-

l J

v f

m y

~

1 9

?

9' c

r

~

The Camissioners.

In the stiff's judgment the ORitL recort does not present any significant new infomation that would change that conclusion.

(Signe0Wisa 1.014 Willian.J. Dircks.

~~

Executive Director for Ociistien's

Enclosures:

1.

Staff's Assessment of ORNL's Graft Preliminary Repont on Pressuri:ed Themal Shock 2.

Latter frtri A. C. Thies, Duke Pcwer Cacany, to P. M. Bemero, NRC, dated October 20, 1981 '

~

l

1. pe l

l l

l k

.5

Enclosure l'

~

~

NRC STAFF ASSESSMENT CF THE ORAFT INTERIM REPORT BY OAK RIDGE NATIONAL LABORATORY ENTITLED, "EVAldATION OF THE THREAT TO PWR VESSEL INTEGRITY POSED BY PRESSURIZED THERMAL SHOCK EVENTS" I.

Purpose The pJrpose of this document is to assess the regulatory implications of the draft Oak Ridge National Laboratory Report, NUREG/CR-2083, entitled, " Evaluation of the Threat to PWR Vessel Integrity Posed by Pressurized Thermal Shock Events",(ORNL report) with respect to the acceptability of continued reactor operation considering the pressurized thermal shock issue. For that purpose, the NRC staff has made a preifminary review of the report, and the staff's conclusions and supporting information are presented here.

II.

Summary and Conclusions The NRC staff finds the ORNL report to be a useful summary of the background and present status of the pressurized thermal shock

~

i ssu e.

As the.ORNL repo.rt states, results cited in the, report are drawn fran previous work and literature sources.

A principal conclusien of the ORNL report is that an overcooling event similar to the most severe transient that has occurred (the Ranc'ho Seco event of March 20,1978) will not pose a threat to the Oconee-1 pressure vessel for several more years.

Another important conclusien of the ORNL report is that if certain events more severe than the Rancho Seco overcooling event were to occur today, and if the reactor pressure vessel were to remain at high pressure or be repressurized, and if fracture-mechanics calculations O

++

immem

• e-e+=

o-e

..e=w

o 2-1

.t believed to be conservative are used, then vessel failure may be predicted for the Oconee-1 vessel. The NRC staff had previously reached this sane conclusion (Refs.1, 2, 3, 4, 5), which is the reason that the pressurized.thennat shock issue is under consider-I ation today.

.The NRC' staff also had previously concluded, and discussed with the Canmission, that the probability of occurrence of pressurized

. overcooling events, more severe than the Rancho Seco event, is sufficiently low that innediate corrective action is not warranted (Ref.1), although longer term (Drrective actions may be required for some plants within a year. The ORNL report does not present any significant new information that.'would change this co,nclusion by the staff.

The. report presents the completed results of analyses for four overcooling transients postulated for Oconee-1. These are: a large break loss-of-coolant accident (LOCA), a main steam line break (MSLB), an overcooling event which act' ally occurred 'on u

March 20,1978 at Rancho Seco, and a postulated overcooling event more severe than the Rancho Seco event, referred to as the runaway feedwater transient (RFT).

In addition, ORNL reviewed partial calculations for a small'6reak' i.0CA, but the calculations were not canpleted for the ORNL report.

Table 8.7 of the ORNL report presents results showing effective full power years before predicted Oconee-1 vessel fail'ure. For the Targe break LOCA and the Rancho Seco transient, previous results are confinned that many years of operation remain before these events wculd present a potential for failure of the pressure vessel.

Time-to-predicted vessel failure results for the small break LOCA are not presented in the ORNL report, but other calculations have been made which indicate this event is not of immediate concern (Refs. 5 and 6) and it will not be. further discussed here.

O

~

r

-'t+'t

- + or e v, -

.g,, p

~,

.3--

.~ )

The time-to-predicted vessel failure results presented in Table 8.7 of the ORNL report for the RFT and the MSLB are the principal focus of this report since they raise the question of whether or not there is an insnediate safety concern at Oconee or other plants.

According to the ORNL report, the MSL3 is the lowest probability event which has been analyzed. The MSLS is stated (small table in Section 3.1) to have an occurrence frequency of S x 10-6 per reactor year. Not mentioned in the report, but apparer,tly included in the quoted occurrence frequency (in order to produce cvercooling conditions sufficiently severe to potentially fail the pressure vessel), is the pr.obabil'ity that'the operator fails to isolate

~

feedwater to the steam generator with the broken line. That human error probability (HEP) has been multiplied by the MSLS occurrence f requency to obtain the estimated frequency of an overcooling event that would challenge the pressure vessel at Oconee-1, (i.e., the estimated frequency of the MSt6 is not stated in the ORNL report, but apparently ORNL assumed the value quoted in WASH-1400 and then used a paritcular HEP to obtain the quoted v.alue of 5 x 10-6 p.er reactor year shown in the ORNL ' report for the overcooling event).

At Babcock and Wilcox (3&W) plants other than Oconee Units 1, 2, or 3, an autanatic feedwater isolation' system and a steam line isolation system. {tre installed. Proper operation of those systems fol, lowing,

a MSLB would prevent an overcooling event severe enough to challenge the pressu're vessel. Therefore, at other B&W plants, estimated f retuency of this event is approximately 5 x 10-6 per reactor year times the probability that autanatic feedwater isolation fails, times the estimated frequency that the steam line isolation system f ail s.

(Details of how the exact' systems vary fran plant-to-plant are not conpletely described in this brief summary.) This canbined estimated frequency for other S&W plants would be below the value stated in the ORNL report of 5 x 10-6 per reactor year.

In the small table in Section 3.1 of the ORNL report, the RFT is stated to have an occurrence frequency of 1.0 per eactor year.

P

..v-g 4

However, additional failures would be necessary to cause.a severe overcooling event as a result of the RFT. Therefore, the statement immediately below the reported occurrence frequency of 1.0 in the table must be considered as a vital qualification of that frequency, i.e., "...for dconee-1.it appears that multiple independent failures a re requi red... " The occurrence frequency for a mild tran'sient initiated by the feedwater systen is indeed close to 1.0 per reactor year sine:e such trarlients are frequent, but such transients are of no consequence to plant safety unless there are subsequent f ailures. The probabilities of all the other failures must be.

cabined in order te arrive at the actual estimated frequency of a severe ' overcooling. event.' That estimated occurrence frequency is believed to be low, as discussed below.

The estimated occurrence frequency of the particular, detailed RFT scenario presented in the report is very low since the total amount of feedwater assumed to be pumped into the steam generators.is censiderably greater than the maximum condensate physically avail-able in the system at, a location where it can, be a source of feed-water. Assuming the amount actually available instead of the ficticiously larger amount would decrease the cooling and make the actual transient less severe.

In addition, feedwater flow rates would preaably be reduced below thost,, assumed in the report, even wnile water is still available in the system to be pumped into the steam generators. This would'nonnally result from loss of the steam sucply to drive the turbine driven pumps as a consequence of flooding of the steam generators which are the source of that steam supply. That is, gross overfeeding of the steam generators might be self-limiting under such extreme conditions. This was not taken into account in the subject report and may be applicable to the MSL3 event as well as to the RFT event.

l iherefore, the NRC staff would expect that an actual RFT would be less severs than the one calf.alated in this report, and it would l

l

..w_.

.._.--.-_.,---.-e

-e v a -

--a~-~~**--

- ~ -.

5-still ' require several failures, including:

the feedwater controller or integrated control system (ICS) must fail; the BTU limiter must fail; and the operator must fail to correctly diagnose the problem and take corrective. action.

(These itens are discussed in further detail in the body of this report.)

The above discussions of overcool.ing event analyses would not be complete without mention of the computer codes used. These fall

-into two categories, fracture mechanics codes and transient codes.

The transient codes are used to calculate the pressure and temperature

<ersus time that is input into the fracture mechanics codes; that is, they do the systems calculations that predict what pressures and temperatures will result, given a;particular hypothetical event. The fracture mechanics codes assume the particular pressure and tenperature versus time history calculated by the transient codes (i.e., that a particular event has occurred). These codes are then used to calculate the probability that the pressure vessel will fail if it has a certain size and shape crack present at a critical location on.the inner surface.

The fracture mechahics code used in.the ORNL report, together with the input data (i.e., materials properties, including fracture toughness and variations of materials ~prtperties with tenperature i-and exposur.e to neutron radiatien) should yield somewhat conservative resul ts. That is, if they differ from reality, it is believed that failures would be predicted when they wculd not in fact occur.

