You are on page 1of 28

651

APPLIED MECHANICS GROUP

A STUDY OF NOZZLES IN PRESSURE VESSELS


UNDER PRESSURE FATIGUE LOADING
By J. Spence, B.Sc., M.Eng., C.Eng., M.I.Mech.E.*,
and W. B. Carlson, B.Sc., D.Sc., C.Eng., F.1.Mech.E.j.

Nozzles in cylindrical vessels have been of special interest to designers for some time and have offered a field
of activity for many research workers. This paper presents some static and fatigue tests on five designs of full
size pressure vessel nozzles manufactured in two materials. Supporting and other published work is reviewed
showing that on the basis of the same maximum stress mild steel vessels give the same fatigue life as low alloy
vessels. When compared on the basis of current codes it is shown that mild steel vessels may have five to ten
times the fatigue life of low alloy vessels unless special precautions are taken.

INTRODUCTION from Mn-Cr-Mo-V plate were denoted D1, D2, D3 and


OVERTHE PAST DECADE there has been an increase in the D4 and that from mild steel, M1. Each vessel contained
number of low alloy steel pressure vessels fabricated. three identical mild steel nozzles, the type of nozzle detail
Moreover many design codes are now allowing design varying from vessel to vessel as follows:
stresses tied to the value of the yield point or proof stress. Vessel D1-Mn-Cr-Ma-V drum with 3 identical
An investigation on Mteen full size steel nozzles in five push-outs attached to mild steel nozzles, Fig. 2.
cylindrical vessels, whose static mess patterns were Vessel D2-Mn-Cr-Mo-V drum with 3 identical
measured and which were subsequently pressure fatigued mild steel set-in nozzles, Fig. 3.
to destruction at about ambient temperature, may there- Vessel D3-Mn-Cr-Mo-V drum with 3 identical
fore be of interest. mild steel set-on nozzles, Fig. 4.
One of the objects of the investigation was to assess the Vessel D4-Mn-Cr-Mo-V drum with 3 identical
relative merits of five designs of large bore connection in mild steel through-welded nozzles with external
thick cylindrical vessels. Each design was tested in tripli- Mn-Cr-Mo-V reinforcing pad, Fig. 5 .
cate in one vessel, four of the vessels being manufactured Vessel M1-Mild steel drum with 3 identical mild
from Mn-Cr-MeV plate steel and one from mild steel. steel push-out nozzles, thicker and slightly different in
Supporting work carried out on small model steel shape from those in vessel D1 above, Fig. 6.
vessels and on rotating bending Wohler type fatigue
specimens is included. Other published work on pressure Representative chemical analyses of the plates are given in
fatigue testing of vessels is reviewed. Table 1.
Each vessel consisted of a 30 in inside diameter cylinder
EXPERIMENTAL VESSELS (19 in thick in the alloy vessels and 1 3in thick in the mild
Fig. 1 shows the layout of each test vessel and Figs 2-6 steel vessel) by 4 ft 0 in long closed at the ends by tori-
give details of the nozzle shapes. The vessels manufactured spherical drumheads 2 in thick, access being provided by a
manway in one drumhead on each vessel. The forged and
The MS.of this paper was received at the Institution on 5th June machined nozzles were 6% in bore x 3 in thick except the
I967 and accepted for publication on 16th November 1967.
* Department of Mechanical Engineering, University of Strathclyde, nozzles in vessel M1 which were l& in thick. Vessels D1,
Glasgow, C I , formerly Research and Development Division, D2, D3 and M1 had no added compensation, but in
Babcock and Wilcow (Operations) Ltd, Renfrew.
t Research and Development Division, Babcock and Wilcox (Opera- vessel D4 external plate type reinforcement of Mn-Cr-
tions) Ltd, Renfrew. Mo-V steel was added to give 100 per cent compensation
Proc Instn Mech Engrs 1967-68 Downloaded from pme.sagepub.com at City University Library on April 25, 2016 V Oiaz
~ Pt I N~ 31
+
658 J. SPENCE AND W. B. CARLSON

Fig. 1. Details of test vessels

LONGITUDINAL SECTION CI RCUMFER ENTl A L SECTION

I Stress in ton/in2 at 2500 lb/in2 gauge

1 Nozzle 1 I Nozzle 2 I Nozzle 3 i Mean


Location 1 Circumferential* I Meridional I Circumferential IMeridionall Circumferential 1 Meridionall 1
Circumferential Meridional

