An A n a l y s i s of a 100 te S t o r a g e Vessel

*T A Smith

The object of this paper is to suggest a method of

reviewing the possibility of cold whole vessel failure and
establish whether a leak before break-condition is more
likely than catastrophic failure.
Fracture mechanics concepts have been used to calculate
critical crack sizes. Stress levels at operating and test
conditions have been calculated using the approaches given
in British Standard 5500:1982 which includes the effects of
bending, shear and stresses around support saddles.

The beneficial effects of stress relief are reflected in

the large crack lengths a stress relieved vessel is able to
sustain. It is shown that vessels in this condition should
leak before a major failure situation can occur.

1. INTRODUCTION

A number of surveys of operational experience from conventional pressure vessels

have been carried out by the Safety and Reliability Directorate of the UKAEA in
conjunction with a group of major engineering insurance companies. Evidence
gathered for these reviews showed that failure did occur in high quality plant.

No major failures of Liquid Petroleum Gas (LPG) storage vessels have been
reported in the UK, and historically the UK has a good record in that there have
been no large releases due to spontaneous tank failure. Nevertheless it was
against the wider background of pressure vessel experience that it was decided
to analyse an LPG storage vessel. The analysis is confined to a static above
ground tank used for the pressurised storage of LPG and of 100 te capacity.

Too many variables are available to designers, constructors and operators to

claim that all storage vessels may be considered alike. Certain recommendations
regarding stress relief inspection requirements, design code application and
material selection do have a common denominator effect in so far as they are
applied. However information to hand is neither sufficiently detailed in
application nor nationally spread in implementation for any firm conclusions to
be drawn about the whole UK LPG industry storage problems on the basis of any
analysis of one vessel.

2. REVIEW OF PRESSURE VESSEL FAILURE EXPERIENCE

Since 1962 the Safety and Reliability Directorate in conjunction with a group of
major engineering and insurance companies, has undertaken surveys of
conventional pressure vessels. (Refs 1.2.3). The principal objective of the
surveys was to obtain data from which to assess the reliability of pressurized
plant built to high standards of construction.
*Safety & Reliability Directorate, UKAEA.

107
 IChemE SYMPOSIUM SERIES No. 93

The surveys were conducted with the objective of identifying failure locations,
numbers, causes, method of detection, and operating conditions from a researched
population with minimum thickness and pressure parameters. These are listed in
the reports.

Evidence gathered for these reviews showed that the majority of failures
manifested themselves as cracks, some of which leaked, but there were also a
number of catastrophic failures which resulted in a large discharge of contents.

Probabilistic failure predictions based on data gathered for the surveys are
possible for generic categories of vessel and could prove useful. Assessment of
individual components or small populations of components are somewhat more
difficult. While data exists from which judgement on failure probability may be
made, other considerations need to be taken into account when analysing the
probable behaviour of a specific vessel category. These can best be solved by a
deterministic approach based on fracture mechanics theories hence the reason for
work described in subsequent sections of this paper.

3. FRACTURE MECHANICS - SOME AVAILABLE METHODS

In recent years fracture mechanics methods have become increasingly important

where the control or prevention of fracture is a paramount consideration.

The main property in fracture studies is the material toughness under various
conditions of stress, mechanical properties and temperature. Material toughness
is a measure of the ability of a material to withstand the stress intensifying
effect at a crack tip. In ductile materials it does this by deforming mainly
elastically in very low stress situations, or plastically as stress levels
increase. At a critical dimension the net section under the crack becomes
unstable and fracture results. There can be, and usually is, a degree of stable
crack growth before the onset of instability. The object of fracture mechanics
is to predict at what size and under what conditions a defect length becomes
critical.

