Documente Academic
Documente Profesional
Documente Cultură
0:38 10:40fl
21=3
when fl
21=3
, 1:35
0:78 when fl
21=3
. 1:35
(
1
There are, besides the warnings by Miller, other concerns
about the above for the following reasons:
With no restriction on f, the above solution gives a
different result when the defect is located at one
crown of the elbow than when it is located at the other.
A unique solution would be obtained if a further limit of
2p/2 #f #p/2 was imposed.
With the expression being applicable to `extended'
defects, it only depends on the bend characteristic l
and crack location and not defect size. Because of the
lack of dependence on crack length, it is not strictly
speaking applicable to short cracks. However it can be
argued that the expression should give conservative
estimates for the latter.
The above expression is supposedly based on the
Grifths results yet it seems to contradict some of his
data. At the crown, f 0 and hence the rst part of
the expression applies, for which case M/M
o
0.38 yet
Miller [17] states that Grifths did two tests on cracks
(at the crown) running the entire length of a 908 bend and
these showed that M/M
o
.0.85.
Clearly, there are grounds for reconsidering the above
solution.
4.1.2. Part-penetrating axial defects
Miller [17] indicates that little evidence exists and gives
the following advice:
There is no difference between external or internal
defects.
Shallow cracks have little effect.
Elbows with surface defects can be treated as made up of
two shells, a through-cracked and a plain shell, one inside
the other. This assumes that if limit load solutions for
through-cracked and uncracked shells are known, then
solutions for part-penetrating cracked elbows can be
derived from:
M 1 2
a
t
M
o
1
a
t
M
`
2
where M
o
is the uncracked bend limit moment (full thick-
ness) and M
`
is the through-wall cracked bend limit
moment (full thickness).
Miller [17] also includes results in a graphical form for a
crack at the crown (f 08), which was considered the
worst case, and shows that a simple linear interpolation
between a=t 0 and a=t 1 can give conservative solutions.
4.1.3. Throughwall circumferential defects
As for axial defects, it is stated in Miller [17] that there is
little effect until c=r . 0:5. For defects extending from the
crown towards the extrados, the dependence on circumfer-
ential angle, 2b, was found to be linear, i.e.
M
M
o
1 2
3b
2p
3
Morton and Ruiz [19], quoted in Miller [17], found that the
bend curvature does not affect its strength and suggest that the
straight pipe expressions could be used. For crown defects:
M
M
o
1 1cosb
2
4
For intrados and extrados defects, Miller [17] quotes the
following general lower bound expression:
M
M
o
cos
b
2
2
sinb
2
5
In Ref. [17] Miller quotes the above expressions the other
way round, i.e. Eq. (4) is given for intrados/extrados cracks
and Eq. (5) for crown cracks. However the equations, as
presented above, are believed to be correct.
4.1.4. Part-penetrating circumferential defects
There are no available experimental results for such
defects, and it is speculated that there are no differences
between internal and external defects. Linear interpolation
between a=t 0 and a=t 1 is recommended, that is Eq. (2)
above could be used as for axial cracks.
4.2. Other sources
In Ref. [7], Zahoor summarises solutions for three
specic cases:
an axial surface aw with internal pressure loading,
K. Yahiaoui et al. / International Journal of Pressure Vessels and Piping 79 (2002) 2736 31
an axial throughwall crack with in-plane bending, and
a circumferential throughwall crack with in-plane
bending.
The proposed solution for axial, throughwall cracks in
bending is:
M
M
o
1 20:15 c=D 6
It is seemingly based on the work by Grifths [8], and is said
to be applicable to extrados, crown or intrados cracks when
c=D # 0:9, l , 0:5 and D=t $ 15.
For circumferential throughwall cracks, Zahoor [7] uses
the data from Grifths' work [8] to produce the following
expression which he states is applicable to short and long
radius elbows:
M
M
o
1 20:2137c=D 20:0485c=D
2
21:056c=d
3
7
Using b c=r and D 2r, this can be rewritten as
M
M
o
1 20:34b=p 20:12b=p
2
24:1b=p
3
8
Eq. (8) is compared with the Miller Eq. (3) in Fig. 4 where it
can be seen that, for crack angles (b=p) of the order of 0.3,
Zahoor gives a limit moment that is 42% greater than Mill-
er's. The agreement is better for longer and shorter cracks.