3
4
30.0

16.5
27.0
0.8
7.2
1-4
1
1______-
1
30.1

16~6
28-8
1
(-0.4
__
4.6

7.2 14.3
29.7 1 1.7
:::: 1
28.6
1.8

7.1

1.2

I 1
-
6 I 276

10.4 ~
3.6

5.4 1 253

10.4
25.1

9.7 1
3.4

5.0 1
26.1

10.1 1 3.4

5.3

10.8 1 8.2 I
I
8.3 6.3 1 11.4 I 9.6 I 10.1 1 ~

8.0
~~

* Circumferential is with respect to the nozzle or shell as the case may be.
Numbers in parentheses refer to gauge locations diametrically opposite.
Fig. 2. Strain gauge locations at push out nozzle on vessel DI
Proc Instn Mech Engrr I96748 Downloaded from pme.sagepub.com at City University Library on April 25, 2016 Vol I82 Pt I No 31
I

Stress in ton/inz a t 2500 Ib/in2 gauge


Nozzle 1 Nozzle 2 Nozzle 3 Mean
Location

2
1
I
26.1
27-4
1I 3.8
1.1 1
31.3
30-5
II 4.9
1.4
1I
Circumferential Meridional Circumferential Meridional Circumferential Meridional Circumferential Meridional

1 27.0
27.1
1I 3.6
3.4 I
28.2
28.3
I
I
4.1
2.0
~-
3 24.5 2.1 23.7 - 20.1 - 22.8 2.1
4 29.1 3.0 26-2 1.1 26.7 0.8 27.4 1.6
5 24.3 3.4 29.9 5.9 22.5 - 25.5 4.7
6 11.6 7.4 1 10.7 6.3 11.3 6.8
7 8.8 9.6 6.2 9.6 9.3 5.0
-
8 4.9 I 1.4 4.3 1 5 3 3.2 I 0.9 4.1 1.2

~ ~~ ~ ~~

Fig. 3. Strain gauge locations at set-in nozzle on vessel 0 2

Downloaded from pme.sagepub.com at City University Library on April 25, 2016


660 J. SPENCE AND W. B. CARLSON
t
Proc Instn Mech Engrs 1967-68
Downloaded from pme.sagepub.com at City University Library on April 25, 2016
Vol182 Pt I No 3
I Stress in ton/in2 at 2500 lb/in2 gauge

Location
I1 Nozzle 1
1
Nozzle 2
1
i Nozzle 3
I
Circumferential Meridional Circumferential Meridional I Circumferential Meridional Circumferential Meridional
Mean
1

1 1
2
3
4
1
I
29.2
21.6
29.2
I
I
4.6
10.0
4.7
1
I
33.2
16.6
29.3
_____-_
i
1
-
7.0

5.1 1
29.9
21.4
31.1 I
4.5
17.5
5-7
j
~

I
30.8
19.9
29.9
I
I
4.6
11.4

5.1
~

1 1 1
5
6 1 28.0

4-1 1 6.8

2.5 I 28.4

3.0 1
7.4

0.9 1
27.9

1.4 1 6.6

0.0 1 28.2

2.9 1
7.0

1.2

8 3.2
--
4.3 3.2 1 5.0 ~ 2.8 4.5 3.0 I 4.6
1
9
10
3.8

7.2
1-3

7-2
3.4

7.2 1-16j 2.8 3.9

7.6
3.0
7.0
3.7

7.4 1 2.4
7.1

12 I 4.2 4.6 4.7 12.0-


Fig. 5. Strain gauge locations at through welded nozzle on vessel 0 4

Downloaded from pme.sagepub.com at City University Library on April 25, 2016


I Stress in ton/in2 at 2500 Ib/inz gauge

$
Location
Nozzle 1 I Nozzle 2
1
1
Nozzle 3
Circumferential Meridionall Circumferential Meridional Circumferential (Meridional Circumferential Meridional
1
i Mean
1 4
1

