ML20150E666

Charging Pump Gas Binding & Cracked Block Evaluation Rept

ML20150E666
Person / Time
Site:	Palo Verde
Issue date:	02/29/1988
From:	ABB COMBUSTION ENGINEERING NUCLEAR FUEL (FORMERLY
To:
Shared Package
ML17303A982	List: ML20150E666 ML20150E671
References
CEN-370(V), TAC-68073, TAC-68074, TAC-68075, NUDOCS 8804010257
Download: ML20150E666 (62)

Text

- _ . - _ _ _ -

CEN 370 (V)

Charging Pump Gas Binding and

.. Cracked Block Evaluation Report Prepared For ARIZONA I NUCLEAR POWER PROJECT PALO VERDE NUCLEAR

GENERATING STATION

. February 1988

! COMBUSTION ENGINEERING R884 les!T:fi8s88:e P DCD

i l

l l

l l

l

. l l

l LEGAL NOTICE THIS REPORT WAS PREPARED AS AN ACCOUNT OF WORK SPONSORED BY COMSUSTION ENGINEERING, INC, NEITHER COMBUSTION ENGINEERING NOR ANY PERSON ACTING ON ITS SEHALF: ,,

A. MAXES ANY WARRANTY OR REPRESENTATION, EXPRESS .OR i

IMPLIED INCLUDING THE WARRANTIES OF FITNESS FOR A PARTICULAR PURPOSE OR MERCHANTABILITY, WITH RESPEC'T TO THE ACCURACY, j COMPLETENESS, OR USEFULNESS OF THE INFORMATIOR CONTAINED IN THl3 ,

REPORT, OR THAT THE USE OF ANY INFORMAYlON, APPA.'8 ATUS, METHOD,

• OR PROCESS DISCLOSED IN THis REPORT M 4Y NOT INFRINGE PRIVATELY i OWNED RIGHTS: OR

8. ASSUMES ANY LIABILITIES WITH RESPECT TO THE USE OF, OR FOR

- DAMAGES RESUl. TING FROM THE USE OF, ANY INFORMATION, APPARATUS, METHOD OR PROCESS OlSCLOSED IN THIS REPORT.

l

\

1 Y

U G

r l

i l . I

)

l l

CEN-370 (V) i l

CHARGING PUMP GAS BINDING AND CRACKED BLOCK EVALUATION REPORT l

l

.. Prepared for i

i ARIZONA NUCLEAR POWER PROJECT ,

I i

PALO VERDE NUCLEAR GENERATING STATION f February 1988 Prepared by

.. C-E Power Systems COMBUSTION ENGINEERING, INC.

l 1000 Prospect Hill Road Windsor, CT 06095 1

4 i

ABSTRACT l 3

i On September 12. 1985 PVNGS Unit 1 experienced a loss of offsite power event

]

during the performance of a load rejection test. A total loss of charging flow was experienced when the charging pumps became gas bound with the Volume l

Control Tank (VCT) hydrogen cover gas. One of the concerns expressed by the  ;

NRC following this event was the potential for damaging the charging pumps due j

,, to gas binding. Based upon the information available within the industry at the time, ANPP informed the NRC that operation of the charging pumps with a i gas or gas / water mixture could result in the generation of internal pressure l spikes sufficient to crack the pump block but the pumps' ability to deliver the roquired flow would not be significantly compromised. The NRC then ,

l requested ANPP to perform an evaluation of the effects of gas binding on an

operating charging pump with a pre-existing cracked block. The first phase of  ;

this evaluation was the performance of a test program to quantity the peak 4

pressures and strains inside the charging pump fluid cylinders during a wide i range of charging pump operating conditions including gas binding. This test a

program has been conducted. The results indicate that gas binding does not r I i l create an operating condition for a charging pump block more severe than  ;

ll normal operations. The purpose of this report is to utilize the test program

] l t

results to evaluate the effects of gas binding on a charging pump block with a l

,. i pre-existing crack.

l 1 'o l I

1 l

1 l

)

, i 1

i 4

l TABLE OF CONTENTS Section Title Paae 1

Abstract i 1.0 PURPOSE 1-1 )

2.0 BACKGROUND

2-1 3.0

SUMMARY

OF RESULTS 3-1 l

3.1 Description of Work Performed 3-1 3.2 Results and Conclusions 3-3 4.0 EVALUATION 4-1 -

4.1 Analysis Methodology 4-2 4.1.1 Crack Initiation and Crack Growth 4-2 Within the Block 4.1.2 Analysis of Through-Wall Crack Growth 4-9 4.2 Confirmation of Analytical Model 4-12 4.3 Leakage Rate Predictions 4-15 4.4 Industry Experience and Analyses 4-22 4.4.1 Previous Gaulin Test 4-22 j 4.4.2 Industry Operating Experience 4-28 l

4.4.3 System Interactions 4-29 4.5 Leakage Monitoring 4-32 5.0 LICENSING EVALUATION 5; I

6.0 REFERENCES

6-1 11 l

1.0 PURPOSE The purpose of this report is to:

a. Provide an evaluation of test data for the effects of gas binding on an Arizona Nuclear Power Project (ANPP) Charging Pump. Test data was obtained from a recent test at the pump manufacturer's

,. test facility (APV Gaulin) and the test results are documented in Reference (2).

b. Demonstrate that a pump with a pre-existing cracked block will not catastrophically fail from the effects of gas binding. .

c. Provide technical justification for resrinding the Palo Verde comitment to declare a charging pump with a cracked block inoperable within 72 hours8.333333e-4 days <br />0.02 hours <br />1.190476e-4 weeks <br />2.7396e-5 months <br /> of the detection of a through-wall crack.

d. Provide ANPP with a basis for monitoring crack growth and the associated leakage trend to avoid the loss of the charging pumps' functional requirements.

i I

1-1 ,

l l

2.0 BACKGROUND

The charging pumps provide charging capability for the Chemical and Volume Control System (CVCS). The charging pumps are three-cylinder (triplex), horizontal, plunger, positive displacement pumps. Each pump contains a packing cooling and lubricating system which recirculates reactor make up water over the packing. Each pump is also provided with vent, drain, and flushing connections to minimize radiation levels during maintenance operations. The rated pump capacity at rated pressure (2735 psig) is 44 gallons per minute (gpm).

A picture of a typical Palo Verde charging pump is provided iq Figure 1 e

and shown schematically in Figures 2A through 2C. Figure 3 shows the charging pump block glometry and identifies the area of concerr..

During normal operation the charging pumps take suction from the Volume Control Tank (VCT) and provide flow into the Reactor Coolant System (RCS). Two of the three available pumps are normally in operation and one letdown control valve is modulated to maintain the pressurizer level. Seal injection water is supplied to the Reactor Coolant Pumps by diverting a portion of the charging flow at a point in the system

. just downstream of the charging pumps.

In the Palo Verde analysis for the natural circulation cooldown event

[ Reference (8)), the charging pumps are credited for delivering flow during a natural circulation cooldown where a flow rate of 32 gpm is required. During this event the charging pumps take suction from the Refueling Water Tank (RWT) via the bypass line around the boric acid 1 2-1 L _.

4 . . .

1 i

i i  !

l l

i l-i ~

m . . .