In Ref. [18], Chattopadhyay et al. carried out FE calcula-
tions of cracked 908 elbows. The cracks were throughwall
circumferential and located at the extrados. Semi-crack
angles of 45, 70 and 908 were investigated. A t to the
data was performed and the following expression for limit
moment calculation was suggested:
M
M
o
1 2C
b
p
9
where b is the crack half angle as in Fig. 1 and C a constant
which was found to be 1.52. This is basically the same as the
Miller Eq. (3) above.
5. Evaluation of data
Detailed results from the Liverpool test and analysis
programme have been presented in Refs. [10,20]. One
contribution [10] addressed the components' global beha-
viour by consideration of overall moment/deection
response while the other [20] attempted to give an insight
into the local behaviour by consideration of strain data
around the crack site. In general, the local behaviour data
was lower than the corresponding global data. Only the
latter is recalled for the evaluation to follow, with a
summary of the data in Table 2 where the results are normal-
ised with respect to the uncracked component values to
show the weakening effect of the cracks.
In Ref. [10], both experimental and FE simulation results
were included and it was found that the FE simulations
(which used large deformation analysis and the true stress
vs. strain curve) predicted the experimental data reasonably
well. A typical comparison is included as Fig. 5. Because of
the similarities between the experimental and FE results,
only the former will be used here to avoid repetition, with
the understanding that the conclusions to come will be
generally applicable to both sets of data.
K. Yahiaoui et al. / International Journal of Pressure Vessels and Piping 79 (2002) 2736 32
Fig. 4. Comparison of Miller's Eq. (3) and Zahoor's Eq. (8) for throughwall
circumferential cracks.
Fig. 5. Moment vs. displacement plots for elbow E6 with circumferential
through-wall crack at the intrados (2b 468, a=t 1:0). (Fig. 5 has been
reproduced from the Journal of Strain Analysis, Vol 35 2000 Issue 1 pages
4757. Fig. 9 by K. Yahiaou, D.N. Moreton and D.G. Moffat by permission
of the Council of the Institution of Mechanical Engineers.)
5.1. Defect-free elbow
As mentioned earlier, expressions for calculating limit
loads of defect free pipe bends have been proposed by
various authors [3,11,12]. All these expressions share the
same form and can be written as:
M
o
Al
n
D
2
tS
y
10
where n can be taken as 2/3 [11,12] or 0.6 [3], and A can be
0.8 [3], 0.94 [11] or 1.04 [12]. For the comparisons below, a
middle of the range expression will be used to avoid repeti-
tion since the main conclusion remains the same whatever
the expression chosen, i.e. n and A are taken to be 2/3 and
0.94, respectively.
In Table 3, the Liverpool data for M
o
on a defect-free
bend (l 0.24) together with the relevant data from the
Greenstreet [5] and Grifths [8] works is summarised. The
results in all cases are normalised by the theoretical limit
moment M
op
of the equivalent straight pipe (D
2
tS
y
). Two
observations can be made about the tabulated data:
(1) Eq. (10) underpredicts the experimental data by over
20% in all cases. This is understandable since Eq. (10)
depends only on the bend characteristic, l and ignores the
constraining effect of the tangent pipe, an effect that
becomes increasingly important for small bend radius ratios.
Grifths [8] demonstrated this effect by testing 1808 return
bend under the same loading and found the percentage
difference to be only about 3%. Next he went on to suggest
the inclusion in Eq. (10) of a strengthening factor of 1.33 for
908 elbows. By applying this factor to his results, Grifths
found the percentage difference to reduce to 3% in one case
and an overprediction of 3.3% in the other.
A similar recommendation was made by Miller [17], for
the following correction factor to be applied to Eq. (10) to
take account of the pipe/bend radius ratio:
M
o
M
op
1 20:36r=R f g 11
for 0 , r=R , 0:67, where M
op
is the limit moment for an
uncracked plain pipe.
