NOVEMEER 1989

PREVENTION QF
CATASTROPHIC FA1LU RE
IN PRESSURE VESSELS
AND PIPINGS

Nordic
liaison committee for
atomic energy
 Nordisk Nordiska Pohjoismainen Nordic
kontaktorgan for kontaktorganet for atomienergia- liaison committee for
atomenergispørgsmål atomenergifrågor yhdyselin atomicenergy

PREVENTION OF
CATASTROPHIC FAILURE
IN PRESSURE VESSELS
AND PIPINGS

Final Report of the NKA-Project MAT 570

Rauno Rintamaa
KimWallin
Kari Ikonen
Kari Torronen
Heli Talja
Heikki Keinånen
Arja Saarenheimo
Fred Nilsson
Matt i Sarkimo
Stig Wåstberg
Christian Debel

November 1989
The present report is available on request from:

Technical Research Centre of Finland (VTT)

Metal Laboratory
Library
P.O.Box 26
SF-02151 Espoo
F i n l a n d

The Swedish Nuclear Power Inspectorate

Library
P.O.Box 27106
S-102 52 Stockholm
S w e d e n

A.S. Veritec
Stig Wastberg
P.O.Box 300
N-1322 Hovik
N o r w a y

Riso National Laboratory

Library
P.O.Box 49
DK-4000 Roskilde
D a n m a r k

ISBN 87-7303-348-0
NORD 1989:75

Grafisk Center Kronjylland, Randers 1989

ABSTRACT

The fracture resistance and integrity of pressure-loaded

components have been assessed in a Nordic research pro-
granune. Experiments were performed to validate the computa-
tional fracture assessment analysis. Two tests were also
conducted on a large decommissioned pressure vessel from an
oil refinery plant.
Different fracture assessment methods were developed and
subsequently applied to the tested components. Interlabora-
tory round robin programmes with the participation of sev-
eral laboratories were arranged to examine elastic-plastic
finite element calculations and fracture mechanics testing.
The transferability of material parameters derived from
small specimens with simple crack geometries to more real-
istic crack geometries in real components has been veri-
fied.

KEY WORDS: Fracture mechanics, numerical analyses,

engineering analyses, leak-before-break,
testing, compact tension specimen, flawed
steel plate, pipe, pressure vessel, acous-
tic emission, austenitic steel, ferritic
steel

This report is a part of the safety research programme

sponsored by NKA, the Nordic Liaison Committee for Atomic
Energy, 1985-1989. The project work has partly been financ-
ed by the Nordic Council of Ministers.
 111

SUMMARY

Cracking and subsequent catastrophic failure in pressure

vessels and piping systems has a significant impact on the
safety and reliability of operating process plants, includ-
ing nuclear and fossil fired power plants. The result has
been high economic losses, increased risks to personnel and
concern regarding safe plant operation. By using validated
fracture mechanics analyses and appropriate material prop-
erties, it is in most cases possible to determine if an
existing crack, due to operation or manufacture, will re-
main static or grow slowly. If a detectable leak occurs
prior to a complete break, catastrophic failure of a pres-
surized component can be avoided.
In order to assure the structural integrity of pressurized
components, reliable knowledge of the relevant material
properties must be available. Valid experimental and com-
putational fracture analysis methods åre then needed. In
addition, because hydrotesting is normally used for integ-
rity assessment of large pressure vessels, a crack can
produce catastrophic failure during a hydrotest. To prevent
this and to locate a crack or cracks, the accuracy of non-
destructive testing methods such as acoustic emission is of
major concern.
To improve the accuracy and validity of experimental and
computational fracture assessment methods, a four year
Nordic research programme was initiated 1985. The aim of
the programme was to clarify how catastrophic failure can
be prevented in pressure vessels and piping systems by
developing the necessary elastic-plastic fracture mechanics
analyses and by providing appropriate experimental data for
their validation.
The reliability of fracture mechanical computation and
materials characterization was. validated by two round robin
programmes on the elastic-plastic finite element calcula-
tion and fracture mechanics testing.
One laboratory each from Denmark, Norway, Sweden and Fin-
land participated in the computational round robin. In this
programme fracture mechanics parameters from one of the
experimental round robin test specimens were calculated
numerically and then compared to the experimental results.
There was little difference in the computational results
from each of the four laboratories. There was also good
agreement between experimental results, values calculated
using two-dimensional plane strain analysis and results
from three-dimensional calculations.
 IV

In the experimental round robin programme, fracture resis-

tance curves for two different materials were determined by
seven testing laboratories. The first material was ferritic
steel taken from the decommissioned pressure vessel that
had been used for the full-scale pressure vessel tests.
The second material was an aluminum alloy. Results indicate
that most of the round robin scatter was due to differences
in test performance between the seven laboratories, al-
though some scatter may have been caused by the macroscopi-
cally inhomogeneous steel. By analyzing the slightly dif-
ferent methods used by individual laboratories in perform-
ing the tests, data scatter was diminished.
One of the major uncertainties in the reliability of frac-
ture assessment analyses involves the transferability of
experimental results which åre normally obtained from small
specimens to geometries and loading conditions encountered
in actual components. To investigate this possible diffi-
culty, the fracture behaviour was assessed of 1) single-
edged notched specimens, 2) three-point bend specimens and
3) surface flawed plates in tension, in bending and in
combined tension and bending loading. Numerous experiments
and numerical calculations were made to this effect. The
material used in all experiments was the steel taken from
the decommissioned pressure vessel mentioned above.
Computational and experimental fracture assessment methods
were verified by component tests. Small straight pipes with
axial and circumferential flaws and a thin-walled pressure
vessel with a circumferential surface flaw were pressurized
to failure. Additionally, a major part of the programme
consisted of two tests using a decommissioned pressure
vessel having dimensions resembling those of a nuclear
reactor pressure vessel.
The engineering fracture assessment methods (Battelle's
limit load method, MPA's method) that were applied gave
reliable and conservative estimates for rupture pressure
and leak-before-break considerations in case of flawed
thin-walled pipes. The factor with the largest influence on
accuracy was the selection of a correct value for the flow
stress of the material. For large pressure vessels the
engineering R6-method developed in Central Electricity
Generating Board, U.K., predicted more stable crack growth
and lower rupture pressures than what was obtained in the
experiments.
Fracture behaviour of the large pressure vessels was simu-
lated more precisely by both elastic-plastic and geometri-
cally nonlinear analyses based on the finite element
method. The calculated strains, stresses and stable crack
growth estimates from the three-dimensional analysis agreed
well with the experimental findings.
 V

This project has produced new insights into the structural

integrity assessment of flawed pressurized components.
Deeper knowledge of the limitation, accuracy and applica-
bility of different fracture assessment methods was obtain-
ed. As a major outcome, extensive experimental data from
the full-scale pressure vessel tests and fracture mechanics
testing was produced for the validation of fracture assess-
ment methods and for their further development. The use of
acoustic emission testing during a hydrotest facilitates
the locating of an existing flaw during the early loading
stages. However, its application for crack growth monitor-
ing requires preliminary fracture mechanics consideration
before the correlation between the acoustic emission event
activity and crack extension can be reliably expressed.
Typical Industries which may benefit from the results åre
process Industries including energy production as well as
off-shore installations. The knowledge and capabilities of
experimental and computational fracture assessment that
were produced in this project can be applied for the design
of new components and when considering the safety and
reliability of main components in operating plants, thereby
reducing the risk of catastrophic failure.
 vi

SAMMANFATTNING

Instabilt brott av tryckkarl och rorledninger fororsakar

betydande ekonomiska forluster och sakerhetsanknytna risker
inom t.ex. energiproduktion och processindustri. Genom
lampligt val av material och ratt dimensionering av kom-
ponenten ar det ofta mojligt att garantera att ett existe-
rande fel, som uppkommit vid fabrikation eller under an-
vandning, inte våxer eller våxer sakta på ett stabilt satt,
fororsakande lackage fore brott. På detta satt kan kata-
strofalt brott av en tryckbårande komponent undvikas.
Såkerhets- och hållfasthetsanalyser av tryckbarande kompo-
nenter kraver pålitlig kunskap angående de relevanta mate-
rialegenskaperna samt noggrannheten av experimentella och
numeriska berakningsmetoder. Dessutom utfors ofta tryckprov
for såkerhetsgranskning av stora tryckkarl. En spricka i
ett dylikt tryckkarl kan fororsaka katastrofalt brott åven
under tryckprovet. For att forhindra detta och for att
lokalisera en spricka eller sprickor bor akustisk emission
provning och annan instrumentering kunna anvandas.
For att forbattra noggranheten av experimentella och nume-
riska hållfasthetsanalyser, initierades ett fyra-årigt Nor-
diskt forskningsprogram, under ledning av Nordiska Kontakt-
organet for Atomenergifågor, år 1984. Det huvudsakliga
tekniska målet i programmet var att klarlågga hur kata-
strofalt brott kan undvikas i tryckkarl och rorledningar.
Den experimentella delen av programmet omfattade provning
av små brottmekaniska provstycken, grova plåtar med sprick-
or, ror och fullskale tryckkarl innefattande en betydande
instrumentering och materialkaraktarisering. Den numeriska
delen bestod av utveckling samt verifikation av brottmeka-
niska analysmetoder och en detaljerad analys av experimen-
ten som utfordes inom programmet.
Tillforlitligheten av brottmekaniska numeriska analyser
samt materialprovning undersoktes via jamforelseprovning. I
den numeriska jåmforelseprovningen analyserades ett av
provstyckena från den experimentella jåmforelseprovningen
av fyra deltagare från Danmark, Norge, Sverige och Finland.
Jåmforelseprovningen var tvådelad. I forstå delen utfordes
två-dimensionella berakningar med antagande av olika spånn-
ingstillstånd. I andra delen utfordes ingående tre-dimen-
sionella berakningar for en basta estimatanalys av prov-
stavens beteende.
 Vil

