ML20236F153

Final Rept Trojan Nuclear Plant SR8 Failure Root Cause Evaluation Steam Condensation-Induced Water Hammer

ML20236F153
Person / Time
Site:	Trojan File:Portland General Electric icon.png
Issue date:	07/17/1987
From:	Lanssen K, Safwat H BECHTEL GROUP, INC.
To:
Shared Package
ML20236F148	List: ML20236F145 ML20236F153 ML20236F168
References
TAC-65471, NUDOCS 8708030230
Download: ML20236F153 (11)

Text

m-

. Trojan'Nuclaer Plant Document Control Desk Socket 50-344 July .27,1987 License NPF-1 Attachment A l

FINAL REPORT i

TROJAN NUCLEAR PLANT SR8 FAILURE ROOT CAUSE EVALUATION STEAM 00NDENSATION-INDUCED WATER HAMMER i

i BECHTEL WESTERN POWER CORPORATION SAN FRANCISCO, CALIFORNIA l

JULY 1987 7 7/F7 ORIGINATOR REVIEWER //tm w [ d 7/f 7/87

, t ,

(-

8708030230 G70727 I PDR ADOCK 05000344 l P PDR

d 1 INTRODUCTION Bechtel's June 11, 1987 report, Evaluation of Main Feed Line Seismic l h Restraint Failure, concluded that a moderate water hammer, either alone l or in combination with other loadings, could have caused the recent Loop B feedwater piping seismic restraint (SR8) failure at the Trojan Nuclear Plant. To further investigate water hammer phenomena, an analysis of various steam condensation-induced water hammers was performed.

A review of industry experience, as reflected in published documentation (such as References 1 through 5) and through discussions with utility and consultant personnel, leads to the conclusion that this type of water hammer M has occurred at other facilities in the past and may well have contributed to the observed restraint failure at Trojan.

Section 2 of this report describes the model used in the analysis along with the assumptions adopted in reducing the actual feedwater piping configuration to a simplified model.

M Section 3 includes descriptions of the cases run, the reasons for choosing the particular cases, and the results.

Section 4 discusses conclusions based on the results in Section 3.

2. ANALYSIS The modeling was based on the simplified general water slug motion model presented in NUREG 0291 (Ref.1). This model was one-dimensional and employed an inelastic (incompressible) fluid water slug causing the steam bubble collapse.

Figure 1 (from Ref.1, page 174) represents a simplified model of a steam void trapped by a water slug. The slug (initially with length Lo at time t = 0) is accelerated by the differential pressure of po - pi, where pn represents the steam generator pressure (or the feedwater pressure acting on the other side of the bubble), and p1 represents the vapor pressure of the condensed steam bubble (at the temperature of the surrounding feedwater). In this analysis the condensation was assumed to occur instantaneously. The resulting impact pressure rose very f ast. For comparison, actual tests at Tihange (Ref.1, page 168) can be seen to reach peak values in approximately one millisecond, and this value can be assumed as the rise time for steam condensation - induced impact pressure throughout this analysis. Although Figure i shows a dead end on the right hand side, a similar water slug and void is expected to exist on the other side of the dead end. This represents the cold auxiliary feedwater (AFW) accelerating through the bubble f rom one side while a slug of hot feedwater (FW) is pushed by the steam generator pressure from the opposite side. Since the AFW slug is much larger than the FW slug, it is likely to accelerate slower. However, for simplicity the two slugs were assumed to accelerate equally fast, and the void size in Figure 1 represented one half of the actual steam bubble size.

TPF 36/2-1

J I

Figure 2 shows the-actual configuration of the feedwater header as it enters _the steam generator and transitions into the ring header. The I first J-tube vent is indicated on top of the 16" header. As normal FW f flow is-interrupted (upon feedwater isolation), with steam generator level below the feedring elevation, the FW inventory will drain from the feedring

,l, through a gap at the~ thermal sleeve. Tests conducted at Trojan in 1975 C 'showed leakrates from the feedring to.be roughly 2 to 3 gpm. The void will .

start to form in the FW inlet header between the 90* elbow outside the ,

steam generacor and the.16" tee inside. When AFW injection starts. .the steam bubble would be trapped between the elbow and the first J-tube. The  !

maximum.. bubble' length was assumed to be from the small end of the '4x16" reducer (the ' dead end.in Fig.1) to the J-tube, or 46". However, since the { '

downstream water slug was assumed to start at the J-tube, the bubble' length was selected.to be 14" (the length of the reducer). ' An average inside pipe