f

! C f. k i-

~~~jj", ,

" N.

.-. <- 4+ '

l

[ ,. .

f.C . -

4h_==._ .

• • y, k ,-

l ,

'? si -

- m '

j I '

I i

l W' ". .

\

l M- t. ., _

i 7--

iE j ',__

t fhs.* *+yo t< <* --

hitM -+ - ~"'

\ '2 FIGURE 1: TYPICAL PALO VERDE CHARGING PUMP l

l j

l l

4 g l 3

.d$4,i ?Jbb

• ~ ~ ~~

ent -r j , 0p i3 k 3qu3 p ,up ,

. i.LI 4 10 j u 41441 ji  !  !

j!,ytt. , "gg  ; i

,a $1 o;,hn p:i ):-l6 y!P; i ,ij i i33 4~ 1 ! g its 4. . < h!

• u ; .t!! !!

-J 8

)1 l1

'! jp e- a o- -c ojppi :/] l r ...  ! n nq l! ;4?*!! ,N $3!b a NS II pg 3

'g.si6jkf 'l 2

- an.g !r!:!!

w

-h;;ibqrsn!!,;sujjjji}; 3  ! iru:lnds I l r(d.'

i h, 3 p' sj u \

n ,si u m !, , #

3 3 u en s - tes38 e . 31 ec as c-el T--- 8 ji d g{ i N \. \ !j'3.3 j  !\- I36 3*a M= .#,$ ; .

hr. '

c' C h" y .S

. eah q,. # --j 96>1 i; 1,; e N/ gQ q 'g *g agg '

.. Ibfii '

.: 4 idis it5 il!!'y e-

L oo A 1 e,d = i:]! [t- Ee !!H s l

9 faQ. ~k'r w'i fi 7 M, KoA[ !ph l c ..,

q _

n

.ga

,i y.!.i

.g , gg- , c3 a, ,_

g,L..g y%  !

-l i i .o 3._, .

3

{~ ~~~

,... . # p' i! 5 ,.. M k , _

I l ls . ,

fbthe 5

. ,! 4 r n ii 4gi jp eegdo.;1 i

!! ,i 'E s

t y.

i

%r

.7.u:qs.,gn.,,g. _j g:

a! ini.r.m 73 e a

M i :  ; i I

hw .- . ri p" =

@e .'aJ sp"a**yI !1i j S. 3 J

dr , /

y d

.= ' [~I E

b;

, 2

![

bq gi k,A (! ~h j! ! s I8 f 5 5 b;

s s, 47 r:mq, i

- il .'.- '

q s

,y  % Ej  ! b;3-.! [g< - c e / J.H! ef g Q I -)

/ '/ /\/ [ , !3If~  ; Q l p N a.

i . > N L e g[3j - - - - - --('

,,* s

1. 7' d

'" u

, g g;g g  :: :n :---l u I

.S

,, w , n et g~}$ 1 ,.,g \ v-1 ,- ;I .

A,vp8M j i i

Sh i

's is m.[ t J:i!!%o;a;*n j lf@x_3]5 L 1 A

5

> r= 1

l9 a 35jg h= $n# $ o 1 ,

II

.i

~ ~%/ t= .

jh*l'+*l 23 ,7

.i

!! .!h[59 h\x l u ' Mi y fl Qu:!

!;'ialily s

13 5 .5li , s; ,,5 ijs -

C\ i1L du aig sl a

d e il s n~i Y .~

is!!!an,s .

.. .son

.. ,!!, i;ioil

[

!j!3l! l[ ' -l5l} 15Q

=

as;:

lh O

rS ee 2

??i'/I!L, -_

2-3

i

I f Jr thf1 kh Ili!IIIk d' I e sci!ith,[MItMi e sli Il,wmi]dmtedDMifaiii ,j 2 l

5U::U t.; !ii ts i rs H i R , rt s a } .} ]. 'l .s![-{ flij 3 :i gji$:

t u!!)iV.1,.i!

7

$1 f o I th;

i 1I!

p l 1

[Ijjid.,!lf'I gf

~

f h!!N:= f 3:,f y U ., j $

,M::Jli!

3 i l.: N :3r

[.}llj 1h i h]. Jiph Qg >I ric ! 19sl

!!d i9p fEl d i-li IS.'p I

ev s! ,

p!bE!N HII?ilWll pfg! '$, L ! e j- j -

a Ih!!!N. f!!hii} lod Ilih l}hjhnj !

!?

; ; ; ;l- ; - -, ,,, , .., ,-!,,-t-t- -....y  % l.. j ;-4 l

[lf 1

' ,,, ,, ,,,, }f C%2:l:!2 .4 . : ;;. .,g(, , . , , 3 ; y; , , , ,l,f,,(,ve  ;,j ,l, [ m mm.

m

... . 2.

uumm >'

.Sen n 3 .(

.. E

, k ,b U z

\ -

/ -

b c:$ i @ o  ;

Co] Co ,lt  !

O o i m.- gg  ;; t o o4

, l

. Z O \

@. O. r-o a y N

.:::t .

u/ ci,p a O p a

@c - x  !,

.  ;; w$ &&&n ;

1 B

! 1e .. I 2 <

6= @@e/ e - !. e ege--@ee/ e g s

s yA q i

g a' Q 'M

\ _ i g

• - 3 n :.

8

..y j

! =

9 M'k

%s\ , , . 1 a

I'% --- u 1

3; P

[

~

m cc i

! w k ., b * .

Sc.!dlP9 i A  %

%N ~i N " '! T is

%~.~[N ip g!j c df !

!b!!!Ni! e' E milwn!!ilh!)lI! 6 i

s!+rnii?Gjij;!?

e m .df y.i

=T lel a9 e

tow i

r- _

gt 2-4 I

( - .

j ,

w I . i

, . . o .

gn i

i ,

i,l t i Iw. P i(;lihff l fIl i

'I l

I I, l'l I

tI l 'i i il

!' h3? g f.,,,s ji l l .' 4 l !l H U!f 9 Hi 3,2.t!$'5 lti:Bis hp5Ol2ji!!j!fi['!i 2 2 :2dQljj T $!- e,

$]lhi iplE2!26@$eohe2@1P2jrI ijlM l l a n!},l{.CtM h

fy,

>s , , t a y i

• Jr .A I ..

- l h

l

! O i a

t

)

. l.' $. 5.' '

• , ,h X i

• 5 t j

• j ! !;

5 3 .

g . s

} u s j $ $ 3 ._ _L. N

-  ! )s <

3 di 11 8 <

l x sj h 3i g! b s W g s e e -  ;  ! s elt 3 ll3 fl. gS ti u s g es ! il -

i h w $I I.l

~

• t s

( g $ 5l $ 3I (( h W l I l h( i*8 l

g. 19 1 3

gi s : 3 1! ; p i 11 m ! !!; :s

)-!' Il li ss ii s li !! ! r' siss}i i s s j :;! n

! ! 5 ! ! ! ! 1 pMp-!O I l: np i 1 1ks!!

a ln i 1 '

N;u!Dwr t

6f l f fg8 f 8 5 5 fl i s g ! ! s s 5E i ss5 -]

8 e -

1

• - - " ~

• - - - e e c c a w

. .i o y g..N -2 -- -

- d- ~ --

• a - a a - - * *

-l- ~ ~~h a

• jpl$n4 a n nlia;li ? q;;;l:ligjjijn nn n m : ii n &

M .

)

e E

i l' U t?>

!i Ill!illiliit!!!i!!i!!i! !!!!!!illi! i!$lil!H!il!lli n

le i  !

a e

8  ;

g o.

de C **

Y ' #

5 g kW h N  !  !

- d I! 3

]

,s 5 N.Jss #8 g 5 3 si

.I

)y ;y *eP -

3

, $ t

. 9 6 5 M' l 1 I f 3 i  ! l h h f bh' ,,

f  ! 5  !

th b $

p ;l ri i i sig

. - >I -

e w h

• N .g 2 n n

a gsi. 1] g.-( I f i i- # r m .$6 a 5 s [4 :

1 0 #!

g i

• 25!>3 2 k 4

i ni! id j = 13

• f 9i

1. i

$ }} h:l1[]oh,hh:*.mit i.,

' E lj{lT E ,1{!] 2'e

.N 1 W ,yI-[!p' se,-ad m 3dlj k5712 gy,3 l 2'yil i i.

J { a $ Y v d  !@wh J

[y. )l {jg - -

1 #

- -  : . v {i-$$ 1 11

$ - g$ad'j g t y ! l y

.- am t 3

9 -

. $ $ s' s I' d :::ist-Ei1.1 UIIIIlYl !I%EbT>$$

O, r}54)2 3 3? ,"*l 4 kTh sf / . 0' T' ' ' '

5YbsN5l(5 5.p!$ W i'( [4, s-- *la!-l-M-l- - -

f sf fl*!--l-l-l- -lch l-l-l-l-l-l- -lbf tf'[-l-l '

i eg.j

_ .t -hbhie{*'(*de{Sf[

<<t , m$h[ *f[tltN)[$h :: ; ' r]b 2-5

' i,

~

r dr<

n%

l

.4 9 y . 4.,

ii 5lke Jyj r- -

l,l x ,=

m h' '

jfE s , g

~~

1 :s.-,_i_ _.._y  ;

=

w= . , g-I p%.b er' i / a i a

-~

~ ~x

<//

1 a

2 >

1,.

l1 # .

)

,. l d:' - L C_ t._.a= j =9 ,

M ,

. - i..

t u,f ,

w o l .

e- e v

i x u s  : I

~

W

~ "

g

}'

mq~

em]?~u %g#-

l f e i 8 E

, i h h i .. z- $!

f*L7Q' w= _ ,

. _ C 9 4

2-6 3% j 3

4 SACK VALVE SORES FRONT m

• PLUNGER SORES b@@

s %s l I

/ j VALVE SORE

~ l i I N 3 I J [

n,'

SUCT10N SECTION A MANIFOLD l %,

, ' CRACK SUCTION g i MANIFOLD L 4::a

, FRONT 1 3

/

p e

SECTION A f!GURE 3: CHARGING PUMP BLOCK GE0 METRY (PAGE 1 0F 2) ,

l l

W c, v,

,i, 1

J, 6

l Crack Initiation Location N s .

, l

~*

• • = ~  ! T . . . - - - _ =

h

,z _

'wAugs; J c

~ ~

hhw A ..b, A

) 4ptrwucs -,

A Q_

-j ' ~' \ \\ \ s ' ,. . ~,,1.

\\\

W

-~

/ L\\ \\ M\M\\ \ '

\ l V Q 6}}){ .

f

_. 1__

\

?:  :+~ l l

N\ h \ .\ \\ \ suction

~ H{O "

N crack hBLocK

^-

-w ~~ nunch " l

-a.wau Qeano s wo sw FIGURE 3: CHARGING PUMP BLOCK GEUMETRY (PAGE 2 0F 2)-

RECIPROCATING CHARGING Pl.ff FLUID CYLINDER SUB-ASSEMBLY X-SECTION, CRACKS OCCUR AT TFE INTERSECTION OF THE HORIZ. P. VERT. BORES -

- AT WE OUTBOARD SIDE (TOWARD THE FRONT CAP) 9 6 & 12 0' CLOCK and PROPAGATE T} ROUGH THE FRONT FACE OF THE BLOCK.

2-78

l l

)

make-up pumps. Some charging flow is diverted through the auxiliary pressurizer spray line to provide RCS depressurization capability. A l

schematic of the Palo Verde Charging / Auxiliary Spray System is shown in Figure 4. l l

l The current charging system design meets the requirements of NRC Branch Technical Position (BTP) 5-1 for plants with Class 2 implementation which permits operator action following a single active failure in the system. If a gas binding event were to occur the charging pumps could be vented using approved gas-venting procedures. These procedures have j been successfully tested [ Reference (9)].

l During the past few years, there have leen several incidents of ,

charging punp block fluid cylinder cracking which has drawn attention to the issue of charging pump reliability and availability. NP.C concerns expressed in rec y.t meetings between the industry and the NRC l

have centered on the overall perception that charging pumps are not l reliable. j On September 12, 1985, Palo Verde Unit i experienced a reactor trip with loss of forced circulation and degraded electrical conditions.

. During this trip, a total loss of charging flow was experienced when the charging pumps became bound with hydrogen from the VCT. This event, coupled with the ongoing problem of cracked pump cylinder blocks at several other nuclear plants, prompted the NRC to focus more I specifically on charging pump operability at the Palo Verde Nuclear Generating Station.

2-8 I

t i

i 1

Volume Control Tank Refueling A

Water Tank lI LT-226(B)

A j > LT-227(B)

CH-203(S)

_ E CH-501(A)

S v

l

' S CH-431 Main Spray CH-514 g  ;

CH-205(S) 7 Auxiliary f,]

s Spray y M

CH-190 CH-524 i Pressurizer (S) CH-VM70 Boric Acid FO-212 CH-239 CH-240 Make-up Pumps CH-536(A) (g) M __

8 N Exchanger Heat -- =

To C M 29 j e Charging N

"***8*

, (S) = Safety Grade Component _

CH-435 (A) = Recieves A Train Power (B) = Recieves B Train Pow Charging Pumps Containment W888 (S) i FIGURE 4: CliARGING AND AUXILIARY SPRAY SYSTEM

. , . . . ~ _ _ _ _ _ _ . -

l l

)

The NRC subsequently re7 tested that Ariz?ria Nuclear Power Project (ANPP) Jake additional conmitments to ensure charging pump operability.

ANPP's response to this request is3s ;rovided in Reference (1) which included a commitment to evaluate the effect of gas binding on an operating charging pump, including one having a pre-existing block crack End, if possible, determine if such a postulated condition would lead to a failure of the pump to deliver the required flow. In

.. addition, ANPP was required to declare a charging pump inoperable if indications of a pump block crack (e.g., boron crystallization or weeping), were observed within 72 hours8.333333e-4 days <br />0.02 hours <br />1.190476e-4 weeks <br />2.7396e-5 months <br /> following discovery of the crack.

The first in a series of responses to charging pump cracking problems at Palo Verde started in May,1986. At this time, the charging pump manufacturer, APV Gaulin, Inc., performed some testing of a Gaulin Triplex Pump, Model MC45. The test showed that internal bore pressure peaks of about 2.4 times operating pressure could be experienced during air injection or starvation. Based on the limited test data, it was postulated that a gas binding event might result in internal pump  !

