N u c l E n g T e c h n o l 4 7 ( 2 0 1 5 ) 3 3 2 e3 3 9

Available online at www.sciencedirect.com

ScienceDirect

journal homepage: http://www.journals.elsevier.com/nuclear-

engineering-and-technology/

Original Article

ESTIMATION OF LEAK RATE THROUGH CIRCUMFERENTIAL

CRACKS IN PIPES IN NUCLEAR POWER PLANTS

JAI HAK PARK a,*, YOUNG KI CHO a, SUN HYE KIM b, and JIN HO LEE b
a
Department of Safety Engineering, Chungbuk National University, Cheongju, Chungbuk 362-763, South Korea
b
Mechanical and Material Assessment Department, Korea Institute of Nuclear Safety, Daejeon 305-338, South Korea

article info abstract

Article history: The leak before break (LBB) concept is widely used in designing pipe lines in nuclear power
Received 5 June 2014 plants. According to the concept, the amount of leaking liquid from a pipe should be more
Received in revised form than the minimum detectable leak rate of a leak detection system before catastrophic
18 November 2014 failure occurs. Therefore, accurate estimation of the leak rate is important to evaluate the
Accepted 18 November 2014 validity of the LBB concept in pipe line design. In this paper, a program was developed to
Available online 22 January 2015 estimate the leak rate through circumferential cracks in pipes in nuclear power plants
using the HenryeFauske flow model and modified HenryeFauske flow model. By using the
Keywords: developed program, the leak rate was calculated for a circumferential crack in a sample
Circumferential Crack pipe, and the effect of the flow model on the leak rate was examined. Treating the crack
HenryeFauske Flow Model morphology parameters as random variables, the statistical behavior of the leak rate was
Leak Rate also examined. As a result, it was found that the crack morphology parameters have a
Pipe strong effect on the leak rate and the statistical behavior of the leak rate can be simulated
Probabilistic Analysis using normally distributed crack morphology parameters.
Copyright © 2015, Published by Elsevier Korea LLC on behalf of Korean Nuclear Society.

1. Introduction based on Henry's homogeneous nonequilibrium critical flow

model modifying the previous EPRI LEAK-01 Code [5]. In 1994,
The leak before break (LBB) concept is widely used in the first version of the SQUIRT program [6] was developed, in
designing pipe lines in nuclear power plants. According to the which the HenryeFauske model [7e9] of thermal-hydraulic
concept, the amount of leaking liquid from a pipe should be behavior was used. The HenryeFauske model allows for
more than the minimum detectable leak rate of a leak detec- nonequilibrium vapor generation rates as the fluid flows
tion system before catastrophic failure occurs [1,2]. Therefore through the crack. The model also considers the pressure
accurate estimation of leak rate is important to evaluate the losses due to friction, bends, and protrusions in the crack
validity of the LBB concept in pipe line design. flow path. The leak rate results obtained using the SQUIRT
Several programs have been developed to evaluate leak program were compared with the experimental data on two-
rates through a crack in a pipe. In 1984, the PICEP program [3,4] phase flow through long tubes, slits, and actual cracked
was developed by Electric Power Research Institute (EPRI) pipes [6]. The HenryeFauske model was also used in the

* Corresponding author.
E-mail address: [email protected] (J.H. Park).
This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://
creativecommons.org/licenses/by-nc/3.0) which permits unrestricted non-commercial use, distribution, and reproduction in any me-
dium, provided the original work is properly cited.
http://dx.doi.org/10.1016/j.net.2014.11.008
1738-5733/Copyright © 2015, Published by Elsevier Korea LLC on behalf of Korean Nuclear Society.
 N u c l E n g T e c h n o l 4 7 ( 2 0 1 5 ) 3 3 2 e3 3 9 333