2 1 28.6
30.3 1 5.3
2.6 I 29.9
31-1 ~
5.4
5.1
29,7
29.5
1 ~
5.9
3.4 30.4
w

5
6
30.7

29.3
8.2
11 3.9

3.7
4.6
1 28.8

28.7
8.2
1.8

5.7
3.7
31.7

26.4
6.2
4.5
I 30.4

28.2 1 3.4

4.2
3.8

10 1 10.3 1 5.7 i 9.3 1 8.7 1 9.1 5.0 I 9.6

Fig. 6. Strain gauge locations at push out nozzle on vessel MI

Downloaded from pme.sagepub.com at City University Library on April 25, 2016


A STUDY OF NOZZLES I N PRESSURE VESSELS UNDER PRESSURE FATIGUE LOADING 663

.I 1 1 1
Table I . Chemical analyses

Material . . _
C
_ _
Si
_
I M n I
~ _ _
S
_ _ _
P
_ ~
Ni 1 Cr Sn
Mn-C-Mo-V drums 0.16 0.10 1.42 0.031 0.018 < 0.15 I 0.38
M.S. drum . .I 0.19 I 0.10 I 0.57 1 0.053 1 0.034 I < 0.15 I < 0.15 1 0.04 I < 0.15 I - 1 0.02

M.S. nozzles . .1 0.19 I 0.10 I 0.74 ~ 0.031 I 0.027 I < 0.15 1 < 0.15 I 0.02 I < 0.15 I - I 0.03

Table 2. Dimensiunal details

Vessel Material Internal Thickness Internal Thickness T/D


diameter of shell, diameter of nozzle,
of shell, Tin of nozzle, t in
Din d in
D1,D2, D3 and D 4 Mn-Cr-Mo-V 30 13 63 4 0.054
MI . . . M.S. I 30 I 1% I 63 1 l-& I 0.065 I 0.162 I 2.5 1 0.225

i.e. the cross-sectional area of material removed to provide and .during testing were switched in turn to a strain
the opening was replaced as reinforcement, distributed in recorder.
this case in the shell, the nozzle and the external plate Pressure was raised in increments to a maximum test
(B.S. 1113 : 1951-the current code when the nozzles pressure of 2900 Ib/in2 gauge, strain gauge readings being
were designed). Vessel and nozzle dimensions together taken at each stage.
with parameters of interest are given in Table 2.
The alloy steel vessels were stress relieved for 2+ h at Fatigue testing
675°C and the mild steel vessel for 3 h at 600-650°C. Hydraulic fatigue testing was carried out with a pressure
Measurements of out-of-circularity were taken at several cycle of zero to 2500 lb/in2 gauge, at a rate of approxi-
sections in each vessel by a special gauge which has been mately 12 c/min. I n the alloy steel vessels most of the
described elsewhere (I)*, the difference between maxi- testing was carried out at 70°C although the early cycles
mum and minimum diameters not exceeding 1 per cent. were applied at ambient temperature. It was not considered
The bending stresses associated with the out-of-circularity that the difference between these two temperatures would
were calculated assuming an unpierced vessel, but the materially affect the fatigue life of the vessels. The mild
values were so low that they have not been considered steel vessel was tested at ambient temperature throughout.
here. The fatigue testing equipment was a Werner and
Phleiderer unit consisting of a 3-throw hydraulic pump,
TESTING higher and lower limit pressure accumulators and a servo-
Testing was in two parts, an initial strain gauge test mechanism controlling the pressure supplied to the vessel
designed to measure the general stress pattern in the under test.
nozzles and confirm the expected stress levels, followed by I n the case of alloy steel vessels testing was continued
a hydraulic pressure fatigue test. until one of the three nozzles failed, so that each result
gave in fact the minimum life of three. In the case of the
Static testing mild steel vessel, when failure occurred the nozzle in
A total of 312 electric resistance strain gauges were question was removed, the vessel patched, and testing
attached to the five vessels in the nozzle areas as shown in resumed until results had been obtained for all three
Figs 2 to 6. Teddington gauges having a 12.5 mm gauge nozzles. Failure was considered to have occurred when a
length x 4 mm width and a 200 Q nominal resistance were crack had propagated through the wall of the vessel or
used in the form of right angle pairs. A temperature nozzle causing leakage.
compensating gauge mounted on a thin piece of plate Following a nozzle failure in any vessel, a magnetic
material was used with every active gauge. These plates crack detection survey was conducted on the three nozzles
were placed adjacent to their corresponding active gauges for signs of other cracking. Only in one case was a crack
and attached in such a way that no strain was transmitted found other than the one causing failure, namely, in
to the compensating gauges. Transformer oil was used as nozzle No. 2 of vessel M1 where a crack was detected
the pressurizing fluid to obviate difficulties of water- diametrically across the nozzle from the failure crack,
proofing. All gauges were connected to switching units i.e. at an identical position in the same nozzle. It was
presumed that all the failure cracks had propagated
* References are given in the Appendix. through the thickness in a small number of cycles.
Proc Instn Mech Engrs 1967-68 Downloaded from pme.sagepub.com at City University Library on April 25, 2016 VO!182 Pt I No 31
664 J. SPENCE AND W. B. CARLSON