pressure spikes in excess of the charging pump design pressure.

Specifically, the limited test data suggested that peak internal l

. i pressures in the range of 7000 psi existed for a relatively short j

., period of time (e.g., milliseconds) during each piston compression l stroke when a gas and water mixture was being pumped, prior to the pump cylinder becoming gas bound. Once the pump became gas bound, the data showed that pressure peaks were reduced due to the compressibility of 2-10

I the gas. These pressure spikes were later attributed to the slow response of the homogenizing valve and are considered not to be representative of an actual gas binding event. The results of this testing are discussed further in Section 4.4.1 of this report.

In March,1987, ANPP comissioned C-E to determine and evaluate the effects of gas binding on an ANPP charging pump. The initial effort 1

included developing a formal test program to quantify the cyclic i pressure loadings during gas binding events. The testing was conducted at the APV Gaulin, Inc. facility on an existing prototype that was equivalent to the Palo Verde pumps. There were a total of 37 tests l performed in this second round of testing. The maximum pressure shown i from the data is about 2850 psia, only 385 psid above the operating l pressure of 2465 psia. The design pressure of the Palo Verde charging pump is 3010 psia.

In addition to tne Gdulin tests, a recent in plant charging pump test at Southern California Edison's (SCE) San Onofre Nuclear Generating Station (SONGS) showed peak cylinder pressures in the range of 2800 psia. During this test, the charging pumps were instrumented with high frequency response pressure transducers. This field data substantiates the test data obtained at Gaulin. The purpose of the test at SCE was

, , to monitor CVCS during normal charging pump operations to determine if there were dynamic pressure wave interactions when various combinations of charging pumps were used.

2-11

Several analyses have been performed for other plants with charging pump block problems. In August, 1984, a crack in a charging pump block of SONGS Unit 2 was discovered. C-E was contracted to undertake a program for examination and analysis of the block (Reference (7)]. The results of the metallurgical and structural analyses led to the hypothesis that the block failure was a cumulative fatigue failure, due to internal pump pressures in excess of 5500 psi. This analyses

,. res' alt, plus previous Gaulin pump data led to the decision to test a pump identical to those at APS to quantify these peak pressure values.

The recent testing, however, has demonstrated that high pressure spikes do not occur.

Another structural evaluation was performed for the Palisades charging pump block [ Reference (5)]. This pump was fcund to be leaking during a regular plant inspection. Upon further investigation, a small crack was found at the surface of the pump at the cylinder bore. This was the same type of failure discovered in the SCE block. C-E was requested to evaluate the effect of this crack on future pump dVailability and structural integrity. In this report, the alternative of clamping the block to inhibit crack growth was introduced and recommended as a means of extending the useful life of the charging pump block. It has been demonstrated, by analysis, that an appropriate

. clamping device will impose compressive forces at the crack tip and therefore inhibit further crack growth. The analysis also demonstrated that catastrophic failure will not occur in the unclamped block.

2-12

More recently (July,1987) an engineering evaluation was performed for Palo Verde to justify continued operation of a cracked charging pump block. The evaluation consisted of a review of previous C-E failure analyses of similar pump blocks, a detailed fracture mechanics analysis, and a seismic evaluation. The results of these analyses indicated that the implementation of a clamping device for an ANPP pump will preclude further crack arowth during normal pump operation. In

, , conjunction with a clamping device, a leakage monitoring system has been suggested as a technique for confirming proper clamping performance. The leakage monitoring system will demonstrate that the ,

crack is not growing if the leakage remains small and constant. The seismic evaluation also demonstrated that incorporating a block clamping device does not affect the original seismic qualifications of the puinp. Also, an analysis currently underway is investigating a ,

i modified design of the block with stronger material properties. '

l 1

• e 2-13 l

3.0

SUMMARY

OF RESULTS

3.1 DESCRIPTION

OF WORK PERFORMED In March of 1987, ANPP comissioned C-E to determine and evaluate the effects of gas binding on an ANPP charging pump. The initial effort included developing a formal test program to quantify the cyclic l

.- pressure loadings during gas binding events. The testing was conducted at the APV Gaulin, Inc. f acility on an existing prototype nuclear charging pump that was equivalent to the Palo Verde pumps, j l

l l

The test loop and tests ere designed so as to approximate the Pajo .

Verde pumping conditions as closely as possible. This included duplicating the hardware and piping arrangement for a number of feet on l l

both the suction and the discharge sides of the pump. A test procedure was developed for conducting the tests. It included a test natrix to obtain data for normal operating conditions and abnormal conditions, i.e., gas binding and low suction pressure (NPSH). Strain gauges and pressure transducers were mcunted in the pump, and additional instrumentation was installed in the loop, for data collection.

A total of 37 tests were performed. Data was reduced for the 9 tests

. which represent normal operation (for baseline comparison) and the most adverse combinations of off-normal operation consistent with the purpose of the test. Furthermore, an automated scan of pump bore pressures was performed for all remaining 28 tests to verify that peak pressures were less than the maximum observed peak pressure. The main 1

l 3-1 4 l

i purpose of the test program was to determine the magnitude of the peak pressure inside the pump cylinder block. The test program and results are described in detail in Reference (2).

In addition to the extensive test program, ANPP commissioned C-E to I

perform an analytical evaluation of a charging pump block with a pre-existing crack during normal operation and during a gas binding

,. event. This analysis applied the test program pressure levels and l resultant strains to a finite element model of the pump block with a typical crack, and predicted the resulting leak rate and rate of crack propagation. This evaluation is described in detail in Section 4.0 of this report and is intended to support justification for relaxing the Palo Verde licensing commitment to declare a charging pump with a j cracked block inoperable within 72 hours8.333333e-4 days <br />0.02 hours <br />1.190476e-4 weeks <br />2.7396e-5 months <br /> of discovery of a cra:k. l In parallel with the above efforts, C-E was contracted to evaluate the l effect of a clamping device on future pump availability and structural l integrity for a cricked charging pump block. This analysis was l performed to determine the potential for crack growth and to evaluate the structural integrity of a clamped block under both normal and off-design (gas binding) operating conditions. C-E concluded that a clamp and continued leakage monitoring would allow the use of the pump

., as an operable pump and should retard further crack propagation. This i

analysis is documented in Reference (6) which has been submitted to the i NRC by ANPP.

3-2 ,

3.2 RESULTS AND CONCLUSIONS The recent ANPP charging pump test program resulted in internal pump bore maximum pressure peaks of 2850 psia, only 385 psid above the operating pressure of 2465 psia. The observed peak pressure is less than the pump design pressure of 3010 psis. The strains measured durIng the test program correlate well with the stresses credicted analytically, which satisfactorily benchmarks the pump block finite 1 element analyses. Previous ASME code analyses of the charging Nimp l

block have demonstrated the structural adequacy of an uncracked block for design pressure conditions.

l The conclusion which can be drawn from the testing program is that gas l binding does not cause the excestive pressure spikes once thought possible. The RCS behaves as a large pressure reservior and will  !

mitigate pressure spikes even if the pulsation dampener is improperly charged. Therefore, gas binding does not create an operating condition for the charging pump block that is more severe than design operating conditions. l l

i The present structural evaluation of a cracked charging pump block l

demonstrates that: -

1. If a crack initiates at the intersection of the pump plunger and i

valve bores, it will propagate to become a through-wall crack with I 3-3 ,

1

consequent leakage in approximately one month (based on a conservative analysis).

2. A large through-wall crack will continue to grow if unrestrained.

It will take approximately one month from the initiation of leakage until reaching the maximum allowable pump leakage rate of

.O 4.4 gpm. That is, the pump will continue to deliver the required

,o flow of at least 39.6 gpm for approximately one month after initial indications of leakage.

