Professional Documents
Culture Documents
The primary objective of this study is to develop a simple methodology which, when coupled with existing observations on pressure vessel behavior, provides an interrelation between nuclear pressure vessel weld integrity and the
parameters of the in-service inspection program, including inspection sample size, frequency and efficiency. The
basic input information on rate of generation and development of weld flaws of different sizes and types is drawn
primarily from published British and German studies taken almost exclusively from welds of non-nuclear pressure
vessels. The input information is varied to reflect differences in weld quality and uncertainty of input data. A modified Markov process is employed and a computer code written to obtain numerical results. If it is assumed that the
quality of nuclear reactor welds are the same as the quality of non-nuclear welds (i.e. the data base), then, based on
the limitations of the model, the predicted critically sized defect concentration is about 50 x 10 -7 per weld at the
end of weld life for welds under both high and low stress if ASME, Section XI, In-Service Inspection Requirements
are applied. Based on the British data and the less stringent inspection standards (compared to Section XI) the estimated number of critically sized defects per weld at the end of weld life is 250 10 -7 and 170 x 10 -7 per weld for
high and low stressed welds, respectively. If it is assumed that the nuclear reactor pressure vessel welds have superior
quality to the non-nuclear welds, then the model predicts an appropriately lower probability of critical defects at the
end of weld life. A variety oi" other sensitivity studies are included in the report. Also, a simple methodology to provide an optimal weld inspection program which is consistent with a minimum cost criteria is outlined. It should be
noted that the results of this study are based o n t h e limitations of the simple model that was used and on a variety of
corresponding assumptions.
Contents
1. I n t r o d u c t i o n
2. Physical d e s c r i p t i o n
3. D e f e c t g e n e r a t i o n , d e v e l o p m e n t a n d r e m o v a l m o d e l
4. D e t e r m i n a t i o n o f i n p u t p a r a m e t e r s
5. Results
Appendix
References
87
88
98
107
117
137
153
1. Introduction
1.1. Objective
T h e p r i m a r y objective o f this s t u d y is to d e v e l o p a simple m e t h o d o l o g y w h i c h , w h e n c o u p l e d w i t h existing observations o n pressure vessel b e h a v i o r , p r o v i d e s a n i n t e r r e l a t i o n b e t w e e n pressure vessel i n t e g r i t y a n d t h e p a r a m e t e r s o f
t h e in-service i n s p e c t i o n p r o g r a m , i n c l u d i n g i n s p e c t i o n sample size, f r e q u e n c y a n d efficiency. A m o d i f i e d M a r k o v
process is e m p l o y e d a n d a c o m p u t e r code w r i t t e n t o o b t a i n n u m e r i c a l results.
* Prepared for the National Science Foundation under Grant GI-39416: 'A General Evaluation Approach to Risk-Benefit for
Large Technological Systems and its Application to Nuclear Power'.
1"Present address: NUS Corporation, Sherman Oaks, California 91403, USA.
88
The basic input information on the rate of development of flaws of different size is drawn primarily from published British and German studies; the input is varied to reflect changes in vessel quality. A simple methodology
to provide an optimal weld inspection program which is consistent with a minimum cost criterion can be
formulated.
1.2. Basis for approach
All modern pressure vessels, whether they are used in nuclear reactor applications or other applications, are composed of welded sections. Experience indicates that under most circumstances when pressure vessel integrity is
lost, it is due to loss of weld integrity rather than loss of integrity of the base metal sections. As a result, this study
will emphasize the potential loss of pressure vessel integrity due to failure of one of its welds.
A pressure vessel weld fails (or is on the verge of failure) when it contains a defect of critical size for the existing conditions, i.e. the defect can propagate rapidly [ 1 - 4 ] . All welds, no matter how perfectly they are made,
initially contain defects [1-7]. Many are small and undetectable, others are large enough to be detected but too
small to require repair, and still others may be large enough to require repair. Of those defects that are large enough
to require repair, most may be detected and repaired; however, some may not be detected and as a result, not repaired. It has been observed that defects can increase in size over time. The exact development mechanism is not
considered in this study which relies on empi.rical observations of defect development. The initial defect size,
defect generation rate and defect development rate are approximated from existing information on weld defects.
The pressure vessels used in nuclear reactors are similar in design and characteristics to some of the pressure
vessels used in other applications [ 1 - 1 0 ] . Although considerable information exists regarding the integrity of nonnuclear pressure vessels, rather less experience is available for nuclear pressure vessels. As a result, to obtain a basis
for approximating nuclear reactor pressure vessel integrity, more expansive data on non-nuclear reactor weld failure has been studied. An attempt has been made to examine data for welds of generally similar type and under
similar stress to the nuclear reactor pressure vessel welds.
1.3. Scope o f the study
This study begins by describing various types of pressure vessels, pressure vessel welds, weld defects and inspection
techniques. A model is formulated to approximate mathematically the process of defect development by assuming
that defects are of discrete sizes and develop at discrete rates.
The model consists of a modified Marker process, used to simulate the defect generation and development proCess, and a modified renewal theory technique, used to simulate defect repair. The model input parameters include
initial defect size; defect generation rate and development rate as a function of defect size; and defect inspection
and repair efficiency as a function of defect size. Input parameters are then selected such that, together with the
model, a fit is obtained to available empirical information. The fit is not unique.
2. Physical description
In this section a physical description of a nuclear reactor pressure vessel and its welds is given [ 1-11 ]. Generation
and growth of defects in welds and eventual weld failure are discussed [ 12-18]. Also, a description of the inspection procedure and inspection technique is given [ 19-29].
2.1. Pressure vessels and vessel welds
There are various types of pressure vessels in existence which include: (a) boiler steam drums [I, 2] ; (b) liquid propane or other gaseous storage tanks [2, 3] ; (c) fired super heaters [3] ; (d) package boilers [3] ; (e) steam generators
[2-6]; and (f) nuclear reactor pressure vessels [4-6]. These pressure vessels are designed to contain pressure
between 100 and 32 500 psi [1-6].
89
Many of the earliest nuclear power plants, including the UK gas-cooled systems and the first US gas cooled reactor
plant, used ferritic steel pressure containment vessels [ 6 - 1 1 ] . The later UK gas-cooled reactors and the French
gas-cooled reactors used concrete vessels [ 6 - 1 1 ] . However, the fight water reactors are more interesting in view of
their larger numbers.
The vessel which serves the important function o f containing the highly pressurized reactor coolant water as it
flows over the nuclear fuel core is usually made up of four major welded segments [7]. These segments include the
top closure head, the nozzle course, the cylinder course and the lowor head [7]. Fig. 1 [1] shows the pieces of a
pressurized water reactor (PWR) vessel before its final assembly. A typical 1000 MW(e) PWR vessel has a 14 ft i.d.
and is 42 ft high with a cylinder wall thickness of 8~ in. [9]. Boiling water reactors (BWRs) have a similar vessel
Fig. 1. Expanded view of PWR vessel showing individual pieces before welding. Cylindrical shell receives highest neutron exposure [ 1 ]
........
CO
ME
HO
ROD
CHANISMS
NTAT ION PORTS
Ll!
HEAD ASSEMBLY
UP
PL
IN1
SU
SLEEVE
CO
5UPP
NN SPRING
qT PIN
ROD GUIDE TUBE
ROD DRIVE SHAFT
ROD CLUSTER
INLET
WN)
UPPE
CORE
~IOZZLE
THER~
:~ADIAL SUPPORT
REAC'I
ACCE~
~RE PLATE
RADIAl
CORE
)PORT COLUMNS
"NTATION
GUIDES
Fig. 2. Schematic view of advanced design pressurized-water reactor vessel and internals [ l ].
90
[9]. The importance of welding is evident from figs I and 2 [1 ]. A typical BWR or PWR pressure vessel contains
about 25 major welds [7]. The volume of each of these welds may vary between 150 and 500 in. 3 [ 1 - 5 ] .
Several processes could conceivably degrade the material properties during service, including temper, strain
aging and irradiation embrittlement. The low alloy SA-533 and SA-508 materials used in nuclear pressure vessels
are relatively insensitive to the first two phenomena [7, 9]. Also, they are given a post-fabrication anneal at
1000F which is expected to make them still less susceptible to both temper and strain aging embrittlement at
peak operating temperatures below 600F [7, 9]. The integrity of both SA-533 and SA-508 steels may be decreased by irradiation, particularly at the belt line where the neutron fluence is highest [ 5 - 1 6 ] .
2.1.2. Nuclear pressure vessel welds
Nuclear reactor pressure vessel welds could be grouped into nine categories (see table 1) [ 5 - 1 6 ] . Each of these
selected groupings have their own defect generation and development potential that are a function of (a) steady
state pressure load; (b) external load; (c) cyclic load; (d) seismic load, etc. For ease of model development and sub-
Table 1.
Comparison of variables affecting disruptive flaw propagation in stratified weld groups [5-16].
Steady External
Cyclicloads
Seismic Fabricability Preoperation Flaw
Post-operation
state
loads
loads
inspectability detection inspectability
pressure
Pressure Thermal
sensitivity
load
1. Dome and shell 5
flanges
2. Cylindrical
shell
3. Bottom dome
4. Bottom dome
penetrations
5. Coolant nozzle 5
openings
6. Combined
5
coolant nozzles
and supports
7. Support skirt
8. Top dome
9. Head bolts
Rating basis: 0, not applicable; 1, of minor significance; 2-4, intermediate rating grades; and 5, could be controlling effect on flaw
propagation.
Notes;
(1) Because of inspection technology limitations, flaws in thick flange forgings are less easily detected than in other vessel base
materials. Probability of critical size flaws is thus increased in the initial structure. Flaw growth due to load cycling should be very
slow. In-serviceinspection access is assumed excellent.
(2) Inspection technique should minimize concern for critical size flaws in the shell courses. Welding methods are conventional
and the highest quality should be attainable. Heat treatment is the least controllable fabrication variable, but quality control
actions should assure effective end results. Final inspection should disclose fabrication and heat treatment defects involving cracking. Cyclic loading effects are infrequent and combined stresses are lower in this .section than in other parts of the vessel. Flaw
growth is not expected to be accelerated by the loading conditions. In-serviceinspection access is good, making flaw detection convenient during post-operation inspection.
91
sequent presentation, this table was simplified. In particular, it was assumed that there m a y exist two major groups
o f welds; those under low steady or cyclic load and those u n d e r moderate or high steady or cyclic load. Those
welds classified as low stress welds include: (1) welds whose direction o f defect development is parallel to the direction o f stress, and (2) defects u n d e r low stress. Welds classified as moderate or high stress have a significant component o f moderate or high stress that is perpendicular to the direction o f defect development. Those velds u n d e r low
stress include main course welds. Welds u n d e r moderate to high stress include nozzle-to-vessel welds [9, 15].
2. 2. I. Pre.service flaws
The pre-service flaws may be in the base metals being fabricated or in the weldments [ 6 - 1 6 ] . This study is concerned with those flaws f o u n d in the w e l d m e n t and heat-affected zone because experience has shown that these
areas are more susceptible to failure than the base metals [5, 6 ] .
In weldments, the pre-service defects that could potentially be f o u n d are entrapped slag, inadequate penetration,
incomplete fusion, porosity, and cracks o f various types that can occur either parallel to or normal to the welding
directions [ 6 - 1 6 ] . Entrapped slags m a y be small, m e d i u m or large and are usually subsurface [ 7 , 9 ] . If they are
small, their appearance is similar to porosity [9]. If they are large, they behave in a way similar to lack o f fusion
or incomplete penetration [9]. Inadequate penetration is similar to incomplete fusion and m a y be small, m e d i u m
or large subsurface defects [10, 11]. T h e y m a y reduce fatigue life and m a y lead to brittle fracture [10]. Porosity
defects may be small, m e d i u m or large [ 11 ] and either subsurface or surface. Subsurface porosity defects are con(3) Initial inspection is convenient. Nozzle opening cutout should give good exposure of material interior conditions. Fabrication
of nozzle inserts is difficult and could introduce some weld flaws, but inspection techniques for nozzles are highly developed.
Cyclic loading in the head is infrequent and combined stresses are low. Flaw propagation rates should be very slow. In-service inspection access to this region is poor, but dome section welds can be examined effectively in more recent head design.