Metallurgical examinations were made of the failure


surfaces after testing.
RESULTS
The majority of the strain/pressure graphs were linear
but a few exhibited non-linearity, notably in vessels D4
and M I where the material in the highest stressed region
was mild steel. Strainlpressure graphs are shown in
Fig. 7 for two of the locations showing non-linearity. At
pressures denoting the onset of non-linearity the stresses
were at least equal to, and usually slightly greater than,
that required by the von Mises yield criterion to predict
onset of yield for the mild steel. At similar high stress
regions in alloy material the stresses were not high enough
to predict yield by the normal yield criteria.
From the slope of the strain pressure graphs, stresses
were calculated using a value of Young’s modulus of
13 400 ton/in2 and Poisson’s ratio of 0-3. A11 stresses or
stress ranges, as the case may be, at 2500 lb/in2 gauge are
tabulated in the tables associated with Figs 2 to 6, along-
side the respective gauge locations, and the stress distribu- PRESSURE-lb/in2 gauge
tions at 2500 lb/in2 gauge are given in Fig. 8. The values
plotted are average values for the three nozzles. No Fig, 7. Typical strain-pressure plots for gauges showing
account was taken of the effect of pressure on the strain non-linearity
gauge readings.
Table 3 summarizes the results of the fatigue testing
and includes the maximum stress range measured on each ally similar D1. Because of different thicknesses a direct
vessel together with the stress concentration factors based comparison cannot be made. It is doubtful if the chamfer
on the mean diameter circumferential stress in the plain on thc crotch of M1 has any significant effect since it
shell. I n no case was there any metallurgical evidence of would tend to reduce rather than increase the stress. The
defects having initiated the fractures, all failures being in values in MI are considered above suspicion since all
regions of high stress. The failures are shown in Figs 9 to gauges in the high stressed region were replaced and a
14. The failure in vessel D4 was at the edge of the rein- second set of gauges accurately reproduced the original
forcement, Figs 12 and 13, where the stress is not ac- readings.
curately known. The maximum measured stress at the
crotch of the vessel D4 has been quoted in the results Relevant results from other tests
although it is probably slightly less than the stress at the The authors’ company has a history of fatigue testing on
failure origin. It will be observed that the stress concentra- various types of component and a few are included as
tion factor (SCF) in vessel M1 is greater than in the nomin- they are relevant to the present subject.

Table 3. Full size vessels, static and dynamic results

II
Vessel Basic shell stress Maximum elastic SCF Number of cycles
at 2500 lb/in2 gauge
in lb/ina
stress range in
Ib/ina I to failure


Experimental Leckie-Penny From Fig. 18
(average of 3)
D1 24400 68 300 2.8 117 085*

1 24400 68 300 2.8 2.7 2.9 I12 776*


D2
D3 ‘ I 24400 I 72800 3.0 108 336*
D4 1 24400 70 600 2.9 138 235*

1- 113 816

I
20 600 70 100 3.4 2.6 2.9 201 138

I I
1 301 141

* Lowest of three.
Proc Instn Mech Engm 1967-68 Vd 182 P t I No 31
Downloaded from pme.sagepub.com at City University Library on April 25, 2016
A STUDY OF NOZZLES IN PRESSURE VESSELS UNDER PRESSURE FATIGUE LOADING 665