The three fracture mechanics approaches used in this analysis have been selected
for the following reasons.

a) The Stress Concentration Theory has been developed from tests on

vessels varying in diameter from 305mm (12 inches) to 2895mm
(9ft 6ins), and from wide plate tests. It has been checked and found
to be in reasonable agreement with other vessel failure data. The
theory has the advantage of only using well established material
properties of yield and ultimate stress levels, elongation and Charpy
'V notch energy values.

b) PD 6493: 1980 Ref 4 is a British Standard Institution document and

should be used in any comparative fracture mechanics exercise. It is
however basically a guide to the sizing of tolerable flaws in welded
joints and employs linear elastic fracture mechanics (LEFM) and
elastic plastic fracture mechanics (EPFM).

c) Numerous techniques are available for extending LEFM into the elastic
plastic regime. Crack opening displacement (COD) measurements are
currently the most widely used of several modifying parameters and
these have been selected for this analysis.

108
 iChemE SYMPOSIUM SERIES No. 93

All fracture mechanics theories are sensitive to a number of input data func-
tions, although in this respect the stress concentration theory is less sensi-
tive than others because it relies on standard material property data which is
readily and regularly established. Other approaches are dependent upon a number
of derived data which include various toughness parameters, energy functions,
material resistance, crack opening displacement measurements, compliance
functions.

When none of these data are available for a vessel under consideration use has
to be made of relevant correlating relationships with known material properties
and data from similar steel.

4. MATERIAL AND MATERIAL PROPERTIES

Steel. selected for the storage of potentially hazardous substances must have
adequate strength, be compatible with tank contents and show little or no
deterioration of properties under service conditions. It must remain ductile
and maintain its resistance to defect propagation under the most arduous of
operating environments, and allow the production of high quality welds. The
mechanical properties of the weld metal and heat affected zone should be at
least as good as that of the parent plate particularly as surveys of
conventional plant show weld and weld affected metal to be the most vulnerable
sites for flaws especially in regions of discontinuities.

Mechanical properties for plate to BS 1501:151 and weld metal were obtained for
yield stress, UTS and elongation together with a Charpy value at -15C

5. FACTORS AFFECTING STRESS ANALYSIS

5.1 Shakedown

Stress levels were calculated for the various loadings to which the vessel is
subjected. These included primary and secondary stresses due to pressure, dead
weight, geometrical discontinuities.

Various factors will arise during the course of testing and operation which will
modify or influence the effect of the stressing system. For example, where peak
stress levels are at the value of the material yield stress the phenomenon known
as shakedown will occur during the first few operating cycles. In particular
the proof pressure test can have a significant effect as illustrated in Fig 2.
Describing this figure it can be seen that on loading up during the pressure
test, if the strain range associated with the linearised peak equivalent elastic
stress is

and if the proof test has an overpressure ratio

then the maximum operating strain range is

109
 IChemE SYMPOSIUM SERIES No. 93

On unloading from the pressure test

Reloading to operating pressure gives

Residual stresses resulting from welding a plate into a structure are essen-
tially of two kinds. There is the stress caused by general shrinkage and fit up
which affects the whole plate. But there is also the more significant stress
caused locally by the cooling contraction of the deposited weld metal and which
is generally considered to be of yield stress magnitude effective over a
distance of the order of the plate thickness.

How residual stresses influence fracture in a ductile material is not fully

understood, particularly where critical crack sizes are long (several times the
plate thickness).

Additional operational stresses may simply be added and may even cause some
defect propagation, but local yielding would relax the stress level before
unstable conditions were reached. This implies a form of mechanical stress
relief and can result from the application of a proof pressure test.

Since the great majority of defects are sited in weld affected material it is
the weld properties which control crack tip behaviour. This is significant
where linear elastic conditions are applied and little, or no, plastic
deformation occurs. However when elastic plastic or general structural collapse
is concerned large areas of the component will be affected and the structural
behaviour of the component will be controlled by the yield point of the parent
plate. Therefore since weld metal comprises only a small fraction of the total
volume of affected material it is the yield stress of parent plate which should
be used in shakedown analysis. For vessels which are subjected to fatigue
loadings, combinations of stresses from all sources should be kept within
minimum plastic limits so that the vessel may 'shakedown' to purely elastic
behaviour and plastic cycling causing incremental collapse can be avoided.