' diameter of 13.564" was selected.

n ,- i; "cinitially, the volume of the steam bubble wa.; calculated by selecting the .

gap between the water surface and the top of pipe and the length. Terminal.

velocity during bubble collapse was calculated using the Reference 1 metnodologyi Both equations . (77), page 177, and (83), page 179, were used, and the results were found to be reasonably close to.each other. Therefore, f? ' '

the two situations illustrated in Figure 3 as (b) and (c) were represen-thtiva of the. conditions studied. Model (b) approximates the steam generator

, level. above the feedring-(af ter a bubble has formed), and Model .(c) corresponds to the case when the' steam generator' level is below the feedring.

The impact pressure at the end of the condensatiet fas calculated using equation (103) on page 191, Va Ph "P 2 where, p = fluid density = 45.496 lb/f t 3 for saturated water at 1100 psia.

This value was used throughout for simplicity resulting in pressures . lower than actual at colder condensation temperatures.

v= terminal velocity at end of bubble condensation.

a = speed of pressure wave propagating through water (= 4500 ft/sec. )

The factor 2 was eliminated to accmt for the two identical bubbles condensing from opposite sides (che bub' ele was assumed to be twice the size g

shown in Fig.1, as discussed above). A transmission factor, S, was applied

'- to account for the changing pipe diameter in the actual pipe configuration

-(through the 14" x 16" reducer). Thus, the hydrodynamic force of the pressure wave was calculated from equation (10), page 57.

Fh=Ph S Ap where Ap = inside crossectional area of the 14" pipe.

l I

l.

TPF36/2-2

l The temperatures, To and T1 , represent saturation values at the pressures, po and pi, respectively. Pressure will be set by the temperature of the water surrounding the steam bubble. If the condensation temperature equals the steam generator temperature (556 F at 1100 psia), To -T1 = 0. This represented one extreme where the steam has heated the adjacent feedwater prior to condensation. The other extreme was represented by the AFW reaching the steam bubb1' aout heating up on the way. This resulted in To -T1=

466'F (assuming AFW temperature = 90 F). Although the Ref.1 methodology indicated the latter temperature difference applied to the Tihange experiments, the actual temperature difference is likely to fall somewhere between the two extremes.

3. RESULTS Table 1 gives a matrix of the cases analyzed, the input variables, the calculated peak pressures, hydrodynamic forces and the corresponding loads at the seismic restraint (SR8). It should be noted that the latter is calculated on the basis of the piping stress analysis done earlier assuming a pressure pulse of 100 psi with a rise time of one millisecond (reported in the June 11, 1987 report ). Each set of runs was made for a selected bubble volume with varying pressure and temperature differentials. The volume was adjusted by varying the gap.

Figure 4 illustrates impact pressure as a function of temperature difference for several bubble sizes. The analysis of a 5 in3 bubble size was specif-ically requested by PGE based on their evaluation according to Govier-Omer correlations (see PGE's June 16, 1987 submittal to the NRC). Figure 4 shows the resulting pressures approaching 400 psi with a large temperature differential. Significantly higher pressures were achievea through rapid condensation of larger steam bubbles.

It should be noted that the cases analyted were selected on the basis of order-of-magnitude pressures thought to have caused the observed support failure as calculated by a piping stress analysis (see Bechtel's Evaluation of Main Feed Line Seismic Restraint Failure, June 11, 1987). A number of combinations of bubble volume and temperature differential would achieve a water hammer of this magnitude as seen in Table 1.

The severity of the water hammer depends to a large extent on the following:

a. The steam bubble volume.

b. The pressure differential acting on the two sides of the water slug.

That is, the difference between the steam generator /AFW pressure and the saturation pressure of the water ,djacent to the steam bubble.

c. The depressurization rate associated with the steam condensation. ,

The last factor has not been analyzed as only an instantaneous depressurization can be assumed in the methodology used. The results are conservative compared with slower depressurtzation rates.

TPF36/2-3

n" t-

4. ~ CONCLUSIONS The results-of the evaluation demonstrate that a number of steam bubble volumes could cause condensation induced water hammers of the magnitude

' needed to damage the SR8 anchorage, and a review of the plant data point to this type of water hammer event. as a likely cause for the observed failure. - Other transient loadings such as those caused by thermal stratification and normal valve closures could have also contributed to the' failure of SR8. These smaller transient stresses could have combined with a condensation induced water hammer to produce loads large enough to cause the observed damage to SR8.