30
N 25
._
-2 20
1

I
v,
15
2
cr
10
I-
* 5
0

+ Inside rneridional.
d Outside meridional.
Inside circumferential.
o Outside circumferential.
Fig. 8. Stress distributions at 2500 Iblin2 gauge
Proc Znstn Mech Engrs 3967-68 Downloaded from pme.sagepub.com at City University Library on April 25, 2016 Yo1 182 Pt 1 No 31
*+
666 J. SPENCE AND W. B. CARLSON

Fig. 9. Fracture at the nozzle in vessel DI


( i ) Torispherical drumhead 3 in and the material was mild steel. This test was carried
A Stress analysis and fatiguetest was carried out on a out because a similar head had seen long service in America
4 ft 14 in i/d 57 in inside 1 in thick (2) and the small knuckle radius gave rise to a higher than
torispherical drumhead. The inside knuckle was usual stress concentration factor. The maximum measured
SCF was approximately 7-25 and the maximum measured
stress range at 336 lb/in2 gauge (on the inside of the
knuckle in the meridional direction) was 60 000 lb/in2.
The head was subjected to a pressure test of 576 lbjin"
gauge for one cycle and thereafter cycled between zero
and 336 lb/in2gauge at 16 c/min until, after 184 000 cycles,
failure occurred at the knuckli, Fig. 15.

(ii) Model drumhead tests


At the same time as the present nozzle tests a number of
small-scale plain torispherical drumheads were pressure
fatigue tested at room temperature. Five steels were
tested :
Mild steel
C-Mn plate steel rolled and tempered before pressing.
Mn-Mo plate steel rolled and tempered before
pressing.
Mn-Cr-Mo-V plate steel rolled and tempered before
pressing.
Mo-B plate steel as received.
Although the tests were on a small scale, normal
manufacturing procedure was followed as far as possible
and all models were stress relieved after pressing and weld-
ing of the heads to the shells. All heads were made using
Fig. 10. Fracture at the nozzle in vessel 0 2 the same die and the stress concentration factor for each
Proc Instn Mech Engrs 1967-68 Downloaded from pme.sagepub.com at City University Library on April 25, 2016 Vol I82 Pt I No 31
A STUDY OF NOZZLES IN PRESSURE VESSELS UNDER PRESSURE FATIGUE LOADING 667

Fig. 11. Fracture at the nozzle in vessel 0 3

Fig. 12. Fracture at the nozzle in vessel 0 4

Proc Iiastn Mech Engrs 1967-68 Downloaded from pme.sagepub.com at City University Library on April 25, 2016 Vol182 Pt I No 31
668 J. SPENCE AND W. B. CARLSON

Fig. 13. Fracture at the nozzle in vessel 0 4

head was about 1.8. Three heads were given a static over- service temperature for boiler drums. The U-notched
pressure test to 3050 Ib/in2 gauge and fatigue tested from specimens had a theoretical elastic SCF of 1.78 based on
0-2000 lb/in2 gauge. Ten were statically overpressure the nominal bending stress. The nominal stress for either
tested to 5000 lb/in2 gauge and fatigue tested at 0-4000 type is referred to d or d, in Fig. 16. All the results are
lb/in2gauge. Although only the mild steel and the Mn-Cr- included in Fig. 16 since several interesting trends can
Mo-V steel results are strictly relevant here, all of the be observed, although only the mild steel and Mn-Cr-
results are given in Table 4 for interest. All failures Mo-V steel results at 20°C will be referred to in later
occurred from the inside surface of the knuckle in the discussion. Between eight and twenty specimens were
circumferential direction consistent with the known used for each S-N curve and scatter was not excessive,
maximum stress region and direction. little difference being noted in scatter between the dif-
ferent steels. Similar small scale results on eight steels
(iii) Rotating bending tests have been presented by Wellinger and Kussmaul (3) but
Plain and U-notch specimens of mild, 1818 stainless, have not been considered here.
Mn-Cr-Mo-V and Cu-Ni-52 Mo steels were machned The amount of information published on small-scale
from $ in thick plate and fatigue tested a t 20°C and 350°C specimen fatigue tests is enormous but in this paper
using rotating bending machines of the cantilever type. attention is restricted to fatigue tests on pressure vessels
350°C was chosen to correspond approximately to the under internal pulsating pressure loading.
Testing of pressure vessels has been carried out at
Southwest Research Institute on full size vessels and at
Ecole Polytechnique on model vessels, under the auspices
of the Pressure Vessel Research Committee and reported
among others by Kooistra et al. (4) ( 5 ) (6) arid by Welter
and Dubuc (7). In Britain the British Welding Research
Association (B.W.R.A.) carried out extensive fatigue
testing mainly on nozzle configurations and these have
been reported by Lane and Rose (8) and Taylor (9). Work
is available in Europe reported by Soete (IO), Hierman
and Stockman (11) and Stockman (12)~although some-
times the test conditions were variable and the stress
ranges unknown. Unpublished work from Germany on
model cylindrical vessels with nozzles has enlarged the
European contribution.
The authors' results and as many of the above as are
relevant have been plotted in Fig. 17 as semi-stress range
against number of cycles to failure. The results in refer-
ence (12)cover a large number of tests in mild steel using
set-on nozzles in cylinders but have been omitted for the
sake of clarity. They cover the same scatter band as that
shown by all the other mild steel results in Fig. 17. A
proposed design curve for fatigue, taken from the A.S.M.E.
code for Nuclear Vessels, has been added. Since the basis
of comparison of multi-axial fatigue tests is a matter for
Fig. 14. Fracture at the nozzle in vessel M I discussion (13) the results have been simply plotted as
Proc Instn Mech Engrs 1947-68 Vol182 Pt I No 31
Downloaded from pme.sagepub.com at City University Library on April 25, 2016
A STUDY OF NOZZLES IN PRESSURE VESSELS UNDER PRESSURE FATIGUE LOADING 669