I den experimentella jåmforelseprovningen deltog sju prov-

ningslaboratorier. Jamforelseprovningen utfordes två ganger
med två olika material. Materialen var en aluminiumlegering
samt ett ferritiskt stål som tagits ur fullskale-tryckkår-
len.
Grova plåtar med kantspricka, trepunkts bojprovstavar och
plattor med ytspricka testades till brott under olika typer
av belastning. Brottbeteendet undersoktes med ett flertal
experimentella metoder. Numeriska beråkningar utfordes for
att verifiera overforingen av resultat med små provstavar
till mera realistiska sprickgeometrier i storre provstavar.
Numeriska och experimentella brottmekaniska beråkningsmeto-
der verifierades genom provning av små raka ror med axiella
och longitudinella sprickor, ett tunnvaggigt tryckkarl med
longitudinell spricka, och utgorande huvuddelen av program-
met, två provningar av stora tryckkarl vilkas dimensioner
motsvarar reaktortryckkarl. Under provens gang kunde brott-
beteendet registreras via en omfattande instrumentering av
tryckkårlen. Instrumenteringen av området intill sprickan
gav realtidsinformation om sprickans beteende. Akustisk
emissionsmåtning (AE) antydde sprickans utveckling redan
vid ett tidigt skede av provet, langt innan sprickan bor-
jat våxa. Proven analyserades med olika ingenjorsmåssiga
berakningsmetoder.
Mer exakta berakningar av brottbeteendet utfordes for de
stora tryckkarlen. De beråknade tojningarna, spånningarna
och mangden stabil spricktillvaxt stamde val overens med
de experimentella resultaten i fallet av tre-dimensionella
berakningar. Fore utforandet av en ingående tre-dimensio-
neli finit elementanalys, maste man forvissa sig om rått-
fardigheten och betydelsen av alla randvillkor som påverkar
det slutliga resultatet.
Projektet har producerat nya kunskaper angående såkerhets-
och hållfasthetsanalyser av tryckbarande komponenter. Genom
anvandande av resultaten från de experimentella jamforel-
seprovningarna kan man utvardera olika osakerhetsfaktorer
vid brottmekanisk provning. Anvandande av akustisk emis-
sion vid tryckprov mojliggor lokalisering av existerande
sprickor i ett tidigt skede. Det mest betydande resultatet
har varit en djupare kunskap betraffande tillgangligheten
och anvåndbarheten av olika brottmekaniska berakningsmeto-
der.
Typiska industrier som gagnas av och som kan dra nytta av
resultaten ar basindustrier for energi, olja, petroleum,
kemi osv. Fortida, framforallt katastrofalt brott i huvud-
komponenten i ett dylikt industriverk fororsakar betydande
ekonomiska forluster samt sakerhetsanknutna risker. En
 VJL11

saker funktion ar nodvåndig under hela komponentens livstid

for en del komponenter såsom tryckkårl i karnkraftverk och
kemiska reaktorer i raffineringsanlaggningar, turbinrotorer,
hb'gtrycksrorledningar med explosivt eller giftigt innehåll.
M.a.o. år forsåkrandet av låckage fore brott -beteendet
viktigt i dylika komponenter och resulterar i betydande
besparingar och okad såkerhet.
 IX

C O N T E N T S
Page
1. INTRODUCTION l

2. RELIABILITY OF FRACTURE MECHANICAL

COMPUTATION AND MATERIALS CHARACTERIZATION 3

2.l Description of numerical J-inteqral round robin

programme 3
2.1.1 Description of the problem and solutions 3
2.1.2 Results and discussion 3
2.2 Description of experimental J-R round robin
programmes 5
2 .2 .1 The first round robin 5
2 .2 .2 The second round robin 7
2 .2 .3 Results and discussion 8
2 .3 Conclusions on the reliabilitv 10
3 FRACTURE ASSESSMENT OF SURFACE FLAWED PLATES
IN TENSION AND BENDING 13
3 .1 Experimental procedure 13
3 .2 Experimental results and discussion 15
4 FRACTURE ASSESSMENT OF FLAWED PRESSURE
COMPONENTS 17
4 .1 Pipe tests 17
4 .1 .1 Experimental procedure 17
4 .1 .2 Analytical calculations 17
4 .1 .3 Results and discussion 18
4.2 Full-scale pressure vessel tests 21
4.2.1 Experimental procedure 21
4.2.2 Analytical and numerical calculations 24
4.2.3 Results and discussion 29
4.3 Conclusions on the applicability of fracture
assessment methods 43
5. RECOMMENDATIONS 45

6. REFERENCES 47
 - l-

l. INTRODUCTION
The present final report summarizes the most important
results of the NKA project MAT 570 elastic-plastic fracture
mechanics. The program plan and results åre described in
more details in the publications and other reports and
working documents [l - 21].
The main objective of the project [1] was to improve the
efficiency and economy of the energy production and process
industry by optimizing the design of pressure retaining
components which would result in fewer and shorter shut-
down periods. The main technical objective was to clarify
how catastrophic failure can be prevented in pressure ves-
sels and pipings by considering particularly the leak be-
fore break concept.
Safety and integrity assessments of pressure vessels can be
performed using complete 3D FEM analysis but the cost may
become unacceptably large. Therefore different engineering
analysis methods for such assessments åre in use today.
Examples of such methods åre R6 and more advanced FAD
methods, Kiefner's method and crack driving force diagrams
in conjunction with EPRI Handbook formulae or even 2D FEM.
The accuracy and validity of the engineering assessment
methods is most appropriately evaluated by comparing the
calculated results with those from experiments. The com-
putational fracture assessment, on the other hånd, is based
on the reliable determination and application of the rele-
vant material property parameter.
The participants in the elastic-plastic fracture project
were Ris^, Dantest and the Jydsk Teknologisk Institut from
Denmark, the Technical Research Centre of Finland (VTT),
Neste Oy and Imatra Power Co. (IVO) from Finland, Veritec/
Veritas Research and SINTEF from Norway, and the Royal
Institute of Technology from Sweden. The project leaders in
each tasks åre given on page 3 of the cover.
The project was funded by the Nordic Council of Ministers,
Technology Development Centre (TEKES), Ministry for Tråde
and Industry in Finland (KTM), Finnish Centre for Radiation
and Nuclear Safety (STUK), Swedish Nuclear Power Inspec-
torate (SKI), Neste Oy, Imatra Power Co. (IVO), Helsinki
Energy Board (HKE) and Technical Research Centre of Finland
(VTT).
 - 3 -

2. RELIABILITY OF FRACTURE MECHANICAL COMPUTATION AND

MATERIALS CHARACTERIZATION

Fracture mechanics is commonly used for assuring the safety

of flawed structures and components. The assessment is
based on the calculation of an appropriate fracture mechan-
ics parameter in a load-carrying component and on the com-
parison of it to the parameter determined by fracture
mechanics testing. To verify computational and experimental
fracture assessment methods, round robin programmes were
arranged on elastic-plastic finite element calculation [4]
and on fracture mechanics testing [5, 6].
2.l Description of numerical J-integral round robin
programme
One of the CT-specimens tested in VTT's Metals Laboratory
for the experimental round robin was chosen to be simulated
in the numerical calculations.
The participants in the numerical J-integral round robin
programme were Kungliga Tekniska Hogskolan in Stockholm,
Ris$ (Denmark), Technical Research Centre of Finland (VTT)
and Veritas Research (Norway).
2.1.1 Description of the problem and the solutions
The numerical round robin for the side-grooved CT-specimens
consisted of two parts. During part one, two-dimensional
finite element calculations using both plane stress and
plane strain were made. For best estimate analysis of the
specimen behavior, three-dimensional analyses were per-
formed. Only two participants performed three-dimensional
analyses, both with 20-noded volume elements. One of the
three-dimensional analyses was made considering the side-
grooves and in the other a smooth specimen with the total
thickness was modelled.
The calculations were made keeping the crack size constant.
The participants were asked to perform the analyses by
imposing the load as a prescribed load point displacement.
As a result of the calculation, the values of load, crack
mouth opening, crack tip opening and J-integral values were
reported. In the case of three-dimensional analysis, J-
integral values were given as variations over the crack
front and as Integrated average values from the crack
front.
2.1.2 Results and discussion
The differences between the two-dimensional results of
different participants were rather small (Fig. 1). The most
remarkable deviations, especially when the plane strain
solutions åre considered, were found in the solution where
linear 4-noded elements were used in spite of second order
 - 4 -

elements. This model was stiffer than the other ones,

though the number of degrees of freedom was the largest.
Good agreement was obtained when the calculated plane
strain results were compared to the net thickness normal-
ized experimental results. Plane stress results were ap-
proximately the same as but slightly less than the effec-
tive thickness normalized experimental result. Three-dimen-
sional results from the two laboratories were very close to
each other and they agreed very well with the experiment
(Fig. 2).
Calculated J-variations along the crack front åre remarkab-
ly different in smooth and side-grooved specimens. When
load, crack tip opening or average J-integral values in-
tegrated from the crack front åre considered, the agreement
between the calculated three-dimensional results and the
experimental ones was very good, though one of the calcula-
tions was made without considering the side-grooves.