(4) Initial inspection is difficult and nozzle installation is awkward. Weld flaws could exist in nozzle-to-ligament welds. Cyclic
loads are small but high peak stresses in nozzle attachments are relieved by shakedown. No significant flaw growth is expected. Inservice inspection is very awkward for ligaments. The weld cross section is small and the welds probably would not contain flaws
of a size that would cause concern for their growth.
(5) Initial inspection access is assumed excellent but heavy nozzle forgings are more prone to flaws than plate. Fabrication
methods for installation nozzle forgings in shell openings are difficult and defects have b~en experienced that were not exposed by
radiography but were found by ultrasonic examination. Combined loads on this zone are the highest in the vessel and include very
high stress concentration at the nozzle-to-sheU junction, thermal cycling pump reactions due to thermal expansion and seismic
forces on the pipe. The high combined stresses make this region the most likely candidate for flaw growth. In-service inspection
access to this area is good, so flaw growth can be exposed,
(6) The same commentary concerning coolant nozzle openings applies here, but loadings are increased due to vessel weight
which is carried by these nozzles.
(7) Initial inspection access is fair. The skirts are fabricated by weld build-up to provide an attachment for the skirt ring. The
quality of the attachment is sensitive to fabrication control. Cyclic loading is infrequent but combined stresses could be severe if
pressure, seismic and thermal loadings have to be considered simultaneously. In-service inspection access in this region is poor.
Because of infrequent cycling, flaw growth rate should be slow and flaw growth is probably not important if the initial quality
control rigorously eliminates fabrication flaws.
(8) Inspection access is excellent, fabrication methods are convenient when few penetrations are used. Load cycling is infrequent.
The main cyclic effects occur during start-up and shutdown for refueling. In-service inspection access is excellent and, if necessary,
the top dome is replaceable. This is the least likely region for catastrophic flaw growth.
(9) Initial inspection is excellent. Bolts must be preloaded to oppose thermal expansion and the loadings control methods are
crude. Redundancy insures against propagation of single failure. In-service inspection frequency makes common mode failures that
destroy structural integrity extremely unlikely.
92
Table 2.
Effect of weld defects on properties [7, 9, 16].
Defect
Size
Location
Overfill
large
to small
surface
Undercut
small
surface
Lack of fusion
large
to small
subsurface
Arc strikes
small
surface
Haz cracks
longitudinal
or transverse
large
to small
surface or
subsurface
Hydrogen
cracks in
stainless
cladding
small
near
surface
Crater pipes
usually
small
subsurface
or surface
Porosity
large
to small
subsurface
surface
Working
temp.
range
brittle
Slag
inclusions
large
to small
usually
subsurface
Hardened
areas in
weld
large
to small
surface or
subsurface
sidered the least dangerous [10]. They have little or no effect on tensile, fatigue and impact properties and are not
expected to grow [10]. The surface porosity defects may grow together under cyclic loads and develop cracks [15,
16]. Longitudinal and transverse cracks are considered to be the most dangerous [ 9 - 1 6 ] and may be small, medium
or large and may be subsurface or surface [16]. Each o f the defect types discussed above may have their own particular initial concentration [5]. A summary o f the defects that could potentially be found in weldments is given
in table 2.
The quality o f the base metal and the weld determine to a large extent the probability o f flaws being initially
present [9, 15]. The initial ~ondition o f such vessel welds is established mainly by the applicable construction code.
The flaws which are likely to be present in a fabricated vessel prior to service are o f two general categories: those
which are within the acceptance standards o f the code under which the vessel is constructed and those larger ones
which escape detection by non-destructive testing techniques, either because o f the limitations o f the inspection
process, the limitations inherent in the material being inspected, failure to apply the inspection process at the proper
stage o f manufacturing, or simply misapplication o f the procedure on the part o f the fabricator. I t is generally concluded that defect sizes which are within the acceptance standards o f the codes and specifications are based on very
93
conservative estimates of their influence on vessel failure [ 19, 20]. There is a continuing research and development
effort throughout the world to improve the detection and discrimination characteristics of non-destruCtive testing
techniques [19, 20].
2.2.2. In-service flaw generation and development
Structural failure usually involves the generation and the slow development of a crack to a critical size at which
time failure occurs by rapid crack propagation if the vessel remains in operation [7, 9]. Almost invariably, such
crack development is the extension of either a crack which initiates in a highly stressed region or a crack-like defect
which remains after fabrication [ 9 - 1 2 ] . In the former case, the crack may initiate due to either a high constant
stress or a cyclical stress with a high mean value [7]. In the latter case, the defect may have been too small to be
detected during examination or may have been overlooked due to errors in examination [9]. In either case, high
stresses are required for subsequent crack extension at temperatures for which the material is ductile unless a
cluster of smaller defects join together [10]. In considering the.types of vessel structural failures possible tn testing or in service, it is convenient to divide the flaw development process into two stages: (1) slow crack development under cyclic or constant load and (2) final failure by rapid crack propagation.
2.2. 2.1. Slow crack growth. This may occur by fatigue or corrosion fatigue, by hydrogen cracking or by stress corrosion cracking. Fatigue and corrosion fatigue damage and crack growth in welds may be caused by cyclic stresses
from pressure, mechanical or thermal loading including transients [6]. Defects which are too small to be detected
by ASME Code examination procedures are not likely to propagate to disruptive failure under conditions encountered in normal reactor vessel service [6, 9]. Larger defects, undetected through error or poor inspection procedure,
might grow sufficiently to cause either leakage or rapid fracture [ 10]. The development rate is frequently stated to
be directly proportional to both the defect size and the amount of stress normal to the direction of development of
the defect [9, 10].
Low yield strength nuclear vessel welds are generally not highly susceptible to hydrogen-induced crack formation
[9]. However, Kussmaul and co-workers in the Federal Republic of Germany (FRG) have reported cases of hydrogen
cracking in the heat-affected zones of welds of 21 NiMoCr 37 steel [12-14, 18].
It is well recognized, both in the FRG and the US that some combination of severe constraint, insufficiently dried
electrodes or other welding materials, insufficient time and duration of preheat or postheat, etc. may lead to hydrogen cracking [9, 12, 18]. Correct welding techniques will minimize the risk of cracking. Proper inspection techniques should detect significant cracks prior to placing the vessel in service, but small or incipient cracks might not
be detected [4, 5]. It is conceivable that even a large crack may remain undetected.
In the FRG, an extensive investigation is proposed to explore the effects of all the important parameters on
hydrogen cracking and embrittlement of 21 NiMoCr 37 [12, 13]. No such cracks were reported in the US as at
the beginning of 1974 [12, 13].
In a weld initially free from hydrogen-induced defects, there appears to be no mechanism by which large
amounts of hydrogen could enter and be retained in the weld and cause subsequent embrittlement or crack formation in service [9, 12, 15]. Corrosion, stress corrosion, localized pitting and crevice corrosion could occur in welds
exposed by localized cladding failure [9, 10], but these are not expected to lead to cracking [9] and the rate of
chemical attack is expected to be so slow that large flaws should not develop [9, 13].
2.2.2.2. Rapid crack propagation. This is assumed to occur in welds by two processes, i.e. cleavage or ductile tearing rupture [7, 9]. Cleavage is the predominant mode of fracture at temperatures below the nil ductility transition
(NDT) [7, 9]. At higher temperatures, increasing microscopic ductility causes the fracture mode to change to the
ductile tear model [9].
In the transition temperature region, from NDT to approximately NDT + 200 F, fracture occurs by a mixture
of cleavage and ductile tear [7, 8].
94
2.2.3. Summary
Prior to reactor service some weld defects exist. After inspection most of the larger defects are removed. After the
reactor becomes operational, several influencing factors may cause the defects to develop. The rate of development
is usually assumed proportional to the amount of stress normal to the direction of development of the defect, the
frequency of cycling and the size of the defect. The defect (if under stress) will continue to develop to criticality
unless it is detected and repaired. Critical defects have been observed although the detailed development process
leading to this has not been identified.
The initial size of the defect, its rate of development and its ultimate size at failure are functions of several parameters including: (a) the particular weld type; (b) the amount of stress on the weld; (c) the frequency and intensity
of the stressing cycles on the weld; (d) the defect type; and (e) the weld and base metal material. It would be very
difficult to develop a theory of defect propagation based on all of these parameters since it may not be possible to
estimate, quantitatively, their influence on crack propagation. Rather, it may be advantageous to develop a theory
based on existing data of actual defect size and growth rate. Owing to a limited history of nuclear pressure vessel
welds, a major portion of the empirical data is extrapolated from a history of several million welds and weld years
on all types of pressure vessel welds that appear applicable [6-16].
The significance of weld defects anticipated in nuclear systems is a function of several factors [7, 9, 12, 15, 16] :
the size, shape, orientation and location of the defect; the type of stresses at the defect location; strength and notch
sensitivity of the weld metal; strength and notch sensitivity of the base metal compared to the weld metal; working temperature; and working environment. Therefore, it is difficult to select a quantitative value for the size of a
typical critical defect.
2.3. Weld inspection techniques and procedures
The weld inspection technique and procedure will now be described briefly for non-nuclear and in detail for nuclear
vessel welds. The purpose of studying the inspection procedure and techniques for non-nuclear pressure vessel welds
is to extrapolate the expected defect concentration at a given time from the number of defects actually found, the
inspection efficiency, and the percentage of the weld inspected. Once the length of the inspection interval is
known, the defect development rate can also be approximated. The nuclear vessel weld inspection techniques and
procedures are examined in order to apply the extrapolated data to nuclear systems.
2.3.1. Inspection techniques and procedures for non-nuclear welds
Most non-nuclear pressure vessel weld defect data comes from information obtained from the British [6, 8, 10, 11 ]
and the German [ 12-14, 18] literature, although some data is available from the US [7, 9, 26].
The inspection of Class I pressure vessel welds in the UK varies from code to code but in general employ the
following pattern [6, 8, 10, 11,20] : (1) during construction 50-100% of all pressure vessel welds are examined by
radiographic techniques; (2) ultrasonic examination of up to 100% of all welds prior to vessel operation is not
generally required except for nuclear pressure vessel welds; (3) visual examination at all stages of construction;
(4) pressure testing prior to vessel operation; and (5) radiography (or recently, ultrasonics as an alternative) of main
welded seams only (this may be 100% of weld or on a sample basis).
Although these codes are suggested, compliance is not compulsory in the UK .[6, 8], and it has therefore been
assumed (in an attempt to be conservative) that only 50% of the welds are initially inspected. The data was collected from 1 July, 1962 to 30 June, 1967.
The choice of techniques'applied in Germany depend upon: (1) the pressure vessel type and operating conditions; (2) the properties of the vessel and its internal and external structure; (3) the kind and size of flaws and
cracks and the kind of other material deteriorations to be detected; and (4) the effectiveness of test materials. The
Germans have two categories of techniques: integral and local [ 19]. Integral methods determine any serious changes
in the weld while local methods are confined to specific portions of a weld. Integral techniques include the overpressurizing test, tests with surveillance specimens, leak tests, and acoustic emission tests. Local techniques include
95
visual, optical, replica, dye penetrant, magnetic particle, eddy current, infra-red (i.e. thermal emission), conductivity, ultrasonic, and transmission techniques.
In Germany it is recommended that 80-100% of the welds be inspected prior to vessel operation [21, 22] and it
is required that all 'accessible' welds be inspected yearly [21,22]. Every four years the vessel should be dismantled
and all welds completely inspected; every eight years overpressurizing tests are required. Equivalent procedures and
techniques may replace one another, the overpressurizing test, particularly, may be replaced by non-destructive
methods. Detailed inspection techniques have been published [21,23, 24].
Based on the number of defects found in welds applicable to nuclear vessel welds, the inspection efficiency, inspection technique and length of the inspection interval; the defect concentration, generation rate and development
rate can be calculated. These parameters could then be applied to nuclear systems.
2. 3.2. Inspection procedures for US nuclear reactor pressure vessel welds
Section XI of the ASME Boiler and Pressure Vessel Code discl~sses the pre-service and in-service inspection of
nuclear reactor pressure vessels and pressure vessel welds [28]. In general, the inspection procedure can be described as follows. During some initial time interval, 0 ~< t ~<8o (i.e. the zeroth inspection interval) either all or a
portion of the welds are inspected with an efficiency of 100% or less. During the next time interval, 8~ < t ~< 8 o
(i.e. the zeroth repair interval) all defects of significant size that were detected during 0 ~< t ~< 8 ~ are repaired.