Industry experience has supported the analytical conclusion that charging pump blocks do not structurally fail in a catastropic n.anner; in all known cases of continued operation with a cracked block, the pump continued to deliver the required flow. As shown in the analysis herein and in the gas binding test program, it would take on the order of a month of continued normal operation (which is no more adverse than operation under gas binding conditions) following initial leakage for an unrestrained crack to grow sufficiently large to require declaration of the pump as inoperable due to failure to deliver the required flow.

During this relatively long period of time of continued operation, periodic leakage ironitoring (as required by Technical Specifications) and visual inspection will provide adequate early warning of crack growth and degrading pump conditions.

The charging pump gas binding and cracked block evaluation discussed herein provides technical justification for relaxation of the APS commitment to declare a charging pump inoperable within 72 hours8.333333e-4 days <br />0.02 hours <br />1.190476e-4 weeks <br />2.7396e-5 months <br /> of the 3-4

l l

discovery of a cracked block. The analysis shows that, although an I

unrestrained crack will continue to grow, it will do so over a sufficiently long period of time (approximately one month) while the. l pump continues to deliver the required flow, that an arbitrary time l commitment to declare tne pump inoperable is not warranted. Monitoring and trending of actual pump leak rates, and comparison to predicted )

leak rates, will provide sufficient time to plan for corrective actions '

1 .. in an orderly manner prior to a prudent declaration of pump inoperability.

I I

i e

h 1

I 4

l 3-5 J .

s

- . . -_-- ,-.,m.- -- ,. ,- _ . . _ , - , -- - ,,x - , .: ,-r ,-r,-.- - -.m,-- - . ,-, , -, -

- < - *,m. --

4.0 EVALUATION The purpose of the structural evaluation described below is to determine the effects of both crack initiation and crack propagation on an ANPP charging pump block during normal and off-normal operating conditions. An estimate of through-crack leakage for various size cracks has also been performed to demonstrate the ability of a pump

.. block with a through wall crack to deliver its required flow for a period of time that significantly exceeds the 72 hours8.333333e-4 days <br />0.02 hours <br />1.190476e-4 weeks <br />2.7396e-5 months <br /> following the discovery of a crack.

In the structural evaluation ef fort, previous charging pump b'inck analyses have been reviewed in conjunction with the test results reported in Reference (2). These analyses have also been applied to the ANPP pump design to address the rescinding of the present 72 hour8.333333e-4 days <br />0.02 hours <br />1.190476e-4 weeks <br />2.7396e-5 months <br /> limit. The structural analysis and evaluation consists of two separate crack mechanisms. In the first case, a small crack is assumed to exist on a pump internal surface. This analysis provides an estimate of the amount of time required for a small flaw to become large enough to reach the block outer surface. In the second case, a through-wall

'. crack is included in a Finite Element model of the block. This analysis provides an estimate of the growth rate of the surface crack.

. The surface crack size is then correlated to leakage rate data from previous work (Reference (5)) to predict the leakage rate for various surface crack sizes. The assumptions and methods for these evaluations are presented in the following section.

4-1

4.1 ANALYSIS METHODOLOGY 4.1.1 Crack Initiation and Crack Growth Within the Block An analysis has been performed on a charging pump block to determine the fatigue life of the block including crack initiation and the propagation of the initial crack until it reaches the outside surface

.- of the block. This stress analysis consisted of a finite element model of an uncracked block to establish the initial stress field within the block.

The stress analyses of the uncracKed Cylinder block Were used to compute the peak stresses at the bore intersection. From these stresses, the fatigue life of the uncracked block was estimated for different internal pressure loadings. Once the operating time of the pump reaches this estimated fatigue life, tne block crack is assumed to have initiated. At this point, the block would still have some residual life available before the crack propagates to the outer i surface. The time required for this initial crack to grow to a size large enough for leakage is the initial evaluation phase for the block.

Ouring this initial period, the initiating crack growth rate depends on the frequency and magnitude of the operating pressure cycles inside the

, cylinder. This part of the analysis results in estimates of the fatigue crack growth rates and the remaining life of the cylinder block for various internal pressures and initial crack sizes. This work also addresses the effect of threshold values of the stress intensity I factors of the material on the fatigue crack growth life of the block.

i l

4-2 I

The metallurgical examination of a San Onofre block indicates that the initial fatigue crack size in the corner region of the intersection of the bores is approximately semi-elliptical in shape and has an approximate radius of 0.032 inches. The fatigue crack resembles a quarter circular corner crack at a hole in a plate. This justified the use of the analytical fracture mechanics solutions available in the literature, Reference (10), for the crack propagation phase of this l

analysis.

For the San Onofre analysis, the equation used to estimate the peak stress intensity factors AK gin the corner region has the relationship:

AKg = Ao'F where "a" is the crack radius, AC is the amplitude of the alternating stress at the corner, F the boundary correction factor, and Q the shape factor for the circular crack. Values of 2.4 and 2.2 are chosen for F and Q respectively. The inherent assumption in the fatigue crack growth analysis is that the crack grows unifomly along the circular boundary and remains circular. Additionally, it is assumed that the applied normal stress A T is unifom in the corner region and equal to the peak stress in the region.

Crack growth estimations require the availability of the material crack propagation rate curve as shown in Figure 5 [ Reference (3)]. The data in Figure 5 has been developed for 316 stainless steel. The pump 4-3

block forging is 304 stainless steel. The crack propagation rate characteristics for these materials are considered to be very similar.

The curve shown in Figure 5 has been used in previous studies with a similar material and in a similar environment. The curve in the figure is well defined at the high propagation rates. In the lower ranges, the curve is extrapolated and is not supported by test data. Threshold levels of the material are especially ill-defined. These are generally

.. assumed to be in the range of 5-7 ksi(UI. The analyses performed here considered this uncertainty by performing a sensitivity analysis to the threshold AKg values over a range of r2-10 ksi E.

For the San Onofre analysis, an analytical solution of the corner crack .

problem was used to compute the stress intensity factors, AK g for various pressure levels ranging from 2250 psi to 6000 psi. For each pressure case, first the AK was g

computed for the assumed initial crack length. Using the material curve of Figure 5 for the crack growth rate da/dN, the incremental number of cycles AN and time AT was then estimated for a small crack extension Aa. The initial crack l

size, "a", was assumed to be 0.010 inches and the crack extension increment, Aa, was typically chosen as .010 inches. The crack size was then revised to a + Aa and the procedure repeated until the incremental life was less than half an hour of the pump operation. A cylinder

., pressurization frequency of 200 cycles per minute was used in the estimation procedure.

4-4

IO'I ,

i ,i , ,

, _: 1 ,

t I i i /i # 1 1 ii

+-

i l 1 i i

/! Ii l ,

II 10 3 X - 316 316a'et03$$'

Q - 316 plate (5 cast

~N~ '

/ i l l I II I .

'/ t i I 'i i l I/ i lI I II 10-. l s

l i , ,

/ I i l I !l I. . i .

. . i

! i i i l /i i i

! i i I 5 /4h l I i iI IO'8 ,

E i .[ N i i i

I i MMI I f

he .-

j 10 '

I I /d .( f i

f g I J I

l

=

1 l/

8 10~7 , , l l l O - i ei  ! .

i g I

( i i i l

i i i i

E 2

/ l

! l

,0. .

/ i i i i I  !

!I i i i  ! , i i

l If 'I I i  !  ! ' i OctoupperboubI 10-, ,/I i i i  ! I i

!b l  !

• ! I . , l l /I '

. i i  ! .

fl  !! l .  ! '

i /i l i !ill

,0. , o I t i

i l i:ri  ! iit'r 10 ICO 1000 Stress intensity Fcctor Range, AK, 41-in ,/2 FIGURE 5:

FATICVE CRACK CRO'#TH RATE 0ATA FOR STAINI.ESS STIEI.S--CONDITIONS OF HIGH R* AND SIMUI.ATED PWR CONDITIONS

• high s'.ress ratio 4-5

The results of the fatigue crack growth analyses for the ANPP block show that the fatigue crack growth life for initial leakage is in general small, of the order of tens or hundreds of hours. These are considerably smaller than the fatigue crack initiation lives of an uncracked block. UFa this data correlating the cyclic stress intensity factor /.

J the cyclic rate of crack extension, the rate of crack growth for the maximum anticipated cyclic pressure of 3010

, psia at 195 cpm is predicted in Figure 6A. In Figure 68, this same data is represented on a rate of growth plot. This curve shows that the rate of crack propagation is very fast, so that a crack, once initiated, will grow to be a large crack in approximately 30 days.