5.2 Bulging

In a component whose wall thickness implies plane stress conditions critical

crack lengths can be several times the wall thickness in length. In (curved)
cylindrical geometries the action of the pressure on the unsupported edges of a
through thickness crack will impose additional stresses at the crack tip due to
the bending effect. The importance of these additional stresses increases with
increasing crack size and pressure, but decreases with increasing diameter until
flat plate conditions apply.

1 10
 IChemE SYMPOSIUM SERIES No. 93

5.3 Fatigue

Another mechanism to be considered is fatigue, and this must be allowed for in

any analysis' in which cyclic loadings can be applied to the component.

6. STRESS LEVELS

It is assumed that the vessel is a horizontal cylindrical unit supported on two

symmetrically placed saddles, one welded to the vessel and the other free. (Fig
1).

Maximum stresses resulting from dead weight and pressure loadings will be those
associated with the vessel acting as a beam and those stresses produced by
reaction at the supports, which are assumed to be rigid for the purposes of this
report. Other areas of concern are those around the larger penetrations such as
the manway penetration.

Design rules for determining stress levels in the saddle areas are contained in
the pressure vessel design code BS 5500:1982.

These rules and formulae have been used to calculate stress levels in the saddle
areas for the tank support system. Future modifications to these design
criteria are likely as a result of work currently being carried out at
Strathclyde University (Ref 5) where tests and calculations suggest that for a
rigid saddle support the stresses predicted by BS 5500 are exceeded by a factor
of 2, while the stress levels for a flexible support system more closely agree
with BS 5500 values. However collapse tests on vessels indicate that even with
rigid supports the design code values are safe.

Further analytical work (Refs 6 and 7) confirm that the maximum stress in
vessels supported by rigid or flexible saddles occurs in the region of the
saddle horn. Results also show that for flexible support saddles there is a
useful built in factor of safety when the design is based on BS 5500. The peak
stress in the vessel is in general higher than values obtained by use of
relevant code formulae but it is restricted to a small area of the vessel and is
mainly a bending stress.

7. FRACTURE MECHANICS PROCEDURES AND CRACK SIZE CALCULATIONS

7.1 Stress Concentration Approach

In the 1960s Irvine and Quirk developed a fracture mechanics theory involving
only conventional mechanical properties, Ref 8.

Consider half the crack length , then if all the load shed by the crack
material is carried by a region "S" at the crack tip then

where S is the width of the stress perturbation at the crack tip which is raised
to an average stress of (Fig 3(A)).

111
 IChemE SYMPOSIUM SERIES No. 93

Fig 3(B) shows a typical plot of experimental results in which' gives a

straight line relationship according to equation (5) rearranged

This is more usually written as

The parameter S' is a material constant which is independent of crack length

but which is affected by temperature.

Considerable effort has been made to establish the dependence of 'S' on values
of Charpy' energy," yield stress and tensile elongation. Ref 9 uses the large
volume of data from several steel specifications to obtain a correlation. This
showS a fairly weak dependence on charpy energy and yield stress since they are
only] Effective to the power 1/3, but the influence of elongation in the value of
'S' is much more significant as it is effective to the power 2.

The material characteristic 'S' may be related to other material properties of

elongation Charpy and yield stress by the formula

Plate to plate variation in mechanical properties also affect calculations of

critical crack size. This is simply illustrated in Fig 4 which shows the effect
of changes in J on values of 'S'. Hence the need to investigate the effect of
scatter in plate properties.