It has been assumed that once a sizeable defect has been detected, there is essentially a 100% probability
that it will be reduced in size so that either it does not merit repair or is too small to be detected. The probability
of not being able to repair a weld defect or replace a weld, given that it is known that the defect is present, is
insignificant [9], assuming proper access for welding. Of course, it is possible that a vessel will continue to be
operated with known defects.
During the time interval 80 ~< t ~< T1 (i.e. the first development and generation cycle) pre-existing defects
develop and new defects are generated. During the first inspection interval, TI ~< t ~< T1 + 8'1, a portion of the
welds are inspected with an efficiency of 100% or less. During the first repair interval, TI + 8'1 < t ~< T1 + 8 l,
essentially all of the defects that were located during T1 ~< t ~< T1 + 8'1 that required repair are repaired. This
process is continued and is shown in fig. 3.
In general, the length of the development and generation cycle is several orders of magnitude longer than the
length of the inspection and repair intervals (i.e. Ti >>8i + ~ , Vi).
The current US nuclear reactor pressure vessel weld inspection procedures are described in detail in the ASME
Boiler and Pressure Vessel Code, Section XI [28]. A summary of the standards is given here.
Nine stratified nuclear reactor pressure vessel weld types have been identified in table 1. Toreduce modeling
complexity, only two types of weld will be considered: those under low steady and cyclic stress and those under
moderate and high cyclic stress.
During the time interval 0 ~< t ~<8~ (i.e. the zeroth inspection) 100% of both the high and low stress welds are
assumed to be examined using at least one surface technique (e.g. dye penetrant) and at least one volume technique
(e.g. ultrasonic).
It is intended that the in-service examinations be performed during normal plant outages such as refueling or
maintenance shutdowns [28]. The inspections are specified in terms of what is required during three successive
generation and development cycles. At least 25% of the examination required during three successive generation
and development cycles shall be completed by the expiration of the first generation and development cycle (with
credit for no more than 33~% if additional examinations are completed) and at least 50% shall have been completed by the expiration of the first two generation and development cycles (with credit for no more than 66~%).
The remaining required examination shall be completed by the end of the third generation and development cycle
[281.
Where the extent of inspections require the examination of all welds during the first, second and third inspection intervals (i.e. for high stress welds) the same portion of the weld inspected by the first inspection interval will
be inspected again during the fourth inspection interval. This rotational basis shall be used throughout the success-
96
~ Ti
i=1
TWELFTH DEVELOPMENTAND
GENERATION CYCLE
11
/ i=~1 Ti+611
11
i =~1 Ti+ ~11
11
i=1
SECONDDEVELOPMENTAND
GENERATION CYCLE
Ti
T 1 + T2
TI + ~1
T1+5 ~
T1
FIRST DEVELOPMENTAND
GENERATION CYCLE
5o | REACTORSTART UP
ire inspection intervals. Where less than all of the welds are required to be inspected during the first, second and
third inspection intervals, a similar portion of the weld not previously inspected (other than the pre-operational
examinations) shall be required in each successive inspection interval [28].
2.3.2.1. L o w steady stress, low cyclic stres~ For pressure-containing welds in the reactor vessel belt line region, at
least 3.33% of the length of each longitudinal weld and 1.67% of the length of each circumferential weld are required to be examined during each inspection interval [28]. When the longitudinal and circumferential weld have
received an exposure to nehtron fluence in excess of 1019 nvt (energy of 1 MeV or above) the length of weld in
the high fluence region to be examined shall be increased to, at least, 50% [28].
For pressure-containing welds in the vessel shell and meridianal and circumferential welds in vessel heads, the
the area to be examined includes weld and base metal for one plate thickness beyond the edge of the weld [28].
The examinations performed during each inspection interval shall cover at least 10% of the length of each longitudinal shell and meridianal head weld, and 5% of the length of each circumferential shell and head weld [28].
97
2.3.2.2. Moderate steady stress, moderate cyclic stress. For pressure-containing welds in the vessel penetration,
the areas subject to examination shall include those pressure-containing welds of reactor control rod penetration
in reactor vessel heads, in the control rod drive housings, at vessel instrumentation connections and at heater connections in pressurizer vessels, among which a weld failure in any single penetration results in conditions that fail
to meet the exclusion criteria. The examinations performed during each inspection interval shall cumulatively
cover at least 8~% of the vessel penetrations [28].
2.3.2.3. High steady stress, high cyclic stress. For nozzle welds, vessel-to-flange welds, and dissimilar metal welds,
the individual examinations performed during each inspection shall cumulatively cover 33~% of each weld [28].
98
ployed, the surface finish, and to some extent, the nature of the defects. It is not unusual to detect a defect which
is as shallow as 0.001 in. [ 15, 17]. Cases have been reported of missing weld defects several inches long and 0.25
in. deep [15, 18].
A volumetric examination is used to indicate the presence of subsurface discontinuities with a method or technique capable of examining the entire volume of the weld contained beneath the surface. One-, two- and threedimensional defects can be detected. Methods, such as radiographic ultrasonic examination, or other newly developed techniques, may be employed provided the method is demonstrated to be capable of detecting subsurface
discontinuities [15, 17, 18, 24].
Radiographic techniques, employing energy sources such as X-rays, gamma rays or thermalized neutrons, may
be utilized with appropriate image-recording techniques such as photographic film or papers, electrostatic systems,
direct image orthicons or image converters [24]. The use of radiographic examination for in-service inspections has
generally not been fully explored for reactor pressure vessel welds because the radiation environment is high enough
to cause film fogging [ 17]. Radiographic inspection is limited in detection capability with regard to cracks; fatigue
cracks especially tend to be located in areas of cross-sectional changes which are difficult to radiograph [17].
The most popular method considered for in-service inspection is ultrasonic testing [ 15, 17, 18, 24]. Its different
principle (reflection from a discontinuous surface) from that of X-rays, plus its capability for application in a multiplicity of directions, makes it a powerful technique.
The model assumes that the weld defects may initially be of five distinct sizes rather than using a continuous distribution of sizes. These five discrete sizes have been assumed because (1) the quality of the available data on defect
sizes is not fine enough to distinguish between a continuous spectrum of sizes or more than five discrete sizes and
(2) the mathematical model used to treat the physical process is considerably more simple for the discrete defect
size.
The quantitative values of the inspection and repair efficiencies are approximated if sufficient data is available,
or otherwise assumed. The assumed data is not completely arbitrary since it is known that the inspection and
repair efficiencies are a function of inspection and repair technique, defect size and the number of prior inspections
and repairs at the defect site. The quantitative values of generation and development rates are approximated from
existing data.
This model uses empirical information to approximate the expected concentration of defects at a particular time
rather than using a 'first principal approach' or a 'mechanistic approach'.
99
Table 3.
Definition of parameters used in model.
Parameter
Definition
Extremely small defects. Defects of 0 size may exist but are either too small to detect or have a less than 50%
probability of being detected even with the most sensitive equipment. Even though these defects do not currently
present a significant problem, they have the potential to generate to defects of the next larger size (size A). The
concentration (or probability) of this defect size does not specifically enter the calculation.
Defects of size A have a minimum probability of 50% of being detected using the most sensitive equipment but
are too small to merit repair. Defects of size A have the potential of developing to defects of size B at a rate of
?~A.
Size B defects are larger than size A defects. If defects of size B are detected during an inspection they are
expected to be repaired (to defects of size A or size 0). If defects of size B are not detected and repaired, they
have the potential of developing to the next largest size defect (i.e. size C).
C is the largest flaw size below 'critical' size. If size C defects are not corrected during the inspection-repair
process, they will develop to size D defects.
These defects are of critical size and unlike defects of sizes 0, A, B or C are assumed to develop very rapidly. It is
assumed that defects of size D constitute a 'failed weld condition" and vessel failure is unavoidable if operation
continues for any appreciable time. Weld defects of critical size are not removed (and placed in smaUer defect
size groups) by repair.
61
a7
Ti
The length of the ith generation and development cycle measured in hours.
Wi
The portion of the weld inspected during the ith inspection measured in percent.
up(.
The probability of detecting a defect of size A during the ith inspection using the/th inspection technique.
HB(j)
The probability of detecting a defect of size B during the ith inspection using the/th inspection technique.
The probability of detecting a defect of size C during the ith inspection using t h e / t h inspection technique.
riP(;)
H~-'(]3
tt iB"+A (/)
l~i ~o ( j )
/~/--,A(j)
The probability of detecting a defect of size D during t h e / t h inspection using the jth inspection technique.
Px(t)
The expected concentration of defects assumed present of size X (where X can be size A, B, C or D) at time t.
Px(t) is measured in defects per weld.
Qi
The constant number of size A defects per weld hour that is assumed to be generated during the ith generation
and development cycle. Qi is called the defect generation rate. Qi is assumed to be independent of the number of
defects present.
kA, i
The rate of development of defects (during the ith generation and development cycle) of distinct size A to defects
of distinct size B. The development rate of each size A flaw to size B is constant and is measured in units of
inverse hours.
kB,i
The rate of development of defects (during the ith generation and development cycle) of distinct size B to defects
of distinct size C. The development rate is constant and is measured in units of inverse hours.
kC,l
The rate of development of defects (during the ith generation and development cycle) of distinct size C to defects
of distinct size D. The development rate is constant and is measured in units of inverse hours.
100
(1)
If the efficiency of detecting a size X defect is a linear function of the distance z from the surface of the weld,
and if the linear attenuation factor is assumed to be oJ then
(2)
HoX(j, z) = HoX(j)e - oz
If the defect'is detected and repair is required, it is assumed that the probability of repairing the defect is about
100%. In other words, the probability of repairing the weld defect (to size 0 or size A) is equal to the probability
that a weld of acceptable quality will initially be manufactured (/> 99.9%). IfHX+V(l~) is the net probability of
repairing a defect of size X to size Y during the 0th repair interval, then
HCo(~,) ~
HoCOA(z) + HoC~(Z).
(3)
101
PHYSICAL PROCESS
0~t6 o
~1 WoHoA(z~
WoHoB(Z)
Wo.Cm
D WoXoO(Z)
Go<t~G 0
@N D
~--w::~;~--r
......
.......................
Wo.o~
...................................
WoHoC~ 0
6<t<T 1
T1 +
61 < t < T 1 +
T2
,oI [~ [~ [~ [B
.o..i IB 1-71 i-q Fq
W1H~(Z)
TI+ ~l<t<Tl+
W1H~(~')
W1H1C(~
")
"O"
D[~
DD
SECONDREPAIR OF WELD
,L. . . . ........
.........
wl Nc~A
I
2,. . . . . .
[
:
jI
w,","~
...................................
WlH1C~ O
.......
'
'l
:
W21.1B~ A
'. . . . . . . . . . . . . . . . . . . . . .'
w~,-,~:~ A !i
.i . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . .
D
i
XC.2...~.
W1H~)(~')
61
oI[~
T 1 + T2 < t < T 1 +T2+ 6~
T1<t<Tl*61
W2H2B ( ~ O
...................................
w2.C~
102
At t = 6 o (immediately following the zeroth repair) the expected concentration of defects of sizes A, B, C and
D assumed present are Ph(8o), PB(6O), Pc(~o) and PD(~O), respectively. (In subsection 4.1 these expected
values are calculated from the actual number of defects found during 0 < t < 6~, the percentage of the weld inspected and the inspection and repair efficiencies.)
From 6o < t < T1 (during the first generation and development cycle) the rate of change in the expected concentration of size A defects is equal to the generation rate per weld hour of size A defects minus the rate of change
in concentration of defects that grow out of size A; the rate of change in the expected concentration of defects
of size B is equal to the rate of change in the concentration of defects that develop into size B from A minus the
rate of change in the concentration that develop out of size B to C; the rate of change in the concentration of
defects of size C is equal to the rate of change in the concentration of defects that develop into size C from B
minus the rate of change in the concentration of defects that develop out of size C to D; and the expected concentration of defects of size D defects is equal to the concentration of defects that develop out of size C to D.