This computation represents a qualitative juogement of the rate of crack growth because the stress assumed is conservative and the pattern of the growing crack is presumed to be circular. The evidence to support this general conclusion is that a part through-wall crack has not been discovered in blocks which have been examined. If the crack extension process were slow it would be more likely that cracking from the bore intersection corners would be discovered when a through-wall

, crack is being investigated.

Although the crack growth is presumed to have a circular shape during the early stages, experience indicates that leaking will occur from a

,, small part of the crack at the piston bore end cap. Therefore the crack must be biased toward running along the piston bore.

4-6 l

\ - _ _ _ _ _ _ _ _ _ _ _ _ .

y 900 -.

2

," 800 -

.. =

8 700 -

y

. 8 4

600 - '

S h3000 - l 53 E~400 -.

S h 300 -

d aE v

200 - t

5 100 -

W D

0--

'Y 0 .01 .02 .03 .04 ,05 .06 .07 .08 .10

.09 i

! I

! INITIALCRACKSIZE(INCH)

I FIGURE 6A: FAT!GUE ANALYSIS TREND FOR BLOCK WITH S"ALL CRACR 4-7  ;

i

0.05 -

A y 0.04 -

. 2

=

g 0.03 -

Cc 4

8 E

g 0.02 - -

w 5

5 v

8 m 0.01 -

1 0.0 i i i i i i i s a 0 .01 .02 .03 .04 .05 .06 07 .0S .09 10 INITIAL CRACK SIZE (INCH)

FIGURE 68: FATIGUE ANALYSIS TRENDS FOR BLOCK WITH SMALL CRACK 4-8 i

i

4 m. >#.. .-. _4__ . .

l

4.1.2 Analysis of Throuch-Wall Crack Growth

!i i The growth of a through-wall crack is more controlled by 't he overall loads in the block than the local stress peak which caused the crack ,

]

' initiation.  !

i The growth of the through-wall crack, starting from a short length on

,. the outsida surface of the block is computed by a finite element analysis which considers the presence of the crack. The analysis in Section 4.1.1 was based on a model of an uncracked block. The growth i e

of the through-wall crack under normal operating pressure of 2465 psia starts slowly and proceeds as shown in Figure 7A. In Figure 78, this f same data is represented on a rate of growth plot. There is an l l

apparent contradiction between the rapid growth of the non through-wall  :

crack (Figure 6A) as it extends toward the outer surface and the slower  !

a

early growth of the through-wall crack (Figure 7A). l

)  ;

1

', This apparent contradiction is caused by the different assumptions

] concerning the location of the crack during the growth process. The j assumption that the small crack grows in a circular manner and is subjected to stresses equal to the intersecting bore peak stress is  !

, excessively conservative and produces a very high crack growth rate. i t

i l ., The crack tip assumed in the through-wall crack anclysis was parallel l to the plunger bore and the crack extension is based on the predicted  !

i  ;

stress field at that location. The behavior predicted by these I l

l evaluations is consistent with available field date The crack growth  !

f process internal to the pump Diock appears to be 4 rapid process since .

! I i

4-9 I l

i L _ . - . ~ - .,

l 4.0 -

1 3.0 -

2

\d O

iE 2.0 -

Il "I

6 v

1.0 -

0.0 , , ,

0 10 20 30 TIME (DAYS)

FIGURE 7A: ANALYSIS RESULTS FCR A BLOCK 'WITh A THROUGH-WALL CRACK 4-10

I I

0.25 -

k i

I l

0.20 -

n oc

.. u w

t

• L

• 0.15 -

i 5

r 3

f i ac.

i A '

p 0.10 -

i d t E

v  !

I 8 i W l

^

0.05 - ,

"5 ,

I i

f i

0.0 i l i I i O.0 1.0 2.0 3.0 4.0 1 .,

CRACK LENGTH (INCH) i I

i  !

FIGURE 7B: ANALYSIS RESULTS FOR A BLOCK WITH A THROUGH-WALL CRACK 4-11  ;

4

e i

these cracks have never be?n observed during nomal pump inspections.

The growth of through-wall cracks appears to be a more gradual process.

Normal plant inspections have detected small leaks but large leakage [

rates that could interfere with pump operation and structural integrity have never been observed. The proper interpretation of the data i presented in Figures 6A, 68, 7A and 78, then, is of a more general qualitative nature and is indicative of the process rate. These actual

,, quantitative results, however, are dependent on analytical assumptions such as crack shape and size and should not be relied upon for detailed  :

results. l 4.2 CONFIRMATION OF ANALYTICAL MODEL i

Recent testing at Gaulin, Inc. has shown that under "worst case" i scenarios of adverse NPSHA conditions or gas ingestion, the internal i i

pump bore peak pressures are within the design pressure conditions for l

the pump. The peak pressures observed during the test were about 2850  !

psia, less than the 3010 psia design pressure. This is different than i

, earlier Gaulin testing which utilized a different discharge system i

which indicated that peak internal pressures of 7000 psia were  !

possible. Since the observed 2850 psia peak pressure is below the 3010 l psia design pressure of the pump, the pump is not subjected to l

., off-normal conditions during gas ingestion, and, therefore, gas ].

ingestion can be treated as a Normal Operation load. The previous l stress analyses of a charging pump block have been utilized for fatigue life deteminations.

1 4-12 l l

~

The recent test provides the opportunity to benchmark these C-E analyses. Strain data at specific locations in the block agree well with the Finite Element analyses of an uncracked block at the same locations. For the case of normal pump operation, the stresses in the region of the bore intersection at: 1) the 12 o' clock position, are 11000 psi (analytical) and 10600 psi (test), and 2) the 6 o' clock position, are 11000 psi (analytical) and 12700 psi (test). The j analytical stresses were calculated in Referenc9 (7) and the test results are presented in detail in the test report (Reference (2)].

For benchmarking purposes, this is a good correlation between analysis and test data. The test results represent strains on the surface of the bore cylinder and the analytical results represent the nearest integration point to the strain gauge which is somewhat below the surface of the cylinder. Considering the stress concentration effects in the corner region, and the differences in the locations of the strain gauge monitoring and stress analysis integration points, the differences in the stress values stated above from the two approaches are considered to be acceptable. The strain gauge locations and corresponding stresses, both, test and analytical results, are shown in Figure 8.

O 4-13

~ ~ ' "'"~

Leads P4 to P3 #I =-+^--

".. A ~ 1--- Mg 1

(.) d' *b ' '

Pres:ure i ha.E ~-

transc. scers l

1. , O x[- '

are located l gp C' with inboard dg -j g difi",'""l y surface of, g,g- -

the cap. -

l; g x

• w, a n ,.m , .

.., .+. ,

I  :

\ ,

Q ' n,'\ '

' -"'K N, M x J _1, i.

}, s

.. . 7 a \ \ 1 t

s

,f j

___3 l.

[

,l NNN s\ n N muez q x i,

}- - - ~ ~

- e \/ NT ;i N .

a N,1 -

A l m 1. '.h y 5 - / . J__ , _ t .] / Af4PP CllP Test e,

8 Strain gages located 1/4" out l~

' ~ ' \, ]' j ,

g8 : _sg s

} board off 1/8" radius. Lead is

,j q

~~ \ ( > at intersection of vertical and a -

7\ -

A L' g  ; horizontal bores.

Strain gage sensor orientation 3

X l @-s-

\ is cirrumferential.

O' N .

k i .

\

Leads 52,54,56 y s' g g Leads S1,53,55 Q

The leads come out through the f,J/ !.jp \ front caps.

lEST/A.lALYSIS C0'1 PARIS 0f1 ,q g Posi ti on ID Test Analysis -

12 o' clock S4 10600 psi 11000 ps-1 ei g,

, , .. q 6 o' clock S3 12700 psi 11000 ps- ' - d'"w%". " *

-=l"g2lll*m 8,2p '

fl0TE: See explanation in $#'

Section 4.3 e, nw mv Gki <

flGURE 8: STRAlfi GAUGE LOCAT10!4E Q

~~

"*I',3 G yj'-] Ei Y = WWE i-E"l"7 - ~~

~~

wJ "'

385~.1.(UI

_r- _,_ , n***.* *>="

4.3 LEAKAGE RATE PREDICTIONS I

The structural analyses described in the preceeding sections were i concerned with characterizing the crack growth in a charging pump block

. from its inception at the intersection of the cylinder bores through its growth to the outer surface of the block. These analyses then characterized the growth of this through-wall crack which would result

.* in an increasing leakage area for the pumped fluid. The purpose of this section of the report is to describa the relationship between the crack growth and the predicted leakage rates. The leakage rate growth curve is then compared to the required pump delivery rates. This correlation between leakage rate and crack growth provices the technical basis for demonstrating that the present requi.rement for declaring a pump inoperable within 72 hours8.333333e-4 days <br />0.02 hours <br />1.190476e-4 weeks <br />2.7396e-5 months <br /> of leakage detection can be rescinded provided a leakage monitoring program is established and implemented whenever a block leak is detected.

Tne analytical results have been sunnarized in Figures 6A, 68, 7A and 2

1

78. These results indicate that a predictable trend for crack growth behavior exists and takes the form of a smooth growth curve. These  !

results demonstrate that the crack growth process will not be i instantaneous and proceed in a step change pattern. These analysis i

. results also demonstrate that the charging pump block will retain its structural integrity even with the occurrence of large cracks. Whereas 4-15 l

. .- - = - -

the structural integrity of the pumps has been demonstrated, the only remaining concern with pump cracking is the ability to deliver the required flow.

The recent test program has confirmed that the structural analyses model is appropriate and that the predicted stress patterns are reasonable. The magnitude of the predicted stress is in good agreement