-o- exp. with net th

-*- exp. with eff. th
-o- participant 1
-«- participant 2
-*• participant 3
-o participant 4

CMOD fmml

Fig. 1. J-integral as a function of load line displace-

ment, results from two-dimensional plane strain
calculations with the experimental ones [4].
 - 5 -

J [kN/m]
400

300-

200-

-D- Experiment
-•- Participants
100 - -D- Participant 2

CMOD [mm]

Fig. 2. Average J-integral from the crack front as func-

tion of load line displacement, results from
three-dimensional calculations with the measured
ones [4].

2.2 Description of experimental J-R round robin proqrammes

Experimental elastic-plastic fracture toughness testing has
during recent years become more and more common in the
Nordic countries. Most experience has been obtained with
the British CTOD test procedure. CTOD is not, however,
often used in nuclear structural integrity assessment. The
usual parameter in nuclear safety assessment is the J-in-
tegral, but the J-integral based J-R-curve test is not
widely used in all Nordic countries. In order to ensure the
reliability of Nordic J-R-curve testing two special Nordic
J-R-curve round robin test programmes were carried out.
2.2.1 The first round robin
The first J-R-curve round robin test programme [5] had
seven Nordic materials testing laboratories participating.
Each laboratory was supplied with material for ten 25 mm
CT-specimens. The laboratories had to manufacture their own
specimens and were free to perform the actual testing in
 - 6 -

any way they found convenient. The only demand was to meet
the requirements in the ASTM JiC-standard E813-81.
The participating laboratories were: Jysk Teknologisk and
Ristf from Denmark, Veritas Research and SINTEF from Norway,
Royal Institute of Technology (KTH) from Sweden and VTT
from Finland (Metals Laboratory and Reactor Laboratory).
The material used in the round robin was a 2 1/4 Cr l Mo
steel taken from a decommissioned reactor pressure vessel
of a Finnish oil refinery plant and used for the full-scale
pressure vessel tests. The six laboratories called labora-
tory l - 6, were supplied with one of 6 pieces cut from the
pressure vessel. The laboratories fabricated 25 mm thick
standard CT test specimens from these plates.
One laboratory (laboratory 7) joined the round robin test-
ing later and fabricated specimens from material extracted
from a place located below the section used for the other
laboratories. The specimens were numbered in the same man-
ner as the others. Laboratory 6 tested ten additional
specimens to perform a key-curve calibration.
These specimens were taken from the same location as
specimens for laboratory 7. Laboratory 6 also tested 105
specimens to get a good multispecimen reference J-R-curve.
Preliminary tests on the material indicated upper shelf
behavior at a temperature of +50 °C. The room temperature
0.2 % proof strength and ultimate strength were
Ri.0.2 = 300 MPa
Ru = 532 MPa.
The round robin test temperature was chosen as +50 °C and
the flow stress to be used in the analysis was estimated to
be 416 MPa.
All laboratories applied essentially single specimen type
methods to evaluate the J-R-curves. Mainly two types of
methods were used, i.e., partial elastic unloading compli-
ance and direct current potential drop. Besides differences
in the testina methods, there were also differences in the
performance and the analysis of the tests. The main dif-
ferences åre presented in Table 1. In order to minimize the
differences, all the raw data was later reanalyzed with the
multiple specimen method as well as a modified key-curve
analysis method [22].
 - 7 -

Table 1. Difference in the laboratories test performance.

Laboratory
1 2 3 4 5 6 7

Method comp. comp. comp. DC-PD DC-PD comp. comp.

Flat bottom
holes yes yes yes no yes yes yes

Close fitted
pins yes yes yes yes yes no no

SZ included
in ia yes yes yes yes no yes ves

ao identified
from comp. comp. comp. phys. phys. .phys. phys.

Rotation correction
of compliance no yes no - - yes yes

Crack growth
correction of J no no yes no no yes yes

Negative crack
growth correction yes no no no no no no

comp. = compliance

2.2.2 The second round robin

The second round robin [6] had the same participants as in
the first round robin with exception of the Reactor Labora-
tory at VTT. The specimen geometry was the same as in the
first round robin. Contrary to the first round robin, each
laboratory was supplied with ready made specimens. Also the
primary test method was fixed as being based on the multi-
ple specimen technique. As a secondary method, each labora-
tory was free to use an additional crack length monitoring
technique so long as the requirements in the ASTM JiC-stan-
dard were not violated.
Two different materials were used in the round robin.
Firstly, a 2 1/4 Cr l Mo steel taken from the reactor pres-
sure vessel used for the second full-scale pressure vessel
test (HC2). Secondly, an aluminum alloy ISO AlSilMg sup-
plied by Norsk Hydro. The aluminum specimens were fabri-
cated by SINTEF and the steel specimens at VTT. The speci-
mens were divided randomly between each laboratory, in
order to limit the effects of possible macroscopic ma-
terial inhomogeneities.
The room temperature tensile properties for the two må-
terials åre presented in Table 2.
 - 8 -

Table 2. Tensile properties of round robin materials.

Material a0.2 [MPa] au [MPa] E [GPa]

2 1/4 Cr 1 Mo 329 522 200

Al Si 1 Mg 318 338 70

The testing temperatures were chosen as +50 °C for the

steel and room temperature for the aluminum alloy.

2.2.3 Results and discussion

The primary results of the first round robin were not at
all promising [5], A centralized key-curve based reanalysis
of the data made the situation somewhat better, but still
the results showed a large amount of scatter. When the
results åre treated by multiple specimen analysis (Fig. 3)
and fitted with a square root expression J = C/Aa the mean
and standard deviation of C is 489 and 58 kJ/ (rn^mm) ,
respectively (Fig. 4). This means that the standard devia-
tion of C is less than 12 %. Such a scatter could well be
attributed to material inhomogeneity considering that the
material had been manufactured and been put in use more
than 20 years ago.
Some questions remained, however, as to how much of the
scatter was due to material inhomogeneity and how much to
test performance variation.
The additional 105 tests performed for the multiple speci-
men reference J-R-curve showed approximately half of the
round robin scatter. This would indicate that part of the
round robin scatter is due to differences in test perfor-
mance between the different laboratories. However,since
105 specimens were extracted from one place whereas the
round robin material was more scattered, it is possible
that the pressure vessel material is macroscopically in-
homogeneous. This could explain the larger amount of scat-
ter in the round robin results. But differences due to the
laboratories could not be ruled out. The primary results
indicated that some laboratories had severe friction prob-
lems in their tests. Such friction will automatically af-
fect the load-displacement behavior causing excessively
high J-values and inflated J-R-curves to be measured.
The results of the second round robin [6] confirmed that
the main source for scatter and systematic differences is
the test performance.
 - 9 -

l
Multispecimen

1000

E
*-^

3 500

x Outer surface specimen

J________l_______l
" 0 1 2 3 4 5
Aa (mm)

Fig. 3. Multiple specimen presentation of first round-

robin results.

/uu l l l 1 1 l l

O
600 h <r f
• A

— 500
Ul
m
U • ^s
£ 400

| 300 1
O Outer surface specimen J = C • /Aa
U 200
Aa m a x > 0.3 mm

100

o l l l 1 1 I 1

LABORATORY

Fig. 4. Square root fit of first round-robin end point

results.
 - 10 -

It is interesting to note that the simple square root fit

yielded less scatter than the more refined power law fit,
for all examined parameters. It was found that the power
law fit parameters had an inverse relationship. The same
trend is to be seen in the ASTM E318-81 JIC and Tm^t. This
indicates that the power law fit and the ASTM E813-81 pro-
cedure does not describe the J-R-curves unambiguously. It
is thought that it might be valuable to further investigate
the possibility of using a simpler and more straight for-
ward fitting procedure such as the square root fit. With
this kind of one parameter fit the description of the J-R-
curve is simple and the result would be closer to a mean
description of the curve. Whether a square root or some
other value of the exponent fits the data better is open
for discussion. The important thing is that a one parameter
fit makes J-R-curve data much more readable.
2.3 Conclusions on the reliabilitv
The scatter of the results of the Nordic finite element
round robin was much smaller than in the round robin [23]
arranged by the European Group on Fracture in 1983. One
explanation that in the European round robin the number of
participants was much larger. The analysis expertise and
capabilities might also have been improved in recent years.
One can draw conclusions on the J-integral testing pro-
cedures by comparing the experimental J-values to calcu-
lated average values from crack front. Experimentally and
numerically derived J-values were in good agreement as were
the loads at a defined load point displacement. This is not
surprising since experimental J-values åre based directly
on the measured load vs. crack mouth opening result. Com-
parable accuracy of both load and J-values confirms the
applicability of the testing method. Another possibility to
consider in the testing procedure is to use the numerical
values calculated for a side-grooved specimen, and to com-
pare average J-integral values with J-integral values cal-
culated from experimental load and CMOD results. When the
calculation is repeated using both the net thickness and
the effective thickness, one can draw conclusions as to,
which thickness value is the most appropriate in elastic-
plastic J-integral testing. The present analysis confirms
the use of net thickness which has been used during experi-
mental measurement (Fig. 5).
 - 11 -