(2) The Liverpool data appear to show the largest percen-
tage difference in Table 3. This is so because different meth-
ods for the determination of the plastic load were used. At
Liverpool, the ASME [9] twice-elastic-slope (TES) deni-
tion has been used consistently throughout the project as
explained in Ref. [10]. Grifths' results [8] are averages
of values obtained by the two least recommended methods,
i.e. the twice-elastic displacement and the tangent inter-
section method. For the test data reported in Table 3,
Grifths could not apply the ASME method because the
results would have fallen outside the test range. Conse-
quently, it can be argued that, had it been possible to
apply the method, then higher values than those included
in Table 3 would have been reported.
The Greenstreet result is also obtained by the tangent
intersection method, which appears to give consistently,
lower values than the ASME TES method from inspection
of some of the load/deection curves included in Ref. [5].
K. Yahiaoui et al. / International Journal of Pressure Vessels and Piping 79 (2002) 2736 33
Table 3
Comparison of experimental plastic load data M
o
/M
op
with Eq. (10)
predictions for uncracked elbows
Source Experimental Eq. (10)
prediction
% difference
{exp 2Eq. (10)/exp}
100
Greenstreet [5] 0.360 0.285 21
Grifths [8] 0.297 0.217 27
0.268 0.208 22
Present work 0.719 0.364 49
Table 4
Comparisons of predictions with Liverpool experimental data
Test elbow reference Experimental data Miller [16] Zahoor [7]]
Predictions (Eq. no.) %
a
Predictions (Eq. no.) %
a
E0 9.01 4.55 (10) 50
E1 8.94 3.14 (1 and 2) 35 N/A
E2 8.54 3.14 (1 and 2) 37 N/A
E2A 7.77 3.14 (1 and 2) 40 N/A
E3 8.49 1.73 (1) 20 4.44 (6) 52
E7 8.73 3.14 (1 and 2) 36 N/A
E8 8.09 2.44 (1 and 2) 30 N/A
E9 7.83 1.73 (1) 22 4.15 (6) 53
E4 7.84 4.12 (2 and 3) 53 N/A
E6 7.46 3.68 (3) 49 4.30 (8) 58
3.57 (5) 48
E10 8.25 3.42 (2 and 3) 41 N/A
E12 7.75 2.85 (2 and 3) 37 N/A
E11 5.89 2.28 (3) 39 3.28 (8) 56
1.97 (5) 33
a
Units are kNm. % expressed as (predicted/experimental) 100%.
5.2. Cracked bends
In the following, only cases related to opening bending
are addressed. For pressure loading, no attempt at assessing
the validity of the Miller recommendations is made for the
following reasons:
1. No cracked pipe bends were tested under internal
pressure.
2. In all (but one) instances, Miller [6,17] states that the
pipe bend curvature had no effect on the limit pressure
and the bend may be regarded as a cylinder. This is
stated for all part-penetrating and throughwall circumfer-
ential cracks, whatever the location and for longitudinal
(axial) cracks located at the crown or extrados. For the
excepted case of axial cracks at the intrados, Miller
recommends the modication of the plain cylinder solu-
tion by use of a multiplier in the form of a radius ratio
function, f(b) as is done for the defect-free bend.
In performing structural integrity assessments of compo-
nents similar to the elbows tested in this programme, a
practicing engineer will use the R6 procedure [2]. Despite
the cautions given in the latter, most organisations will not
have the resources to experiment or numerically simulate
the cases. Instead, the relevant limit loads will inevitably be
obtained from the solutions compiled by Miller [6,17] or by
use of the advice given therein. For R6 users, an assessment
of the integrity of a structure can lead to complications if the
limit load is not objectively determined.
The validity of the Miller and Zahoor predictions against
the experimental results are assessed in Table 4. In Table 3
the data were normalised while in Table 4 the actual
moments are quoted. This is deliberate since, for the
Table 4 results, there are two interacting effects: (a) the
conservatism of the uncracked bend predictions, and
(b) the presence of the crack. Should the reader wish to
attempt to separate these two effects, the 49% difference
quoted in Table 3 could be used to do this. In fairness to
Miller, the following points are made: (a) because of the
limited data available at the time of the compilation, Miller
had no choice but to err on the safe side in formulating his
advice; (b) Miller only addresses extended defects (in the
expressions cited above, there is no dependence on crack
length); and (c) for part-penetrating defects, the approach
used by Miller was fundamentally conservative, being based
on the lower bound theorem of plasticity.