.a with the measured strains from the test program. In addition, the test data demonstrated that during all normal and off-normal operating modes, the internal pressure peaks within the charging pump block are below the design pressure of the pump.

These finite elrent analyses have been performed for SCE, Palisades 1

and APS (Referances 5, 6 and 7). Although these charging pump blocks are slightly different in design, there is enough similarity between

.i them to establish a trend for cracked charging pump behavior. Figure 7A shows a plot of crack length versus time of a block with a through wall crack. This curve was developed in the Palisades analyses (Reference (5)]. The Palisades analysis was the first charging pump analysis where cracks were modelled using finite element methods. In this case two crack sizes were considered beyond the baseline (i.e., no crack) case, a block with a one inch through-wall crack, and another

• . with a two inch through wall crack. In the Palisades analysis, the crack opening aret.s (C0A) were calculated from the displaced shape of the cracked block model. The COAs were then converted to leakage rates

• based on previous pipe break analyses as well as experimental data developed for Leak Before Break (LBB) piping studies.

4-16

- .. . . . - _ . - . - . . - _ = _ . - . - . - . .. ._ . . _ . _

l The relationship between crack length and leakage flowrates is shown in ,

Figure 9. The leakage flow rates typically calculated in piping leak before break analyses range from 200 to 1000 gpm per square inch of surface crack area. Since the crack is growing from the inside, the leakage flow per surface area will tend to be higher than if the crack I were growing from the outside. Therefore the rate is anticipated to be j 1

in the high end of the range. The rate for the pump, however, will be

• 1ess than for a piping system because the pressure is cycling and not constant as preserved in piping evaluations. This would reduce the 2 ,

rate on the order of a factor of two. Therefore, a rate of 500 gpe/in was selected and used in Figure 9.

j The leakage rate was measured for the Palisades block crack. The crack I

was estimated to be 0.75 inches long on the surface of the block and  !

1 leaked at a rate of about .2 gpe. This data is plotted in Figure 9 in

]

)

order to evaluate the selection of the leak rate chosen for the l analysis. The proximity of the data point to the curve demonstrate that the leak rate chosen is appropriate. ,

The ANPP block is stiffer than the Palisades block so that its leakage rate curve will be to the right of the Palisades curve in Figure 9.  !

l The location of the ANPP curve is based on the crack opening area

)*. computed for an assumed 2.8 inch long surface crack and the same leakage rate of 500 gpe/in2used to determine the Palisades curve.

l f l

] 4-17 I

l i

3.0 -

2.5 .

EST! MATED APS PALI $ ACES

~

1 .5 .

b 2

- 1.0 .

l l

i

.5 PAL!$ADES ACTUAL LEAKING CRACK 0 1 2 3 4 5 Crack length (lech)

., FIGURE 9: LEAKAGE FLOW VS. CRACK LENGTH 4-18

Figure 10 is also constructed from the Palisades analysis. The equivalent Arizona pump curve will be similar but off-set to the right of the Palisades curve due to the stiffer ANPP block. The analytical I results (estimates) presented in Figures 9 and 10 are consistent with available experimental data and state-of-the-art methods for mathematically modeling crack propogation. It is acknowledged that precise quantification of leakage flow for a given through-wall crack

, length (or leakage flow from a through-wall crack which has propogated for a specified period of time) can only be determined through field measurement (leakage monitoring program). The important features of the analytical results are: l t

a) The order of magnitude of the leakage flow estimates (Provided in Figures 9 and 10) are supported by experimental  !

evidence, b) The curves presented in Figure 9 sho- that a stiffer charging i

pump block will have a smaller crack opening area and '

leakage flow when two blocks, having Nual cylinder pressures ,

and equal length through wall cracks, are compared.

c) The curve presented in Figure 10 shows that an increasing

., leakage flow from a propogating through-wall crack is a slow and smooth progression (i.e. days as opposed to hours) until '

the slope of the curve is in the range of 0.1 to 0.5 gpm per ;

l day.  !

4-19 l

}

i 4,0- '

3.0 -

n

_b 8

d 2. 0 -

W 5

5 a

1. 0 -

0.0 , , , i 0 10 20 30 TIME (CAYS)

FIGURE 10: LEAKME FLOW vs TIME 4-20

These results can now be compared with the ANPP charging pump flow requirements.

The minimum flow requirement for the charging pumps is established by Technical Specification Surveillance Requirement 4.1.2.4 (Reference to Section XI of the ASME Boiler and Pressure Vessel Code). The minimum flow requirement (Required Action Level) is 39.6 gpm. The pump rating

.. is documented as 44 gpm. A leakage rate of 4.4 gpm would be considered an upper limit for continued delivery of required flow. The Technical Specification minimum flow requirement of 39.6 gpm is conservative with respect to the safety analysis required flow. The safety analysis applicable to the charging pumps is the natural circulation cooldown analysis provided in Reference 8. This analysis assumes a charging flow rate of 32 gpm during the natural circulation cooldown. For the purpose of

• this analysis, the minimum flow was conservatively assumed to be 40 gpm which corresponds to a 4 gpm leakrate. From Figure 9, 4 gpm is estimated to correspond to a maximum allowable surface crack length greater than 3 inches.

Figure 10 is a cross-plot of Figures 7 and 9 which shows a plot of I leakage flow versus time for the Palisades analysis. From this curve,  !

, an estimate of the time remaining to minimum required flow (40 gpm),

from initial detection, can be obtained. Given a postulated 4 gpm

].

leakage rate, it would take approximately 25 to 30 days to declare the pump inoperable based on the Palisades analysis. The Arizona block, f being slightly stiffer, could remain operable for a longer time.

j l

! 4-21 f

The leakage rate curve in Figure 10 is the basis for the development of a program which addresses charging pump operability. Once a leak is observed, a leakage monitoring system is a reliable indication of crack growth and of the pumps ability to deliver its required flow. A constant leakage rate does in fact demonstrate a stable condition where pre-existing cracks are stable and are not growing. The implementation of a leakage monitoring system will justify rescinding the present

, practice of declaring a pump inoperable within 72 hours8.333333e-4 days <br />0.02 hours <br />1.190476e-4 weeks <br />2.7396e-5 months <br /> after first observing a pump leak.

4.4 INDUSTRY EXPERIENCE ANO ANALYSES 4.4.1 Previous Tests Performed by APV Gaulin. Inc.

The purpose of this section is to discuss tests conducted by A)V Gaulin, Inc. two years ago on a similar reciprocating positive displacement pump.

During these test runs, APV Gaulin Inc. observed short duration l (approximately 70 milliseconds) pressure spikes during the compression

. ' stroke in a pumping chamber. These spikes were of magnitudes of two to three times the back pressure set point of the pump liquid end. This discussion will describe those tests and compare the configurations of the test system with those of the charging pump tests conducted recently l

., for APS [ Reference (2)).

A brief description of the test system configuration that produced these pressure spikes is provided. This test was conducted on a triplex reciprocating positive displacement pump (APV Gaulin Model MC45) 4-22 i

l

similar in design to the triplex pump used in the Chemical Volume and Control System (CVCS) at APS. The suction of the pump was connected to an atmospherically vented storage tank. The suction arrangement provided the pump with sufficient supply pressure above the net positive suction head required to obtain full pumping capacity. In addition, a fitting was installed in the pump suction line to allow for the introduction of air into the supply water.

One of two different types of back pressure regulating valves were attached directly to the discharge side of the fluid end. The back pressure regulating valve maintains high pump discharge pressure (approximately 2000-3000 psia; originally designed as a homogenizing valve for the food processing industry). A mechanical (spring loaded) homogenizing / backpressure regulating valve was used in the first series of tests (See Figure 11). The back pressure maintained by this style of valve is adjusted by changing the preload on the valve spring. The valve regulates internal pump pressure by preventing flow until the internal pump pressure on the underside of the disc overcomes the spring force that tends to hold the disc shut. This type of regulating valve reacts very quickly to changes in pump internal pressure. The second type of back pressure regulating valve is similar, except the valve is actuated by a hydraulic cylinder acting against the same valve

., disc and seat, and there is no spring in the design (See Figure 12).

i, 4-23 i

i

O i

178-Velve Sedy 125-Vse.e 29--Velve Guide l

I43--V. Bad Paddos

, s 498--Vetse had Cay pens III-V. some Gemhee 24-Ves.e s p. ' =< --

+ _.  :- ._

%Ahxhme -- -. A SOS-Vetse med Wester g

W//////W l

36 - Vsese Spdes 745-V.Gunde Goebes y7 125 _ - r mue ia__V e med