J [kN/m]
300

-B- J (kN/m)
-»- J-net(kN/m)
-«- J-eff(kN/m)
200 -

100 -

CMOD [mm]

Fig. 5. Comparison of average J-integral from the crack

front in the side-grooved specimen with the values
calculated from load vs. load point displacement
result as a function of load line displacement
[4].
 - 13 -

3. FRACTURE ASSESSMENT OF SURFACE FLAWED PLATES IN TENSION

AND BENDING

3.l Experimental procedure

The most widely used parameter for characterizing the crack
tip state under elasto-plastic conditions is certainly the
J-integral. Numerous experimental investigations have been
carried out to explore its relevance to crack growth de-
scription. However, the majority of these studies have been
concerned with a few geometries, suitable for laboratory
work and materials testing. Much research remains in order
to verify the transferability of results obtained from such
specimens to geometries encountered in actual failure as-
sessment situations.
The main object of the investigations [7] at the Department
of Solid Mechanics at KTH was to compare results obtained
from typical laboratory specimens with those obtained from
experimentation on larger specimens with more realistic
crack configurations. Another question that has not been
sufficiently studied is the effect of more complicated
loading situations. In most experimental investigations
only one loading system is applied so that at least ap-
proximately a state of proportional loading prevails. In
the present work load history effects of this kind åre
investigated for three-dimensional crack configurations.
The material used in all experiments was the 2 1/4 Cr l Mo
steel taken from the decommissioned pressure vessel used
for the full scale testing of this project. All tests here
were performed at room temperature (~ 21 °C). The specimens
were prepared so that in all cases the crack front was
perpendicular to the rolling direction.
The two-dimensional test specimens were of two båsic types,
a single-edged notched tensile specimen and a three-point
bend specimen. In order to study the effects of side-
grooves, some of the specimens were grooved. The thickness
of the specimens was varied, the normal alternative had a
thickness of 25 mm, while a variant with a thickness of
50 mm was also tested. In order to maximize the difference
between the tensile and bend specimens, the tensile ones
were loaded by damping the edges with jaws that prohibited
rotational movements. As a result of this constraint a
reaction in form of a bending moment tending to close the
crack was present besides the tensile load. This moment was
determined by strain measurements on gauges placed along a
line near the edge of the specimen.
The loading of the specimens was done in accordance with
ASTM E-813 with partial unloadings used to facilitate the
crack growth determination by the compliance technique. J
was evaluated for the experiments from the experimentally
observed data. The experimental determination of J for the
 - 14 -

bend specimens was performed according to ASTM E-813. For

the SEN-specimens there is no standardized procedure for
experimental J-evaluation and a procedure developed by
Kaiser for determination of J under combined tension and
bending was used.
A series of four experiments with a more complex crack geo-
metry and loading was performed. The specimen is shown in
Fig. 6 and is a 50 mm thick plate that is somewhat curved,
the midpoint of the cracked section being offset a distance
e from the midpoints of the ends of the specimen.

Fig. 6. The specimen for the surface crack testing.

The object of this design is to induce a bending moment
tending to open the crack by tensile loading of the speci-
men . The specimen ends åre welded to circular thick plates
which in turn åre mounted by screws along the periphery to
a servo-hydraulic testing machine. In addition to the ten-
sile loading, the specimen is mounted to the machine with a
initial slant angle 90 which induces an initial bending
moment. Thus, the specimen is subjected to two loading
systems. The applied tension can be regarded as a primary
loading which induces both tensile and increasing bending
loads. The initial displacement can be considered as secon-
dary bending load. By varying the offset e and the angle 60
different M-N histories can be obtained where M and N åre
the moment and the membrane loads respectively as referred
to the cracked section. One of the objects of this inves-
tigation is to study whether the path in the M-N-diagram
has any influence on the crack growth initiation behavior.
In the course planning, a simple beam model of the system
was considered. The M-N histories for different e and 60
were obtained by running this model with the program system
 - 15 -

ABAQUS taking plastic behavior and large deformation ef-

fects into account. In addition to these four experiments
two tests were made with the offset e equal to zero. One of
these were tested under pure tensile loading while the
other was tested under three-point bending.
The crack was produced by first machining a semi-circular
slit perpendicular to the surface. After fatigue loading in
bending the cracks assumed approximatively semielliptical
shapes. The depth was kept around 25 mm and the surface
length around 85 mm.
During the testing the specimens were instrumented with
strain-gauges at different locations for measurement of
loads and crack length. The crack length measurements were
made by partial unloadings and strain measurements in the
vicinity of the crack. In addition a color-marking techni-
que was employed. Different colours were injected into the
crack at some occasions during the loading process.
3.2 Experimental results and discussion
In the two-dimensional tests stable crack growth generally
occurred. The agreement between the tensile and bend tests
without side-grooves was fairly good in view of the dif-
ficulties in obtaining reliable crack lengths for the ten-
sile tests. The side-grooved bend tests, however, differed
from the smooth tests in that the JR-values were markedly
lower than for the smooth specimens.
In the six experiments with surface cracks, unstable crack
growth followed immediately after initiation of crack
growth in all experiments but one. The result of the previ-
ously mentioned beam type calculations for the experiments
with the offset e greater than zero åre shown in Fig. 7.
together with the behavior observed during the actual ex-
periments . As can be seen the agreement between the pre-
dicted and the observed behavior is good, except perhaps
at the highest load levels where plastic effects relax the
bending moment very rapidly. An observation that can be
made from the figure is that the path to the critical point
in the M-N space is of minor importance. The presence of a
secondary bending of the present nature here seems to have
little importance for crack growth initiation under duetile
conditions.
An evaluation of tests by the R6-method was performed using
global collapse load solutions and the Kic-value measured
from the two-dimensional tests with side-grooves. The R6
option l was assumed and the results from the evaluation
åre shown in Fig. 8. As can be seen from the figure some
scatter is evident and also that some points fall below the
assessment line. By comparing experiments 5 and 6 it seems
that the R6 procedure of accounting for secondary stresses
is not satisfactory. From Fig. 7 it is seen that these two
 - 16 -

specimens failed at approximately the same instantaneous

loading conditions although the initial bending was very
different. In the R6-procedure this initial difference
causes the points for 5 and 6 to fall widely apart in the
assessment diagram.
In addition to this evaluation, more accurate three-dimen-
sional elasto-plastic FEM-calculations with the program
system ABAQUS åre in progress.

Crackdim
a \mm
48.8
46.5
44.5
45.0
49.2
50.0

——— Experiment
— — Beammodell (FEM)
X Failure
O Initiation

2.50 3.50 4.50

Force [MN]

Fig. 7. Moment versus force histories for complex testing

of flawed plates [7].

Fig. 8. Failure assessment by the R6-method [7]

 - 17 -

4. FRACTURE ASSESSMENT OF FLAWED PRESSURE COMPONENTS

To verify computational and experimental methods used for
fracture and catastrophic failure assessment, pipes and
full-scale pressure vessels were tested up to rupture.
4.1 Pipe tests
Pipe tests [9, 10] were performed by Imatra Power Co's
Energy Laboratory in Helsinki during years 1985 and 1986. A
total of 122 straight small diameter and thin-walled pipes
representing two materials were tested. All pipe materials
were ductile. The first test pipes were without flaw. The
flawed pipes were made by machining axial and circumferen-
tial flaws in the inner and outer surface of the pipe
walls. Circumferential flaws were made also by welding two
pipes together giving a flaw geometry more similar to a
natural sharp crack than in cases where the flaws were
machined. The radius of the machine crack bottom was about
1.1 mm in most cases. The load was pure internal pressure,
but in some pipes also a bending moment relative to the
internal pressure was superposed. Some tests were performed
with pure bending moment load only.
4.1.1 Experimental procedure
The length of the pipes was about half a meter and the
pipes were closed at the ends by flat plates. The pipes
were pressurized by water, but because of the small volume
of water and rapid decrease of the pressure after a burst
of the flaw, nitrogen was also often introduced into the
pipes before the water. In this way it was possible to
simulate the stability of a flaw in long steam containing
pipes, where the pressure decrease is not rapid. Because of
the limited instrumentation, only the burst pressure of
each pipe was measured. In those cases with leaks, the leak
area was measured after test completion. Stress strain
curves for each materials were determined after pressure
testing.
4.1.2 Analytical calculations
Analytical considerations [8] were done by applying the
Battelle's limit load method [24] and also in some cases
MPA's method [25]. In addition to the rupture pressure, the
stability of the burst surface crack was estimated and
compared to the experimental findings. In cases where the
critical pressure, i.e. the pressure at which a crack be-
came unstable, of the surface crack was greater than that
of a trough-the-wall crack, catastrophic failure occurred.
In the absence of a pressure drop, the crack would extend
unstably at both ends.
 - 18 -

4.1.3 Results and discussion

From the results it can be concluded that if the material
is ductile, as it was in all the pipe tests, one can ac-
curately estimate the rupture load. The stability of a
burst crack can also be determined by using Battelle's and
MPA's limit load methods. Some of the results åre presented
in Figs. 9 - 1 3 , where the lines indicate rupture loading
ratios (stress in the ligament normalized by the flow
stress of a material) åre shown as a function of crack
geometry (crack length C/radius of a pipe R). The lines for
through-the-wall crack indicate the boundary of instability
for the through-wall cracks of same length as the surface
crack. More extensive description of the piping programme
has been presented in the report by Ikonen et al. [8].