Detailed evaluation of the mechanics of materials aspects of the ORNL report is difficult because of insufficient infonnation for l

l many of the values used. For example, Table 7.1 (page 7-1) lists paraneters which must be known in order to set up OCA-1 for a thermal shock analysis but many of them were given only by inference (e.g., alpha, E, Poisson's ratio, yield and ultimate strengths) or G

Ge

-e

+

+

w eb,-

tm.,%,N M g., g ome-

.- e. gee e,-

..ee..

j ge and K,).

With respect to the by reference to documents (e.g., K g

heat transfer coefficient, the values given in the text (page 8-4:

2 1000 BTU /hr-ft -F*) and in Table 8.1 (page 3-6: either 200 or 330 2

BTU /hr-ft -F') are contradictory.

The ORNL reports on the HSST thennal shock experiments (TSE) in the past have failed to follow the dictates of the ASME Section XI reconmendations for analysis and toughness (K and K,) deter-7 g

minations. The conclusions given in the ORNL reports, that TSE calculations and obsenations were in agreement, relied on a somewhat circular,argum'ent of me'chanical property detenninations.

In response to a specific NRR request, reanalysis of TSE-Sa by.0RNL in accordance with the ASME C6de showed the original analysis to result in crack extension over-estimates (Ret. g). The NRR inter-pretation of the reanalysis led to the conclusion that the Code method would have added about 45'F of conservatism to the prediction.

Section 7 of the ORNL report (and, to a lesser extent, other sections)

' sets forth arguments of conservatism and non-conservatism in a mixed, non-rigorous way. While it is true that several uncer-tainties exist, most of them are recognized and accounted for by selecting individual values c snservatively. Because the, draft ORNL report does not present the results in teens of a sensitivity analysis, 'the everall degree' of conservatism in the fracture-mechanics code results cannot be evaluated quantitatively.

With respect to the systens codes used to calculate the primary systen coolant behavior, the 'IRT code used in many of the calculations is not believed to be appropriate for such calculations for several reasons.

It does not contain a realistic model for flows in the primary. system once significant voiding has occurred.

Instead, it uses flows that are input as a table and therefore are invariable.

When significant voiding occurs, as it does in the RFT event presented, l

I )

ll I'

the code continues to assume primary system coolant flow to the steam generators, where it is cooled and then returned to the primary system thereby making the pressure vessel overcooling event worse.

In actuality, it is predict?d that for certain cases the voids would collect in the -system-high points.

If sufficient voids collect at the top of the hot 'eg inverted U-bend, a natural circulation flow interruption (or " vapor lock," to use a common analogy) would occur and overcooling would be greatly decreased.

In this way, the code would tend to predict events more severe than the actual event. The code also has a different cause for inaccuracy in an unknown direction.' That is, the code fails to conserve mass and energy throughout the calculation, by observed snounts as much as 25%. The energy or mass flowing out of one vclume does not equal the total anount of energy or mass received at all other volumes as a result of the flow, as it must in reality. This discrepancy can result in errors that will vary in magnitude and direction (i.e., errors due to this latter inaccuracy can be conservative or non-conservative by varying snounts).

In conclusion, the ORNL report states that overcooling events similar to the most severe event that has occurred (Rancho Seco) will not pese a threat to the Oconee-1 vessel for several more ye a rs. The report then goes on to demonstrate that it is possible to postulate more severe events (with correspondingly icwer occurrence l

probabilities) for which vessel failure is predicted.

The staff's judgment is that the occurrence frequency of a severe overcooling transient that would threaten the integrity of the Oconee-1 vessel is sufficiently low that time is avafiable for the staff to carefully evaluate the condition of the vessel and propose solutions to the problem before 62rther regulatory action is needed.

%6 t

~

=

L r~ -~

-m

~

~-

_ ~ _

~~

III. Discuss' ion The sections below provide a more conprehensive summary of the preliminary review of the ORNL report by the NRC staff.

A.

The Overall Reoort fra a Systems Viewooint (1) Runaway Feedwater Transients (RFT) t Of the general classes of transients identified, the most canplex fran a system and controk viewpoint is the Runaway Feedwater Transient (RFT). T'he Feedwater Control subsystem of the Integrated Control System (ICS) is" designed to naintain a total feedwater flow equal to the feedwater flow demand. The flow in the feedwater system is control. led by the ICS in the autanatic mode by using input signals and monitoring process parameters, or it can be controlled by the operator through. the ICS at the Loop Feedwater Demand level or at the F'eedwater Valve Position or Feedwater Pump Speed level. The operator can intenene at a'ny time, therefore he can be an initiator and/or a tenninator of an overecoling transient.

The cause of the RFT may be internal or external to the ICS (canponent failure or" operator error). The severity of the overcooling transient for Oconee-1 can be. reduced or tenninated in one of three ways:

(1) The operator can take' control of the FW pumps or valves, if his indication (i.e., instrumentation) is coerating and he diagnoses the prcblem correctly, in which case the event can be tenninated quickly. However. if the event is not diagnosed correctly, he may make the overcooling more severe.

(2) The STU limiter can limit FW demand since it continuously calculates the BTUs or energy contained in the steam generator.

(3) The hi-level limit is a fixed setpoint which limits the liquid level in the steam genera-t ors. The hi-level trip (present only on Oconee Units 1, 2, and 3

)

D a

Od u

e+

w g

Jng r weas4

-4D-b M

A p

he

-g-4 e

h e

g

1

-g.

but not on other S&W plants) can trf* FW pumps at its independent fixed setpoint. All of these potential mitigation actions were implicitly assumed to fail in the runaway feedwater transient in the ORNL report.

(2) Main Steam Line Break (MSLB)

The Oconee Units 1, 2, and 3, and Rancho Seco do not have main steam isolation valves (MSIVs), whereas all other B&W units do have MSIVs. The three units at Oconee do not have automatic FW isola -

tion, whereas all ether B&W units do have FW isolation. All B&W units except the Oconee units have sone type of. main steam line._

break logic that will isolate FW to the faulted steam generator (at some plants both steam generators) and at some plants the logic will also shut the MSIVs. At most units the cooldown transient would be terminated by the steam line break logic (MSIV closure and/or feedwater isolation will terminate the cooldown, with a delay for SG dryout,in the case of only feedwater isolation).

It is, however, necessary to take credit for the coerator terminating or throttling the-HPI pumps at a iater time in the event, At Oconee-1 with no MSIVs or.. automatic FW isolation, operator e.

action must terminate the MSLS overcooling transient by closing the FW valves and allowing the s' team generator to steam dry. No credit was assumed in the ORNL report for operator action to limit feedwater flew for the MSL3 accident.

(3) Main Feedwater Svstem as it $ould Goerate for a RFT and MSL3 In the calculations of the ORNL report, full amin feedwater fic< is assumed throughout these events. With one steam generator flooded and the other steam generator isolated, as in the MSLS, the operator will probably not be able to maintain the turbine driven MFW pumps O

%e 2e ees

=.-e en % -

<gge e e a ese-

>waa h M d

"*#W**

5Meea N

in a nJnning condition because the primary steam source has been The condensi a pumps and condensate booster pumps do not t

1ost.

have enough head to riaintain full flow for these events. Multiple failures must occur in the ICS to prevent automatic runback and trip of the MFW train. Without'a sufficient supply of feedwater (condensate in the hotwell) the MFW pumps will eventually lose suction. As pointed out in the Sumary and Conclusion section

bove, the ORNL report does assume more than the actually available amount of condensate for the RFT event. All of these reasons why MFW may be lost or reduced were ignored in the ORNL report analyses, thereby making the. overc'coling more severe for the RFT and MSLS accidents in the report.

In addition, the feedwater temperature was assumed to ramp down to the hot well tadperature within one minute after MFW pump. trip, a consenative assumption. The likelihood of such behavior is extremely small since multiple failures of various systems would have to occur.

The report acknowledges that plant design modifications have al ready been made which will reduce the likelihood of excessive feedwater transients at Ocenee-1. No attempt was made to determi.ne the effect of these modifications on the olant'r susceptibility to such trans'ients, i.e., no cr' edit was given for the decreased expected frequency -of these transients resulting from the modifications.

.B.

The Overall Report fran a Probabfif ty and Risk Assessment IMA1 'liewcoint (Probability of Transient and Accident Secuences)

(1) Sumary We have concluded that the occurrence frequencies estimated in the ORNL report for the types of initiating events analyzed a.re..reas' enable. and in fact are conservative for the 'iSL3 and

RFT when capared to estimates in use by the NRC staff.