_ io - m e4.hase FIGURE 11 SPRIfG LOADED IMCVPRESSURE REGULATITG VALVE SINGLE STAGE DYNANT HOMOGENIIMG VALVE MODELS M18-30-45-75 ' MANTON.GAUUN AGG. CO., NGC

) V B14 o.w= a.-asam no.-e=1 me.-mar == .

j ..n er a w unis meer me i

J i

^^

- p

'nmmmmw -

~~

FLOW ,,

.f -

i' 3 U \ /.A I /

$,I l

h*""j"" * /

cwun y\ 1 .

7

, k ,

6 Q _

g b 6 4 3 St h 20 15 16 lT 16 El FIGURE 12 TYPICAL APV GAULIN, INC.

HYDRAULIC VALVE ASSEMBLY HYDRALK_ICALLY OPERATED I%CKPRESSURE VALVE -

- USED Iri MODIFIED TEST LOOP AT TERMINATI0f4 OF DISOMRGE PIPItJG IN CIMRGING PUMP TEST AT GAULIN FOR ANPP i

j' The pressure to the hydraulic cylinder is controlled by a pressure reducing valve. The force from the piston of the hydraulic cylinder is transmitted to the valve disc holding it shut until the internal pump  !

pressure is greater than the piston force and the valve opens.

i J Several test runs were conducted with the spring-loaded pressure >

regulating valve to measure pump performance. A displacement transducer was attached to one of the pump plungers to measure its  !

relative position, and a hig5 frequency pressure transducer was mounted in the corresponding front cap of the plunger bore. Pump position and pressure transducer outputs were connected to x and y inputs on an oscilloscope to allow observation of plunger bore pressure and plunger -

position. Having observed normal pump operation, several additional

( test runs were performed while injecting air into the suction of the pump, resulting in starvation of the pug. While the initial pressure rise of the compression stroke was delayed (up to 90% of the compression stroke) when the plunger bore was filled with the air / water i mixture, the pressure reached 2300 psi (the normal back pressure set

. point) without causing any abnorsal pressure spikes as the pressure increased in the plunger bore. An additional test was performed at an air injection rate that caused the pump to become air bound. Air

~

., binding caused the pump outflow to drop significantly and the pump discharge pressure to fall to less than a few hundred pounds per square inch. It was concluded from these test runs that air injection could I

not cause hig., internal plunger bore pressures.

1 I

i 4-26 i  !

l t

f

4 A second set of pump tests was conducted with the hydraulic operating back pressure valve. Similar base line tests were run using the same data acquisition equipment. Plunger bore pressure traces appeared normal i i

for the initial runs without air injection. Subsequent test runs  ;

injecting air into the pump suction produced short duration pressure  !

spikes as the pressure initially increased in the plutiger bore during  !

the compression stroke. The maximum observed pressure spike was approximately 7000 psi.

Tht: concern that these pressure spikes may occur in the pumps used in the CVCS systems were partially the reason why the most recent ADS charging ,

pump tests were initiated. The test was configured to determine whether these pressure spikes in fact do occur, and to quantify the I values for further stress analysis. The test also was designed to evaluate the causes of these pressure spikes and to make recommendations for eliminating or reducing the effects or the causes.

Further analysis of the previous APV Gaulin, Inc. tests indicato that (

the cause of the pressure spikes was the slow response of the nycraulically operated back pressure valve which allowed the plunger to produce pressure spikes when air was in the pumping chamber, j i

i i , The conditions allowing these high pressure spikes to be generated are  !

not present in the CVCS system at Palo Verde. The pump discharges

]

j against the RCS back pressure. All attempts to induce the sinjlated Palo Verde system to make the pump generate significant pressure spikes above design pressure were unsuccessful. These attempts included operation both within design operating conditions and also off-design operating i ,

i conditions. t J t

(

4-27 i i

t

i I

4.4.2 Industry Ooeratino Exoerience The purpose of this section is to provide information on actual operating I experience of positive displacement charging pumps in C-E designed NS$$'s. This section will cover the methods used to detect cracked  ;

charging pump blocks and the operating experience of charging pumps with known block cracks.  ;

Through-wall cracks have been detected in charging pump blocks by one I of several means. Many block cracks were visually detected by plant (

operators or maintenance personnel af ter observing very small (e.g., 20 drops per minute) amounts of leakage or dripping from the charging pump. Many of these cracks were so small that they were hard to see even when using dye-penetrant techniques. Several leaking pumps have been identified during close inspection of plant equipment after the j RCS water inventory balance surveillance requirements indicated  !

leakage. Several utilities have also discovered cracks in charging  !

pump blocks after maintenance personnel mistakenly replaced end cap I gaskets in the belief that the small amount of observed leakage had been caused by a leaking gasket when it was in fact due to a crack in the block. All three of the aforementioned crack detection methods

, have identified charging puso cracks while the leakage was very small.

Of the forty cracked blocks reported by utilities with C-E NSSS's, all N of these pumps were still capable of delivering the required flow at l

full discharge pressure. Operating experience indicates that through-wall crack leakage has always been observed prior to pump i degradation that would cause the pump to be declared inoperable on a  ;

pump performance basis.

4-28

Prolonged operating experience of charging pumps with identified through-wall cracks is very limited. Most utilities elect to replace a charging pump block soon af ter a through-wall crack is detected. On average, the charging pumps seem to operate for approximately 9000 hours0.104 days <br />2.5 hours <br />0.0149 weeks <br />0.00342 months <br /> before these fatigue induced cracks in the block initiate and propagate through the wall and are observed leaking. One utility did continue to run a charging pump with an identified through-wall crack. Over the course of a week it was found that the leakage gradually increased. At this point, the utility decided to secure the pump even though it still was delivering adequate flow. A second utility installed a custom designed compressive clamp on a cracked block to allow continued use of this pump while a replacement block was being manufactured. This pump was operated during hot functional testing for several weeks. After the clamp was installed the crack leakage was significantly reduced. A temporary leakage collection system was installed to help quantify the crack leakage rate. in addition, administrative controls were established to inspect the pug and measure crack leakage at frequent intervals. The leakage rate did not increase during the time the charging pump was operated with the clamp installed indicating that the clasping device was effective in arresting the crack.

4.4.3 System Interactions

• • The recent ANPP charging pump gas binding test program used a test loop at APV Gaulin containing one charging pump with its associated suction stabilizer and pulsation dampener. System piping on the discharge side of the pump was reproduced to scale through the first system check 4-29 li

t valve. An additional seventy feet of piping was included before the backpressure valve. The back pressure valve was used to maintain the ,

loop pressure at 2250 psia which is representative of RCS pressure at ,

the ANPP plants. Since the backpressure regulating valve can only i

maintain systein pressure when adequate flow is provided % rough the i system piping, a surge volume was installed just upstream of the valve to maintain system pressure. This was necessary when air injection and low

.. NPSH runs caused the piping discharge flow to decrease. Maintaining system line pressure accurately simulates actual CVCS conditions downstream of the charging pump discharge check valve since RCS pressure would maintain this section of piping at full pressure even if the pumps were cavitating or air bound. Additionally, the surge volume ,

simulated the potential for reverse flow that could exist if the other i

charging pumps in the CVCS were operating at a higher pressure (e.g.,

if only one pump were cavitating or air bound). This setup did cause [

j the check valve in the discharge line to close when pump discharge pressure dropped off. i

}

Harmonics between multiple operating pumps could not be investigated in l this test configuration. A recent test by a utility that has a similar l CVCS layout and identical Gaulin pumps prevides information on potential  !

harmonic interaction. This utility installed high frequency response

.. pressure transducers on each charging pump. Normal pump operating  !

I information was collected for each pung run alone. After this data was l collected, potential hannonic interactions were studied. The utility ran every combination of two charging pumps to see if the discharge pressure transducers could detect any beat frequencies of random high 4-30

)

1

1 i

pressure occurrences. None were observed even though these pump .

combinations were run for significant periods of time. This utility l concluded that there were no pressure spikes in excess of 2800 psia a

occurring in their plant. j 4

I .