Mathematically, during the first cycle:
dPA(t)/dt = Q1 - XA,IPA(t),
dPB(t)/dt= XA,1PA(t) -- XB,IPB(t),
dPc(t)/dt- XBAPB(t)-- hc,lPc(t) and dPD(t)/dt = hc,lPc(t).
(4, 5)
(6, 7)
Subscript 1 on the generation and develop.ment rates corresponds to the first generation and development cycle.
The solutions to the simultaneous equations (4)-(7) during the first generation and development cycle are,
respectively,
PA(6O < t <~ TI) = (QI/XA,1) ~1 -- exp(--hA,1 t )) +PA(t = 60) exp(--hA, lt),
(8)
(9)
1 - exp(--hA, lt )
+
1 -- exp(--hB, lt )
Pc(tSo < t ~< T1) = hA, 1XB,IQ 1 LXA,I(XB,1 _ XA,1)(XC,1 _ hA, l ) XB,I(XA, l _ ha,l)(hC, l _ XB,1)
1 - exp(-~h=, It)
exp(--hB, lt)
exp(--XA, I t)
(hB,1 -- hA, l)(hc,1 - hA, l )
exp(-hc,xt)
(10)
103
+ P c ( t S o ) {1 - exp(--Xc,lt)} + ?,8,1Xc,IPB(8O) [
+ (1/hC,l) (1 - e x p ( - X c , l t))]
(XB,~ - XC,~)
+eD(t -
60).
(11)
At t = T1 + 81 (immediately after the first inspection and repair interval) the expected number of defects per
weld of size A is equal to the number of defects per weld that were added to A from repair of defects of size B and
size C. The expected number of defects per weld of size B at t = TI + 81 is equal to the number of defects per weld
of size B at t = TI minus those that were removed from size B and repaired to size A or size D defects. The expected
number of defects per weld of size C at t = Tl + 81 is equal to the number of defects per weld at t = TI minus those
that were removed from size C and repaired to size A or size D defects. The expected number of defects per weld of
size D at t = TI + 51 is approximately equal to the number of size D defects per weld at t = T1 since repair of
size D defects is not assumed to be allowed. Mathematically,
PA(t = TI + 81) = P A ( t = T1) +H~-~A(y,)W1PB(t = T1) + H C - ' A ( y , ) W I P c ( t = T1)
because
HB-*A(E)WIPB(t = TI ) + HC-*A(E)WIP(t = T1 ) ~ e A ( t = TI ),
(14)
(15)
(16)
and
(17)
During T1 + 81 ~ t ~ T 1 + T 2 (the second generation and development cycle) the expected number of defects
per weld can easily be calculated since P A (t = T + 61 ), PB( t = TI + 81 ), P c ( t = T1 + 61) and PD(t = T1 + 61 ) were
calculated in eqs (13)-(16), respectively, and the generation and development of eqs ( 8 ) - ( 1 1 ) may be adjusted to
reflect new initial conditions at t = T1 + 61. Mathematically,
P A ( T , + 6 , <<.t<<.T, + T2) = Q2 l[ 1 - exp {--hA,2(t- T 1 ) ) |] + P A ( t + T1 + 8 , ) e x p { - - h A , z ( t - T1)),
[
XA,2
J
(18)
(~.a,2 -- XA,2)
(19)
1 - exp (-),B,2(t - r l ) ) ___
1 - exp(--),A,2(t -- r l ) )
+
Pc(T1 + 8 1 ~< t ~< TI + T2) = XA,2?,a,2Q2 [XA,2(XB,2 -- )'A,2)ff'C,2 -- XA,2) ~'B,2(~'A,2 -- ~'a,2)(XC,2 -- XB,2)
1 - e x p { - ; ~ , 2 ( t - Tt)}
] + XA,2XB,2PA(t = Tt + 61) [
exp {--XA,2(t - r l ) }
XA,2) (~C,2
~A,2)
104
f
I
I
+
~3
I "~
~I
IH
I
+
%/
4+
.{.
II
II
d;
1
e~
e~
'~
,~~,
11
eg
i
.<
irl
K.A.
Solomon
e t al., P r e s s u r e
105
vessel integrity
?
I
,<
,<
~i ~.
,<
-I?
e~
t ~" ' ~ l
~ r
eD
~7
I
,<
e<
I
e~
f~
,<
~J
,<
ill
I
sY
<
,<
I
e~
L
e~
,<
106
A,2
+ hB,ZPB(t
- Tl )I
hB,2&2
= T1 + 6 1)
exP l-k,20
hB,2)
+ @A,2
exP 6hB,P
(t -
&,2
Tl)}
+ t -
(lhB,2)[1
XB,2)
+ XA,2
exP{-hS,2(t
hB,2)&,2
(AA,2
%@=Tl
(l/hB,2)@
k,?@A,Z
hB,2&,2
@C,2
TINI f
-
XB,2)
XA,2)@B,2
(AA,2
hB,2)
WI1
hA,2)
T,))]
k,2)
TINI
T,Nl
-
AC,21
61)
-exP{-k,Z(f-
XA,2)
&,2)0B,2
@B,2
T,Nl
AA,Z)@C,Z
k,2)@B,2
th~,2k,2P~(t=Tl
t(1/k,2)[l
exP{--hA,f@-
exP{-k,a(t-
(20)
1
- exp{-&(t
exP{--hA,2@
(l/k,2)[1
ts,)[l-exP{-~,2(t-T1)}l
-exP{-hB,2(t-
TIN
Tl))
hB,2)
T,Nl + f - (1/k,d[l
= T1 + 6 I )
exP{-hB,2(t-
&,2
(1/&,2)[1 -
(l/AA,2)[1
ct -
t-
hB,2)
+~,)exP{-k,20-
hA,Z@B,Z
hB,2 k,ZPAct
+ (lhB,Z)[l
+PC(t=Tl
k,2)
@C,2
+T2)=h~,2h~,2&,2e2
hB,2@A,2
+ exP {-kc,2
and
PD(T, +61 GtGT1
Tl)I
k,2)(hB,2
kc,21
T&i
tPDtt=T,
f61j,
(W
where
(23)
and where the matrices used in eqs (22) and (23) are defined in table 4.
To calculate the number of defects present during any generation and development interval the following information is required: (a) the initial concentration of defects of size A, B, C and D; (b) the percentage of the weld
inspected during the zeroth, first, second, etc. inspections; (c) the inspection efficiency of detecting size A, B, C
D defects
rates
during the first, second, third, etc. cycles.
107
Table 5.
Definition of parameters for cost constraints
Parameter
Definition
CTOT
Total cost of inspecting all pressure vessel welds throughout the lifetime of the vessel.
CF
Total fixed cost of inspecting all pressure vessel welds throughout the lifetime of the vessel.
CV
~(i, J3
Total variable cost of inspecting all pressure vessel welds throughout the lifetime of the vessel.
Variable cost of inspecting per ith number inspection, per volume (or surface area) of weld, per/th technique.
Total volume or surface area of weld.
Portion of volume (or surface area) of weld inspected during the ith inspection using the/'th inspection
technique.
wi,j
CD
Total length of time needed during the ith inspection for the/th sequential technique beyond the length of
time that the reactor would normally be down for other, unrelated reasons.
Cost per unit of downtime.
i = 1 , 2 . . . . ,n
j=1,2 ..... m
costs and the variable costs. Once the decision to inspect is made. the fixed cost is invariant with respect to the inspection parameters (i.e. the percentage of the weld inspected, efficiency of inspection, length of the inspection
intervals, inspection techniques, etc.).
The variable cost associated with the inspection of a weld is a function of the cost of the particular inspection
technique, the number of different techniques, the number of inspections, the size of the area or volume of weld
insepcted, and the cost per unit of additional downtime required to perform the inspection beyond the time that
the reactor would be down for other purposes. The variable cost may also be a function of other parameters that
are currently undefined. Based on the above assumptions, the total cost Ca'oa- is equal to the sum of the fixed cost
C~ and the variable cost Cv,
Ca'oy --- CF + Cv,
CF = constant,
(24)
and
tl
rr/
rt
(26)
i=1 j=l
Equation (22) may be combined with eq. (26) to produce five non-linear, simultaneous, first order differential
equations (see table 5 for parametric definitions). Using non-linear, dynamic programming an optimal inspection
procedure can, in principle, be developed.
108
Because of the rather limited nuclear pressure vessel experience and the relatively extensive experience in nonnuclear pressure vessels, much of the data has been derived from the latter source. When evaluating the data, considerable care is taken to ensure that all data (non-nuclear and nuclear) is applicable to nuclear vessels. The data
represents many vessel years of study in several countries including the US [7, 9], the UK [6, 8, 10, 11], Australia [6, 15], the FRG [12-14, 18], France [16], Belgium [16, and Austria [16]. Conservative assumptions are
made when the data is in doubt.
(27)
i{
PB(5O) = ~
FB(5o)
(28)
(29)
1 I FD(5) /~--IFD(~0).
PD(5O)= ~ [woHD(Z)
(30)
109
The number of defects classified a particular size which were actually found during the zeroth inspection (i.e.
0 < t < 8~) are outlined in table 6 together with the number of defects per weld expected present after the zeroth
inspection and repair interval (i.e. t = 80). The rather wide range in the number of defects of a particular classification expected present is due in part to the uncertainty in the magnitude of the model parameters. The significant
discrepancy in the range of the number of defects of a particular size expected present using the UK data [10] and
the German data [11-14] may be due in part to the assumption that only 50% of the UK welds are inspected
during 0 < t ~< ~ whereas 80% are inspected during this interval in Germany. The assumption concerning the
number of welds inspected in the UK may be valid and is due to the fact that weld inspection is suggested but not
enforced. The assumed 50% inspection area reflects an upper estimate in the actual number of defects of critical
size present. The British may, indeed, inspect 100% or nearly 100% of the weld during 0 < t <~ 8~, but due to insufficient evidence about the portion of welds inspected, the above assumption was made. Another conceivable
reason for the discrepancy between expected number of defects present in the UK and Germany may be due, in
part, to the different written descriptions of critically-sized defects. The British may classify a defect as being
critical while the Germans may classify the same defect as subcritical. Other conceivable reasons for this discrepancy may be a significant difference in inspection procedures, inspection efficiencies and weld quality that were
not apparent to the observer.
The Germans did not record size A defects. The range of expected size A defect concentration based on British
data varies between 4 x 10 -4 and 6 x 10 -4 defects per high stressed weld and between 3 x 10 -4 and 5 10 -9
defects per low stressed weld. Since the welds were not under stress prior to t = 8~, there is no reason to suspect
that these two ranges would differ. Owing to lack of sufficient documentation by the British, the expected defect
t
concentration at t = 8 o of size B and C are equal and in fact are slightly smaller (by about 20%) than the range in
the expected defect concentration of size B derived from the German data. The Germans record approximately
0.5-1 x 10 -6 defects per weld of size C. Using the British data, the range in defects of size D varies between about
2 and 15 x 10 -7 defects per weld. However, using the German data this number varies between about 1 and 7 x 10 -a
defects per weld of size D (significantly different from the British).
Commercial LWRs in the US have an accumulated total of only 125 reactor years of operation as of late 1973.
The limited number of reactor years and the discrepancy in reactor vessel design does not permit relevant statistical inferences of failure probabilities. There have been several specific instances of non-descriptive defects found
in the vessel welds due to errors in design, fabrication or construction. The instances where operating conditions
contributed to the generation or growth of defects will be discussed in a later section. The Elk River Reactor (ERR),
Monticello [9] and Japanese JPDR BWR [9] reactor vessels all used a weld overlay cladding of stainless steel
similar to type 308. Cracking was observed in the ERR cladding prior to start-up, and enhanced cracking was observed after one year of operation. Similar behavior was observed at JPDR and Monticello.Table 6 summarizes the
results outlined above. Although some additional experience is available from 1974, it would be almost meaningless to calculate the expected defect concentration for this limited data. Table 6 also summarizes the defect data on
US non-nuclear vessels built since 1962.