Caparison of the ORNL report and NRC estimates is given in th'e following table. The ORNL report does not appear to distinguish (as done below) between the probability of the initiating event and the resulting overecoling event probability.

Estimated Frequency.Per Reactor Year ORNL NRC Pressurized Pressurized ~ ~

Initiating Overcooling Initiating Overcooling Event Event-Resul ting Event Event Resulting

-fra Initiating fra Initiating Event

• Event

• 3x10*f(B&W)*10~4Q&W)

~

RFT 1

Not Stated 6x10~ (CE&W) <2x10 (CE&W)

Large Not Stated 5x10-6 1x10-4 3x10-6 MSLS d

3x10~#

1x10 Small 3x10-4 Not Stated LOCA Large 1x10"#

Not Stated 1x10~4 Not Stated

~

LOCA Rancho Not Stated Not Statid 3x10-f(B&W)$10~3 (B&W) 6x10" (CE&W) s10" (CE1W)

Seco

• Equal to frfue...f of initiating event times probacility of accitional f ailures and/or error probabilities as discussed in text.

Overcooling transients at pressure in PWRs result from small break LOCAs, main steam line breaks, or feedwater transients, only if additional failures, either hardware-or human-related, occur subsequent to the initiating event. Fra a PRA viewpoint, f

we believe that a more realistic way to analyze an wercooling f

transient at pressure is to consider it as a secuence of events. Using event tree methedology, the overcooling transient.

l 1

~

sequence is represeitted by a set of event trees.

Each event tree in the set has event headings, corresponding to a different initiating event, a specific assumption made in the analysis about feedwater' flow rate and/or temperature, or a postulated f ailure (hardware-or human-related). Rigorous detennination of the estimated frequency of occurrence for each event

~

sequence thus generated would involve assigning an estimated occurrence frequency to each event and canbining them to obtain the estimated event sequence frequency. Such an effort was beyond the limited scope of this review. However, within the past year, the 3RC staf? has made simplified analyses to obtain estimates of the frequency of overcooling transients which are summarized in the above table. To date, we are not aware of any subsequent analysis, including tre suMeet ORNL

' report, that would cause us to alter those estimates.

The integrated NRR/RES task action plan being prepared for the technical resolution of the pressurized.thennal shock issue includes a rigo'rous PRA anaTysis of the overcooling transient event sequences such as that discussed above.

(2) Discussion Speci'fic camnents regarding each of the classes of initiating events are as follows:

(a) Rancho Seco Event: The most serious pressur'ized overcooling event was. that at Rancho Seco en March 20, 1978, in which the coolant temperature dropped from 550*F to 280*F in about I hour while the system pressure first dropped, then returned to near its original value. Based on this experience, an occurrence frequency of 3 x 10~2 per e

i

reactor year was estimated for a B&W plant to experience an overcooling transient as severe or more severe than the Rancho Seco event (as described in M. A. Taylor's memorandum of October 29, 1980, Ref. 3).

Since this occurrence and the occurrence of two less severe events, operators have received special training in transient response. Babcock & Wilcox plants have-added a back-up power supply to the non-nuclear instru-mentation bus, whose failure initiated the three transients abov e.

The NRR staff examined the impact of the improved power supply and operator training and suggested that these improvements might have reduced the frequency to 10-3 per reactor year for an overcooling transient as severe as the Rancho Seco event for B&W plants. For more severe events, such as the RFT, that might challenge the Oconee-1 vessel if they were to occur today, the staff estimates that their frequency is 10~4 per reactor year for B&W pl, ants.

The cperating experience of CE and Westinghouse plants has also been examined. There have been no events like the Rancho Seco transient, but there have been sane p recu rsors. These are events which typically led to secondary steen dump valves or steam bypass valves sticking open, but which did not result in steam flcws large enough to produce very severe overcooling transients.

The most severe of these transients occurred at Arkansas Nuclear One-2 (a CE plant) on December 27, 1978, where a main steam relief valve lifted and failed to reset, thereby causing the reactor coolant temperature to drop by 107*F in 52 rc*nutes. This lack of severe overcooling events at CE and W plants plus the greater thermal inertia 6e e

-w

.a.=.-e m e

+ w e,

-w w,w,,

• _g*g,,,

• wa.-,

e.

.e me

\\

a of most W and CE plants, leads the staff to estimate an RFT occurrence frequency of a factor of 5 lower than for B&W plants 2 x 10-5 per reactor year. Also, the staff estimates the frequency of a large steam line break or its equivalent to be no greater than about 10-4 pr reactor year, and for a pressurized overcooling event-resulting fran a MSL3 severe encugh to challenge the Oconee-1 vessel if it were to occur today, the estimate is a factor of 30 lower, 3 x 10-6 per reactor year.

These _ estimated frequencies are summarized in the above table. There may be a factor of 10 uncertainty associated with these. estimates.

b.

Small Break LOCA: The ORNL report does not provide canplete calculations for the small break LOCA. However, in a simplified analysis of an overcooling event ' initiated by a sina11 break LOCA, (i.e., between 2" and 6" equivalent diameter) the NRC staff (Ref. 5) obtained an estimated

. f frequency of occurrence of 1 x 10-* per reactor year.

This result was based on an assumed occurrence of 3 x 10-4 per reactor year for the LOCA event and an operator human error' probability of 3 x 10-2 (operator i

failure to throttle or tenninate safety injection pumps),

c.

Main Steam Line Break: The table on page 3-4 of the ORNL report gives an estimated frequency of occurrence for an overecoling event resulting frem a MSLB of 5 x 10-6 er reactor year.

Reference 5 contains the results of a simplified analysis by the NRC staff of the probability of occurrence of an overcooling transient caused by a MSL3 and subsequent operator error. These results are summarized as follows.

• • • • h-tr e -w 4 *

= =....,

,,,,sD_

For the case of a large MSLS-initiated overcooling transient, the estimated frequency of occurrence was 3 x 10-6 per reactor year. This result was based on an

~

assumed frequency of occurrence of 1 x 10~4 per reactor year for a large MSL3 and an operator HEP of 3 x 10~2 per demand (failure to tenninate feedwater flow to the steam generator (SG) with the broken line and/or failure to close the main steam isolation valves to that SG).

d.

Runaway Feedwater Transient (RFT): The RFT analyzed in the ORNL repoFt assumes multiple failures subsequent to an initiating event. A better description might be given by the tenn " overfeed trans'ient." Such transients usually arise from other transients which initially empty the steam generator (s), such as, in this case, a stuck-open bypass valve. Following this, a loss of automatic feedwater control or a manual error coupled with the failure of the operator to diagnose the situation and -

take appropriate corrective action would result in excessive feedwater being supplied to the steam generator.

The NRC staff has perform'ed a review of Licensee Event Repor.ts (LERs) regarding overeco. ing, events. Based on l

this review, a frequency of occurrence of overfeid transients

'of 3 x 10~1 per reactor year was estimated for 5&W plants.

The corresponding estimate for Westinghouse and CE plants was S x 10~2 per reactor year. A realistic estimate of the frequency of occurrence of an RFT must consider the frequency of occurrence' of the initiating event and further indeoendent failures (e.g., failure of the steam generator high level MFW pump trip) and/or continued inappropriate operator actions which exacerbate the transient. The inclusion of multiple failures, both human and hardware-related, requires analysis of an O

69 i

l 1

I

. entire spectrum of RFTs. The NRC staff's action ' plan for resolution of the pressurized thermal shock issue includes such a complex analysis. The occurrence frequencies are believed to be lov,, but quantitative results are not now available. However, a preliminary estimate is given in the above table.

C.

The Overall Report fran a Systems Code Viewooint The NRC staff considers the use of the IRT corouter program to evaluate the response of the primary system to severe overcooling transients to be inappropriate, since this class of events is well outside the range of. the ptogram's capabilities.

IRT is capable of handling mild or intermediate transients 'which do not result in void fonnation in the primary system.

IRT does not adequately cor.erve mass and energy.

The following critir.a1 items demonstrate t. e shortcomings of IRT for use in analyses of severe overcooling transients:

(1) Flow Distribution:

IRT does not solve the momentum equation. The input data specifies the primary flow.

The heat renoval rate is dependent on the flow. For the cases presented, a natural circulation flow taken from the Oconee FSAR is input.

(2) Voids in the Primary System:

In the IRT calculations, void fonnatien is allowed only in the reactor vessel upper head region. The effect of voiding is not properly treated in IRT, which is a honogeneous equilibrium calculation. For certain of the cases presented in the ORNL report, the upper head region is voided at about 100 After this time, additional voids are incorrectly secones.