PRAISE program [10,11], which was developed in order to  

So  SLc
evaluate the leak and loss-of-coolant accident (LOCA) XE ¼ : (4)
Sgc  SLc
probabilities of pipes in nuclear power plants.
Collier et al. [12] compared the calculated leak rates from Here So is the entropy at the crack entrance plane, SLc is the
the HenryeFauske model with the measured leak rates over entropy of liquid at the crack exit plane, Sgc is the entropy of
five orders of magnitude in flow rate using simulated cracks saturated vapor at the crack exit plane, and N is defined by:
and intergranular stress corrosion cracks in stainless steel
pipes. They found that the analytical model agrees relatively N ¼ 20XE for XE < 0:05
(5)
N ¼ 1:0 for XE  0:05:
well with the mean value of the measured leak rate. They
also observed significant scatter in the experimental leak The constant B in Eq. (3) is given by 0.523 [7] and L is the
rate data. They mentioned that this scatter is because of length of the flow path. DH is the hydraulic diameter defined
partial plugging of the flow area by particulates. by:
Rahman et al. [2] introduced a modified HenryeFauske flow
4  area
model. In the previous model, the surface roughness is DH ¼ ; (6)
wetted perimeter
assumed to be constant. In the new model, however, the
surface roughness is assumed to be a function of crack Here, area is the cross-sectional area of the flow path. If the
opening displacement (COD). Depending on whether COD is cross-section of the flow path is a circle with diameter D, DH
large or small, the surface roughness is assumed to be large becomes equal to D. If the cross-section of the flow path is a
or small also. The number of turns and actual length of the crack with length 2b, DH¼area/b.
flow path are also assumed to be a function of COD. The In Eq. (2) pe, pf, pk, pa, and paa are the pressure losses due to
modified HenryeFauske model was implemented in the entrance effects, friction, bends and protrusions in the flow
PRO-LOCA program (Battelle, Columbus, USA) [13], which is path, phase change acceleration, and area change acceleration
a probabilistic fracture mechanics program used to estimate respectively. Each of these terms is expressed by the following
the frequencies of LOCA. equations.
A program was made to evaluate the leak rate through The pressure loss due to entrance effects, pe, is given by:
circumferential cracks in pipes using the HenryeFauske flow
model and the modified HenryeFauske flow model. The G2o yLo
pe ¼ ; (7)
2C2D
calculated leak rate and pressure loss results from the two
flow models were compared and discussed. Considering crack where CD is the coefficient of discharge. A value of CD ¼ 0.95 is
morphology parameters, such as surface roughness and the recommended for tight cracks with CODs < 0.15 mm. For
number of turns along the flow path, as random variables, the cracks with larger CODs, a coefficient of discharge between
distribution characteristics of the leak rate were examined. 0.62 and 0.95 should be used. CD ¼ 0.95 was used in this study.
The pressure loss due to friction, pf, is given by:

2. Flow and COA models   2

L G 

pf ¼ f 1  X yL þ Xyg ; (8)
DH 2
2.1. HenryeFauske flow model
where f is the friction factor, X is the fluid quality, and a bar on
the variable means the average value in the region. The flow
Mass flux through a crack in a pipe can be calculated using the
path can be divided into two ranges of L/DH > 12 and 0 < L/
HenryeFauske flow model given as the following Eqs. (1) and
DH < 12. The range L/DH > 12 corresponds to the two-phase
(2) [6-9]:
flow region with liquid and gas and the range 0 < L/DH < 12

 1 corresponds to the one-phase flow region with only liquid.
j Gc ; pc ¼ G2c    ¼0 (1)
Xc ygc

 ygc  yLc N dX E Thus Eq. (8) can be expressed as follows:
go pc dp
c
  2

 L Gc 
 G2
U Gc ; pc ¼ pc þ pe þ pa þ pf þ pk þ paa  po ¼ 0: (2) pf ¼ f  12 1  X yL þ Xyg þ 12f o yLo : (9)
DH 2 2
Here the subscripts o and c mean the values at the crack
Considering the relationAoGo¼AcGc, Eq. (9) becomes:
entrance plane and at the crack exit plane respectively. G is
  2  2
mass flux, p is pressure, vgc and vLc are specific volumes of L Gc 
 Ac
pf ¼ f  12 1  X yL þ Xyg þ 6f G2c yLo : (10)
saturated vapor and saturated liquid at exit pressure, and go is DH 2 Ao
the isentropic expansion coefficient. In Eqs. (1) and (2), mass
Based on the PRAISE program [10] the friction factor f is
flux at crack exit plane, Gc, and pressure at crack exit plane, pc,
given by:
are unknowns. After solving the equations, the leak rate
through a crack can be obtained by multiplying Gc by the crack    2
DH
opening area at crack exit plane, Ac. f ¼ C1 log þ C2 ; (11)
2m
In Eq. (1), Xc is the nonequilibrium vapor generation rate
given by: where m is the surface roughness and has a value of 6.20 mm
(0.0002441 inch) in SCC growth and 40.0 mm (0.0015748 inch) in
Xc ¼ NXE f1  exp½  BðL=DH  12Þg; (3) fatigue crack growth. The coefficients C1 and C2 are given by
where: [10]:
334 N u c l E n g T e c h n o l 4 7 ( 2 0 1 5 ) 3 3 2 e3 3 9