The following conditions appear to have been present at Trojan'during the last operating cycle.which would have increased the potential for producing a steam condensation-induced water hammer in the feedwater system:

a. Steam generator water level dropped below the feedring elevation on several occasions (including during the August 6,1986 plant trip),

per plant data.

b. A feedwater flow interruption of approximately 30 seconds (after loss of main feedwater) is sufficient time to create a void, and a steam bubble, of well over 5 in3 at the top of the 16" inlet header.

The size of the void will depend on the back leakage through the feedwater check valves in addition to the feedring leakage at the thermal sleeve.

c. Upon injection of cold AFW, the resulting temperature difference between the steam temperature and the adjacent feedwater temperature is between 0 and 300 F.

It should be noted that a water hammer of the type analyzed, is very l

unpredictable in nature. Coincident pressure, temperature and void '

conditions must be present and turbulence must be conducive to create a slug on both sides of a trapped steam bubble. Numerous tests have been performed  !

with seemingly the right conditions, without resulting in significant i water hammer effects. (The 1975 tests at Trojan included large feedring voids, but AFW'was introduced gradually and no water hammer was noted.)  ! '

Therefore, it is difficult to establish the size of the steam bubble associated with the restraint failure observed at Trojan. However, it is judged that results of the calculations reflected in this report are f j

bounding the water hammer event that is likely to have occurred.

In summary, the conditions appear to have been present at ' Trojan to f acilitate a steam condensation-induced water hammer of sufficient magnitude l to cause failure of SR8.

TPF36/2-4

_____________-____A

i

5. REFERENCES

1. J. A. Block, et. al., An Evaluation of PWR Steam Generator Water Hammer, NUREG-0291.

2. R. L. Chapman, et. al., Compilation of Data Concerning Known and Suspected Water Hammer Events in Nuclear Power Plants, NUREG/CR-2059.

3. R. A. Uffer, et. al., Evaluation of Water Hanmer Events in Light Water Reactor Plants, NUREG/CR-2781. q

4. A. W. Serkiz, Evaluation of Water Hammer Experience in Nuclear Power Plants, NUREG-0927.

5. R. W. Braddy, et. al., Bechtel Power Corp., Water Hammer in Feedwate r Lines, October,1974.

i l

l TPF36/2-5

(

////// 1 p 7 /= Pi .

O V T / /./ / // / , './ /,

x L FIG UR E 1. A"GENER AL SLUG IN A STRAIGHT PIPE.

De g

m .,.

l l

'= * *\ '

. 19 4 i

. d., .

. s I .

• '%;5;,=.

• . . e.

.ff: .

.. .g

j. ' *h 4**.

l .

l

~).

y-... y .,

.'A -

_m. .

4

k. _

' :E.t .'.3.'2 "f

~ f?.-l.' 5?S'$bSi?*Y fp .,

~ -

.4.*(4, -

/ u..,.

=

f(,' -

~ o+ a< _ _ ,T.T. 30'at -

r; .- .;,m,- .,.8.,, s ..\

'#. \

4:-

\ . . \ ,*. x\ i

\ s. l T ( g\, q. . - d,c ,

<=e_m , I ll

~

v  ! > >  !>

M '/  :

5 '/*l ,l 5

...r.,

>/ , *

.N:.'# '

! D .*

. 4 .. . ., .'

. . . "im* *f . . rJ.g. / .

i

.j

.- W Z.,.....g.Y,,...

!4 ..

y

,,4

...f .,., s..!,. f g . s. .

/~ ^ \

[. .

7 i F

's, N x_

Y.t.:, t- g N .. - i s .

\ 3.....

E Wir2" 7

• N~.***= N'/

vmw oo.ac. a.

FIGURE 2

d

, I

. l l

M " '

. / / _v.,

,- /j /

T- yaX s*

• 4 I

l (a) SLUG FED BY A RESERVOIR. PIPE INITI ALLY EMPTY. J

\

• M l

/ = xi + x = V, .j j ed

/= r., =:

(b). SLUG FED BY A RESERVOIR. PIPE PARTLY FULL "*

INITI ALLY.

L M 'H l 1 t_.

/ L/ ud f i

= y =:

(c) NO RESERVOIR. INITI AL LENGTH L o. PIPE PARTLY FULL INITIALLY.