.me

~-+*'C**--ee.

g,7

4

--17 assumed.to be hmogeneously mixed throughout the primary system. The assumed primary system flows in cases involving assumption of single-phase natural circulation flow are therefore incorrect.

In fact, it is expected that the voids will collect in the pipes leading to the steam generators and interrupt the circulation, de-coupling the secondary system fra the primary system and removing the heat sink. The loss of the heat sink will st0p the cooldown.

(3) Energy B,alance: For tRe " runaway feedwater" transient at 1000 seconds, the discrepancy in the energy balance amounts to 25t (Ref. 7). 8erockhaven National Laboratory (BNL) estimates that this corresponds to a temperature error of approximately 307.

It is not known whether the error is conservative or non-conservative. For the MSL3 overcooling transient, the discrepancy is negligible.

The NRC st'aff asked Los Alamos Nat$cnal Laboratory (LANL) to perfons TRAC code calculations of MSL3 and " runaway feedwater" overcooling transients for conditions similar to those perfonned by SNL with IRT. The results fran the TRAC calculations for the MSLB show that the temperatures calculated by the two codes are in agreement with one another (Ref. 8). The pressures differ, 'however, because of the difference in mcdeling of the pressurizer. A non-equilibrium model is used in IRT while TRAC uses an equilibrium model. Actual pressures would probably lie between results from the two codes.

TRAC and IRT differ greatly in their modeling capabilities of various phenmena. For example, two phase flow irt the primary is modeled in TRAC and not in IRT (except in the

-- ~

upper head and pressurizer). An agreement between the results calculated by the two codes could be expected to occur only if single liquid phase exists. in the primary.

In summary, IRT is not an appropriate program to use for severe overcooling transients. The treatnent of momentum (flow) and void fonnation are key elements in the transient behavi or. The results presented after 100 seconds are highly suspect.

The staff believes the overcooling rates calculated are conservetive. However, it is not possible to quantify the amount of conservatism.

D.

The Overall Recort from a Fracture Mechanics Viewooint (1), Codes The use of the OCA-1 canputer code to calculate the stress intensity (using assumed flaw morphology) 1.s acceptable". Nevertheless, both the ORNL and the NRC staff agree that it is important to modify the code at the earliest possible date to include the tenperature-dependence of material parameters (such as the elastic modulus, coefficients of thermal expansivity and conductivity), rather than using average values. Also the possibility of crack arrest in materials of high toughness (relatively high temperatures, upper-shelf energy levels) has not been addressed.

Since advanced elastic-plastic fracture mechanics concepts are required in treating this matter, the lack of a solution in the ORNL report is not surprising.

However, accurate calculation of the crack arrest behavior will hinge on the treatment of the high tenperature, high toughness aspects.

~

l In general, the fracture mechanics calculations in the ORNL report were performed. in a manner with which the NRC staff agrees, but j

u l

L t

n---

.~..m--

--r.

s.

,w.-4ei.

. T

~

refinements to the OCA-1 computer code have been suggested for sane time and improvements could be made today. The OCA-1 code can providt results which can be judged accurate; the results will be T

conservative when conservative inputs (lower bound KIc, NDT per R.G. 1.99r etc.) are used as was done by ORNL.

4 (2) Warm Prestressing

' Warm prestressing (W85) is the tenn applied to a phenomenon which can limit the extent of total crack advance during sane overcooling transients. WPS has been demonstrated in the laboratory with small

~

specimens and in a 1arge thick-walled cylinder during an unpressurized

~

thermal shock experiment.

In general, the 3RC staff believes that considerable caution must be used if any credit is taken for the effects of warm prestress in analyses of the pressurized thermal shock problem. The draft ORNL report does show results both with and without WPS (Table 8.6) although the connents in Table 8.7 indicate ' hat WPS is effective t

only for a large-break LOCA. There are so many detailed variations in the postulated accident scenarios, involving turning pumps on er off and tapping several water searces, that the time variations in K are quite uncertain. Only with a rather smooth c'h'a'nge in K7

~

u 7

relative to the toughness, KIc, (which also varies with time) can the benefits of WPS be assumed with confidence.'

E.

Fluence Uncertainty The ORNL report states that the uncertainty in fluence estimates can be as great as 250%. However, contrary to the ORNL estimate of 250%, the staff believes that current fluence estimates can be made to within 220% provided that one uses: (a) a well calibrated and benchmarked transport code and (b) measured values of the neutron flux and its distribution.

O 66 m

F 4, - - - -

9

--m,**-w

-,e--s

. 20 --

The staff has a threefold progran for code calibration and benchmarking using:

(a) the results of the PCA experiment (b) the surveillance capsule results fran Maine Yankee and Fort Calhoun, and (c) the surveillance results fran ANO-1. Consequently, the staff expects future fluence calculations to be used in longer tenn resolution of this issue, to be within !20% instead of the ORNL report's stated uncertainty of !50%.

MD e

9 e

e

=

0 4

ee i

21 -

References' 1.

Board Motification - Themal Shock to PWR Reactor Pressure Vessels (BN-81-06), May 8, 1981.

2.

Themal Shock to PWR Reactor, April 28, 1981, Memorandum for Harold Denton fran D. Eisenhut..

3.;

Insights on Overcooling Transients in Plants with the B&W NSSS, October 29, 1980, Memorandum for S. Fabic from M. Taylor.

4.

Pressurized Themil Shock, Comnission Paper, SECY-81-286, May 4, 1981.

5.

Frequency of Excessive Cooldown Even 2 Challenging Vessel Integrity, Ashok C. Thadani to Gus Lainas, Memorandum dated April 21, 1981.

6.

Sumary of Meeting with the Babcock & Wilcox, Westinghouse, and Cembustion Engineering Owners Groups on July 28, 29, and 30,1981, respectively, concerning pressurized themal shock to reactor pressure vessels (RPV), Docket No. (All Operating PWRs), August 14, 1981.

7.

October 11, 1981 letter frmt M. Levine to S. Fabic, " Mass and Energy Non-Conservatism in Overcooling Calcuations.with IRT."

l 8.

Personal ecmmunication to N. Zube, RES/NRC, frcm LANL, abcut October 15, 1981.

9.

January 6,1981 letter fran G. D. Whitman to Warren S. Hazelton.

l e

9

%e m

O f

• 9

l l

~

i i

DUKE POWER-COMPANY CHAmt.orrc. N. C. 2S242 iro4: 37s.4a4e i

A.c.twiss se e. v.c p.s.io e

Oneoucinnne ane Tannoussesone

.0ctober 20, 1981 1

'Mr. Robert M. Barnero, Director Division of Risk. Analysis Office of Nuclear Regulatory Research U. S. Nuclear Regulatory Commission Washington, D. C. 20555 ORNI. Evaluation of the nreat to PWR Vessel

Subject:

Integrity Rosed by Rressurized normal Shock Pressure; Draf t Interim Report

Dear Mr. Bernero:

Duke Power Company appreciates the opportunity to provide comments on the As you are aware, Duke has provided certain specific subject document.

technical information regarding the Ocones Unic 1 reactor vessel in an effort to assist the NRC. in the completion of this evaluation. Duke engineers have reviewed the subject document and consider chat the evaluation contains sig-nificant deficiencias in the area of thermal-hydraulics conditions and repre-sents unrealistic transient conditions. De application of these transients to the Oconee i specific material properties results in misleading and mean-Our more salient concerns are in the ingless calculated vessel lifetime.

following paragraphs with additional details provided in the attached.

ne evaluation-of the reactor vessel thermal shock issue is extremely complex and requires a thorough understanding of several highly technical disciplines.

Among the technical areas involved are instrumentnion and controls, systems analysis, reactor vessel materials, non-destructive examination techniques, In linear elastic fracture mechanics, and probabilistic risk assessment.

order to do a meaningful evaluation, these technical areas need to interact in a coordinated nanner; the results of one area cannot be input into subse-4 quant analyses without a thorough understanding of the basis of the input.

nis document does not indicate that any coordinated effort was attempted by the various organizations involved to assure that the results provided were the document tends to imply that the individual tasks realistic. In fact, were performed independently of each other with the end result being a totally disjointed document that is not suitable for understanding and communicating the real perspective of the issue.

1 One of the principal.sechanissa contributing to the occurrence of pressurized reactor vessel fracture is the creation of certain unique temperature-pressure time histories at the reactor vessel, the calculation of which would require insights into plant design f eatures, system f ailures and effects, plant par-n e fracture mechanics analyses formance, constraints, and transient behavior, embodied in this ' docume h6 BJJU7tqz4

\\

e N

Mr. Robert M. Bernero, Director October 20, 1981 Page 2 thermal-hydraulic accident conditions and not germana to the real plant.