DH  
C1 ¼ 3:39; C2 ¼ 0:86 for  27:74 mG  mL d d
2m m ¼ mL þ  0:1 for 0:1 <  10: (18)
(12) 9:9 mG mG
DH
C1 ¼ 2:0; C2 ¼ 1:74 for > 27:74
2m d
m ¼ mG for 10 < ;
The pressure loss due to bends and protrusions in the crack mG
path, pk, is given by: where mL and mG are the local and global surface roughness,
2
respectively, and d is COD at the crack center.
G 
 The number of 90 turns in the flow path is also assumed to
pk ¼ ðev Þ 1  X yL þ Xyg (13)
2 be a function of das follows [2]:
where ev is the total loss coefficient over the crack flow path
d
length. The variable ev can be determined experimentally by nt ¼ ntL for 0 <  0:1
mG
defining:
 
ev ¼ e½L; (14) ntL d d
nt ¼ ntL   0:1 for 0:1 <  10 (19)
11 mG mG
where e is the number of velocity heads lost per unit flow path
length for a given type of crack. The experimental data from a
d
fatigue crack in a girth weld suggests a value of e ¼ 6 velocity nt ¼ 0:1ntL for 10 < ;
mG
heads per mm of crack flow path. For SCC crack growth, a
value of e ¼ 3 velocity heads per mm of flow path is appro- where ntL is the local number of turns in the flow path.
priate. G is the mean value of mass flux given by: Because one 90 turn corresponds to one velocity head loss, nt
has the same meaning as e in Eq. (14).
Ao Go þ Ac Gc As the flow path is not perpendicular to the pipe surface
G¼ : (15)
Ao þ Ac and not straight, the real flow path length is longer than the
The pressure loss due to phase change acceleration, pa, is wall thickness. The real path length, La, can be obtained by
given by: multiplying the wall thickness, t, by a correction factor K as
follows:
2
pa ¼ GT ð1  Xc ÞyLc þ Xc ygc  yLc (16)
La ¼ Kt: (20)
where GT is the mean value of mass flux in the two-phase
The correction factor K is also given as a function of d as
region of the flow path.
follows:
The pressure loss due to area change acceleration, paa, is
given as follows [6]: d
K ¼ KGL for 0 <  0:1
"   2 # mG
G2 
2
G2 yLo Ac Ac
paa ¼ c  þ c ð1  XÞyLc þ Xygc  
2 Ai Ao 2 KGL  KG d d
"  2 # (17) K ¼ KGL   0:1 for 0:1 <  10 (21)
Ac 9:9 mG mG
 1 ;
Ai
d
where Ai is the cross-sectional area at the plane where the K ¼ KG for 10 < :
mG
two-phase flow starts, i.e., where L/DH¼12.
In the program it is assumed that the cross-sectional area
is constant along the flow path. Then Go¼Gc and paa becomes 0. If the pipe wall thickness to hydraulic-diameter ratio, L/DH,
Considering Eqs. (7), (10), (13), and (16), it can be noticed that is larger than 30, the HenryeFauske two-phase flow model can
the pressure losses, pe, pf, pk, and pa, can be expressed as a be used. Here the surface roughness, number of turns, and
function of Gc. Then from Eq. (2), pc can be expressed as a actual length of the flow path are assumed to be functions of
function of Gc. Substituting the relation into Eq. (1), we can COD. When (L/DH)<30, the model needs to be modified. The
get an equation with only one unknown variable, Gc. PRO-LOCA program uses the following modified Hen-
ryeFauske flow model [13]. If (L/DH)  30, the HenryeFauske
two-phase model is used to calculate the mass flux Gc,
2.2. Modified HenryeFauske flow model calculate pc/po and the mass flux for L/DH¼30, and let the
values be (pc/po)1 and (Gc)1 respectively. When 4.6(L/DH)<30,
Rahman et al. [2] introduced a model modifying the pc/po is assumed to be (pc/po)1 in the region (L/DH) < 4.6, pc/po
HenryeFauske flow model. In their model the surface is assumed to increase linearly with L/DH from 0 to (pc/
roughness is assumed to be a function of COD at the crack po)1, when 12  (L/DH)<30, the mass flux is assumed to be
center. The number of turns and actual length of the flow (Gc)1. When (L/DH)  4.6, the leak rate is obtained using an
path are also assumed to be a function of COD. orifice-type flow equation where the fluid properties are
The surface roughness is assumed to be a function of COD evaluated at the average pressure, (poþpc)/2. In the region
as follows [2]: 4.6  (L/DH) < 12, the mass flux is assumed to increase
d linearly with L/DH from the mass flux at L/DH ¼ 4.6 to the
m ¼ mL for 0 <  0:1
mG mass flux at L/DH ¼ 12.
 N u c l E n g T e c h n o l 4 7 ( 2 0 1 5 ) 3 3 2 e3 3 9 335