Where bulging can take place failure can occur at significantly lower values of
internal pressure. The actual effects can be calculated by the formula

where the constant K is proportional to the third power of the material yield
stress. Results of tests carried out on cylinders ranging in diameter from 12.0
inches to 9 ft 6 inches are plotted in Fig 5. The results indicate that the
relationship between K and is given by the equation

Fig 6 shows the effect of the additional bending stresses due to bulging and
illustrates that failure pressures are in fact lower where bulging can occur.
Since the component diameter 'D' appears in the denominator of the bulging
effect equation given above, and is also squared, it is obvious that as a
cylinder or sphere increases in diameter the bulging effect tends to vanish and
flat plate equations may be applied. In the case of partial penetration cracks
the bulging action is a latent effect until snap through of the remaining
ligament occurs. In the particular case of the 100 te vessel it was found that
the bulging could be neglected.

112
 IChemE SYMPOSIUM SERIES No. 93

Table 1 lists crack sizes using equation 7 and typical material properties for
weld metal and plate to BS1501:151. Plate crack sizes at operating and test
pressures respectively are 3.3 and 2.2 inches (84mm and 56mm) while for
unstress-relieved conditions after shakedown the crack size is 2.1 inches
(53mm). The equivalent figures for weld metal are 3.6 and 2.6 inches (92mm and
66mm) reducing to 1.0 inches after shakedown (25.4mm).

7.2 British Standard PD 6493:1980

PD 6493:1980 gives guidance for carrying out an engineering critical assessment

of defects found in fusion welded joints. The methods proposed are applicable
to welds in a number of structural steels of thicknesses 10mm and above, and are
aimed at establishing tolerable defect sizes.

It will be assumed that stresses are due to static loads and act in a direction
perpendicular to the defect. Dynamic effects are not considered. Since the
tank under consideration has been stress relieved residual stresses should be
minimal.

Typical Yield stress, ultimate stress and elongation values are available.
Charpy impact values show considerable scatter; lower bound values will
therefore be used.

No fracture mechanics data in terms of fracture toughness (KIC) or K for plane

stress conditions, crack opening displacement (COD) or the J integral contour
are available. However formulae have been derived correlating some of these
factors with formal mechanical properties and these will be used where required.

Several fracture 'mechanics theories are available - the document lists 12

methods - but there is no universally accepted approach for assessing material
behaviour in the plastic regime.

Many empirical correlations exist between fracture toughness as denoted by K

and Charpy energy values One widely used correlation is the Rolfe Novak
and Barsom approach which relates K to Charpy and material yield stress
in the following form

It is recommended however that this correlation be used for Charpy values above
the transition temperature and for steels with relatively high (110-246 KSI)
yield points. At low energy absorption levels little variation appears to exist
between slow bend and dynamic Charpy values and a correlation for below
transition temperature values for steels with a yield stress range 36-50 KSI is
given by

where for structural mild steels A can be taken as 5. Another general tran-
sition range correlation for steels whose yield points vary between 39 and 246
KSI is

1 13
 IChemE SYMPOSIUM SERIES No. 93

Ref 10 lists 13 Fracture Toughness V Charpy Energy correlations for K and K

values for a wide variety of steels and for above and below Charpy transition
temperatures. Selection of an appropriate correlating formula is complicated by
the fact that several are based on pre-cracked Charpy and slow bend test
specimen data and these data are not available.

Selection of the most appropriate formula must be subjective because of

uncertainties in basic data and applicability. It would be prudent therefore to
selct the correlation which gives the lowest values of K since at low energy
levels this is unlikely to be markedly different from K . Fig 7 shows Equation
11 to be considerably more pessimistic than Equation 12.

From mill tests a typical figure for the yield stress of BS 1501:151
material

Yield stress for weld metal is somewhat higher at

where C is obtained from PD 6493 Fig 14 and is dependent upon the ratio of
applied stress to yield stress

Applied stress = membrane stress + bending stress + Q + F

Ref 11 gives a value of 1.8 for a machined fillet stress concentration factor
(K ). A welded fillet such as the geometry at the saddle attachment would
attract a higher value than this, say 2.5. (BS 5500:1982 (2.3.3)).

Stresses included under the symbol "Q" are self equilibrating and result from
mechanical loads differential thermal expansions etc and for the purpose of this
analysis are assumed to have a negligible input to the overall applied stress.