The major deficiencias in situation, especially for the Oconee reactors.

the thermal-hydraulic analyses are identified in the attachment to this Portions of the ORNL report have also very appropriately discussed letter.

Yet the limitations and deficiencies in. the thermal-hydraulic analyses.

fracture mechanics calculations were done for these extraneous and irrelevant accident conditions.

4 The subject document is inconsistent within itself, which can cause signifi-

)

~

The' report was originally intended to be an cant interpretation dilemmas.

evaluation of the B&W NSSS design and its susceptibility to pressurized thermal However, the document contains statements which make it unclear as shock.

to whether or not 'the intended purpose was achieved as noted by the following.

In Chapter 1.0, it is stated that although Oconee 1 was selected for the initial

study,

"... thermal-hydraulic behavior needs to be further evaluated as recoquended later in this report and because there are special control systens provisions in Oconee-1 limiting transients, more analysis,needs to be done before.their results are applied tu Oconee-1 or generalized.to other plants."

This is further elaborated upon by the following from Chapter 5.0.

"All the current simulations possess limitations which give concern These for the realism of the thernal-hydraulic predictions.

inherent in the codes and also' result linitations are, in part, from modeling deficiencies and questionable input assumptions..."

in Chapter 8.0 vichout qualification:

And yet the following statements occur "A summary of results for the five over:coling accidents analyzed Table 8.7 indicates the is presented in Tables 3.6 and 8.7.

total number of II?!s that.a 3&W-type reactor can operate before the overcooling transients considered would likely result in I

vessel failure."

i j

and also,

...the inclusion of cladding in the analysis will also result in smaller threshold fluences. Thus, in this rec ect the results in Table 8.6 and 8.7 are somewhat optimistic."

l We consides..these latter two statements as misleading and inapproprista con-sidering the significant limitations of the study.

i E

-..-m5..6

..wg.

n

_,m - -

o 4

5 Mr. Robert M. Bernero, Director October 20, 1981 Page 3 1

i An. additional concern is that the subject document does not sufficiently address significant programs currently in progress that address the areas l

of vessel material properties that are supported not only by Duke Power, but This is also by other utilities that own plants with the B&W NSSS design.

particularly su.prising because by letter dated May 12, 1981, J. Mattimoe, to the 4

SMUD, on behalf of the B&W Cuners Group, submitted a letter report Staff outlining such programs that had been completed and those still under-By failing to recognize the other ongoing studies on this issue, the vay.

This is incorrect.

report implies that it is "the best,available information."

It should be noted that certain branches within the NRC Staff are aware of these programs.

The' evaluation of the reactor vessel fluence aspect, the interpretation of the Oconee reactor vessel material parameters, and the fracture mechanics It is calculations contained in this report have also several limitations.

apparent that the chapter.on fracture mechanics. calculations contains several I

pessimistic presumptions and opinions based on unsubstantiated data and

}

limited information. A technical report of this nature should be based on i

an objective analysis.

yurther, the document f ails to address two important items ubich are associated One is the enhanced inservice examination of the reacto:

with this issue.

vessel beltline region welds in order to achieve a higher confidence level in selection of inicial flav size. As the NRC Staff is aware, such an enhanced examination was performed on the Oconee 1 reactor vessel during the current outage, using an ultrasonic technique with a stand-off distance that allows Not only were all results within ASME cede-detection of near-surfa'ce flaws.

allowable, but also they were smaller than those sizes critical to the ther.nal d

d rather

'All indications were considered to be pre-service in uce shock issue.

The second item is thermal annealling, which is briefly than service induced.

While the technique used in

=entioned, and then only in a positive sense.

centro 11ed conditions may seem premising, extensive work and effort will be i

required to perfect a technique suitable for use on an irradiated ?WR reactor It is misleading to state that such a technique is currently practical, vessel.

i particularly when solely based on a personal communication and pre 1%hary laboratory results.

As in the case of many other severe accidents, reactor vessel thermal shock cannot be envisioned to be forgivdng to all bounding and overly conservative In order to obtain meaningful conclusions of the severity of j

assumptions.

the problem, it is necessary to analyze systematically accident conditions by considering relevant initiating events, mechanistic system failures, and credible operator actions and by utill ing phenomenological models and methods that taka into account realistic system boundary conditions and plant performance constraints." Duke has recognized that the reactor vessel thermal shock issue r

e

,-,9 w

...,e 3

,p,

..%,,-e.-.-

e.,

e

.ww-41*e 1

,.n ew'. E ' " *

-e e e

Mr. Robert M. 3ernero, Director October 20, 1981 Page 4 is a very important issue that requires careful study and timely resolution, and the way to approach the issue is by means of a cogent and systematic analysis of relevant accidents and by consideration of plant specific features Duke has been both in regard to system capabilities and vessel parameters.

fervently working on such an effort, and it is our hope that when this work is complaced, the necessary perspective on this natter will be obtained.

In summary, the report in its present form is not suitable for-understanding and communicating the real perspective of the issue.

In. fact, it could unduly distract attention from the' orderly' efforts now being pursued on the resolution of the issue. Accord'ingly, we ask that the report be modified significantly taking into account our comments or be withdrawn from general release.

Very truly yours, A. C. Thies RLG/php Attachment ec: Mr. R. C. Kryter Instrumentation and Controls Div1'sion Oak Ridge National Laboratory P. O. Box I Cak Ridge, Tennessee 37830 S

be e

i e "

-*w w, - w w -.

4 DUKE. POWER COMPANY Detailed Comments on ORNL Draft Interim Report Evaluation of the Threat to FWR V'essel Integrity Posed by.?ressurized Thermal Shock Events i

1 Chapter 1.0 page 1-2, 3rd paragraph:

4

~

the need to perform " realistic systems i

..We agree with the statement about i

analyses to determine appropriate input temperature and pressure transients for the vassel integrity studies, and (to evaluate accurately] the mechanical integrity of the pressure vessel" through plant specific studies. However, the analyses conducted thus far fall short of this goal, as recognized on 2

page 1-3:

"...because thermal-hydraulic behavior needs to be further evaluated as recommended later in this report and because there are special control system provisions in Oconee 1 limiting transients, more analysis needs to be done before their results are applied to Oconee 1 or generalized to other plants."

~

Chacter 2.0

~

i A clear and consistant definition of the " runaway feedvater transient" is The thermal-hydraulic analyses utilized in this report consider necessary.

this transient to consist cf an unmitigated main feedvater overf eed transient event with a concurrent f ailure of the turbine bypass valve system fellowing a reactor trip transtant. However, the probability discussion of Section 3.1 apparently visualizes this accident as a more general secondary system upset condition which includes stema generator overfeed transients, steam generster pressure contrcl malfunctions, and events involving failures in feedvater flow control' and SG pressure control functions.

l Chaoter 3.0 a

l page 3-1:

i In order to obtain the real perspective of the safety significance of this probles, one needs to consider the probability of occurrence of a break in the reactor vessel at the correct location and of sufficient size to com-promise adequate core cooling capability as a result of crack initiation sad This probability is composed of several (possibly independent) l propagation.

(1) the probability that a break large enough would probabilities, including 4

occur given ' that the fracture mechanics calculations predict a chrough-vall crack propagation, (2). the probability that a through-vall crack propagstion would occur given the specific pressure-temperature condition (this probability P

4 m

a e

pg tt0 w

• -~ -,

+

r

,-y---g

--s

is depende'nt on the probability that fisus of certain unique size and c f enta-at the location of minimum tion capable of through-wall propagation exist material strength), and (3) the probability that the potential transient events produce the' pressure-temperature conditions necessary for unarrested crack propagation.

page 3-1. Table and 3rd paragraph:

l l

No basis is provided for the assigned probability of a runaway feedwater transient (RFT) *. The value provided is arbitrary and is not based on any review of operating experience or quantitative assessment of probability of RFT that causes severe overcooling conditions. The RTT character 1ted by a frequency of occurrence of 1/Ry represents a general secondary system upset condition of an ove: cooling nature and not the accident created in the sub--

sequent sections of the report.

~

~

page 3-3, 1st paragraph:

2 The ETPY results provided in this paragraph are not valid due to the inherent errors and limitations of the thermal-hydraulic conditions utilized. Further-more as discussed in detail later, the assumed fluence race per EFPY is inaccurate.

l page 3-3, 2nd paragraph:

The basis of this statement is not apparent. ' Figure 5-4 shows predictedWithin temperature response for,all transients including RTT and MSL3 (IRT).

j 600 secs for RTT and 250 secs for the MSL3,-primary coolant temperature is predicted to be below 200 F.