2.3. COA and COD equations RambergeOsgood stressestrain relationship is given by the
following equation:
To obtain the leak rate through cracks, a solution for the crack  n
opening area (COA) is necessary. In the PRAISE program the e s s
¼ þa ; (25)
e0 s0 s0
following COA solution was used [10]:
where e0¼s0/E and s0 is the reference stress, which usually has
Zb
4 the same value as yield strength. Other COD solutions when
A¼ JðxÞdx; (22)
s bending moment is applied with or without axial load can be
0
found in [14].
where J(x) is the applied J integral expressed as a function of If the cross-sectional shape of the flow path is assumed to
the half crack length x, s is applied stress, and b is the half be elliptical, COA can be calculated from COD using the
crack length at which COA is obtained. following equation:
COA can be obtained using the elastic plastic crack opening
p
displacement (COD) solutions. For a circumferential through- A¼ db; (26)
2
wall crack under axial tension load, COD is expressed by the
following Eq. (23) [14]: where b is the half crack length.

   n
P P
d ¼ f2 þ ae0 pR$H2 $ ; (23)
tE P0 3. Program development
where:
A program was developed to estimate the leak rate through a
  " (  1:5  3 )#
qe qe qe circumferential crack in a pipe using the HenryeFauske flow
f2 ¼ 2 $ 1 þ A 4:55 þ 47:0
p p p model and modified HenryeFauske flow model. In order to
solve Eqs. (1) and (2), the thermodynamic properties of water
2      2 3 should be known. For this purpose, the program developed by
F2t
6 $ n1
nþ1
$ ss0t 7 Riemer et al. [15] was used.
6 b
7
qe ¼ q61 þ  2 7 In order to obtain the probabilistic distribution characteris-
4 5
1 þ P0 P
tics of the leak rate, the Monte Carlo simulation method was
used. In this method, the crack morphology parameters given in
"  1:5  4:24 # Table 1 were treated as normally distributed random variables
q q
Ft ¼ 1 þ A 5:3303 þ 18:773 (24) and new values of parameters were generated in each leak
p p
rate calculation. The cumulative distribution function was
   0:25 obtained using the simulated leak rate values for the given
R R crack length. In the program, the normally distributed
A ¼ 0:125  0:25 for 5   10
t t parameters were generated using the algorithm proposed by
Box and Muller [16]. The program was written in Cþþ.
   0:25
R R After developing the program, the obtained leak rate was
A ¼ 0:4  3:0 for 10   20
t t compared with the leak rate from the PRAISE program in order
to check its accuracy. In this case, the HenryeFauske flow
P model and the COA solution of Eq. (22) were used. It was found
st ¼
2pRt that the difference between the two leak rate results was < 1%.

P0 ¼ 2s0 Rt½p  q  2 arcsinð0:5 sin qÞ:

Here P is the applied axial load. H2 is a constant depending 4. Numerical results

on q/p, n, and R/t. R, t, and q are the pipe mean radius, wall
thickness, and crack half-angle, respectively. a, s0, e0, and n 4.1. Comparison of COA solutions
are constants in the RambergeOsgood stressestrain rela-
tionship. b ¼ 2 for plane stress and b ¼ 6 for plane strain crack In the analysis, the pipe material is assumed to be ASME
tip condition. The value of b ¼ 2 was used in this analysis. The SA351 CF8M. The material properties and pipe geometries

Table 1 e Mean and standard deviation of crack morphology parameters [13].