PRESSURE LOADING. CIRCUMFERENTIAL FLAWS

P I P E SIZE 60.3 • 2.9 MM

o stable tests
• unstable tests
O stable & unstable
tests

Fig. 9. Results from tests, where pipe material was Fe

37B, flaw was circumferential and pipes were
loaded by pure internal pressure. The line of the
through crack divides the area into stable and
unstable areas [8].
 - 19 -

PRESSURE L O A D I N G , C I R C U M F E R E N T I A L FLAWS
P1PE SIZE 76.1 • 3.65 MM

O stable tests
• unstable tests

Fig. 10. Results from tests, where pipe material was AISI
304, flaw was circumferential and pipes were
loaded by pure internal pressure. The line of the
through crack divides the area into stable and
unstable areas [8].

PRESSURE LOAOING. A X I A L FLAWS

PIPE SIZE 60.3 " 2.9 MM

O stable tests
• unstable tests
stable & unstable
tests

Fig. 11. Results from tests, where pipe material was FE

37B, flaw was axial and pipes were loaded by pure
internal pressure. The line of the through crack
divides the area into stable and unstable areas
[8].
 - 20 -

PRESSURE L O A D 1 N G , AX1AL FLAWS

P1PE S I Z E 7 6 - l • 3 . 6 5 MM

O stable tests
• unstable tests

Pig. 12. Results from tests, where pipe material was AISI
304, flaw was axial and pipes were loaded by pure
internal pressure. The line of the through crack
divides the area into stable and unstable areas
[8].

B E N D I N G MOMENT LOADING. CIRCUMFERENTIAL FLAWS

P I P E SIZE 60.3 • 2.9 MM

stable tests
unstable tests

CEGFND
O THR. CRACK
A A/T-0.2
A/T-O.<
X A/T-0.6

Fig. 13. Results from tests, where pipe material was FE

37B, flaw was circumferential and pipes were
loaded by bending moment. The line of the through
crack divides the area into stable and unstable
areas [8].
 - 21 -

4.2 Full-scale pressure vessel tests

Theoretical and experimental fracture assessment methods
were verified by performing three tests with full-scale
pressure vessels. The first test [11] was carried out in
Espoo at VTT in winter 1986, the second [12] and third one
[20] in Skoldvik at Neste Oy's Oil Refinery Plant in Sep-
tember 1986 and in August 1988. The experiments were plan-
ned and conducted by VTT.
4.2.1 Experimental procedure
Pressure vessels

For verifying the experimental arrangements and instrumen-

tation a preheater (PH-test) from a coal-fired power plant
was tested first [11]. The vessel had been in operation
from 1963 to 1984; the design pressure and temperature were
6 bar and 250 °C. Before the test the vessel was mechani-
cally degraded by a circumferential notch (Fig. 14). The
vessel was tested at 20 °C using water for pressurization.
The material used in the preheater was a boiler vessel
steel (Hil).
The main full-scale pressure vessel tests consisted of two
destructive hydrotests [12, 20] carried out on a large
decommissioned pressure vessel (Fig. 15) which had been
used as a hydrogen cracking reactor in Neste Oy's oil refi-
nery plant from 1965; the design pressure and temperature
were 144 bar and 471 °C. The vessel was fabricated of hot-
formed and welded (axial weldments) rings and two hot-
formed heads. The material used was a heat resistant high
strength steel 2.25 % Cr l % Mo.
The wall thickness (without cladding) of the rings and
heads was 152 mm and 60 mm, respectively. The inside wall
of the vessel was cladded with a ~7 mm thick stainless
steel layer which was partly separated from the base metal
during the operation. Hence the cladding was thought not to
influence the fracture behaviour of the vessel.
For fracture mechanics purposes, an artificial but sharp
axial flaw was introduced into the inner wall of the ves-
sel. The flaw was prepared by applying a special grinding
and welding technigue in three stages. In the first stage
rough machining was done by carbon arc gouging to produce a
deep prenotch. In the second stage the prenotch was deepen-
ed by a grinding wheel to produce a smaller notch (depth
20 mm, width 12 mm). In the last stage this notch was
filled by three weld deposits. Immediately after welding, a
long sharp crack was produced by hot cracking in the middle
of the weld. The dimensions of the flaws in the HC1 and HC2
tests åre given in Fig. 15.
 - 22 -

LOCATION OF CIRCUMFERENTIAL NOTCH

NOTCH DIMENSIONS (PH-TEST)

29 = 90"
a = 4 mm
t = 6.5 mm a =90"
R = 350 mm r = 0.1 mm
b = 4 mm

Fig. 14. PH pressure vessel and dimensions of the circum-

ferential surface flaw produced into the vessel
wall.

The same pressure vessel was used for both tests. After the
HCl-test, the vessel was repaired by installing a repair
piece from the sister vessel. Welding was done by applying
temper bead technigue and performed by Rauma-Repola
Mantyluoto Works.
 - 23 -

Plan view on
top head

HG 1-TEST
.Circumferential weld
• Top head of vessel
Ring 2a Ring 3

HC2-TEST
Location of transducer

Section B-B

Fig. 15. Decommissioned HC-reactor pressure vessel of an

oil refinery plant and dimensions of the axial
surface flaw produced into the pressure vessel.

Test instrumentation

The instrumentation used during the tests allowed the

determination of membrane stresses (strain gauges), crack
opening displacement (LVDT, omega-type clip gauges), vessel
ovalization (LVDT), crack initiation and crack growth
(strain gauges, DC potential drop technique and acoustic
emission), pressure and temperature [12, 20].
The potential difference across the crack, near the crack
faces, as well as GOD were measured at three positions.
 - 24 -

The acoustic emission (AE) instrumentation of the pressure

vessels [17, 18] was planned so as to monitor the entire
vessel, as in a normal hydrotest, and also to more precise-
ly monitor the defeet area by using extra transducers. The
main task of the AE system during the measurement phase was
to store all raw emission data (arrival time and amplitude
of AE events on each channel) on floppy discs. The AE ac-
tivity on each channel (total of 64 channels) as well as
the results of real-time analysis were shown on the com-
puter terminal during the test period. Afterwards a more
comprehensive and time-demanding analysis was carried out
by the system computer to obtain the final results. The
mean resonance frequency of the transducers used was about
100 kHz. Preamplifiers gave an amplification of 60 dB. The
sensitivity of all transducers was checked and yielded the
pulse velocity of 3.0 mm/)j.s, which was used for AE source
location calculations.
4.2.2 Analytical and numerical calculations
Rupture pressure estimation of the preheater pressure
vessel (PH-test)

The rupture pressure and leak-before-break condition was

calculated by using the Battelle's method. The yielding
stress of the material was 296 MPa and the ultimate
strength 455 MPa. The flow stress was calculated as an
average of the yielding stress and the ultimate strength
giving thus a flow stress value of 376 MPa.
Precalculations of the first test with HC reactor pressure
vessel (HCl-test)

The aim of the precalculations [11, 12] was to choose suit-

able flaw dimensions and estimate the leak-before-break
condition in the test. Precalculations were performed using
the Battelle's method [24], tangent method [27] and R6-
method [26]. As a result of the calculations made by using
the Battelle's method, a flaw with the length of 1.6 m and
the depth of 110 mm of was recommended. Due to fabrication
problems, however, crack depths greater than 100 mm were
not possible. To keep the burst pressure below about 200
bar a flaw with the length of 2.5 m and the depth of 100 mm
was produced.
The unstable crack extension in the length direction after
ligament rupture was also estimated by taking into account
the pressure drop in the vessel and the enlargement of the
leak area.
 - 25 -

Precalculations were performed by Technical Research Centre

of Finland (VTT), Veritas Research (VR, Norway) and Fraun-
hofer - Institut fiir Werkstoffmechanik (IWM, Federal Repub-
lic of Germany).
Postcalculations of the first test with HC reactor pressure
vessel (HCl-test)

The postcalculations [14, 15, 19] were elastic-plastic and

geometrically nonlinear finite element analyses. They were
needed because in engineering analyses it was not possible
to consider the effect of the brittle weld intersecting the
flaw nor the effect of the maintenance deck which was not
totally removed before the test. The analyses were carried
out using the 84-version of the ADINA code and the J-in-
tegral values were calculated using a post-processor pro-
gram VTTVIRT.
Due to the different material properties and the premature
failure of the ligament (Fig. 21), three-dimensional ef-
fects were essential. Because of two symmetry planes, only
one quarter of the pressure vessel had to be modelled, Fig.
16. The length of the model was limited to 3.5 m in the
axial direction. The model shown in Fig. 16 contained over
16 000 degrees of freedom. There were 20-noded and 15-noded
three-dimensional volume elements and one beam element
which was needed to describe the effect of the maintenance
deck [19]. The crack tip was modelled by collapsed elements
causing an 1/r-singularity in the strains near the crack
front.
Two different analyses with stationary cracks were carried
out. The aim of the first one with original surface crack
geometry was to locate where the crack starts to grow and
to estimate the amount of the stable crack growth along the
artificial crack front. The aim of the second analysis with
the through-the-wall crack geometry was to study whether a
narrow through-the-wall crack in the surface crack ligament
tends to grow in the axial direction, i.e. into the base
metal. Also the shape of the crack at the end of the test
was roughly assessed.
 - 26 -

Fig. 16. Finite element model used for the three-dimen-

sional fracture assessment of the HCl-test [19].