This figure would tend to indicate that the 0

0 rather predicted tenparature is well below 212 7 during most of the transient j

than well above 212 F at the time of predicted failure.

Chapter 4.0

• ~*

page 4-1, 2nd paragraph:

I The neutron power signal obtained from the The last sentence is incorrect.

RPS can modify main feedwater demand if its mismatch with the ICS reactor demand level exceeds a set tolerance only if the cenditions in the steam generator oermit, i.e., BTU lbnits, high and low S.G. lev'el limits override.

Loop A-and 3 steam generator feedwater demands are reduced to tero in 15-20 seconds following reactor trip due to the combined actions of cross limits and 3TU limits. Tripping of RC pumps due to HPI actuation also requires that the operator verify the reactor has tripped.

N' hen the reactor trips, the i

ICS controls feedwater flow as described above.

page 4-1, 3rd paragraph':

The section is entitled " Reactor Protection System." The integrated control system (ICS) discussed in the paragraph is not part of the RFS and _ should be separated out.

.9 4

-w--w

+e-.

e w.

---s..

<--,-e.

'r, -. -N

l 3

A reactor trip will not only occur upon turbine trip, but also will occur on loss of main feedvater.

The RPS low pressure trip is 1800 psi.

l page 4-2, 3rd' paragraph:

Only cuo LPI The Low Pressure Injection Systen to incorrectly described.The third pump can be pumps are started automatically.

aligned to either A or 3 crain.

Although the core flood tanks (accumulators) are mencioned in Sections 4.4.2, 4.4.3, there is no description in the system description paragraph.

4 pass 4-2, 6ch paragraph-The LPI System is actuated when the primary system f alls below 500 psi.

Substantial flow, however, by this systen,; could occur only when the systen pressure falls below 200 psi, l

page 4-2, 7th paragraph:

While the discussion of the main feedvater control is fairly il only briefly discusses features of the ICS that tend to limit the potenc a for an. evercooling event.

Further, there is no mention of the E'mergency Feedvater System and its contro totally independent of the ICS.

i and instrumentation, which, in f act, art page 4-5, 2nd paragraph:

The =ain feedvacer pumps are supplied, The sentence is incorrect as vricten.

water from the condenser hoc well chrough three condensate booster and three hoewell pumps.

nakeup to the hoewell, not directly to the feedvater pumps.

page 4-5, 4th paragraph:

3 gallons, the Althcugh the maximum inventorh from all sources is 295 x 10 Although conden-142;000 gallons in the hoewell.

actual usable inventory issate makeup to the hotwell can be achieved from ch i

is 192 000 condensace available for an unconcrolled main feedvater flow event gallons (or for 3 minutes at full flow rate).

pass 4-5, 5ch paragraph:

The ocal feedvacer demand vill run back The first secpoint is incorrect.

at a maximum rate. of 20% per minuce to track generated megavacts following demand.

reactor trip-if the conditions in the ' steam senerator vil e

i t-Q.

Y y--

--y w

we

--t-

-+e r e - - - -eyen---

--yw-+w,

--wa-

4 l

will reduee the demand to whatever value is approprisce.

The second setpoint is partially described correctly; the following should be added. The feedvater valves.will transfer to emergency level control which compares the actual level in the steam generator ca a 50% level setpoint.

This circuitry will either open 'or close the startup valve as appropriate with the pumps controlling on D/P.

In addition to the listed trips, each main feedwater pump will trip on low suction pressure or on overspeed.

page 4-6:

An attempt is made to represent functionally the main feedwater portion cf the ICS. This figure should be redrawn to represent more accurately the control system. As a minimdb the level limiter should be moved above this controller and another~ controller added to control the startup valve on less of all RC pumps.

page 4-7:

For single control failures occurring below the manual control points alsa, the high level trip of the main feedvater pumps will be available to mitigate r

the events.

No discussion of the availability of instrumentation and controls is presented.

A description of the present system was provided co OKNL (copy of. July 23, 1981 letter of William O. Parker, Jr. to NRC) and yet no mention is made of the nultiple instrwnentation available to the operator.

The first sentence on page 4-7 should be changed as follows: This reviev

-divided the main feedwater porrion of the ICS into three general areas, as shown in Figure 4-3.

..,y.

page 4-7, Ind persgraph:

In the second sentence, manual control is required fo11'owing ICS failure.

Sections 4.5.4 through 4.9 use the term excessive 'feedwater on numerous occasions' with no attempt to define the amount of excess. Someone who does know the system may not understand the differences and in fact could not interpret excessive feedvater to mean the hypothetical runaway feedwater transient. This should be clarified in future reports.

The last sentence should be changed as follows: It should be noted that with-the steam generator high level trip for the feedvater pumps, failure of out a startup level signal to a " low" condition can result in an overfeed of one steam generator.

page 4-8:

~

InTable4-1,It should be noted that several indicated failures cause over-feed to only one. steam generator.-

-_m.

,,.k-,

w,

[__,._.-

4

-.v

5-1.,

The Oconee 1 avant sequences referred to were submitted to the Staff in July 1981 as part of the Abnormal Transient Operating Guideline Program. These i

are currently under review by the Staff.

\\

page 4-10,'Section 4.7:

Overcooling transients are, alerted to the operator by numerous alarns and are easily recognized by decreasing temperatures and the causs id'entified by steam generator conditions.

The continuance of main feedwater at 100% flow race requires multiple ICS failures and f ailures o.f other flow limiting functions or deliberate operator action to open feedwater valvts to both steam generators and to disable certain trip functions. Even then, the condition can persist only for a short duration because it is self-limiting (due to high SG pressure conditions or due to rapidly diminishing inventory).

page 4-11, 2nd and 5th paragraphs:

Based on a detailed review of the IRT and TRAC calculations, we believe that to characterize them as being "approximately bounding" is overly optimistic.

Chapter 5.0 Of the four Oconee events, only two events can be considered as representative initiating events of the general secondary system upset condition category of events of interest in reactor vessel overcooling,. These two

• events are the 1/4/7A switchyard isolation event of,Oconee 2 and the November 10, 1979 loss of ICS power event of Oconee 3.

In the Oconee 2 transient the overcooling' was caused by excessive steam load combined with a high initial design pre-scribed steam generator level, which has subsequently been reduced. For the 4

Oconee 3 event also, the major contributor was excessive steam lead (auxiliary steam drawdown and partially open turbine bypass valve) with sone minor con-tribution frem overfeeding one steam generator. In both cases the primary ~

system cooldown was limited to 4200F, and even if the operator had f ailed to take action the transient would have progressed only to a modest overcooling

~

event and not of the severity calculated to occur in the present analyses.

The third Oconee event (June 13, 1975 event in Oconee 3) involved a stuck-open

?cKV, and the actual charmal-hydraulic transient behavior was milder than the calculated small break LOCA transient. The fourth event involved a temporary undercooling in Oconee 1 on December 14, 1978. During this type of an event, the primary rystem undergoes a rapid but finite cooling of the primary side when normal cooling is reestablished. The prtnary system ecoldown is limited to 520 - $40 7 and as such is not different from typical reactor trip events as f ar as overcooling events of interest for reactor vessel integrity are concerned.

It is worthwhila to exmaine the operating history of the Oconee reactors with regard to the occurrence of the "RFT" event, which is characteri:ed by the failure of the cein feedwater flow control system to run back f eedwater flow f

af ter a reactor viip, followed by the f ailure of the SG high level trip of I

the MWP's 'an'd': enet.Trent stuck-open failure of the T3V System, and not con,

r

• • v-'**ae i

e,

• e e e yg e 9

• g-t 97 9;ig g-cm e *-

-==*=*y===y

,- = = = = > -

• e one..