Crack morphology variable Corrosion fatigue IGSCC PWSCCebase PWSCCeweld
Mean SD Mean SD Mean SD Mean SD
mL (mm) 8.814 2.972 4.70 3.937 10.62 9.870 16.86 13.57
mG (mm) 40.51 17.65 80.0 39.01 92.67 65.26 113.9 90.97
nL (mm1) 6.730 8.070 28.2 18.90 8.043 2.043 5.940 4.540
KG 1.017 0.0163 1.07 0.100 1.060 0.095 1.009 0.011
KGL 1.060 0.0300 1.33 0.170 1.327 0.249 1.243 0.079
336 N u c l E n g T e c h n o l 4 7 ( 2 0 1 5 ) 3 3 2 e3 3 9

Table 2 e Geometry and material data used in the Modified Henry–Fauske model
100
analysis. 6
Henry–Fauske model
with PRAISE COA equation
Geometries and material properties Input data
80
Yield strength (MPa) 201.1

Leak rate (kg/s)

Ultimate strength (MPa) 529.2
4
Elastic modulus (GPa) 215 60

(gpm)
a 12.22
n 4.84 40
Outer diameter of pipe (mm) 878.8
2
Thickness of pipe (mm) 71.1
20

used are given in Table 2. In Table 2, a and n are parameters in 0 0

the RambergeOsgood relationship. Normal operating 0 100 200 300

pressure is assumed to be 15.51 MPa. The applied axial Half crack length (mm)
stress from deadweight is 14.34 MPa and the stress from
Fig. 2 e Comparison of leak rates obtained from
deadweight and restraint of thermal expansion is 59.2 MPa.
HenryeFauske flow model and modified HenryeFauske
Fig. 1 shows two COA solutions for a through-wall
flow model.
circumferential obtained using Eqs. (22) and (26). When the
half crack length, b, is small, the two equations give similar
COA values. As b increases, however, the COA from Eq. (26)
becomes larger than that from Eq. (22).
Fig. 3 shows variation of normalized pressure loss terms as
the half crack length increases for the HenryeFauske model.
4.2. Comparison of two flow models The pressure loss terms are normalized with the crack
entrance pressure, po. In the Fig. 3, pt is the total pressure
Mass flux cGc was calculated for a through-wall circumferential loss. The pressure loss due to entrance effects, pe, was
crack with crack length 2b using the HenryeFauske model excluded because the term was too small in comparison to
described in earlier. Multiplying the obtained mass flux by COA, the other terms. When the half crack length is small, pf is
the leak rate was obtained as a function of half crack length. Eq. the dominant pressure loss. As the half crack length
(22) was used for COA calculation in order to get a similar leak increases, however, pf decreases rapidly to a small value.
rate as the PRAISE program. The leak rate was also obtained However, pk increases rapidly and becomes the dominant
using the modified HenryeFauske model described earlier pressure loss term.
and the handbook COA solution of Eq. (26) was used as an Fig. 4 shows variation of normalized pressure loss terms for
improved COA solution. The two leak rate results were the modified HenryeFauske model. The model shows a
plotted in Fig. 2 as a function of the half crack length. It can similar variation trend in each pressure loss term compared
be noted that both results show a similar leak rate when the with the previous result of Fig. 3. When the half crack length
half crack length, b, is small. However, the leak rate from the is short, pf is the dominant pressure loss. However, pk
modified HenryeFauske model becomes much larger than becomes the dominant pressure loss term as the half crack
that from the original HenryeFauske model as b increases. length increases. It can be noted that the total pressure loss

1,500
HB equation 1.0
PRAISE equation

0.8
1,000
Normalized pressure loss
COA (mm )
2

0.6 pf/po
pk/po
pa/po
500 0.4
pt/po

0.2

0
0 100 200 300
0.0
Half crack length (mm) 0 40 80 120 160
Half crack length (mm)
Fig. 1 e Comparison of COA solutions obtained from
PRAISE equation and handbook solution. COA, crack Fig. 3 e Variation of normalized pressure loss terms in
opening area. HenryeFauske flow model.
 N u c l E n g T e c h n o l 4 7 ( 2 0 1 5 ) 3 3 2 e3 3 9 337

1.0

0.8
Normalized pressure loss

0.6

pf/po
0.4 pk/po
pa/po
pt/po
0.2

0.0
0 50 100 150 200
Half crack length (mm)
Fig. 5 e Probability density function of leak rate for crack
Fig. 4 e Variation of normalized pressure loss terms in morphology variables of corrosion fatigue when the half
modified HenryeFauske flow model. crack length is 50.8 mm.