Using the procedures outlined earlier gives an

Applied stress = 347 MN/m2 (50 KSl)

Equation 13 gives values of a which are listed in Table 1. These crack sizes
are acceptance sizes and a guide to the sensitivity requirement of inspection
techniques and attract a safety factor of between 2 and 10 (Ref 12).

7.3 Elastic Plastic Fracture Mechanics

114
 IChemE SYMPOSIUM SERIES No. 93

condition to control fracture. Defect sizes may be several times material

thickness in length and relatively large plastic zones are formed in front of
the developing crack tip. Because of the inherent ductility of carbon and
carbon manganese steels they are able to withstand considerable deformation
under normal loading rates and temperatures seen in service. Linear elastic
considerations are upset and the limits of their validity exceeded when the
effects of defects extend beyond the materials elastic regime into the plastic
region.

Two of the more generally used parameters for extending LEFM are the 'J' contour
integral and the crack opening displacement (COD) measurement.

A value for the 'J' integral can be estimated experimentally and is a function
of the area under the load extension curve for cracked specimens. In an
elastic-plastic condition it is a measure of potential energy change in the
system for an increment of crack extension.

A more widely used technique is that of the COD measurement which is a measure
of the deformation at a crack tip at the instant of crack propagation. There is
a considerable volume of COD data and several correlations have been formulated
with other mechanical properties such as Charpy 'V notch energy and material
yield stress. It is perhaps because of this access to a greater data base that
COD has become such a generally used concept.

The technique for extending fracture mechanics into the elastic plastic
condition used below is the one of crack opening displacement (COD) measurement
(o). For plane stress non workhardening conditions Dugiale Ref 13 has developed
a relationship combining crack length , yield stress applied stress ,
Youngs modulus (E), and COD so that

Burdekin and Dawes have derived a correlation of COD and yield strain to
calculate a critical crack length

Cotton (Ref 13) has modified the Burdekin approach for elastic plastic
conditions and derived

Figures 16.6 through to 16.11 of Ref (13) plot COD V temperature for various
steels, C Mn, weld metal, and heat affected zone (HAZ) material. Test data are
at temperatures between -20C and -150C. Considerable scatter is evident but a
lower bound COD value for parent plate material extrapolated to +10C is 0.66mm
(25 x 10 inches). For weld metal and HAZ material CODs at 10C are
respectively 0.2mm (8.4 x 10 ) and 0.19mm (7.8 x 10~ inches). A further
extrapolated value for mild steel plate gives 0.45mm (18 x 10~ inches) at
+10C.

Ref (14) gives a value of for BS 1501:151 material as 1.15mm (45 x 10-3
inches) as ascertained by 3 point bend tests on material machined from test
vessels. This figure was reasonably constant over a temperature range from +8C

1 15
 IChemE SYMPOSIUM SERIES No. 93

to +90C. A further reference (Ref 15) plots COD V temperature for mild steel
wide plate tests. Fig 3.11 of this reference gives 0.75mm (30 x 10 inches) as
the COD value at 10C.

Refs (14) (15) and (16) show that CODs for BS 1501:151 as measured during vessel
tests are generally greater than the values obtained by laboratory bend tests.

Table 1 lists crack sizes for a range of COD values.

8. FATIGUE ASSESSMENT

In fabricated components such as the vessel being considered in this paper,

critical flaw sizes will be much larger than any defect which may be initially
undetected. The difference in size between the initial defect and a critical
size defect will depend to a large extent upon the quality of manufacture and
inspection coverage.

If we assume that the component will not fail by brittle fracture with little or
no extension of an original defect, it is necessary to calculate the number of
operational cycles necessary for a small crack to grow to a critical size.

Several steps are necessary for this exercise.

1. Assume an initial flaw size -

2. Calculate a critical crack size using fracture mechanics

parameters - a
cr
3. Determine cyclic stress range in terms of K

4. Apply standard crack growth rate formula - the Paris equation

5. Integrate or use iterative procedure between a and a to

obtain number of cycles.