The foreign vessel statistics, particularly the German data, span a relatively long period and reflect the improvements in material processes, fabrication practices and design codes which occurred during the latter part of the
period covered. Moreover, from the large number of vessels covered, it is evident that most of the boilers were small
and that the studies probably included many heating boilers, for which the average level of quality would be expected to be lower than for central station boilers. (This is especially true of the UK data.) It is therefore probable
that the statistics tend to over-estimate the expected defect concentration for large fossil-fueled boilers [9, 10] and
may partially account for the discrepancy between the UK and German data.
Many of the reported defects present prior to start-up were in regions, such as tubes, that have no counterpart
in a reactor vessel. These failures, where adequately identified, have been eliminated from consideration as not
pertinent to reactor vessels.
Based on closer tolerances in the design, materials and inspection during fabrication of nuclear vessels (and their
welds) over non-nuclear vessels (and their welds), it would be valid to assume that the initial concentration of
110
?
x
~0'4
5c~
!
06
?
x
c~c~
!
oo~.
~("l
o%
c~c~
c5
~O
e4
T
X
O~
d
z
~
<
O~
I
0
~ O
0 0
~ 0
~ 0
X~'4
'4 oO
~o
~ Q
w~ ~D
O0
.=.
oo
0
111
defects in non-nuclear vessels ":s larger than the concetration of defects in [nuclear vessels for each o f the four
defect sizes [9, 10, 13]. The limited data on nuclear vessels does not allow us to calculate any quantitative value
o f defect concentration prior to vessel operation. However, the non-nuclear data provides an upper bound on
these concentrations.
4. 2. Percentage o f welds inspected and the corresponding efficiency
In this section the percentage o f the weld examined during the zeroth inspection W0, during the first inspection
WI ; during the second inspection W2 , and so on, will be discussed. The results are summarized in table 7. Foreign
weld inspection standards are generally- more relaxed than US standards [19, 20, 27]. Also, standards for inspecting
non-nuclear pressure vessel welds are more relaxed than standards for nuclear welds [9, 23, 27, 28].
There are two purposes for estimating the percentage o f the weld inspected. First, it is necessary to determine
what percentage of the welds are inspected in the past to estimate the expected number o f defects that could have
been found had all welds been inspected, Secondly, it is necessary to estimate what percentage o f the weld is
currently being inspected by US vessel standards. This value is taken directly from ASME Ill and XII [27]. For all
vessels (foreign and domestic, nuclear and non-nuclear) it is expected that 50% ~< 1t/0 ~< 100% where W0 = 100%
is usually the case [19, 20, 27].
For US nuclear vessel welds under high stress W1 = I2 = W3 . . . . .
W12 = 33~% where every fourth inspection
examines over the same area (non-overlapping technique). For US nuclear vessel welds under low stress W1 = I92
.....
Wl2 where 2% < Wi < 5% depending on weld [27]. It is extremely difficult to estimate the portion of the
weld inspected in foreign vessels. A conservative estimate would assume that only 20% o f the high stress welds are
Table 7.
Percentage of weld inspected (some values are assumed).
Country
Wo(%)
Ii
Comment
(i = I, 2, 3 ..... 12)
(%)
Us(4O,S4)
Nuclear
high stresswelds
I00
331
100
2-5
Non-nuclear
high stress welds
low stress welds
100
100
33t
2-5
All vessels
high stress welds (?)
low stress welds (?)
50-100
50-100
20-30
1-5
FR GO6-aa,42,43,46,49)
All vessels
high stress welds (?)
low stress welds (?)
50-100
50-100
20-30
1-5
Other vessels09,40,41 )
All vessels
high stress welds (?)
low stress welds (?)
50-100
50-100
20-30
1-5
UK(32,34,3S,44)
112
Table 8.
Range of assumed inspection efficiencies as a function of defect classification [8, 10-19, 21, 24 ].
Technique
Stress
wave
i = 0 50-70
i > 0 50-70
70-90
70-90
90-95
90-95
96-99
96-99
45-65
45-65
Pressure
wave
i=0 i> 0 -
99.9
99.9
.
.
Ultrasonic
i = 0 50-70
i > 0 50-70
70-96
70-96
70-96
70-96
90-98
90-98
98-99.9 volume
98-99.9 technique
0-60
60-80
80-85
85-90
0-60
60-80
80-85
85-90
surface
technique
Dye
i = 0 15-60
penetrant i > 0 20-60
70-95
70-95
90-96
90-96
96-98
96-98
15-55
20-60
70-95
70-95
90-95
90-95
95-96
95-96
surface
technique
Leakage
99.9
99.9
volume
technique
40-70
50-80
70-85
70-90
98-99
98-99
0-30
10-40
40-60
50-75
70-80
70-85
98-95
94-98
volume
technique
Visual
70-90
70-90
.
.
.
.
90-95
90-95
96-99
96-99
Comments
.
.
effective for
large defects
exam
Radiograph i = 0
0-40
i > 0 10-50
volume
technique
inspected at each interval and only 1% o f the low stress welds [27, 2 8 ] . Since the British do not follow a uniform
inspection code, it is n o t k n o w n if they use a non-overlapping inspection technique for their high stress welds as is
done in the US. As a conservative assumption it has been supposed that the British do not use a non-overlapping
technique.
A range o f assumed inspection efficiencies per technique for each defect size classification is listed in table 8.
The value H X ( 6 ) is the efficiency o f the ]th technique in detecting (and l o c a t i n g ) a size X defect during the ith
inspection a t t e m p t .
113
equal to the number of size A defects per weld hour that are generated. The development rates during the ith generation and development cycle are proportional to the number of defects per weld that grow from a defect of a particular size to a defect of the next largest size. The generation and development rates are constant during generation
and development cycle, but may change during subsequent cycles.
From the data, the expected concentration of defects of sizes A, B, C and D present immediately following the
zeroth repair has been calculated in subsection 4.1 to be PA(60), PB(60), Pc(6o) and PD(6O), respectively. These
values were calculated from the number of defects actually found during 0 ~< t ~< 6 o, the number of welds inspected,
the inspection efficiency and the percentage of welds inspected. Using the available data, the number of defects
actually found during the first inspection (Tt ~< TI + 6~), during the second inspection (Tt + 7"2 ~ t ~< Tt + 6~),
and so on, are known. Also, the number of welds inspected, the inspection efficiency, the percentage of welds inspected, and the length of the inspection interval are known during subsequent inspections. As a result, the
expected concentration of defects at t = Tt + 6 x, t = Tt + T2 + 62, etc. can be calculated.
The values of Qt, ~.A,l, ~,B,1 and hc, 1 are calculated by solving eqs (8)-(11) simultaneously at t = Tt. The
values OfPA(t = 6o),PB(t = 60), Pc(t = bo),PD(t = 6o),eA(t = Tl),eB(t = Tl),Pc(t = Tl) and PD(t = Tt)have
been calculated from the existing data (see subsection 4.1). The resulting four transcendental equations are solved
for the four unknowns, Qt, XA,1, XB,1 and ?,c,1Table 9.
Number o f defects found during the first five years of vessel service [ 10].
Age
(yr)
Tl
Assumed
size
classification
Number
found
Description
5
3
5
5
3
B
C
B
B
B
1
1
1
1
1
3
3
3
5
3
B
C
C
B
B
1
1
1
1
1
3
3
3
3
3
B
B
D
C
D
1
1
1
1
1
3
3
3
3
3
B
B
C
B
C
1
1
1
1
1
5
3
5
3
5
B
A
A
A
A
1
9
14
18
22
114
The values o f Q2, ~kA,2, ~kB,2 and hc,2 are calculated by solving eqs (18)--(21) simultaneously at t = TI + / ' 2 .
The values O f P A ( t = TI + 8 1 ) , P B ( t = Tl + 8 1 ) , P c ( t = T1 + ~ I ) , P D ( t = TI + 8 1 ) , P A ( t = Tl + T2),PB(t = TI + T2),
Pc(t = Tt + 7'2) a n d P D ( t = Tt + T2) are k n o w n .
In general, the values o f Qn, hA,n, ~kB,nand hc,n are calculated by solving the m a t r i x e q u a t i o n (22) at t = Y-n=l ti
where the values o f
t=
t=
"=
PB t = ~ T i + S n - I
+Sn-I
PB
t=
PC t = ~
i=1
and
PD
Ti+Sn-I
PD t = ~
i=1
Ti+Sn-I
=
i=1
Table 10.
Number of defects found during the second five years of vessel service [10].
Age
(yr)
T1 + T2
Assumed
size
classification
Number
found
Description
9
7
9
8
7
B
B
C
B
B
1
1
1
1
1
9
6
6
6
6
B
C
C
C
C
1
1
1
1
1
7
7
9
9
7
B
B
B
B
B
1
1
1
1
1
'low stressweld"
'high stress weld'
'high stress weld'
'low stress weld'
'low stressweld'
8
6
8
10
6
C
B
B
B
C
1
1
1
1
1
~8
~ 10
8
~ 10
A
A
A
A
12
16
26
28
115
008
)t~Z
I+
"".~
I I I l [
I I 1 1 1
o
I
~,.q.
o
Iw.,
e...
~:~x
0091
<
~gg
r~
e..
o
I ~
o
I
"~
oq.~
<-,<
b.o
e.
o
~x
llll
t + l b
OZ6 <
OOZ
r~
I
I l l l l
p.l
o
I
q o"~ j
r,
o
" ~
c:w
o% t ~
116
016 (
l l l l l
=,
~C x
o
I
"G
~C x
(::w
008I
,<
IA
0
'~
v
I
.o
e-
o
I
<-o
,<
',D
I l l l l
00~ I
P.
o_
I
~D
<::w
117
An alternative method of obtaining ~c,i is suggested which consists of modifying a formula on crack growth in
other metals [4]. It is assumed that the development of flaws in weld metals and other metals obey the same physical laws. No empirical formula has been derived for the development rate of defects in welds, but a formula does
exist for defect development in 7075-76 aluminum [4]. I f N is the number of stressing cycles over a range of
+OKst.and a is defect size, then
da/dn = a' {o(rr) l/2a} a[~, _ o0r) 1/2a} '
(31)
where a and/3 are material constants. For simplicity, it is assumed that the specimen experiences a constant rate of
cycling so that the rate of growth may be expressed as da/dt.
The length of time taken for one defect of size 0 to grow to size D is equal to
A
a(a)
T = jf a(a)
~ 3'2 dt + jf~ a(a)
_ ~ d3'2
t + J ~ - -1"~ a(a)
q 7 5 c l t_+ ~ ~_al/---------~a,.
_
(32)
Also,
B
1
f a(a) a/2
XA, t. =AJ~---~-~ dt,
dt
and
1 f a(a) 3/2
;kc, 1 = j ~ _ ~ d t .
(33-35)
The value of at. and/3t, can be obtained by solving eqs (33) and (34) simultaneously for each generation and development interval. The value of )~c,t. can then be calculated.
The above methodology is applied to the German and US data and the results are displayed in tables 11 and 12
for low and high stress welds respectively. Note that the values derived are not unique and depend on assumptions
made with regard to inspection efficiency, generation rate of new size A defects, etc.
5. Results
The mathematical models developed in section 3 are combined with the data of section 4 to yield a range of estimates on the probability that a nuclear reactor pressure vessel weld contains a defect of critical size. The number
of defects of a given size predicted is compared with existing statistics-on defects. A set of model parameters is
selected such that the predictions of the model are consistent with existing statistics. The more sensitive input
parameters are determined. Finally, an optimal inspection procedure is suggested. This procedure is based on
several assumptions regarding inspection and downtime costs. The calculations are performed for defects under
both low and high stress since the development rates and inspection procedures for these two categories are
distinctly different.
118
VV
VVV
VV
VVV
0 0 ~ 0 0
0 0 0
0
"'.0
,<
0
"el
0
~') .It- it% to)
0
0
<
d~
,m
c5
8X
x
-t--
O0
w~
"WD ~O "It"
~,...,
-it.
119
It would be interesting to compare the predictions o f the model with the actual number o f defects found. In
particular, the UK located three defects of critical size (i.e. size D), in approximately 181 000 low stress welds
(during 1.3 x 101 low stress weld hours). Since size D defects are considered to be in the failed state, the probability o f finding a size D defect is assumed to be essentially equal to the probability o f one existing. According
to the British data, the expected D size defect concentration is 1.7 x I 0 - s defects per low stress weld after
1.3 x 10 I weld hours.