Precalculations of the second test with HC reactor pressure

vessel (HC2-test)

The flaw geometry for the second test was chosen on the
basis of the engineering analyses for the first test. The
flaw was deeper and shorter than in the first test so that
a leak before break situation was expected. The aim of the
precalculations was to investigate the phenomena in the
test. Precalculations [20] were performed by Technical
 - 27 -

Research Centre of Finland (VTT) and Institut fur Werk-

stoffmechanik (IWM, Federal Republic of Germany). Precal-
culations were performed using the Battelle's method and
R6-method.
Postcalculations of the second test with HC reactor

pressure vessel (HC2-test)

Postcalculations [21] were at first aimed to consider the

effect of residual welding stresses, which were not con-
sidered in the pre-test calculations. The postanalyses were
done by using engineering methods and two-dimensional
finite element models.
Due to the welding of the repair piece (size 2 x 3.5 m)
into the vessel used for the HCl-test, a crack mouth open-
ing displacement (COD) of -0.8 mm and hoop strain of 1800
microstrains in the ligament at the middle of the crack
length were measured after the welding procedure (Fig. 17).
This crack closure was compensated according to the COD
measurements during the test. The pressure range which
exceeded the COD value of 0.8 mm, was assumed to cause the
rupture when the Battelle's method was applied.
Two-dimensional FEM-analyses were carried out in order to
simulate the vessel behaviour due to the residual stresses
and to estimate the effect of the repair welding. The main
purpose of these calculations was to find correct boundary
conditions for later three-dimensional analysis.
 - 28 -

CRACK MOUTH OPENING DISPLACEMENT

DURING WELDING
REPAIR PIECE
WELD
THICKNESS

150
WELD
THICKNESS
(mm)
100

50

-1.0 -0.5 O 0.5

CMOD (mm)

150 -

100
UJ
Z

o
I

UJ
50

-600 -400 -200 O 200 tOO 600 800 1000 1200 1100 1600 1800 2000
b)
S T R A I N (ym/m)

Fig. 17. Generation of the crack mouth opening displacement

(a) and ligament hoop strain (b) during the repair
welding of the vessel before the HC2-test [20].
 - 29 -

4.2.3 Results and discussion

Rupture behaviour

The failure mechanism of the thin-walled PH pressure vessel

as well as HC reactor pressure vessel was through plastic
collapse of the ligament. In the PH test where a mechanical
notch was used, the notch did not grow but the ligament was
only necked and finally ruptured at 64 bar pressure. Due to
the pure plastic collapse of the notch ligament, the real
rupture pressure agreed quite well with the calculated
value of 58 bar. The crack extension in the length direc-
tion was only about one millimeter, and hence the LBB con-
dition was valid, even if the Battelle's method indicated
slightly unstable crack extension. Similar fracture behav-
iour was found in the HC2-test as shown in Fig. 18.
In the HC2-test, in addition to extensive necking in the
outer surface (Fig. 19), stable crack growth and in some
local regions brittle crack extension was preceded before
ductile tearing of the ligament at the ultimate pressure of
189 bar. The crack was also extended in a brittle manner at
one end of the crack, as can be seen from Fig. 20.
In the HCl-test, when the pressure of 170 bar was reached,
a local leak occurred in the half-way area of the crack
length, where one of the circumferential welds of the ves-
sel was located. After the leak the pressure could be in-
creased further up to 186 bar, at which point the leak rate
became equal to the capacity of the pumps. The reason for
the local leak was the weldment with lower toughness than
parent metal. Outside of the leak area stable crack growth
of l - 2 mm was detected.
 - 30 -

Fig. 18. An overview of the leak region in the HC2-pressure

vessel test [20].
 - 31 -

Crack ligament

Inner surface

Fig. 19. Profile of the vessel wall after the HC2-test.

Fig. 20. Brittle crack extension at one end of the crack in

the HC2-test.
 - 32 -

Circumferential weld

Outer
surface

Precrack

Inner
surface

Fig. 21. Fracture contour of local rupture in the crack

ligament of the HCl-test [12].

Experimental detection of crack initiation and growth

The instrumentation of the crack field area by strain gau-

ges, GOD sensors, PD sensors and AE detectors was aimed to
detect crack initiation and crack growth in each test, ex-
cept PD in the HC2-test [12, 20].
According to the measurements, outer surface hoop strains
in the crack ligament were localized in all tests. The
strain exactly opposite to the crack tip (K3) was first
compressive reaching maximum compressive value in the 75-
85 bar pressure range. The strains at the distance of about
one ligament thickness from the crack plane (K4 and K5)
were increasingly tensile with rising pressure but showed
sharply accelerated increase at the same pressure where the
strain on the ligament reached its maximum compressive
value. This phenomenon was independent whether an axial or
circumferential part-through-wall crack was considered.
 - 33 -

10000

-2000
50 100 150 200
PRESSURE (bar)

Fig. 22. Typical crack ligament straining in all flawed

pressure vessels with axial or circumferential
surface flaw (as an example HC2-test) [20].

By applying the acoustic emission monitoring system the

flaws could be located already in early stages of the test
in spite of heavy background noise level which was mainly
due to the partially loosened cladding in the tested ves-
sel.
The AE testing revealed some trends which could be corre-
lated with the other experimental measurements. In that
phase of pressurization where strain and GOD values were
linear, the rate of AE events was very high as can be seen
from Fig. 23. In the HC2-test this range corresponded ap-
proximately to the pressure of 60 bar. In the pressure
range exceeding 60 bar, the ligament started to plastify.
Simultaneously the AE event rate along the crack ligament
decreased to a relatively low value and remained low until
the pressure of 150 bar was exceeded. Approximately at this
pressure, the rest of the ligament (elastic zone in the
ligament close to the outer surface) started to plastify
causing an increase of the AE event rate. The same tendency
was not seen at the crack end regions, where plastifica-
tion was found to be small and the AE rate was low.
Amplitude distribution of the registered and located AE
events (Fig. 24) revealed that the events with high ampli-
tude were coming between the pressure of 150 bar and the
rupture pressure, which was 189 bar for the HC2-test.
 - 34 -

AE ACTIV1TY ALONG THE CRACK LENGTH

LOCATED EVENTS/BAR
12.00 T

10.00 -

8.00 -

6.00 --

4.00 --

2.00 -

0.00
20 40 60 80 100 120 140 160 180 200

Fig. 23. Acoustic emission activity of the HC2-test using

events located not f urther than 150 mm on both
sides of the crack line [20]. The total crack
length is divided into five sections (length 450
mm) numbered from l to 5 to demonstrate AE event
rate along the crack length. Sec 3 corresponds to
the midsection and sec l & 5 to the ends of the
flaw.
 - 35 -

SEC1 SEC2 SEC5

1-50 bar

38 44 50 55 60 38 44 50 55 60 38 44 50 55 60 38 44 50 55 60 38 44 50 55 60
Amplitude

50-100 bar

44 50 55 60 38 44 50 55 60 38 44 50 55 60 38 44 50 55 60 38 44 50 55 60

156-179 bar

38 44 50 55 60 38 44 50 55 60 38 44 50 55 60 38 44 50 55 60 38 44 50 55 60

182-189 bar

44 50 55 60 38 44 50 55 60 38 44 50 55 60 38 44 50 55 60 38 44 50 55 60

Fig. 24. Amplitude distribution of the AE-events registered

in five sections along the crack length in dif-
ferent pressure ranges of the HC2-test [20].
 - 36 -

Deformation of pressure vessel circumference

Because of the flaw, the vessel wall was subjected to bend-

ing. The flaw area deflected to the inside of the vessel
causing the ovalization of the vessel as shown in Fig. 25.
The diameter as measured parallel to the crack decreased
while the diameter measured perpendicular to the crack
plane increased.
Calculated hoop strains from a three-dimensional analysis
along the outer and inner surfaces of the vessel circum-
ference agreed well with the measured ones [15, 19].
Measured and calculated hoop strains along the inner sur-
face were higher than those along the outer surface at 90°
(Fig. 26a). At 45° (Fig. 26b), however, the situation was
reversed indicating significant influence of the crack on
the stress field. In Fig. 26 the 0° reference plane is the
crack surface plane.

T«
7^//^^ \
] l !l ! !
^"^ ' ' '''"' l , > >, l rn^i-AU \\^ \
_^_ i i > : \ \ \ \7HWl
A
~" -"^-ii:id"""nifl
- —-ULJl
1

Fig. 25. Deformations of the pressure vessel in the HC1-

test (exaggerated) [15].
 - 37 -

1000

3
S?
K
E-
OT
(X
O
O

a)
80 120
PRESSURE(BAR)

3
g
K 500-
VI
p-,
O
o
LEGEND
0 = D 1 2 MEAS
A = D 3 2 MEAS
x = D12 3D CALC
+ = D32 3D CALC
0 = D12 2D CALC
® = D32 2D CALC

80 120 160
b) PRESSURE (BAR)

Fig. 26. Calculated and measured inner surface (D30, D32)

and outer surface (D10, D12) hoop strains along
the pressure vessel circumference, at the orienta-
tion of the 45° in a) and 90° in b), and crossing
the crack ligament at one third of the crack
length (HCl-test) [15].