These calculations require a knowledge of the basic laws of crack growth and the
material variables contained in the constant 'C and the index 'n' of the crack
growth formula.

A large volume of work has been done to establish the constants 'C and 'n' and
as with other basic data considerable variation exists.

PD 6493:1980 gives C = 1.7 x 10~ and n = 4 for propagation rate per cycle and
a 99% probability of survival - the units being mm/cycle and K in MN-3 /2.
12
Fulmer Research Institute, (Ref 17) gives C = 10 and n = 4 . 0 for "structural
steels" - units being in m/cycle.

AGR Materials Programme (Ref 18) gives and n = 3.5 for

"structural steels" units m/cycle.

116
 IChemE SYMPOSIUM SERIES No. 93

Assuming a partial penetration defect has perforated the vessel wall with a
length equal to the wall thickness and for conservatism using the edge crack
constant 1.12, the number of cycles for it to grow to twice the wall thickness
is given by

where, is the increment of crack growth and a is the average crack size

between two increments of growth.

Using the PD 6493:1980 constants (as they give the greatest growth per cycle)

N = 150 cycles

Using the Barsom and Rolfe constants which give the lowest crack growth per
cycle, gives a value of N in excess of 3 x 10 cycles. Fig C.2.1 of BSS
5500:1980 gives a value for N of approx 2.5 x 10 cycles obtained from smooth
specimens. For approximately 35% of operating time the tank will be about half
full, ie will contain between 40 and 60 tons. 40% of its time it will contain
between 60 and 100 tons and for the remaining 25% of its time less than 40 tons.
If these figures taken from actual field record are typical the number of full
stress range cycles seen during the tank lifetime will be relatively small, say
350 during service life. It is apparent, therefore that despite the difference
in numbers of cycles the use of different constants will produce against the
number of cycles the vessel will see, fatigue is unlikely to be a significant
problem as a failure mode.

9. CONCLUSIONS

1) Table 1 summarises the critical (and tolerable) crack sizes obtained

by the three methods used in this paper. It should be noted that a
reserve factor of between 2 and 10 is built into the PD 6493:1980
figures.

2) There is reasonable agreement on critical crack sizes between the

two fracture mechanics approaches adopted. PD6493:1980 gives defect
sizes which are essentially a guide to the sensitivity required of
inspection techniques. Given that these sizes attract a safety
factor of up to 10 these figures again show reasonable agreement
with the other methods used.

3) A fatigue assessment has been made to estimate the number of cycles

required to increase a defect from a length equal to the vessel wall
thickness to a length equal to double the wall thickness.

Selecting data which give the maximum and minimum crack growth per
cycle gives the number of cycles for 2a=t to 2a=2t as

3
The equivalent figure from BS5500 is 2.5 x 10 . Providing
inspection ensures that no major defect will exist initially,
fatigue should not be a significant problem as a failure mode.

1 17
 IChemE SYMPOSIUM SERIES No. 93

4) In a relatively thin walled pressure vessel manufactured from a

readily weldable low strength ductile material like BS 1501:151
critical sizes of through the wall cracks will be several times the
wall thickness in length. Under normal circumstances the vessel
contents will leak from such defects before disruptive failure can
occur, but local chilling may defeat this condition particularly in
unstress relieved vessels.

For partial penetration defects propagating under operating

conditions, when the crack depth plus its associated "plastic" zone
exceeds the vessel wall thickness snap through of the remaining
ligament will occur when the stress on it reaches the UTS. If at
this point the resulting through thickness crack is sub-critical
leakage will result.

5) In the unstress-relieved condition critical crack lengths can be so

short (25.4 mm) that a leak before break condition could be
excluded.

6) For the future in order to minimise the need to deduce required

fracture mechanics parameters, it is necessary to establish the
relevant properties from actual vessel material. Transition curves
from Charpy tests or at least sufficient test results to establish
upper and lower shelf values are required.