In order to demonstrate the scattering behavior, the leak

from this model is less than that from the HenryeFauske rate values obtained from the analysis are plotted in Fig. 8.
model. Open symbols represent the leak rates obtained when the
crack morphology parameters were treated as random
4.3. Distribution of leak rate variables. The crack morphology data for SCC fatigue in
Table 1 were used. In Fig. 8 and 10 leak rate results are
As Collier et al. [12] indicated, the leak rate model agrees plotted for each half crack length and the error bars are also
relatively well with the measured flow rate if the mean plotted to represent the mean and the standard deviation of
value is considered. However, the measured flow rate shows the 10 obtained leak rate values. The solid curve represents
significant scatter. In order to simulate the scattering the leak rate when constant mean values were used for
characteristics of the measured flow rate data, the crack crack morphology parameters. It can be seen that the
morphology parameters should be treated as random simulated leak rate exhibits significant scatter.
variables. The leak rates were also obtained when the crack
In order to examine the statistical distribution character- morphology data for primary water stress corrosion cracking
istics of leak rate, 1,000 leak rate values were generated for the (PWSCC) weld in Table 1 were used. The results are plotted in
given crack geometries and material properties using a Monte Fig. 9. It was found that this simulated leak rate also exhibits
Carlo simulation. In the simulation, the crack morphology significant scatter. The statistical properties are given in
parameters were assumed to be normally distributed random Table 4.
variables. The mean and standard deviation values for crack If crack morphology parameters are considered as random
morphology parameters are given in Table 1. The modified variables we can demonstrate significant scatter in the leak
HenryeFauske model and the handbook COA solution of Eq. rate. If we estimate the leak rate with constant crack
(26) were used in the calculation. Geometric, material, and morphology parameters, a large discrepancy between the
loading data were the same as the previous analysis. estimated value and the observed value will be expected.
The first crack morphology parameters for corrosion When leak rate estimation is necessary for LBB or LOCA
fatigue in Table 1 were used. Fig. 5 shows the cumulative
density function of the leak rate when the half crack length
is 50.8 mm. The 1st, 5th, 95th, and 99th percentiles and the Table 3 e Characteristics of distribution of leak rate for
mean value are also indicated in Fig. 5. The line crack morphology parameters of corrosion fatigue.
corresponding to the constant parameter indicates the leak Properties Leak rate for corrosion fatigue (kg/s)
rate obtained when constant crack morphology parameters b ¼ 50.8 mm b ¼ 101.6 mm b ¼ 152.4 mm
were used in the calculation. Several important statistical
Mean 0.0753 0.509 1.939
properties are given in Table 3. If we compare the median
Standard 0.0320 0.233 0.752
value of 0.0654 kg/s with the 5th percentile of 0.0424 kg/s deviation
and the 95th percentile of 0.1389 kg/s, it can be noted that 5th percentile 0.0424 0.238 0.971
the leak rate is widely distributed. The probability density 10th percentile 0.0451 0.263 1.169
functions and statistical properties when the half crack Median 0.0654 0.456 1.787
lengths are 101.6 mm and 152.4 mm are also given in Figs. 6 90th percentile 0.1241 0.813 2.85
95th percentile 0.1389 0.966 3.55
and 7 and Table 3.
338 N u c l E n g T e c h n o l 4 7 ( 2 0 1 5 ) 3 3 2 e3 3 9

Fig. 9 e Leak rates for normally distributed crack

Fig. 6 e Probability density function of leak rate for crack
morphology variables of PWSCC weld.
morphology variables of corrosion fatigue when the half
crack length is 101.6 mm.