EXPECTEDNUMBER
OF DEFECTSPERWELD
PAlt)
PB(t)
104 _
101 --
103
100
12
i0.1
Pc(t)
10.3 !_
PDlt)
10-6
'
PD
TO TO+ T 1
2 -
i~o'i
101
too
14!
17 -
PA
PB
PC
I0-2
~0.3
10.1
10"4
10.2 :
IO"5
o 'u ~ .
10-6
10-6 - L
.I
/ ~ ~ :
10.6
10"9 / ' ~
"-
oI
-,"
p4"'
'
I
100 000
3O0000
200,000
TIME (HOURS)
Fig. 5. The expected number of defects per weld of sizes A, B, C and D as a function of time. Base run for low stress welds.
EXPECTEDNUMBEROF
DEFECTSPERWELD
PA(t)
PB(t)
Pc(t)
PD(t)
/
104 -
101 -
11~'
1o
101
10.2
100
10.3 J
10-1
10"4
10.2
10.5
10"3
10"6 h
TO
PD
PA
,.
PB
PC
10"5
10"6
0
100,000
20(] ,000
300000
TIME (HOURS)
Fig. 6. The expected number of defects per weld of sizes A, B, C and D as a function of time. Shortened inspection intervals for
low stress welds.
120
EXPECTED NUMBER
OF DEFECTS PER WELD
PA(t)
PB(t)
Pc(t)
PD(t)
PC
104 .
101
10 3
10 0
102 -
10-1 _
101
10-2
100
10.3
10-I
10-4
10-2
10-5
10 .3 _
10 -6
To
10-4 _
L ~
PD
T 0+ T 1
10-7 _
PA
PB
~"
p ~ d,m
10.6
10.8 ~6
10"6
."'1
10-9
'
1
100,000
" """ I,
200,000
300000
TIME (HOURS)
Fig. 7. The expected number of defects per weld of sizes A, B, C and D as a function of time. No defects initially present and no
inspections for low stress welds, except for zeroth inspection.
EXPECTED NUMBER
OF DEFECTS
PER WELD
PA(t)
PB(t)
104 --
101 -
10 3
100
102 -
10-I _
101 -100 --
10-2
Pc(t)
10"3J--
PD(t)
,, t
T O TO+ T 1
10-4 ~-
10-3
10-5 ~-
10-1
10-4
10-2 L
10-5
10-6 L_
2
iz. 0
m,zp am
%
PA
10-7
~o
~
~d
Bo ~
b~
10.8
~l,md
pd
e,"
10-9
I
0
100,000
200 00O
TIME (HOURS)
300000
Fig. 8. The expected number of defects per weld of sizes A, B, C and D as a function of time. Inspection area increased for low
stress welds.
121
102
PB(t)
PC( t )
PD ( t )
10 . 2 1 -
X0 TO+ T 1
lO
10-1
lO"3 I -
,"
1o-S
1oI -
Il i
PD
"
.,,0,1- I
10- 4 1 -
"
/,.,,r"-I
,.,:,L I A-I
PC
.l_,.t/!,, "
"
10-2 lO.3
i..
.
n l
1o.6L_
10"5 --
0r
0
n,
i ,
100,000
...
20(] ,000
TIME (HOURS:
300,000
Fig. 9. The expected number of defects per weld of sizes A, B, C and D as a function of time. Inspection area considerably
increased for low stressed welds.
EXPECTED NUMBER
OF DEFECTS
PER WELD
PA(t)
PB(t)
lO 3
Pc(t)
PD(t)
T O T0+T 1
~T,
4x10- 5 ~- 4x10-9 --
101
Pc
PD
~'
~'
~,~," PB
o.o,
10"1
I-
10-21_
0.01
I
~
Lf'] i
,,,~,,,, , ~ o g
T-
100 000
..
I"
,,~,6
P'*
20(1000
TIME (HOURS)
I
300000
Fig. 10. The expected number of defects per weld of sizes A, B, C and D as a function of time. No defects initially present for
low stress welds.
122
EXPECTEDNUMBEROF
DEFECTSPERWELD
PA(t) PB(t) Pc(t) PD(t)
~,~'~ ~
1012
103I
100 ~-
PD
Pc
102~- 10"1 ,
108
104
PB
?
.~,,.
100~
10"3?
10"1~
10"4
PA
10-2[-- 10-5
I
100,000
200,000
TIME(HOURS)
Fig. I 1. The expected number of defects per weld of sizes A, B, C and D as a function of time. Increased generation rate for low
stress welds.
PD(t= 350,000HOURS)
a
,.J
ILl
~=r~ -ln-5
mm
mqQ.
N
~ lO-5
u. I-..
o5
10-7
10-5
10-4
10-3
10-2
10"1
100 PA(t- 0 HOURS)
NUMBEROFDEFECTSPERWELDOFSIZE "A'" AT t = 0 HRS
Fig. 12. The number of defects of size D per weld at t --- 350 000 hr as a function of the number of size A defects per weld initially
present. All other parameters are fixed as for base run, low stress welds.
The UK data is based almost exclusively on non-nuclear pressure vessel welds which are assumed to be fabricated using less stringent standards than nuclear pressure vessels [10, 23, 24] and have lower inspection requirements [10, 23, 26]. Also, ~he non-nuclear vessels are under a much wider range of pressures [9, 10]. Hence, one
might assume that there is a distinct difference in weld quality between the vessel welds studied by Phillip and
Warwick and those used in nuclear pressure vessels.
By the end of weld life, there is approximately a 170 x 10-7/6.3 x 10 --/ or about a 27 times higher concentration of D size defects existing in the UK non-nuclear low stress welds than the concentration predicted for US
nuclear vessel low stress welds using the model and the assumed base data.
123
There are a variety of combinations of input parameters that could be changed to increase the predicted defect
concentrations in nuclear vessel low stress welds to 27 times their predicted concentration at the end o f weld
life; there is no unique combination. F r o m fig. 15 it can be learned that one such combination could be obtained
by keeping the percentage o f welds inspected, the frequency of inspection, and efficiency of inspection fixed at
the base run value and increasing the generation rate to 4.1 x 10 -7 defects/weld hr and increasing XA,i to
4.1 x lO-9/hr; Xn, i to 4.1 x 10-S/hr; 3.c, / to 4.1 x 10-7/hr.
If ASME Code Section XI Inspection Standards [28] were applied to the UK welds instead o f the assumed
standards [10, 23, 26], then the expected number o f size D defects per weld at the end of weld life would be
reduced to approximately 60 x 10 -7 .
The reason for the discrepancy between the predicted number o f defects of critical size using (1) the model
with base run parameters (i.e. the assumed US welds) and (2) the number o f defects of critical size that would
have been found by the British [10] had the US inspections applied [28] is due to the assumption [9] that the
~ 10-5
10.6
m=
z ~ 10-7
10-7
10"6
10-5
10"4
10-3
10-2
PB(t = 0)
Fig. 13. The number of defects of size D per weld at t = 350 000 hr as a function of the number of size B defects per wold initially
present. All other parameters are fixed as for base run low stress welds.
Z ~ 10"7
10-7
I
10-6
10-5
10-4
10-3
10-2
NUMBER OF DEFECTS PER WELD OF SIZE "C'" AT t = 0 HOURS
Pc(t = 0)
Fig. 14. The number of defects of size D per weld at t = 350 000 hr as a function of the number of size C defects per weld initially
present. All other parameters are fixed as for base run, low stress welds.
124
(LOW STRESS
wE. ,
m ~ 10-5 ~_
~,.
10"61-
|
10.7
I
10-8
10-9
I
I
I
10-7
10-6
10-11
1040
10.9
10"8
1040
10"9
10"8
10"7
~'B
10-9
I
10"8
I
10.7
I
10.6
~C
hA
Fig. 15. The n u m b e r o f defects o f size D per weld at t = 350 000 hr as a function o f generation and development rates, low stress
welds.
~ 10,6
g~
~i
o:
lo-7
o
10-8
10-9 _
I
20
I
40
I
6O
I
8O
I
100
(PERCENT) W
PERCENT OF WELD INSPECTED PER INSPECTION
Fig. 16. The n u m b e r of defects of size D per weld at t = 350 000 hr as a function o f the percentage o f the weld inspected for low
stress welds.
125
10.5
I
]
A.
~- _~ lo-6
8
~ " 10-7
10.9
4O
I
h
I
[
5O
6O
I
7O:
80
85
90
I
95
9O
92.5
95
i
I
97.5
I
80
H~('~ ) (PERCENT)
I
100
HB("~') (PERCENT)
I
100
HC(E;) (PERCENT)
99.8
99.9
I
100
HiD(~') (PERCENT)
EFFICIENCY OF INSPECTION
Fig. 17. The number of defects of size D per weld at t = 350 000 hr as a function of the percentage of the weld inspected for low
stress welds.
g~
10"6
J..
1"71
I1,,,I
[
I
I
I
I
I
1
5
10
15
20
2S
30
35
NUMBER OF INSPECTIONS, EQUALLY SPACED
OURING 40 YEARS
I
40
Fig. 18. The number of defects of size D at t = 350 000 hr as a function of the number of inspections. For low stress weld.
126
welds used in US nuclear vessels are assumed to have ten times the integrity of the welds that were used in the UK
non-nuclear vessels. If this assumption were removed and if ASME Code Section XI [28] inspection standards
were used, then the expected number of size D defects per weld at the end of weld life using the model would be
approximately 60 x 10 -7.
However, for the purpose of this study, it is assumed [9] that the US nuclear reactor pressure vessel welds have
ten times the integrity.
From fig. 16 it can be learned that if only the percentage of welds inspected is changed, then the predicted
concentration of defects using the model can be increased by a factor of 27 by reducing the percentage of welds
inspected to less than 2% per inspection. From fig. 17 it can be learned that if the inspection efficiency is the
only parameter to change, then it would have to be reduced considerably to increase the predicted concentrations
of D size defects by a factor of 27. If the only variable parameter is the number of inspections (fig. 18) then the
predicted concentration of D size defects could be increased by a factor of 27 by greatly decreasing n. There are
a variety of other combinations of model parameters that could increase the predicted defect concentration by a
factor of 27 including changing the initial defect concentration.
The expected number of defects per weld of size A is a monotonically increasing, exponentially behaved
function. The rate of increase appears to be continuous because the number of defects repaired from sizes B and
C to size A is insignificant compared to the number already present in A, i.e.
Therefore, the expected number of size A defects during any time in the interval 0 < t < 350 000 hr can be
approximated from eq. (8) as
PA(t) = (Q/XA)(1 - exp(--XAt)} + P A ( t = rio) exp(--XAt),
where Q1 = Q2 Qn = Q2 and XA, 1 = XA, 2 . . . . .
A defects is
dPA(t)/dt~- Q
(36)
exp(--XAt)--PA(t=~o)XAexp(--XAt),
~Kexp(-XAt),
--K(1-XAt),
(37-39)
(40)
As t increases, 1 - ~,At decreases, and d P A ( t ) / d t decreases as well. Hence, the number of size A defects increases
at a decreasing rate.
The expected number of defects per weld of sizes B and C increases with time as a decreasing rate as well. The
increase is exponentially behaved but is non-monotonic due to the fact that defects of sizes B and C are removed
at periodic inspections, and thus their number decreases in a discontinuous fashion. The expected number of
defects per weld of size D increases monotonically with time at a decreasing rate. Defects of size D are not allowed
to be repaired.
The effect of doubling the number of equally spaced inspection intervals, keeping all other base parameters
fixed (fig. 6) is to decrease the expected number of defects per weld of size D by. about 60% at t = 350 000 hr
(40 yr). The effect of hating only a zeroth inspection, assuming that no defects of sizes A, B, C and D are initially
present, and keeping all other base parameters constant, is to increase P o ( t = 350 000 hr) to 1.3 x 10 - 6 defects
per weld (about twice what it was for the base case). The comparison is not too surprising since for the base case,
only 5% of the welds are inspected with less than 100% efficiency during each of the 12 inspections. Hence, at
the end of 350 000 hr only about 50% of the weld defects were corrected. In fig. 7, 0% of the welds were repaired
subsequent to t = 0. In fig. 8 twice the normal inspection area is inspected (all other base parameters are kept
constant) and the result is that P D ~ 4.1 x 10 -7 defects per weld (about 65% of the value Of PD for the base run).