Conventional mechanical properties of yield and UTS stress levels at

upper and lower operational temperature are required.

Crack opening displacement measurements on crack material over an

appropriate range of temperatures are also required. BS 5762:1979
describes an approved test procedure.

Fatigue data which would establish the constant and index in the
Paris crack growth formulae would be useful additional information.

10. ACKNOWLEDGEMENTS

The work reported in this paper was carried out under contract for the Health
and Safety Executive and is more widely covered in the Safety and Reliability
Report, SRD R314, "An Analysis of a 100 te Propane Storage Vessel".

The views expressed are those of the author and do not necessarily reflect the
views or policy of the Health and Safety Executive.

11. REFERENCES

1) PHILLIPS C A G and WARWICK R G. A Survey of Defects in Pressure Vessels

Built to High Standards of Construction and its Relevance to Nuclear
Primary Circuits. AHSB(S)162. 1968.

2) SMITH T A and WARWICK R G. The Second Survey of Defects in Pressure

Vessels Built to High Standards of Construction and its Relevance to
Nuclear Primary Circuits SRD R30. 1974.

3) SMITH T A and WARWICK R G. A Survey of Defects in Pressure Vessels in

the UK for the Period 1962-1978 and its Relevance to Nuclear Primary
Circuits. SRD R203. Dec 1981.

118
 IChemE SYMPOSIUM SERIES No. 93

4) British Standard Institution, Guidance on some Methods for the Derivation

of Acceptance Levels for Defects in Fusion Welded Joints. PD 6493:1980.

5) BLENKIN, R. The Design and Manufacture of Supports for Horizontal

Vessels, Journal of Stain Analysis, Vol 12, No 3, 1982.

6) STANLEY, P AND MABELSON A R. A Three Dimensional Photoelastic Study of

the Stresses in Horizontal Saddle Supported Liquid Filled Storage
Vessels, Journal- of Strain Analysis, Vol 17, No 3, 1982.

7) TOOTH A S, DUTHIE G, WHITE G C and CARMICHAEL J. Stresses in Horizontal

Storage Vessels. A Comparison of Theory and Experiment, Journal of
Strain Analysis, Vol 17, No 3, 1982.

8) IRVINE W H. An Elasto-Plastic Theory of Fracture and its Application to

the Analysis of' Experimental Data, SRD R48, 1977.

9) QUIRK A. Fracture Toughness Characterisation of Steel from Conventional

Tensile and Impact Data, SRD R89, April 1978.

10) ROBERTS R and NEWTON C. Interpretive Report on Small Scale Test

Correlations with K Data. W R C Bulletin 265, Feb 1981, ISSN
C
0043-2336.

11) PETERSON R F. Stress Concentration Design Factors.

12) MILNE J, HARRISON R P, DOWLING A R. Evaluation of a Flaw Tolerance: Use

of Sensitvity Analysis on the Input Data on Constant to Rigid Safety
Factors, CEGB/I Mech E, 1980.

13) ROLFE S T, BARSOM J M. Fracture and Fatigue Control in Structures.

Applications of Fracture Mechanics, 1977.

14) The Application of COD Measurements to Large Scale Test Behaviour. 'A
COWAN, N KIRBY. Symposium on Practical Fracture Mechanics for Structural
Steels.

15) An Assessment of the COD Test. Report of a Collaborative Programme

Undertaken by the NDACCS CODA Panel Symposium on Practical Fracture
Mechanics for Structural Steels.

16) Computational Modelling of Microstructural Fracture Processes in A 533B

Pressure Vessel Steel. EPRI Document No NP 1398.

17) Private Communication. Fulmer Research Institute.

18) AGR Materials Test Programme, UKAEA Internal Document

119
IChemE SYMPOSIUM SERIES No. 93

120
IChemE SYMPOSIUM SERIES No. 93

121
IChemE SYMPOSIUM SERIES No. 93

122
IChemE SYMPOSIUM SERIES No. 93

123