j.. The test loop for the most recent testing performed at Gaulin was  !

t configured in such a way as to model the CVCS at APS. The charging pump,

.. operated as if it were connected to a very large pressure reservoir i (RCS). Any fluctuation in pump perfomance during low NPSH or ,

air injection runs would cause the discharge system check valve to f 4

operate due to the surge volume pressure as it would in the plant due  ;

t I to the backpressure provided by the RCS. In addition, recent .

f instrumented tests at a plant with a similar CVCS arrangement indicate  !

l that there are no significant harmonic interactions between operating l

i cnarging pumps.

)

r i

1 k

. i r

I t

l i

4-31  !

\

l

4.5 LEAKAGE MONITORING J

A leak rate calculation has been performed by correlating the crack size r versus leakage rate for r .st charging pump data at other plants and by c

considering pipe crack size versus leak data from previous C-E testing  ;

and literature searches. These correlations have agreed with field tatt ,

data for previously evaluated charging pump block cracks. Continued

.. leakage monitoring and visual inspection is the key to continued l declaration of pump operability for both nomal and off-design operating

)

conditions. ,

The predicted leakage flow versus time curve of Figure 10 can be used ,

to provide a comparison between predicted and actual pump performance with a leaking cracked block. Trending actual crack leakage rate over time will indicate when the actual leakage rate begins to increase 4

rapidly, providing sufficient early warning of impending significant P

, pump performance degradation. Although the maximum permissible leakage

! rats is 4.4 gpa for a technical specification minima delivered flow j . requirement (required action level per ASM Section XI) of 39.3 gps, {

i ,

the key to early warning of impending degradation is the point at which 1

\

\

l, the slope of the actual leakage rate curve increases significantly (i.e., the "knee" of the curve; a slope in the range of 0.1-0.5 gpm per l

'. day would be indicative of this "knee").

4-32

i

(

l 5.0 LICENSING EVALVATJAN F

l

, This report, along with the related testing program, provide further i

i i support for the conclusion that the PVNGS charging pumps comply with the requirements of BTP RS8 5-1 for Class 2 plants. In the event that a charging pump with a pre-existing cracked block becomes gas bound, the  ;

pump can be vented and restored to operation. This analysis and testing l l .. effort has proven that gas binding will not lead to the inability to

) deliver required flow and that the pump internal pressures associated with gas binding are within the design pressures for the pump.

Therefore, the charging pumps at PVNGS can be expected to deliver their required flow following gas binding events. .

2 k

The information presented in this report supports a finding that there is no technical justification for having to declare a charging purp  ;

inoperable because of indications of a cracked block. The declaration of  !

inoperability should only be made in the context of the charging pump's  !

ability to deliver the minimuo flow rate 'pecified in the plant technical l

specifications. The analytical and empirical data presented herF i

, indicates that operating charging pumps with cracked blocks can survive a ,

l t

, gas binding event, can continue to deliver the necessary minimum flow rate for extended periods of time, and are not subject to catastrophic

,.

• failure. Satisfying the current charging pump technical specifications

] limiting conditions for operation, therefore, provides adequate assurance i

4 A

5-1 1

)

f 4 t that the charging pumps are performing their intended functions. It is recomended, therefore, that the comitment made by APS in Reference (1) f I for Palo Verde Unit 2, requiring that APS declare a charging pump f inoperable within 72 hours8.333333e-4 days <br />0.02 hours <br />1.190476e-4 weeks <br />2.7396e-5 months <br /> of an apparent through-wall crack, be deleted.

l Operation of the facility in accordance with the proposed change does not involve a significant increase in the probability or consequences of any +

,, accident previously evaluated. The operability of the charging pumps at ,

j Palo Verde is assured by the requirements of plant Technical 1 Specification 3/4.1.2. This limiting condition for operation specifies .

I that the minimum number of charging pumps capable of supplying the i

required flow rate will be available to the plant operators. The analytical and empirical data presented in this report concludes that a charging pump with a cracked block can continue to satisfy the flow l requirements for an extended period of time even following a gas binding h event, it was, furthermore, found that cracked blocks do not f ail 1

i catastrophically but tend to leak over many hours (days) of continued  ;

pump operation. As a result of these findings, the need for a  :

] declaration of inoperability of the first indication of a charging pump I a

! cracked block is unnecessary. Deleting this comitment will not affect I

i, the current requirements specified in the technical specifications and  !

will not, therefore, involve c significantly increase in the probability [

j .. or consequences of any accident previously evaluated. l I

i l

l 4

l 5-2 I I

l

}

l In addition, administrative controls on visual inspection and leakage  ;

monitoring of the charging pumps provides a meaningful method of i determining the existence of a crar.k and pump leakage, anj early warning t

of impending significant pump flow degradation.

I i

Operation of the facility in accordance with the proposed change does not -

create the possioliity of a new or different kind of accident from any

.e accident previously evaluated. The initial reason for requiring that the charging pumps be removed from service within 72 hours8.333333e-4 days <br />0.02 hours <br />1.190476e-4 weeks <br />2.7396e-5 months <br /> of indication of an apparent through-wall crack was a concern that the charging pump could  :

catastrophically fail during a postulated gas binding event. This report concludes that, during a gas binding event, the peak internal pressures ,

are less than pump design pressure. A gas binding event, therefore, can be treated as a normal opdrational load. In conclusion this change does not creat2 the possibility of a new or different kind of accident from those previously f. valuated.

Operation of the facility in accordance with the proposed change does not involve a significant reduction in the margin of safety. The intent of the Technical Specification Survalilance Requirement 4.1.2.4 (ASME Section XI test) is to ensure that the charging pump can deliver at least 39.6 gpa to the RCS. Eliminating the commitment requiring a charging 4 pump be declared inoperable within 72 hours8.333333e-4 days <br />0.02 hours <br />1.190476e-4 weeks <br />2.7396e-5 months <br /> of an apparent through-wall crack will not affect the charging pug's ability to perform its design function. Industry experience and analysis has shown that charging pumps do not fail catastrophically and that it will take approximately one 5-3

month for a through-wall crack to propagate to a si:e that prevents the pump from being capable of delivering its minimum flow requirement. The Action Statement contained in the technical specification will ensure that an adequate margin of safety exists in the event of a large crack.

Administrative controls regarding visual inspection and leakage monitoring of the charging pumps will also provide further assurance that

,e the charging pump minimum flow rate of 39.6 gpm is satisfied. In conclusion, the elimination of the comitment to declare a pump inoperable within 72 hours8.333333e-4 days <br />0.02 hours <br />1.190476e-4 weeks <br />2.7396e-5 months <br /> of an apparent through-wall crack does not cause a significant reduction in the margin of safety for this plant.

3 1

5-4 g g ____ _ _ _ _ _ _ _ _ _ - - - - _ _ - _ _

i 6.0 REFEREMCES

1. E. E. Van Brunt, Jr. (ANPP) to G. W. Knighton (HRC), Palo Verde Nuclear Generating Station Unit 2, Dor.ket No. STN-50-529, Cht*ging Pump Opevability", ANPP-34127-EEVB/TFQ, November 29, 1985.

6

2. CEN-369(V), "Reciprocating Charging rump Test Report", Combustion  ;

,, Engineering, January 1988.

l

3. Raju, Newman, "Stress Intensity Factors for Two Syneetrical l Comerrades".

4. NUREG/CR-1319. "Cold Leg Integrity Evaluation, Final Report",

Battelle, Columbus Laboratory, February 1980.  !

5. "Structural Evaluation of Palisades Charging Pump Block", Combustion  ;

Engineering, July 1986.  !

l

6. CEN-360(V), "Structural and Seismic Evaluation of Palo Verde Charging Pumps Block", Combustion Engineering, June 1987.

7. CEN-306(S), "Metallurgical and Structural Evaluation of a Charging '

8, Pump Block", Combustion Engineering, May 1985.

l l

l l

6-1 l

b t

! 8. J.G. Haynes (ANPP) to G.W. Knighton (NRC) dPVNGS Natural Circulation [

l Cooldown fest Report," ANPP-40069, dated February 9,1987.  ;

9. E. E. Van Brunt, Jr., (ANPP) to G. W. Knighton (NRC), "Palo Verde Nuclear Generating Station Unit 2, Docket No. SIN-50-529, Venting of j the Charging Pumps," ANPP-35286-EEVB/BJA/98.05, February 24, 1986. f l

> 10. H. L. Ewalds, R. J. H. Wanhill, "Fracture Mechanics," Edward Arnold, t Maatschappij Publishers, 1985.  ;

r t ,

l l

I

. I I

l l

6-2  ;

l i

. .- - - - - ~ __ -- - ._ __ .----.--w_