Fig. 15. Torispherical drumhead failure

functions of the maximum stress. No attempt has been flush) in cylinders, except some of the B.W.R.A. work (9)
made to correlate the results with curves derived from which utilized spherical vessels with circular holes, and the
simpler data (14)(15) but enough information has been authors’ drumheads. It would have been possible to
included to allow anyone interested to do so. Table 5 include fatigue results on other types of geometry and
summarizes approximate material data. A few liberties loading as Ruiz and Bickell(16) have done, but these have
have been taken with some of the data to allow it to be used not been considered adductive to the following discussion.
and individual points from some authors may be slightly The complications of comparing results from such dif-
misplaced due to the difficulty of allocating accurate peak ferent sources are many, but qualifying remarks will be
stress ranges, particularly in model vessels. However, the kept to a minimum to ensure clarity even at the expense
errors involved are not likely to affect the general picture. of completeness. The surface condition, physical size
The results given in Fig. 17 relate to nozzles (often and/or stress gradient would need to be considered for a
Proc h s t n Mech Engrs 1967-68 Downloaded from pme.sagepub.com at City University Library on April 25, 2016 Vol I82 Pt I No 31
*
670 1. SPENCE AND W. B. CARLSON

Table 4
(a) Model drumhead fatigue tests.
I I

1
I

Material Ref. No. Basic shell stress, Maximum stress, I NO. of cycles
Ib/ina range to failure
lb/ina
Mild steel. . 1* 18 000 36 000 2.0 414 000
669 000
2f 19 000 36 000 1.9
3 36 000 57 400 1.6 63 400
4 36 000 53 800 1.5 57 300

C-Mn steel 1* 19 000 34 300 1.8 1 462000


42 500
2
3
36 000
35 500
58 000
64 000
1.6
1.8 1 38900
Mn-Mo steel . 1
j 36 500
36000 1 58 500
53800
1.6
1.5 I 71 100
85200
' 1
I
I
Mn-Cr-Mo-V steel 1 36 000 I 57 400 1a6 72 300
.I 2 36 000 1 64 500 I 1.8 38600
Mo-B steel . 1 35 500
36 000
53 200
57 800
I 1.5
1.6 I 84 300
65900

* Pressurized at 3050 lb/inz gauge and fatigued at 2000 Ib/in2 gauge. Remainder pressurized at 5000 Ib/in2 gauge and fatigued at 4000 lb/ina
gauge.
(b) Model drumhead, chemical composition.
- ~ -
Material
_______
Mild steel . .
C-Mn steel .
Mn-Mo steel .
Mn-Cr-Mo-V steel .
Mo-B steel .
I