Fracture behaviour assessment

The rupture pressure in the PH test was estimated only by

Battelle's method. The calculated value of 58 bar agreed
 - 38 -

well with the measured rupture pressure of 64 [11]. Battel-

le's method predicted the case to be a catastrophic fail-
ure. The test result was a leak without break because of
small energy content in the system.
Precalculations of the HC1 according to the Battelle's
method resulted that the dimensions of the flaw should have
led to a catastrophic failure at the pressure of 194 bar.
The estimates for the unstable crack growth at through-the-
wall crack ends varied between 1.4 and 3.0 mm. In the test
no unstable crack growth occured due to the relatively
brittle weld in the middle of the pressure vessel which
allowed a significant leak before the critical burst pres-
sure could be achieved.
The results of the preanalyses [11-13] åre summarized in
Tables 3 and 4 and compared to the experimental findings in
Fig. 27.

Table 3. Results of the preanalyses of the HCl-test.

Crack initia- Stable crack Rupture

tion pressure growth in pressure
thickness
direction
(bar) (mm) (bar)

Battelle's f ormula 200

EPRI method : 2D 110 8 140
3D 195 9 200
R6 method: IWM 122 5 133
VR 173 1.5 173
VTT 135 - -
R6 and EPRI : IWM 141 10 189
EPRI method : 2D 90 2 130

Table 4. Results of the preanalyses of the HC2-test.

Crack Stable crack Rupture

initiation growth in thick- pressure
pressure ness direction
(bar) (mm) (bar)

Battelle's method - - 130

R6-method IWM 102 2.4 108
VTT 124 2 125
 - 39 -

TEST (60°C) PREDICTIONS HC1 HC2

200 • f
HC2 ,.„.. Battelle O
P no l
(VTT) EPRI <l
-- End of test
OL/
180 - (VTT) _
D
< ) Leak in weld
R6 (VR)
_ 160 - Design
Crack initiation pressure
(0 (471 °C)
a /, ^X/VVX/UQ/

*-* -i A n / j ' / / / / / / / /
<i (^ Rattallo
tjdllcllc

y// y/////^
Pop-in in weld . QR6/EPRI
ESSURE

(VTT)
_
4
C

O T R6 (IWM)
^
i <VTT) 0
i^
M
O

R6
R6
(VTT)
tS (IWM)
DC EPRI T
°- 100 - 2D
R6
(VTT)
(IWM)
O
EPRI (Weld)
80 ~ (VTT)
Rupture

60 - 1 Initiation
4O

Fig. 27. Coraparison of measured and calculated crack in-

itiation and ultimate rupture pressures in the
tests HC1 and HC2 with full-scale chemical reactor
pressure vessel.

In order to ascertain the accuracy and applicability of

the FE-analysis on the fracture behaviour assessment, cal-
culated strains and crack opening displacements were com-
pared to the experimental ones [15, 19].
Adjacent to a crack cross-section at the 1/4 crack length
position the measured and calculated hoop strains in the
ligament had a similar trend, but there were some differen-
ces between the absoluts values (Fig. 28). Near the liga-
ment, the outer surface hoop strains had a high gradient as
a function of the distance from the crack plane, hence
small inaccuracies in gage locations may be one reason for
the differences between calculated and measured strains.
The measured and calculated crack mouth opening displace-
ment (COD) values at one quarter of the crack length were
in good agreement (Fig. 29). The calculated CTOD value,
which is also shown in Fig. 29, reached a value of l mm
corresponding roughly the JIC at a pressure of 150 bar. The
J-values along the crack front åre presented in Fig. 30 at
different pressure levels. The maximum J-value occurred in
 - 40 -

8000-

K
E-
c/)
PH
o
o

-4000
ao 120
PRESSURE (BAR)

Fig. 28. Calculated and measured outer surface hoop strains

of the ligament cross-plane located at one quarter
of the crack length (HCl-test) [15, 19].

O
O

O 40 80 120 160 200

PRESSURE (BAR)

Fig. 29, Calculated and measured crack mouth opening dis-

placements at the 1/4 (COD2) of the crack length
as well as calculated crack tip opening displace-
ments (CTOD) at the COD2 position (HCl-test) [15,
19].

the weld and was about 10 % higher than the corresponding

value in the neighboring base material. The higher yield
strength of the weld material caused higher J-integral
 - 41 -

values in the weld. The JTC value of 378 kN/m was reached
at 150 bar, which was found to be the approximate pressure
for the onset of stable crack extension in the thickness
direction in the HCl-test.
On the basis of the material JR-curve and the calculated J-
values one can estimate stable crack growth at the end of
the test (186 bar). According to the analysis [15] pre-
sented in Fig. 30, the estimated crack growth in the base
material was 3.5 mm in the vicinity of the weld and 2.5 mm
at one quarter of the crack length, and at the crack's end
no crack growth in axial direction was expected. Fracture
surface investigations [11] indicated about 2 mm crack
growth over a length of approximately 500 mm on both sides
of the leak. The J-level in the weld is much higher than
weld material J-R-curve values, which could indicate insta-
bility or at least considerable crack growth through the
weld. A very sharp peak in the J-distribution along the
front of the trough-the-wall crack occurred in the crossing
of the initial and new crack fronts. It showed that the
through-the-wall crack tends to grow much faster near the
inner surface, where it should extend considerably into the
base material as supported by the shape of the experimental
through-the-wall crack (Fig. 21). Estimated and measured
crack growths åre compared in Fig. 31. If the crack had
been located entirely in the base material, it would have
presumably penetrated the wall along most of its length
thus causing a catastrophic failure. The failure pressure
would probably have been slightly over 200 bar.

LEGEND
o = JMEAN P50
a = JMEAN P100
+ = JMEAN P150
x = JMEAN P185
o = JMEAN P190

0.25 0.50 0.75 1 2 3 4

XABS (M) CRACK EXTENSION (mm)

Fig. 30. Distribution of the J-integral value along the

surface crack front at different pressure levels
and determination of stable crack growth based on
the application of the material's fracture resis-
tance curve (HCl-test).
 - 42 -

Weld / Outer surface

X
Inner surface

Fig. 31. Estimated and measured crack front location in the

local leak region (HCl-test) [19].

In spite of the complicated geometry, the overall agreement

between experimental and calculated displacement and strain
results was good. Conservative estimates were obtained for
stable crack growth. Extensive three-dimensional analyses
were necessary to explain the failure behavior in the test.
In the precalculcations of the HC2-test [20] the residual
stresses due to repair welding were not taken into account
which was probably the reason for the large discrepancy
between the actual rupture pressure (189 bar) and the aver-
age estimate rupture pressure (120 bar). Because the pres-
sure vessel in the HC2-test behaved primarily elastically,
i.e. the plastic zone near the flaw was small with respect
to the vessel size, the crack closure (COD-value of -0.8
mm) after welding can be compensated for by adding the
pressure load which corresponds to the COD value of 0.8 mm.
During the test the COD value of 0.8 mm was reached at the
pressure of 60 bar which meant that the COD value was ac-
tually zero at this loading level. In applying the Battel-
le's method it can be thought that the portion of the total
pressure exceeding 60 bar causes the final rupture. Thus,
adding 60 bar to the value of 130 bar evaluated in Table 4
without the welding residual stresses, a value of 190 bar
is obtained, which is the same as the ultimate rupture
pressure achieved in the test.
 - 43 -

The effect of welding residual stresses on the fracture

behaviour is currently being investigated by two-dimensio-
nal finite element analysis [21]. The emphasis of this
analysis is to assess correct solutions for appropriate
boundary conditions which åre essential in three-dimensio-
nal analysis.
4.3 Conclusions on the applicability of fracture assessment
methods
The fracture behaviour of flawed pressure-containing com-
ponents can be assessed in different ways. The proper in-
strumentation of the flaw ligament region by strain gauges
gives early information on the possible extension of the
flaw. Additional instrumentation with an acoustic emission
monitoring system can locate the flaw already in early
loading stages. The application of AE testing for crack
growth monitoring requires preliminary fracture mechanics
considerations of the case concerned, before the correla-
tion between the AE event activity and crack extension can
be reliably expressed.
The engineering methods which were applied on the cylindri-
cal pressure vessels, were Battelle's limit load method and
the R6-method. The finite element method calculations con-
sisted of two- and three-dimensional linear, elastic-plas-
tic and both elastic-plastic and geometrically nonlinear
analyses.
If the material is ductile and the geometry and loading
situation åre simple, Battelle's limit load method gives
the ultimate rupture pressure with reasonable accuracy and
low cost. The R6-method was useful because it could be used
to estimate the degree of stable stable crack growth prior
to rupture when the J-R-curve of the material was known.
In the case of a realistic finite flaw in a simple struc-
ture, as in a cylindrical vessel, three-dimensional effects
åre essential and three-dimensional finite element calcula-
tions åre necessary. Even with the complicated geometry in
the first test (HCl-test) of a decommissioned reactor pres-
sure vessel, the overall agreement of displacement and
strain results between the experimental and numerical
values from three-dimensional finite element analyses was
good. Conservative estimates were obtained for stable crack
growth. Extensive three-dimensional analyses were necessary
to explain the failure behaviour of the vessel in the test.
 - 45 -