~.

sidering a.ny operator actions. The'three Oconee units combined have now accumulated 23 reactor years of operat ion, during which time 186 reactor trip events have occurred. Our review of these reactor trip events indicates that in'all cases the feedwater was run back, either pronpely or with acceptable delay, after the reactor trip and,,did not represent a perpetual full flow condition. Furthermore, the SG high level. trip of the main feedwater pumps have been challenged nine times as a result of moderate overfeed conditions due to slow feedvater runback or during loss of ICS power events. In all cases successful trip of the system occurred as designed. With regard to the turbine bypass valve system, we have had no instances in which all the turbine l

bypass valves stuck open. Although we have had a few instances involving excessive steam loads and/or partial failure of the T37 System, these events produced only modest overcooling of the primary system. In all cases successful and' timely operator action has been found to occur. Additionally, it should be pointed Jut that design changes have been made and operating precedures have been written to prevent / reduce the probability of steam generator overfeeds

~ (RFT). The present responsd to all Three of the overfeeds listed in Appendix 5 would be a trip of the' main feedwater pumps which would automatically initiate auxiliary feedwater. Auxiliary Feedwater would maintain steam generator level at 25" (240" if all the' RC pumps trip), thereby preventing both steam generator dryout and overfill. Operator confusion would not result on loss of ICS power since adequate backup instrumentation and controls and emergency procedures are available.

i page 5-7, 1st paragraph:

Fluid mixing between the EPI and cold leg is of minimal importance during overcocling transients..The tempera:ure on the downcomer R7 is affected pri-l marily by the temperature of the flaid at the veld location of interest and 4

the fluid flow rate which govern the heat transf er coefficient.

page 3-7, 5th paragraph:

Flow distribution is important in determining the iate of heat removal from 4

the RV wall and thus the temperature gradient in the wall. The assumption of an arbitrary flow affects not-only the thermal-hydraulic calculation but also the heat transfer frca che wall.

page 5-7:

Additional deficiencies in the thermal-hydraulic predictions beyctd those identified in Section 5.3.2 are evident and are is follow:

It is inappropriate to use the IRT code for any external and released A.

This is evident applications since the code is still under development.

by the fact that the code does not have a momentum equation and therefore all the flow races,in the analyses are non-mechanistic. In addition, the code has not been widely used "in the industry and its capabilities have not been demonstrated.

The justification for using IRT to simulate a 3&'4 configuration has not 3.

been established. Once a code has been verified (this has not been comp aced l

e

• O 4

9 O- -,

h* ' * *._

y

- * =

6en.s

7

.t

^

' r IKT), the nodalization of the system being modeled musc be qualified by comparison with data from the system being modeled. There is no indica-cion that this has been done using IRT on a B&W plant.

~

a example of this is the apparent failure to conside,r reactor vessel l

upper head circulation flow in the analyses and, siso, the failure to consider the feedwater injection location and the pre-heating in the steam generator.

C.

There is no indication that the analysts had the necessary intimate familiar-icy with the Oconee plant to set up a realistic and appropriate set of i

boundary conditiocs for a simulation. Overcooling transients are strongly i

affected by boundary conditions. Without a realistic set of conditions, -

. che transient response will not represent the true response, and the results are essentially meaningless. A lot of simulation experience in terms of plant system familiarity and knowledge of code capability and limitations are essential.

Some examples of boundary condition errors are:

I

~

~

a.

Incorrect feedwater flow.

b.

Incorrect turbine bypass setpoint and capacity.

c.

Incorrect EPI actuation setpoint and' flow versus pressure.

d.

Feedwater enthalpy versus integrated flow delivered.

e.

There are no secondary steam relief valt es or atmospheric dumps.

f.

Caission of control system responses or sdditional assumed failures that are not identified, e.g., high 3G level trip of both main j

feedwater pumps.

i D.

It is very misleading to labe.' a particular analysis without explicitly identifying the failure assump:icas made in the analysis. As an example, the 13LT analysis labeled, " Turbine Trip", is actually a turbine trip vich j

a f ailure of the main feedvatar to run back, with a f ailure of the high SG level trip of. the main adwater pumps, with a failure of the turbine bypass valves on both steam generators, assuming a rapid ' decrease in feed-water temperature, and assuming a f ailure of the operator 'to terminate the overcooling or perform any other mitigative action. The assumptions l

which determine the transient response should not be lost in the ge eration of picts of results, and neither should the limitations of the code utiliced.

Chaeter 6.0 page 6-1, Secticu 6.1, last paragraph:

Although the variations in Table 6.1 do occur in source parameters, they are not necessarily uncertainties in the. calculation of fluence. Many of these-items are accounted for in the calculational procedure. For example, eye ~a and cycle-to-cycle core power distributions are averaged over the capsule irradiation period with the use of PDQ generated power distribution data at i

selected time intervals during fuel cycles.

The basis of.t.hese statements is experience in the analysis of 12 capsules from 8 B&W reactors.

+

i o

5.

cy 2.-

,)

d ^

.'y

.,,p,

4 l

page 6-1, 4-2. Section 6.2.1, 1st paragraph:

l Although this procedure was used to calculate fluence from Oconee 1 espsules i

OCIF and OC,IE, an improved procedure is presently being used which incorporates scattering cross sections directly into the r-G the capsule geometry and P3 reactor model, thereby eliminating the need for corrective factors.

page 6-3. Table 6.2:

An important step ves omitted, that' of normalization of calculated flux to flux derived from measured dosimeter' activities.

This table should read:

j. Calculate capsule flux (E>l ME7) by multiplying the value from the r-G model times the P /P1 and capsult perturbation factors and cLaes an axial 3

shape factor based.on the axial power shape in a peripheral fuel assembly.

k.

Obtain a normalization ' factor from the ratio of flux (E>l MIV) derived from dosimeter reactions to calculated flux (E>1 MEV) in the capsule.

l 1.

Perform an axial. 2-D, P, r-z calculation.

1 1

Correct flux values from the r-e model with the P /P, capsula perturbation, 3 1 m.

axial shape, and normalization factors.

For veld locations, displacement factors frem the :-z model and r-9 model n.

are applied to the vessel flux (or fluence).*

page 6-5, 6-6, Section, 6.2.3, last paragraph:

• ' spread in normalizing factors is misleading with respect to caldulational uncertainties because only fission reaction data frem the OCIE capsule were uced to calculate fluence. Data from CCIF vere discounced because of suspected errors in activity measurements. This was the first capsula anily:ed at 3&W

~

and such large ' discrepancies have not been observed in any subsequent espsule analysis.

page 6-6, Section 6.3, first paragraph:

The uncer'tainty evaluation in 3AW-1485, was primarily based on conservative estimates with relatively little experience. Thus values of 1 30% for predicted beltline region fluence and i 50% for certain veld locations. vere reported.

Since then, 3&W has participated in the Blind Test, a calculational benchnark sponsored by the Light Water Reactor 1,ressure y_essel Deosimetry improvement l

Program (NRC funded) and the OCIT fluence calculation 'has been checked by another phase of the LWRPVDIP.

(R. L. Simons at HEDL did the analysis.) The I

Blind Test indicated that the 3&W transport calculational procedure would produce a f ast flux (E>l MEV) that deviated i 5% from a notsalized capsule location to vessel surface and T/4 locations. The HEDL calculation of capsula fluence was 5% greater than the 3&W calculated value. In addition, analyses of 12 capsules i

from 8 3&W 'rehttors have consistently shown I/C values within i 10% for fission l

e

((~

'T, p

,e4y,,9

, g.,ew 9e 3

w n-

+v

=

-9

~

reactions. Based on these developments, recent estimates of fluence uncer-l for t tainties are 110% at the capsule, t 15% in the vesse fluence in pending to capsule irradiation periods, and i 18% for predicted Comparable values for vessels in reactors without cap the future.

and 1 21%.

values in the ' absence of a detailed uncertsinty analysis.'

t 18:

50% value reported in 3AW-1485 for veld locations was intended.co in i l and azimuchal displacement 3

The 1 the added uncertainty (above 130%) of using ax aApparently, this was A displace-4 in the ORNL ment f actor of.89 (as is used for the critical veld locationwhen compared to the seitline factors.

analysis) cannot be in error more chan -12When statistically combined wit i

fluence.

uncertainty.

this would result in a 1 32 pages 6-6, 6-8, Section-6.4:

i l

there has been no verification of the 3&W calculat ona In fact, 36W has The implication that procedure for the de' termination of fluence'is incorrect.

Vessel Dosimetry successfully participated in the Light Water Reactor Pressure d the Improvement Program to benchmark both the calculational procedure an dostmeter measurement technique.

page 6-7, Table 6.5:

fast flux Data in this table apparently.are based on extrapolation of theTo obtain mo dure described in 3AW-1,485.

averaged over cycles 1 an'd 2.

tion (in tLne) should be based on a predictive proce of-For Oconee 1, the use of this procedure is particularly important becausespo a conversion to an 18-month fuel cycle in cycle 6 with a corre in ex-core f ast flux of approximately 30%.