Table 5. Material properties for collected results


1
Source Material U.T.S., Yield, Elongation , Reduction
I lb/inz lb/inz per cent of area,
I I per cent
I --
1 -
B.andW. . I M.S. 1 60 500 38 000 36
~~ ~

B.W.R.A. (8) (9)


' I M.S.
Supertough
Mn-Cr-hlo-V
~
-
91 000
114 000
38700
66000
93 000
1I 24
18 6
___ ~

, - - ~ _ _
S.M.R.I. (6) and Ecole Polytechnique (7)

-
I
I
A.201
A.302
T-1
I

I
55000
95 500
113600
35 300
77 000
103 600 20
- ~
60
_ _
64

-
German (unpublished). , . ~

~
St.35
Cu-Ni-60Mo
--I-
1 74500
89000 -1 37 000
56 500 1 38
25
-
63
56
-
. . 1
Belgian (10) M.S. vessel
nozzles , 64000
58 600
43 000
36 000
i
36
36 -

Proc Instn Mech Engrs 1967-68 Vol I82 Pt 1 No 3I


Downloaded from pme.sagepub.com at City University Library on April 25, 2016
671
A STUDY OF NOZZLES IN PRESSURE VESSELS UNDER PRESSURE FATIGUE LOADING

80 000

70 000

N 60000
._
\
-
n
I
2 50000
4
LL

In
Vl

40000
v)
-
5
In
_1 30000
<
-
z
5z
20 000

10 000

Fig. 16. Rotating-bending fatigue tests; M.S., Mn-Cr-Mo-V, Cu-Ni-52 Mo, S.S.

detailed comparison. The T-1 model vessels had the nozzle bearing in mind the limitations of many theoretical shell-
crotch ground and polished and the SCF used was a mean centreline methods, though not all (25), when dealing with
value of the measurements reported. In the authors' model a sudden change in geometry as in the case of nozzles.
tests and in the German unpublished work the stresses To obtain a check on Rose's suggestion the available
were calculated. All other stresses were obtained by strain literature has been studied. No general trends common to
gauge techniques although it is not certain that the maxi- all the results are apparent except rather obvious ones
mum stress was measured in every case. This is particularly which help little. However, by restricting the study to
so in the Belgian results where some oblique and offset nozzles of identical shape, with the maximum stress at the
nozzles were used. In a nozzle of type C (60" obliquity) same location, tested by one organization, reported in a
for instance it would be difficult to ensure that the maxi- similar fashion and therefore directly comparable, it was
mum stress was being measured. possible to present the results and sketch in a graph
similar in form to the theoretical type used by Leckie and
Penny (24). Flush constant thickness nozzles under in-
DISCUSSION
ternal pressure tested by B.W.R.A. over the past ten years
Stress levels or so were used and are shown in Fig. 18, together with a
For the purpose of the present investigation the stress theoretical line produced by Eringen et al. (26). The
ranges quoted in Table 3 have been extrapolated slightly experiments cover a fair range of variables from tjT = 0.2
from the measured values as shown in Fig. 8. Reliable to 2.3, rjR = 0.2 to 1.0, R j T = 7.7 to 50 and Fig. 18 has
theories are not yet available to predict values for nozzles been found convenient as a guide when checking SCFs
in cylinders and accordingly the subject has occupied for nozzles in cylinders. The Leckie-Penny curves for
many experimental researchers (17)-(21). Approximate flush nozzles in spheres have been superimposed for
methods of calculation have been available (21) for some comparison.
time and Rose (22) suggested that the theories readily It can be seen that treating the cylinder as a sphere is a
available for nozzles in spheres (23) (24) could be used fair approximation although at low values of ( r / R ) 1 / R / T
directly as a measure of the stress levels at nozzles in it underestimates the SCF (as expected, 2.0 compared with
cylinders, although the maximum stress in the two cases 2-5) and at high values it could overesumate the SCF,
might be at different locations. This has become usual especially for thick nozzles, tlT > 0.5.
practice for want of something better and is included in The experimental points refer to the crotch on the longi-
the British code for nuclear vessels. It is perhaps worth tudinal axis of the cylinder. Theoretical predictions (26)
Proc Instn Meeh Engrs 1967-68 Downloaded from pme.sagepub.com at City University Library on April 25, 2016 Vol182 Pt 1 No 31
672 J. SPENCE AND W. B. CARLSON