In fig. 9 100% of the weld is inspected during each inspection (with all other base parameters kept constant). The
127
expected number of defects per weld o f size D at t = 350 000 hr is about 1.0 x 10 -s (less than 2% o f the value
o f P D for the base run). The discontinuity in P c ( t ) and PD(t) is due to the large percentage of C and D size
defects repaired.
When zero defects o f sizes A, B, C and D are initially present, PD ~-- 4 x 10 -9 defects per weld (fig. 10) assuming all other input parameters are the same as the base run. In other words, complete elimination of all size
defects at t = 0 has the effect of decreasing the D concentration at t = 350 000 hr by a factor of over 100. When
the generation rate is grossly increased by a factor of 108, the expected number of defects per weld of size D at
t = 350 000 hr is 5 x 10 - 2 (fig. 11).
Figure 12 displays the sensitivity of PD to the number of defects per weld o f size A initially present. All of the
other parameters are the same as the base run.
Figures 13 and 14 display the sensitivity Of PD to the number of defects per weld o f sizes B and C initially
present, respectively. The value Of PD is most sensitive to the initial number of size C defects and least sensitive
to the initial number of size A defects. The number of defects per weld o f size D at t = 350 000 hr as a function
of Q, )~A, )~B and )'c is shown in fig. 15. The value Of PD is extre~nely sensitive to the generation and development rates.
Figure 16 displays the sensitivity Of PD to the percentage of the weld inspected per inspection. Between
lgi = 0% and Wi = 100% the expected defect concentration o f size D defects decreases by about three orders of
magnitude. The rate o f decrease of defect concentration is largest at very small values of Wi, and the rate decreases
monotonically thereafter.
The expected defect concentration o f size D defects at t = 350 000 hr is not extremely sensitive to the inspection
efficiency (fig. 17), in the range of efficiencies considered. The value of PD is displayed in fig. 18 as a function of
the number of inspections where the inspections are assumed to be equally spaced. The value Of PD decreases
monotonically at a decreasing rate as n increases. The decrease is over one order of magnitude between n = 1 and
n = 40. When n = 40 (and where W l = W2 . . . . .
W4o = 5%) the welds are inspected twice, completely, subsequent to t = 8 o hr with an efficiency o f less than 100%.
The most sensitive parameters are the generation and development rates. The second most sensitive parameters
are the initial defect concentrations. The two most sensitive groups of parameters are due to weld quality and
stressing conditions. With regard to more controllable parameters (i.e. those involved in inspections) the most
sensitive parameter is the percentage of the weld inspected and the second most sensitive parameter is the number
of inspections in 350 000 hr. Theqeast sensitive input parameter is the inspection efficiency, since the assumed
efficiency is only allowed to be perturbed by about -+10% or less of its initial value and as a result has little influence on the expected number of defects.
128
V ~ V V
~.sg z
~.
..~ o ~
O 0
,,~
x
o
~':~
x
m. m.m.
~.
eJ
ddod
<
0 0 0 0
~..,0
q~q
~ q q
o.
~o
g=
~g
o.
.It.
129
inspection, over sixtimes the portion of the weld is inspected for high stress welds than for low stress welds, the
large number of size A defects and the larger development rates cause PB and Pc to be 64 and 12% larger for high
stress welds, respectively. The value of Pz) for high stress welds is estimated to be about 20% less than for low
~tress welds using the base data.
EXPECTED NUMBER OF
DEFECTS PER WELD
PA(t)
PB( t )
PC ( t )
PD ( t )
B
104 --
lO1 _
lO-3-
lO~/
103 --
100 --
T O T O + T1
I
102 --
10-1 --
104 -
10-7
10-2 --
101
PD
2
e~0
Ti
PA
PB
PC
lOo
I0"3
10-5 I-
10.8
i
10"1
104
10-2
10-5
I
le 4s
10-9
I
100,000
200,000
TIME (HOURS)
300,00U
Fig. 19. The expected number of defects per weld o f sizes A, B, C and D as a function of time. Base run for high stress welds.
EXPECTEO NUMBER OF
DEFECTS PERWELD
PA(t)
104 -
PB(t)
101 -
Pc(t)
10-3 -
PD(t)
10-6 -
I
102
104
10-31
"Q ....g:
10-1
104!
10-2
10,5
10.5
10"8
10-6
10-9
~PApB
"""i
10-7
10"2 /
100
.....
11111.111111
..J
""
PC
PJ
21111,111111
TIME (HOURS)
300,000
Fig. 20. The expected number of defects per weld o f sizes A, B, C and D as a function of time. Base run for very high stress welds.
130
EXPECTED NUMBER OF
DEFECTSPER WELD
PA(I)
Ps(t)
104 --
101 -
103
100
lO2
to-1
101
10-2
100
Pc(t)
10-3 p
/
PD(t)
10"6 __
10-4 p
10"7
10"3
10-5 ~-
10"s
10-1
104
10-2
i0.5 I
10-6 L_
PD
illJ r
PA
PB
PC
pjd
10.9 I -,
#o d
f'
p#'
I
300.000
lOO,OOO
TIMEIHOURS)
Fig. 21. The expected number of defects per weld of sizes A, B, C and D as a function of time. Shorthand inspection intervals for
high stress welds.
EXPECTED NUMBER
OF DEFECTSPER WELD
PA( t )
PB( t )
PC( t )
PD( t )
100-
10"4 F
10"1 -
,o.sL/
10"3
,o,
"
I,,,1,111 ~,1 ~ ! ' I l l l
pBPD
. . . . _.
P,
,o r_
10o i
lO3-
10.4
lO-71-
10"2
10"5 -
10"5
10"9
.,#~.,"7.,, ,'~-
I
0
100,000
J
200,000
TIME (HOURS)
I
300,000
Fig. 22. The expected number of defects per weld of sizes A, B, C and D as a function of time. Only zeroth inspection, no defects
initially present, for high stress welds.
131
8"ii,l
~ 1o-5
10.7
10-5
10-4
10-3
10-2
10-1
100
PA(t = 0 HOURS)
Fig. 23. The number of defects of size D per weld at t = 350 000 hr as a function of the number of size A defects per weld initially
present. All other parameters are fixed as for base run, high stress welds.
~"
O 10.5
10 .6
,.o,
X
I
10-6
10.7
I
10-5
I
10 4
I
10-3
I
10-2
PB (t = 0 HOURS)
Fig. 24. The number of defects of size D per weld at t = 350 000 hr as a function of the number of size B defects per weld initially
present. All other parameters are fixed as for base run, high stress welds.
(n
re
10.5
1"6 ~-Ud W
I-o.
rJu~
cr,
Xw
w 10.7
10-6
10-5
10-4
10-3
10-2
Pc(t = 0l
Fig. 25. The n u m b e r of defects per weld of size D at t = 350 000 hr as a function of the number of size C defects per weld initially
present. All other parameters are fixed as for base run, high stress welds.
132
10"4
"ilo-E
10-6
10-7
10"9
10-6
10-11
10-10"
10-10
10-9
10-9
10"8
10-7
10"6
J
10-9
I
10"8
10"8
10-7
10-7
10"6
xA
xB
~'c
Fig. 26. The number of defects of size D per weld at t = 350 000 hr as a function of generation and development rates, high stress
welds.
~ S los
E ~ lO~
8 ag
~' lo7
10+8
0
I
20
I
40
l
60
I
80
l
100
(PERCENT) W
PERCENT OF WELD INPECTEO PER INSPECTION
Fig. 27. The number of defects of size D per weld at t = 350 000 hr as a function of the percentage of the weld inspected, for high
stress welds.
13 3
8-,
10.6
F0~ 0
~ 10"71
10"8 I
I
60
i
t
40
I
50
70
80
HiA(~) (PERCENT)
80
I
85
90
;i
95
100
HiB(I:) (PERCENT)
I
92.5
!
95
iI
i
i
97,5
i
100
HC(~) (PERCENT)
90
I
99.8
100
HiD(2:) (PERCENT)
99.9
EFFICIENCY
Fig. 28. The number of defects of size D per weld a~ t = 350 000 hr as a function of the percentage o f the weld inspected, for high
stress welds.
PD(t - 350,000 HOURS)
ii
< 10.7
,,I
10
15
20
25
30
35
NUMBER OF INSPECTIONS, EQUALLY
SPACED DURING 40 YEARS
40
Fig. 29. The number of defects of size D per weld at t = 350 000 hr as a function of the number of inspections, for high stress
wolds.
134
It would be interesting to compare the predictions of the model with the actual number of defects found in
high stress welds. In particular, the UK [ I 0] located three defects of critical size in approximately 120 000 high
stress welds (during 8.8 x 109 high weld stress hours). Since size D defects are considered to be in the failed state,
the probability of finding a size D defect is essentially equal to the probability of one existing. According to UK
data, the expected size D defect concentrations by the end of the weld life is 2.5 x 10 -s defects per high stress weld.
The UK data is based almost exclusively on non-nuclear pressure vessel welds. Since the non-nuclear pressure
vessel welds are assumed to be manufactured to less stringent standards and undergo less stringent maintenance
[10, 19, 20, 27], one would expect the nuclear welds to be less likely to have critically sized defects than the nonnuclear welds [9]. In fact, the UK non-nuclear high stress welds have about 250 x 10-7/5.1 x 10 -7 or approximately a 49 times higher expected concentration of size D defects than the concentrations predicted for US
nuclear vessel high stress welds using the model and the assumed base data.
There are several combinations of input parameters in the model that could be changed to increase the predicted defect concentration in nuclear vessel high stress welds to 49 times their predicted concentration at the
end of weld life. There is no unique combination. For example, increasing the concentration of size D defects at
t = 0 to 1 defect per high stress weld, keeping all other base run parameters fixed, will increase the expected size
D defect concentration at t = 350 000 hr by a factor of about 49 (see fig. 23).
Increasing the concentration of size B defects at t = 0 to 5 x 10 -3 defects per high stress weld, keeping all
other base run parameters fixed, will increase the expected size D defect concentration at t = 350 000 hr by a
factor of about 49 (see fig. 24).
There are a variety of other combinations of input parameters that could be varied in such a way as to increase
the expected concentration of size D defects at t = 350 000 hr (end of weld life) by a factor of 49. A combination
of weld integrity parameters (i.e. initial distribution of defect sizes, defect generation rates and defect growth
rates), are used in the model to simulate the number of defects of size D per weld that the UK found during the
weld life (i.e. 250 x 10 -7 size D defects in high stress welds). The assumed UK inspection standards are used. If
the ASME Section XI standards [27] are applied, then the expected number of size D defects per high stress weld
is reduced to about 48 x 10 -7.
As in the case with the low stress welds it has been assumed [9] that the US nuclear reactor pressure vessel
welds have ten times the integrity of the UK non-nuclear welds [10]. If this assumption were not made, then the
number of critically sized defects per weld (at the end of weld life) that would be predicted by the model for
nuclear vessel welds would equal about 50 x 10 -7 which is approximately equal to the number that would have
existed in the UK welds [10] had they used the ASME Code [30] during inspection. For the purpose of this
study it has been supposed that the assumption [9] (i.e. the US nuclear reactor welds have ten times the integrity)
is valid.
In fig. 20 the generation and development rates are increased by a factor of ten over the base case for low
stress welds. The result is that PD = 9.3 x 10 - 7 defects/weld (about 48% larger than for low stress welds). The
effect of the increased generation and development rates is apparent.
When the number of equally spaced inspections is doubled from 12 to 24, the value OfPD = 2.5 x 10 -7
defects/weld or about one-half of the value for Po for the base run using high stress weld data (fig. 21). When zero
defects are initially present and there are no subsequent inspections, then PD = 1.0 x 10 -4 defects/weld (fig. 22).
Figures 23-25 display the expected D size defect concentration at t = 350 000 hr as a function of the initial
concentration of size A, B and C defects, respectively. The value o f P D is most sensitive to the initial concentration
of size C defects and least sensitive to the initial concentration of size A defects.
The value Of PD as a function of generation and development rates is illustrated in fig. 26. The value of Po is
extremely sensitive to the generation and development rates especially for the larger rates.