Table 4 e Characteristics of distribution of leak rate for

crack morphology parameters of PWSCC weld.
Properties Leak rate for PWSCC weld (kg/s)
b ¼ 50.8 mm b ¼ 101.6 mm b ¼ 152.4 mm
Mean 0.0675 0.385 1.242
Standard dev. 0.0270 0.1861 0.632
5th percentile 0.0431 0.217 0.638
10th percentile 0.0455 0.234 0.694
Median 0.0592 0.318 1.021
90th percentile 0.0979 0.660 2.12
95th percentile 0.1322 0.799 2.50

analysis, the leak rate should be obtained as a probability

distribution function. In order to improve the accuracy of leak
rate analysis, accurate crack morphology parameters should
be known.
Fig. 7 e Probability density function of leak rate for crack
morphology variables of corrosion fatigue when the half
crack length is 152.4 mm.
5. Conclusion

(1) A program was developed in order to obtain the leak rate

through a circumferential crack using the HenryeFauske
model and modified HenryeFauske model; (2) the modified
HenryeFauske model using an improved COA solution gives a
similar leak rate to the HenryeFauske model when the crack
length is short, but the model gives a larger leak rate than the
HenryeFauske model as the crack length increases; (3) sig-
nificant scatter was demonstrated in the estimated leak rates
when the crack morphology parameters are treated as nor-
mally distributed random variables; and (4) by using the crack
morphology parameters for SCC fatigue and PWSCC weld,
cumulative density functions were obtained for leak rate.

Conflicts of interest
Fig. 8 e Leak rates for normally distributed crack
morphology variables of corrosion fatigue. All authors have no conflicts of interest to declare.
 N u c l E n g T e c h n o l 4 7 ( 2 0 1 5 ) 3 3 2 e3 3 9 339

Acknowledgments [8] R.E. Henry, H.K. Fauske, Two-phase critical flow at low
qualities, part 1: experimental, Nucl. Sci. Eng. 41 (1970)
79e91.
This work was supported by the Korea Institute of Nuclear
[9] R.E. Henry, H.K. Fauske, Two-phase critical flow at low
Safety, Daejeon, Korea.
qualities, part II: analysis, Nucl. Sci. Eng. 41 (1970) 92e98.
[10] D.O. Harris, D.D. Dedhia, S.C. Lu, Theoretical and User's
Manual for Pc-PRAISE, Failure Analysis Associates, Inc.,
references Lawrence Livermore National Laboratory, 1992. USNRC
Report NUREG/CR-5864.
[11] D.O. Harris, D. Dedhia, WinPRAISE 07; Expanded PRAISE Code
[1] P.M. Scott, R.J. Olson, G.M. Wilkowski, Development of in Windows, Structural Integrity Associates, Inc, San Jose,
Technical Basis for Leak-before-break Evaluation Procedures, USA, 2007.
USNRC, 2002. NUREG/CR-6765. [12] R.P. Collier, F.B. Stulen, M.E. Mayfield, D.B. Pape, P.M. Scott,
[2] S. Rahman, N. Ghadiali, D. Paul, G. Wilkowski, Probabilistic Two-phase Flow through Intergranular Stress Corrosion
Pipe Fracture Evaluations for Leak-rate-detection Crack and Resulting Acoustic Emission, Battelle, Columbus,
Applications, USNRC, 1995. NUREG/CR-6004. USA, 1984. EPRI Report, NP-3540-LD.
[3] D.M. Norris, A. Okamoto, B. Chexal, T. Griesbach, PICEP: Pipe [13] PRO-LOCA-GUI/PRO-LOCA User's guide (Version 3.5.32),
Crack Evaluation Program, 1984. EPRI Report EPRI NP-3596-SR. Battelle, Columbus, USA, 2009.
[4] H.S. Mehta, N.T. Patel, S. Ranganath, Application of the Leak- [14] A. Zahoor, Ductile Fracture Handbook, vol. 1, 1989. EPRI
before-break Approach to BWR Piping, 1986. EPRI Report Report NP-6301-D.
NP-4991. [15] D.H. Riemer, H.R. Jacobs, R.F. Boehm, A Computer Program
[5] D. Abdollahian, B. Chexal, Calculation of Leak Rates through for Determining the Thermodynamic Properties of Water,
Cracks in Pipes and Tubes, 1983. EPRI Report NP-3395. University of Utah, Salt Lake City, USA, 1976. Report DGE/
[6] D.D. Paul, J. Ahmad, P.M. Scott, L.E. Flanigan, G.M. Wilkowski, 1549e2.
Evaluation and Refinement of Leak-rate Estimation Models, [16] G.E.P. Box, M.E. Muller, A note on the generation of random
1994. NUREG/CR-5128, Rev. 1. normal deviates, Ann. Math. Stat. 29 (1958) 610e611.
[7] R.E. Henry, The two-phase critical discharge of initially
saturated or subcooled liquid, Nucl. Sci. Eng. 41 (1970)
336e342.