6. Discussion
Throughout the course of the Liverpool research
programme, it has been noted that short radius piping
elbows under opening bending are rather tolerant of
crack-like defects. The quantitative evidence to support
this view will be summarised below. From a qualitative
standpoint, Figs. 6 and 7 show the two elbows with the
largest throughwall cracks on completion of testing. In
both cases, the defect mouth width before testing was
1 mm. It can be seen that for the axially cracked component
E9 (Fig. 6) the crack has opened due to a mixture of fracture
mechanics modes I and III. As can be seen in Table 2, even
with such a large throughwall crack, the TES plastic load
was 87% of the uncracked value. For the circumferentially
cracked component E11 (Fig. 7) the crack opening mode is
clearly mode I with, in this case, a TES plastic load of 65%
of the uncracked value (Table 2). In both of these compo-
nents, and indeed in all other elbows tested, signicant crack
propagation did not occur until after the TES plastic load
had been reached. This is illustrated in Fig. 5 for component
E6 where the TES plastic load is about 7.5 kNm and crack
propagation clearly commences at about 10 kNm.
Considering the Liverpool results presented in Table 2,
two observations were made in Ref. [10] and are repeated
here:
1. The FE calculations predict the experimental results
reasonably well. This is evidence to conrm that the
K. Yahiaoui et al. / International Journal of Pressure Vessels and Piping 79 (2002) 2736 34
Fig. 6. Elbow with axial through-wall crack after testing. Fig. 7. Elbow with circumferential through-wall crack after testing.
FE simulation method used has been reliably validated
and further FE calculations on different bend geometries
and similar defect sizes need not be supported by more
expensive testing.
2. Compared with the uncracked geometry, the defects
considered in the Liverpool programme do not appear
to cause a dramatic reduction in plastic loads except for
the case when the defect is throughwall, circumferential
at the intrados and of length in excess of a pipe radius.
The data from Table 2 is used to support this observation
in Fig. 8.
The comparison between the Liverpool experimental data
and the predictions by existing solutions is summarised in
Table 4. The Zahoor predictions, for the cases for which
they are applicable, are consistent and are in the range
5060% of the experimental results. The Miller predictions,
on the other hand, are consistently much lower and vary
from as low as 20% up to 53% of the experimental values.
The lowest underpredictions are surprisingly for the
throughwall defects, components E3, E9 and E11. The
exception to this is throughwall component E6 where the
Miller prediction is about 50% of the experimental value.
For completeness, it is recalled that all Liverpool test
components were of the short radius type. While some theo-
retical works do not distinguish between long and short
radius geometries, there are instances where distinctions
between the two are made. The Grifths 1.33 strengthening
factor mentioned in Section 5.1 above or the modied Miller
solution (Eq. (11)) are examples. However, perhaps the
most serious evidence for this distinction to be made can be
appreciated by reference to the Grifths results included in
Fig. 9. There are two comments to make on the latter gure:
1. The short radius results (curve R/r 2) are similar to the
K. Yahiaoui et al. / International Journal of Pressure Vessels and Piping 79 (2002) 2736 35
Fig. 8. Summary of experimental data: (a) axial defects at the crown; (b) circumferential defects at the intrados.
Fig. 9. Effect of axial defect length on limit moment (from Grifths [8]).
Liverpool results and consequently the observations made
earlier are conrmed.
2. The long radius geometry (curve R/r 3) appears to be
more vulnerable when the defects are of length one pipe
diameter or longer.
7. Conclusions
Considering the low limit moments predicted by existing
methods compared with the experimental plastic moment
data now available for short radius elbows, it can be safely
concluded that the former are unduly conservative. The
present experimental evidence, together with the results
obtained by Grifths for the short radius case, indicate
that such elbows can tolerate, in most instances, large
defects with little reduction in limit load capacity. Careful
examination of existing and present experimental data indi-
cate that, from a limit load point of view:
1. Axial defects with opening bending, whatever the size or
location, are not detrimental to short radius elbows.
2. Long throughwall axial defects at the crown can cause
some reductions in limit loads under closing bending.
3. Long throughwall circumferential defects at the intrados
(under opening bending) or at the extrados (under closing
bending) are the most critical defects. It is recommended
that this type of defect needs to be focussed on in any
future investigations.