The value Of PD varies between 2.1 x 10 -4 and 4.3 x 10 -s as the percentage of the weld inspected per
inspection (i.e. W), varies between 0 and 100% (fig. 27). The value Of PD is especially sensitive to changes in t
for small values of I~. The value of I is more influential in changing the value Of PD for high stress welds than
for low stress welds.
135
The sensitivity Of PD to changes in efficiency is depicted in fig. 28. Increasing the efficiency of detecting a
size A, B, C and D defect by 20, 10, 5 and 0.1%, respectively, reduces the value Of PD by less than 40%.
Figure 29 illustrates the expected concentration of size D defects at t = 350 000 hours as a function of the
number of equally spaced inspections n during that time period. The value of PD is more sensitive to changes in
n for small values of n. In general, Po is more sensitive to change s in input parameters for high stress welds than
for low stress welds.
_o
,,o,
75%
~ 25% -
10
20
30
NUMBER OF EQ4JALLY SPACED
INSPECTIONS IN 40 YEARS
40
Fig. 30. Recommended inspection procedures to achieve a minimum value OfPD(t = 350 000 hr) within a given cost constraint.
136
!
dfe~
_=
t~
~o
o
0
e~
=.
137
Fig. 30 illustrates the combinations of W and n that give this result (see table 15 for input parameters). It was
assumed that WI = W2 = W3 . . . . .
Wn and that the inspections were equally spaced.
This method can be extended. For example, if Cv < $1 000 000 then PD = 0.6 x l0 -7 defects per weld and a
new set o f n and W values are obtained.
Appendix
This appendix presents a portion of the raw data used in this study. The data is extracted from ref. [10]. Table A1
lists and describes the weld defects found in approximately 240 000 UK welds immediately prior to vessel usage.
Table A2 lists and describes defects found during service.
138
<
<
"~=
"0
es~
e.
~ ~'~= ~
o
0
:
,ff~
~.
Do
t'~'q"
f~
>
r~
~x~.~
~1~'~
~ x
~,~ 0
.
u
u
.8
.~. o~
t~o
.~
e"
o~
0
~3
*d
x
x
r~
x
oo
,q.
Urn,
x
~l ~
x
139
140
<
<
.<
<
<
t,4
0~,,,
..
o~
'~
O'
::
~
~E
0
u
"-
C
to
e~
~E
t~
.q-
'q.
t~
,q-
~D
e~..,
f~
x
0
e~
o
u
~E
e',x. ~
to
~x~
to
141
<
<
<
<
<
~d
4~
u
.) , ~
~n
~N
et
Lt~
e~
e~
~
~
~,~
~x
*
x.~
. ~
.~.
x,~ ~
, ~
. ~
142
~N
.j=.
o~
~o
.,~
*6
i
.=.
.~$=
.~
0
~
0
.=_
e~
O
_,g
8
I
.q
[...
(',i.
'1
,t
.=.
.=.
.=_
E~
m
,,...i
t~
oO
e~
."M
e..
E
eL.',~
i!
t~
~o~
~
x
.~" ~J
oo
~E
e~
,q.
~ =
143
s:
<
<
<
<
>.
oJ
O3
o~
e*
-d
-d
0
o~
O3
~.~
.=..~
"6
~
u~
,.~ .~ ~
f~
e~
e~
O3
ee
~ o~
O3
,q.
oO
~o
~o
t~
oo
o o tt~
~D oO
o~
oo
e~
x=
~ x
~D
o o . ~.
q ("4
o3 ,-~
p,
~.~
,q
o eq
~/3
~x
144
z.
e~
~'~
~1~ ~
X ~
eJ
,!
e-'
~l ~,
e.o
E
I
N
f.,i
et
._=
tt~
ao
r~
-6
0
m
t..
~
~l ~
.~
.,-
.~
145
.<
z=
z..
0
~ fl:
t~
~_~
e
~ o, ~
~ .~~
~'~
~.~
~ ~.~
"0
e 0
. ~
~
,q.
et
. 0 0 0
.~o
~o
~o
co
NO~
c~
,
~E
.~
~.
~x
~ x
~.=
x ~J~
2~
~'~ x
c~
o.~
146
LL~
w.q
r~
"0
t.,
--~
e~
8~
0
e~
t~
0
0"0
~ .~ o
0
o~
e;
t~
0
0
.<
t~
t~
[...
fq.
t,t
.=.
~o
to) tr~
~ o
~D
~D
p~
e~
t.
,.~ . i ~
~x
0
. e.
0
e~
el)
,g~
c~
e~
"0
"0
e~
e
e~
e)
-~
147
<
<
z.
0~..=
~=
~.>
o~
~1
.~
,~.~
~ 2~
.~.
::1 o
.~. "~ ~
~~
~.
s~
u.,
0
~.~ o
.~
~o~
.N "E
~ ~-
e~
~d
C~
I,,1
el.
t'~.
('1
m
gb
N$
~o
,=,,i
e~
......
o8
E~
~
. ~ . ~
~x
0
e.,
-s
.'~
~ "0 x
~'~
"
0
-I ~ "~
~x~
~0
~x
=g=2
N...
148
EL
=o .~ .o
0
o ~.~
e~
~ ~.~
2'~
"~
~ '~
S o=
~ ~ .~-
o .o .~
~
o
n~
CD
,.0~,0
--
~'~
~
~..~
x
~ -I~.
r~
149
Zp.
"-
e,
E . ao
~
o '
~
~-.~
O . ~ . ~ e ~
~_~ ~~.~ ~
,.
~ 0~ ~ ' ~ , ~~
"~
.~
. ~
.~ o.,~ ,..
~~.~~
~
,~.~
,,.I,
~.~
"Ex
~ ~'~
"cJ
E-IN
150
.d
c~
m
s~a
cd
~o
..
~o~
.~ o
o~
q)
c~m ~
.=.
~3
~o
C~
8~
t~
.~
03
o)
t~
~x
o
o~
~1 ~
~4
e~
[-
151
.<
z~
>
3
c~
0
~.~~ *
,~
~ - ~ N ~ -~e ~.~
..
I~" 0
"<.-~
c~ ~ 1
.'~'~
. .
c~
..o
~'~
3
E
c~
~ 3
C,
el.
e-
el.
~x
E~
_~>
0
a~
~10
U'3
152
e~
o~
CD
fq
fq
"c)
o
~1~ - I ~ - I e~
e~
c~
153
References
[ 1 ] L.E. Steele, Neutron irradiation embrittlement of reactor pressure-vessel steels, Reprint from At. Energy Rev. 7 (2), Vienna
(1969).
[2] L.E. Steele, Radiation embiittlement of reactor pressure vessel steels, Nuci. Eng. Des. 3 (1966) 287-298.
[3] J.R. Hawthorne and L.E. Steele, Initial evaluations of metallurgical variables as possible factors controlling the radiation
sensitivity of structural steels, Naval Research Laboratory, Report, 6420. Washington, D.C. (0000).
[4] T.W. Graham, The consideration of initial flaw size distribution and crack detection probabilities in fracture mechanics,
M.S. Thesis, UCLA (1973).
[5] USAEC, Ultrasonic Investigation, Topical Report, 15, AT(11-1)-2154, USAEC, Washington, D.C., Sept. (1973).
[6 ] R. O'Neil, Safety and reliability aspects of in-service inspection of pressure vessels, Safety and Reliability Directorate,
UKAEA, Risley, UK, June (1973).
[7 ] P.F. Packman, Status of nondestructive inspection techniques with a special reference to welding defects, US-Japan
Seminar on the Significance of Defects in Welded Structures, Tokyo, Japan, 15-19 Oct., 1973.
[8] R. O'Neil and G.M. Jordan, Safety and reliability requirements for periodic inspection of pressure vessels in nuclear industry,
Safety and Reliability Directorate, UKAEA, Risiey, UK, Mar. (1972).
[9] W.R. Stratton (of USAEC, Washington, D.C.), Letter and Appendix (Re: Integrity of Reactor Vessels for LWRs, Wash 1285)
to Dixy Lee Ray, Chairwoman, U.S. Atomic Energy Commission, Washington, D.C., 14 Jan., 1974.
[ 10] C.A.G. Phillips and R.G. Warwick, A survey of defects in pressure vessels, AHSB(S) R162, Health and Safety Branch,
UKAEA, Risiey, UK (1968).
[11] R. O'Neil and G.M. Jordan, Safety and reliability requirements for pressure vessel inspection, Conference on Periodic
Inspection of Pressure Vessels, London, 9-11 May, 1973.
[ 12] O. Kellermann, Present views on recurring inspection of reactor pressure vessels in the Federal Republic of Germany, IAEA
No. 81, IHEA Panel on Recurring Inspection of Reactor Steel Pressure Vessels, Pilsen, Germany, Oct. (1966).
[13] O. Kellerman, E. Kraegeloh, K. Kussmaul and D. Sturm, Considerations about the reliability of nuclear pressure vessels status and research planning, Paper 1-2, Pressure Vessel Technology, Part I, Design and Analysis, Second International
Conference for Pressure Vessel Technology, San Antonio, Texas, 1-4 Oct., 1973.
[14] O. KeUermann (Director, IRS, Federal Republic of Germany), Letter and Appendices (Re: Periodic Inspection of Pressure
Vessels) to R.F. Fraley, USAEC, Washington, D.C., 2 Jan., 1974.
[15] Gustof ~)stberg (Ab Atomenergi, Nykoping, Sweden), Letter and Appendices (Re: Pressure Vessel Weld Defects and
Inspection Procedures) to Spencer Bush, USAEC, Washington, D.C., 23 Jan., 1974.
[ 16] Spencer Bush (of USAEC, Washington, D.C.), Letter (Re: Pressure Vessel Weld Defects and Inspection Procedures) to
Gustof (3stherg, Ab Atomenergi, Nykoping, Sweden, 6 Feb., 1974.
[ 17] R.D. Wylie, In-service inspection of steel nuclear reactor pressure vessels, Reprint from At. Energy Rev., Vienna (1970).
[ 18] O. Kellermann, Recurring inspection of reactor pressure vessels in the Federal Republic of Germany, Report from The
Institut Ftir Reaktorsicherheit, FRG (1974).
[19] R. Sku61derbrand, The application of pressure vessel codes to nuclear reactor systems, Reprint from 150/TC85/SC3/WG6/27,
FRG (1968).
[20] Technical Report of the British Engine Boiler and Electrical Insurance Co. Ltd, Vol. II, Sept. (1954).
[21 ] Joint Task Group of the German Atomic Forum of the Institute for Reactor Safety and the Technical Surveillance Agencies,
Recommendations for Periodic Safety Inspections of Nuclear Power Plants, Berlin, Germany, 1968.
[22] K. Kussmaul (of the Laboratory for Testing of Materials, University of Stuttgart, Stuttgart, Germany), Beobachtungen an
Hochleistungs-Kesseltrommeln, Mitt. Ves B, Apr. (1967).
[23] W. Schoch (of the Laboratory for Testing of Materials, University of Stuttgart, Stuttgart, Germany), Berich uber die
Sufgatretenen Schaden an Kesseltrommeln, Mett. Ves B, Apr. (1969).
[24] G. Mieze, Analysis of German pressure vessel and boiler drum statistics, Third CREST Meeting of Specialists on the Reliability of Mechanical Components and Systems for Nuclear Reactor Safety, Denmark, 24-26 Sept., 1969.
[25 ] T.W. Graham and A.S. Tetelman, Materials Department, School of Engineering and Applied Science, UCLA, to appear in a
future issue of AIAA.
[26] P. Godbout, Nuclear power plant reliability, Ph.D. Dissertation, UCLA (1973).
[27] USAEL Technical report on analysis of pressure vessel statistics from fossil-fueled plant service and assessment of reactor
vessel reliability in nuclear power plant service, WASH-1318/UC-78, USAEC, Washington, D.C., May (1974).
[28] ASME, Rules for in-service inspection of nuclear reactor coolant systems, ASME Boiler and Pressure Vessel Code, Section XI,
New York, Jan. (1970).
[29] ORNL, The integrity of reactor pressure vessels, ORNL-NSIC-15/OC-80, Oak Ridge National Lab., Oak Ridge, Tennessee, May
(1966).
[30] K.A. Solomon, Reliability techniques applied to nuclear power plant systems, Ph.D. Dissertation, UCLA (1974).