5. RECOMMENDATIONS
The test programme was comprised of tests with two large
pressure vessels, semi-scale pressure vessel wall, one
small pressure vessel, CT-specimens and 122 small pipes.
The theoretical methods applied on the analysis of pressure
vessels, specimens and pipes can roughly be divided into so
called engineering methods and into methods based on the
finite element method.
The engineering methods which have been applied on the
cylindrical pressure vessels and pipes, have been
Battelle's limit load method, MPA's method (Material-
priifungsanstalt, Stuttgart), the R6-method and failure
assessment diagram. The finite element method calculations
have consisted of two- and three-dimensional linear, elas-
tic-plastic and both elastic-plastic and geometrically non-
linear analyses.
The following conclusions concerning the engineering
methods can be drawn: If the material is ductile, as it was
in the performed tests, and the geometry and loading situa-
tion åre simple, i.e. cylindrical geometry and axial or
circumferential crack orientation, Battelle's and MPA's
limit load methods give the bursting load and the stability
of a burst crack with low costs and quite reasonable ac-
curacy. One of the biggest problems is how to choose the
flow stress for the calculations. The smallest value is the
yield stress of the material and the highest possible value
is the true ultimate stress. One recommendation might be to
make a test by pressurizing a cracked cylinder and deter-
mine the flow stress from the burst pressure. Certainly the
flow stress determined in this way depends on the geometri-
cal dimensions and their ratios, and probably even on the
crack geometry.
The R6-method is useful because it can be used to estimate
stable crack growth before bursting if the J-R-curve of the
material is known. The R6-method is also useful, when the
material is not fully ductile.
Conclusions of the application of the finite element method
on cracked pressure vessel and pipe analysis åre as fol-
lows: In case of a realistic finitely long flaw even in a
simple structure, as in a cylindrical vessel, three-dimen-
sional effects åre essential and three-dimensional finite
element calculations åre necessary. This is usually very
expensive but gives very detailed information, e.g., stres-
ses and strains at every interesting point and J-integral
distributions along the crack front at different load
levels. The amount of stable crack growth and the fracture
behavior can be assessed based on the calculated J-varia-
tions and the measured J-R-curves. In more complicated
cases such as would be caused by a more complicated geomet-
ry, combined loading, or multiple materials (as in cladded
 - 46 -

pressure vessels or weldments), extensive three-dimensional

finite element analyses åre possibly the only way to ex-
plain the failure behavior of a pressure vessel or piping.
The proper consideration of boundary conditions and availa-
bility of the relevant material parameter data åre very
essential for achieving good results. Automatizing of the
finite element mesh generation and access to super com-
puters greatly decreases the time needed for such computa-
tions.
 - 47 -

6. R E F E R E N C E S
6.1 Publications and documents produced in the proiect
1. Torronen, K. (ed.), Prevention of catastrophic failure
in pressure vessels and piping. Progranune Plan. Espoo
1985. Technical Research Centre of Finland, Metals
Laboratory. VTT-MET B-84. 18 p.
2. Ikonen, K. (ed.), Leak-before-break concept in nuclear
reactor pressure vessels and pipings. Espoo 1985.
Technical Research Centre of Finland (VTT). Research
Report 368. 164 p. + App. (in Finnish).
3. Ikonen, K. (ed.), Computation methods of the LBB con-
dition. Espoo 1985. Technical Research Centre of Fin-
land (VTT), Nuclear Engineering Laboratory. Report
178 p. + App. (in Finnish).
4. Tal ja, H., Nordic numerical round robin for a side-
grooved CT-specimen. Espoo 1989. Technical Research
Centre of Finland (VTT), Research Report, (in press).
5. Wallin, K., Final results of the Nordic J-R curve test
programme. Espoo 1987. Technical Research Centre of
Finland (VTT). Research Report 510. 29 p + App. 8 p.
6. Wallin, K., Final results of the second Nordic J-R
curve test programme. Espoo 1989. Technical Research
Centre of Finland, Research Report, (in press).
7. Faleskog, J., Zaremba, K., Oberg, H. & Nilsson, F.,
Numerical evaluation of two- and three-dimensional
elastoplastic crack growth initiation experiments.
Royal Instituts of Technology, Dept. of Solids
Mechanics. Stockholm 1989. Report 101 (in press).
8. Talja, H. & Ikonen, K., Computational failure assess-
ment of pressure vessels and pipings. Espoo 1989.
Technical Research Centre of Finland. Research report
602. 109 p. (in Finnish).
9. Kukkola, T., Intermediate research report, Rev 1.
25.5.1987. Imatra Power Company, V-LBB-PUKO-A6 rev.l.
10. Kukkola, T., Leak-before-break continuation experi-
ments. Test Report. 6.4.1987. Imatra Power Company,
V-LBB-PUKO-A6.

11. Rintamaa, R., Keinånen, H., Torronen, K., Talja H.,

Saarenheimo A. & Ikonen, K. Fracture behavious of
large scale pressure vessels in the hydrotest. Proc.
of SMIRT Postconference Seminar no. 2, Davos, Switzer-
land, August 24 - 25, 1987. Published also in Interna-
 - 48 -

tional Journal of Pressure Vessel and Piping

34(1988), p. 265 - 291).
12. Rintamaa, R., Torronen, K., Keinånen, H., Sarkimo, M.,
Sundell, H., Talja, H. & Ikonen, K., Prevention of
catastrophic failure of pressure vessels and piping.
Results of pressure test with a large vessel (HC1-
test). Espoo 1988. Technical Research Centre of Fin-
land, Research Report 515. 51 p.
13. Torronen, K., Rintamaa, R., Talja, H. & Ikonen, K.
Evaluation of analysis methods for prevention of cata-
strophic failure in pressure vessels. In: Folker H.
Wittman (ed.). Structural Mechanics in Reactor Tech-
nology, Vol. G., p. 269 - 274.
14. Rintamaa, R., Keinanen, H., Torronen, K., Talja, H.,
Saarenheimo, A., & Ikonen, K. Verification of analy-
sis for fracture behaviour estimation of pressure
vessels. Proc. of the 13. MPA-Seminar, Stuttgart,
October 8-9, 1987, 21. (Published also in Nuclear
Engineering and Design 112 (1989), p. 299 - 309).
15. Saarenheimo, A., Talja, H., Rintamaa, R., Keinanen,
H. Se Torronen, K., Verification of the nonlinear frac-
ture analysis by a large scale pressure vessel test.
IAEA Specialists' Meeting on Large Scale Testing, May
25 - 27th, 1988, MPA Stuttgart, Fed. Rep. Germany. To
be published in Nuclear Engineering and Design.
16. Rintamaa, R., Wallin, K., Torronen, K., Talja, H.,
Saarenheimo, A., & Ikonen, K., Destructive hydrotest
of a large scale pressure vessel with a flaw. Publish-
ed in Pressure Vessel Technology (eds. Cengdian and
Nichols.) Pergamon Press. Oxford 1988. Volume 2
Materials & Fabrication. P. 903 - 917).
17. Sarkimo, M., Acoustic emission monitoring of large
vessel at pressurization until rupture. 9th World
Conference on non-destructive Testing. WCNDT:
Amsterdam, The Netherlands 23 - 28 April, 1989. Inter-
national Committee on non-Destructive Testing. 6 s.
18. Sarkimo, M., Acoustic emission results in pressure
test to failure of a large-size pressure vessel. Fin-
tech-seminar: Advances in non-destructive testing.
Espoo 1988. Technical Research Centre of Finland. VTT-
Symposium 93. p 176 - 190.
19. Saarenheimo, A. Talja, H., Ikonen, K., Rintamaa, R.,
Keinanen, H. & Torronen, K., Fracture behaviour simu-
lation of a flawed full-scale pressure vessel. To be
published in International Journal of Pressure Vessel
and Pipings.
 - 49 -

20. Keinanen, H., Rintamaa, R., Oberg, T. & Tal ja, H.

Evaluation of catastrophic failure risk in pressure
vessels - Results of the HC2 pressure vessel test.
Espoo 1989. Technical Research Centre of Finland,
Research Report (in press).
21. Keinanen, H., Rintamaa, R., Oberg, T., Talja, H. &
Saarenheimo, A., Comparison of experimental and com-
putational results of the pressure vessel test HC2.
Espoo 1989. Technical Research Centre of Finland,
Research Report (in press).
6.2 Other references
22. Wallin, K., A simplified key-curve approach for J-R-
curve testing of CT-specimens. Espoo 1987. Technical
Research Centre of Finland. VTT Research Report 503.
22 p.
23. Larsson, L. H., EGF Numerical Round Robins on EPFM.
Proc. of the Third Conference on Numerical Methods in
Fracture Mechanics. Swansea 1984. P. 775 - 814.
24. Rodabaugh, E. C., Comments on the leak-before-break
concept for nuclear power plant piping systems, Oak
Ridge National Laboratory, NUREG/CR-4305(1985). 43 p.
25. Julisch, P., Stoppier, W. & Sturm, D., Exclusion of
rupture for safety-relevant pipings systems by com-
ponent tests and simple calculation. Trans. 8th Int.
Conf. on Structural Mechanics in Reactor Technology,
Brussels 19 - 23 August 1985. North-Holland Publishing
Compavy, Amsterdam 1985. Paper G3/8. 5 p.
26. Ainsworth, R. A., Dowling, A. R., Milne, I. & Stewart,
A. T., Background to and Validity of CEGB Report
R/H/R6 - Revision 3. Int. J. Pres. Ves. & Piping
32(1988), p. 105 - 196.
27. Kumar, V., German, M. D. & Shih, C. F., An engineer-
ing approach for elastic-plastic fracture analysis,
EPRI Report NP-1931, Palo Alto, USA (July 1981).
169 p.