SEVER AL IDEAll2ED ONE-DIMENSION AL MODELS OF SLUG M OTION FIGURE 3 i

- - - - - - - .__-_-_.._-____-_-___N

p. ..

• 1 l

d () --

O n

(

.N -

er e a g

4 <> --

Q - o v

0 <

Ci n a 9 W

d-H, r,

w# f

6. i o

t' c <, --

m - o o

f

• H ,

.O.

O hy & D - N Dd C

- o.

o b

• O g

. ,D %O - o a w H s N g g Q u_

o p .e fDW

~

O \

Q \ P Q- \ [

L A '\ C

\(' L 5 ~

O C

o N \ '

-O \ +

E-

\D tm.

\D I I 1 a i i i i t i I 6 i f-M. N.

w w

Cn. C. i. c. t9 + . m. N.

w w w

• O C C C C C C O O LS ,

1 (spuusnon) o pied saa.::d

Table.1 Simplified Model Results

c. Delta Impact Hydro- Load a press. temp. velocity p~ essure dynamic 0 SR8 s P T V1 PP Force _~

e psi F ft/s psi hips hips Sat 1 (Ini ti al void volume of 5 i n^3 , with gap O.109 in) 9.66 2001 '

161.0 20 3.24 150.96 19.47 2002 239.0 30 3.94 183.93 23.72 11.77 1003 437.4 60 5.33 248.81 32.09 15.92 1004' 646.8 100 6.48 302.57 39.02 19.36 1005 952.4 200 7.87 367.16 47.05 23.50 1006 1025.7 250 8.17 381.00 49.14 24.39 1007 , 1066.3 300 8.33 '388.50 50.10 24.86 1008 1086.7 350 8.40 392.19 50.58 25.10

_ 1009 1095.6 400 .B.44 393.79 50.78 25.20 1010 i 1098.3 437 8.45 394.28 50.85 ,

25.23 Sat 2 (Initial voi d ' volume of 10 in^3 , with gap O.174 in) 20v1 1 1 51.0 20 4.58 213.53 27.54 13.67 2002 239.0 30 5.58 260.16 33.55 16.65 2003 437.4 60 7.54 351.94 45.39 22.52 2004 -646.8 100 9.17 427.98 55.19 27.39 2005 952.4 200 11.13 519.35 66.98 33.24 . 1 2006 1025.7 250 11.55 538.95 69.50 34.49 ]

2007 1066.3 300 11.78 549.53 '70.87 35.17 2008 1086.7 350 11.89 554.76 71.54 35.50 i 2009 1095.6 400 11.94 557.02 71.83 , 35.65 2010 1098.3 437 11.95 557.71 71.92 l 35.69 5et 3 (Initial void volume of 20 in^3 , with gap O.277 in)

SuO2 '

161.0 20 '

6.47 301.90 '

38.93 l 19.02 3002 239.0 30 7.88 367.83 47.44 23.54 3003 437.4 60 10.66 497.60 64.17 '31.85 3004 646.8 100 *2.97 605.10 78.03 38.73 3005 ?52.4 200 15.74 734.29 94.69 46.99 3006 1025.7 250 16.33 762.00 98.27 4E.77 3007 1066.3 300 16.65 776.96 100.20 49.73 3008 1086.7 350 16.81 784.35 101.15 50.20 3009 1095.6 400 16.85 787.55 101.56 50.40 .

3010 1098.3 437 , 16.90 788.52 101.69 , 50.47 5et 4 (Initial vo2d volume of 50 in^3 , with gap O.513 in) 4001 '

161.0 20 10.23 477.35 61.56 30.55 4002 1 239.0 30 12.46 581.60 75.00 37.22 l 4003 '

437.4 60 ,

16.86 786.77 101.46 50.35 4004 646.8 100 20.50 956.75 123.38 61.23  !

4005 952.4 200 24.88 1161.01 149.73 74.30 4006 1025.7 250 25.82 1204.83 155.38 77.11 4007 1066.3 300 26.33 1228.48 158.43 78.62 4008 1086.7 350 26.58 1240.17 159.93 79.37 4009 1095.6 400 26.69 1245.22 160.59 79.69 l 4010 1098.3 437 2e.72 1246.77 160.78 79.79 Initial sl ug length (each) =32 in Total void length =28 in Finn 1 slug length (each) =46 in