4. Axial or circumferential, part-penetrating defects, of
the sizes investigated, do not signicantly weaken
short radius elbows under opening bending. This
comment is based solely on the limited Liverpool
results because no other similar data exist in the open
literature. It is recommended that `substantially' larger
part-penetrating, external and internal, defects be given
some consideration in future to conrm, or otherwise, the
above observation.
Acknowledgements
The work reported here was supported by British Energy
Ltd on behalf of the IMC (British Energy and BNFL). The
close collaboration with colleagues at British Energy and
the UK Nuclear Installations Inspectorate on many aspects
of the work is gladly acknowledged.
References
[1] Dowling AR, Townley CHA. The effect of defects on structural fail-
ure: a two-criteria approach. Int J Pres Ves and Piping 1975;3:77
107.
[2] Assessment of the Integrity of Structures Containing Defects, British
Energy Generation Ltd, Document Ref R6, Revision 4, 2001.
[3] Spence J, Findlay GE. Limit loads for pipe bends under in-plane
bending. Proc 2nd Int Conf on Pressure Vessel Technology, San
Antonio, 1973. p. 3939.
[4] Rodabaugh EC, Moore SE. Evaluation of the plastic characteristics of
piping products in relations to ASME Code Criteria, NUREG Report,
CR-0261, ORNL, 1978.
[5] Greenstreet WL. Experimental study of plastic responses of pipe
elbows, ORNL-NUREG Report 1978:24.
[6] Miller AG. Review of limit loads of structures containing defects. Int
J Pres Ves and Piping 1988;32:197327.
[7] Zahoor A. Ductile Fracture Handbook. Vols. 13, Electric Power
Research Institute, EPRI NP-6301-D/N14, Palo Alto, California,
USA, 1989, 1990 and 1991.
[8] Grifths JE. The effect of cracks on the limit load of pipe bends under
in-plane bending: experimental study. Int J of Mech Sci
1979;21(2):11930.
[9] ASME Boiler and Pressure Vessel Code, Section III, 1995.
[10] Yahiaoui K, Moffat DG, Moreton DN. Piping elbows with cracks.
Part 2: global nite element and experimental plastic loads under
opening bending. J Strain Anal 2000;35(1):4757.
[11] Calladine CR. Limit analysis of curved tubes. J Mech Engng Sci
1974;17(2):8587.
[12] Goodall IW. Lower bound limit analysis of curved tubes loaded by
combined internal pressure and in-plane bending moment, CEGB
Report, RD/B/N4360, August 1978.
[13] Shalaby MA, Younan MYA. Limit loads for pipe elbows with internal
pressure under in-plane closing moment. ASME, J Press Ves Technol
1998;120:3542.
[14] Shalaby MA, Younan MYA. Limit loads for pipe elbows subjected to
in-plane opening bending moments. ASME, J Press Ves Technol
1998;121:1723.
[15] Chattopadhyay J, Nathani DK, Dutta BK, Kushwaha HS. Closed form
collapse moment equations of elbows under combined internal pres-
sure and in-plane bending moment. ASME, J Press Ves Technol
2000;122(4):4316.
[16] Yahiaoui K, Moffat DG, Moreton DN. Damage assessment of piping
elbows loaded by steady internal pressure and dynamic in-plane or
out-of-plane bending, Proc 8th Int Conf on Pressure Vessel Technol.
Vol. 1, Montreal, ASME, 1996. p. 36175.
[17] Miller AG. The plastic collapse of cracked pipe bends under internal
pressure or in-plane bending, CEGB Report, TPRD/B/0806/R86,
December 1986.
[18] Chattopadhyay J, Dutta BK, Kushwaha HS, Mahajan SC, Kakodkar
A. Limit load analysis and safety assessment of an elbow with a
circumferential crack under a bending moment. Int J Pres Ves and
Piping 1995;62:10917.
[19] Morton J, Ruiz C. Through cracks in pipes and pipe bends. Proc
IMechE Conference, Paper C89/80, 1980. p. 21722.
[20] Yahiaoui K, Moreton DN, Moffat DG. Local nite element and
experimental limit loads of cracked piping elbows under opening
bending. Strain, J BSSM 2000;36(4):17586.
K. Yahiaoui et al. / International Journal of Pressure Vessels and Piping 79 (2002) 2736 36