Outside of

'InsiAle of 3T/4_

RV Wall

• /4_

RV. Wall _

1.053+17 2.20E+18 1.22E+18 2.86E+17 2.67E+18 1.48E+18 3.47E+17 1.27E+17 3.26E+18 1.81E+18 4.23E+17 1.55E+17 3.66E+18 2.03E+18 4.75E+17 1.74E+17 O 1 vill be the 3 asis is assumption that relative effect of 18-month cycle in AN same in Oconee 1.

l but the calculations have not been made.

Chancer 7.0_

page 7-1, Section 7.1, Table 7.1:

h data are Detailed descriptions of all data used and certification that sucAlso,' error analys appropriase ice those analysis have not been provided.

for input date has not been provided.

l

• 1 e

+g

page 7-3, first paragraph, next to last sentence:

The basis of the statement that uncertainty is not large in the parameters included 1n,ASHI Section III is not provided.

e i

first paragraph, last sentence:,

Tnis - sentence - conflicts with : the previous - stat ement. Data should 'de co ' support this-position. Explanation should be given as to how such data relate to the data which are used to evaluate vessel integrity. Also, data for the KIR uncertainty in the determination of RT'DT should be provided.

h second paragraph, first two sentences:

These two sentences appear to be in conflict. They should be clarified and supported with actual data to substantiate the ooinion expressed in this para-graph.

second paragraph, last sentence:

The reference to support this statenant should be provided.

third paragraph, fir'st sentence:

The reference to support this statement should be provided.

third paragraph, last sentence:

This statement does not recognize ch's Oconee Unit 1 and the 3&W Owners Group Research Program which is in progress and is generating this data.

page 7-4, second paragraph:

This paragraph appears to express an ooinion which should be based on sdund data. Since the statement is made that Regulatory Guide 1.99 is "not exc'es-sively conservative" for Oconee I weld metals, the data should be either pre-seated or ref erenced so that a better definition of " excessively conservative" can be better understood.

second paragraph, last sentence:

Over 35 surveillance capsules have been removed from power reactors and the data support the conservatism of Regulatory Guide 1.99.

As for irradiation programs at test reactors, most of these are complaced and the data are available, page 7-5, third paragraph:

The scacement' regarding reduction in upper-shelf energy of weld e adding is aisleading. No mention is made of the fact that these fluences are well above that expected at EOL of any operating pWR. At the fluence levels predicted, the claddin.g is expected to lessen the degree of crack propagation.

/

4 v

~

11 i

page 7-5, Ref erence 8:

~

This information should be included in the analysis and sheuld not be stated I

as a reference since it represents a significant reduction in EOL fluence, j

Chapter 8.0 page 8-2, first paragraph:

It is stated that if the temperature of a major portion of the coolant in the -

primary system is above 2120F, the opening due to crack propagation may be excessive and core coo ing not ma nta ne.

It is interesting to note that RFT l

i i d 0

and MSL3 transient show bulk temperatura decreasing to below 200 F.

a

^

Even with the considerable errors in the transients analyses provided, it could be postulated that the copious amounts of LP injection available would be more than sufficient to maintain the core cool, in much the same way as it is predicted to occur Euring a postulated L3LOCA..

d page 8-2, second paragraph:

It should be noted, again, that only the vessel material properties are approxi-mately representative of Oconee 1.

The accidents analyted are not at all representative of Oconme 1 or any other plant with B&W NSSS.

fourth paragraph:

Tor overcooling events, little if any vent valve flow will occur because there is minimal diff erential pressure between the core outlet and inlet. The thermal analysis is dependent on the downcomer temperature and the flow conditions.

~

It is not apparent what flow conditions were assumed in the thermal-hydraulic calculations and thus what was assumed in thermal analysis of the KV vall.

page 8-3, second paragraph:

s-The ORNL analysis ignores axial gradient in fluence. Axial gradient of fluence i

6 is Meiliced in the B&W analyses.

The assemption of a pre-existent long sharp crack is unrealistically conserva-This is particularly true for Oconee 1 which recently underwent a 100%

tive.

examination of beltline region welds with no cracks indicated.

page 8, third - fif th' paragraphs:

No basis is provided to support the assumption that the fluid-film heat-transfer coefficient is 1000 5tu/hr-f t2,oF.

It is stated that this corresponds to full-flow conditions, but i-is not stated what flow conditions were actually assumed in the transient calculations. It is inconsistent to assume one mode of system transfer coefficient based f operation during the transient calculation and a heat on a different mode of operation. This is particularly important in that severe overcooling transients may interrupt RCS flow and thus reduce heat transfer; and vic'h RC pumps -assumed running, a finite amount of heat is in fact added to the RCS.

O

~

• =*

=e s

e t

9-

__a

page 8-6, Table 8.1:

The CRNL analysis should have utili:ed actual veld parameters inasmuch as is that of the base metal rather these data, vere provided by Duk,s.

The RTNDT than the veld and the chemistry is that of a hypothetical veld metal.

page 8-13:

It is inappropriate to perform the fracture mechanics analyses of the IRT steam line break of RTT cases with all their known deficiencies and atypical-ities. The analysis, discussion and results for these two esses should be deleted.

page 8-16, Section 8.4, last paragraph:

n/ca /EFPY is appropriate for the critical veld 2

19 The fluence rate of.046 x 10 location (lower long wald at inner surface of vessel) for the first 4 ETPY (through cycle 5). Thereaf ter, because of the conversion eg9theig-month fuel cycle, a better value for >4EFFT expcsure is.033 x 10 n/cm /ETPY at the critical weld location.

Also, as noted previously, the i 50% uncertainty does not apply to velds close to the beltline region. Based on the initial analysis at the time RAW-1485 was written, this value vould be 132:; more recent estimates indicate about i 20%. These are estimates because a detailed uncertainty analysis has not been performed.

page 8-17, Table 8.6:

The colu=n with WPS and without WPS appears to have the line item designations WPS should provide additional time to fract re in all cases excepe reversed.Table 8.7 appears to have this relationship correctly iden'tified.

L3LOCA.

page 8-19. Table 8.7:

The revised values of fluence /IFPY to account for the 18-month fuel cycle vill lengthen the threshold time calculation for times 1E17T.

a Threshold Time

'(E7?T 3 Comments and Qualifications 26 WPS not assumed effective (44E7P if it vare).

V 26 Through-vall crack predicted.

19 2

10 n/cm /ETPY for the inicial4EIPTand0.033x10{nratevalueof0.046x b3ased on fluence accumulati 9 n/cm /IFPY thereaf ter. This value may have an uncertainty of as much as t 323.

o

^

4 o.e v e-n

.-9--

O

...e t.

i Chapter 9.0 page 9-1,

-2, Mitigative Measures:

In view of the inherent weaknes's ' contained in the report, it is considered that identifying the need for any changes is premature.

As a point of clarification, increasing the BWST camperature would have essentially no effects on reducing the severity of an overcooling transient.

3ecause a S3LOCA was not performed, it is not clear what basis there is for stating that an increased BWST temperature would reduce the degree of over-cooling caused by actuation of HPI. In fact, with vent valve flew and plant-specific analyses, the thermal shock concern for $3LCCA is minimized.

page 9-2, sixth paragraph:

~

~

The practicality of is-place annealling is overstated. While the principle may have been demonstrated under controlled laboratory conditions, extensive work is necessary to implement in-place annealling on in operating reactor Extensive evaluations are necessary to demonstrate the acceptability vessel.

750 -850*7 if a plant and its support systems 0

of annealling at temperatures of were designed to lesser camperatures.

page 9-2, Section 9.2.2, General Changes:

An additional change that is presently occurring in 3&W plants that will sig-nificactly lover the fluence on the reactor vessel vall, is the result of goin's to 18-monrh L3P reload cycles.. Once-ourned fuel is loadad on the peripherv of the core thus lowering the peripheral fuel assembly's power and the co'eres-ponding leakage flux or fluence to the vessel. The results of this are noted in the references of Chapter 7 Reference 8.

Chapter 10.0 page 10-1, Section 10.1:

See previous comments on fluence analysis.

Section Concluding Remarks:

and are leading to uncertainty and lack All rsmarks are negative in con's.

Yet, scacement is made that, "Nenetheless, of confidence in the final resu tn.

for all their shortcomings, the analyses at hand are the best presently avail-able on a nonproprietary basis, and... (merit} a great deal more study using refined techniques."

page 10-2, Section 10.2:

The ispiteation hare is that fluence calculations in general have uncertainties in the range of + 30% - 150%. The uncertainty in the Oconee 1 fluence calcula-

)

tions is such smaller.as discussed in the comments on Chapter 8.0.

o

---m,.....

_