Large number o f
mtld steel t e s t s

N - cycles
Fig. 17. Comparison of pressure vessel fatigue results

Fig. 18. Comparison of theoretical SCFs for spheres with experimental SCFs for cylinders
Proz Instn Mech Engrs 1967-68 Downloaded from pme.sagepub.com at City University Library on April 25, 2016
Vol 182 Pt I No-31
A STUDY OF NOZZLES IN PRESSURE VESSELS UNDER PRESSURE FATIGUE LOADING 673

have indicated that for large nozzles the maximum stress useful, since Fig. 19 shows that when vessels are designed
may not occur at this point. Consequently at high r/R on the above basic design stress criteria, the high yield
4 R I T values if rlR is large it may be considered wise to vessels give lower lives than mild steel vessels. For ex-
continue to use the spherical values. Events are changing ample, at a semi-stress range of twice the design stress
rapidly in this field and it may well be that proper curves (i.e. a SCF of about 4) the T-1 full size vessel survived
for the cylinder case will become available in the near about 4000 cycles compared with the A.201 vessels which
future. could be expected to last 60 000 cycles. At most stress
Values of calculated SCF have been added to Table 3 levels there seems to be a factor of between 5 and 10 on the
for both the Leckie-Penny sphere type calculation and numbers of cycles for mild steel as compared with alloy
the present empirical approach. It can be seen that the steels, depending on the source of thejnformation. The
empirical values are marginally better. only case which does not follow this pattern is the American
model vessels which show that the T-1 results are a few
times better than mild steel, but it is known that the T-1
Fatigue considerations models were ground and polished.
It is noticeable that vessel D4 had a slightly higher fatigue Codes which recognize this position demand a proper
life than vessels D1, D2 and D3. This may have been partly stress and fatigue analysis. Particularly good in this respect
due to the fact that the material in the highest stress region is the ASME section I11 which gives a detailed method of
was mild steel and partly due to the failure being at the analysis and B.S. 3915 which also calls for a fatigue analy-
outer periphery of the reinforcement where conditions sis. B.S. 1515 makes tentative recommendations and may
were unknown. In Fig. 17 it can be seen that the present be sufficient in some applications. Where codes do not
full size and model tests form two groups and unex- require a detailed stress analysis (e.g. use area replacement
pectedly the models are the lower. Interesting as they are with consequent high SCFs) or a fatigue analysis, the
in themselves the results can be treated with the remainder fatigue life of an alloy vessel may be significantly lower than
of the results in Fig. 17 which summarizes the majority its mild steel counterpart under the same service conditions
of the known work in the Western hemisphere on fatigue unless special precautions are taken.
testing of vessels both model and full size. The concept of shakedown (28) has been proposed as a
It need hardly be pointed out that the alloy steels design criterion, With current design stress levels of 2/3
showed little or no better fatigue behaviour than the mild yield stress this requires a semi-stress range factor of
steel vessels either on the models or the full size. With the approximately 1.5 (see discussion to reference (28)). It
possible exception of the T-1 model vessels the improve- can be seen that designing on shakedown (or some fraction
ments were marginal. This type of conclusion has been thereof) would not distinguish between steels of different
made often on small scale specimens and Signes et al. (27) yield values. Shakedown design may well safeguard against
have put forward a tentative explanation applicable to incremental collapse and low cycle fatigue, but in the case
small scale fillet welded specimens. They suggested that of alloy steels does not automatically provide a safeguard
fatigue cracking initiates from small defects at the weld against fatigue in the 5000-10 000 cycle region.
material and that the magnitude of the stress concentration At reasonable stress levels most results for even the
associated with the acuity of the defects explains the high yield materials gave more than 10 000 cycles. This
similarity in measured fatigue strengths of welded seems hopeful from an operational point of view since
materials of different static strengths. It is doubtful if this many plants will receive less than 7000 cycles (one per day
explanation could directly apply to the vessels considered for 20 years) during their lifetime, but it should be remem-
here since in many cases the fatigue crack did not com- bered that all of the tests discussed were relatively short
mence at the weld material. For the present purpose it is term where corrosion had little effect. Over 20 years
sufficient to note that there was little improvement of alloy corrosion would considerably affect the fatigue life. Any
over mild steel in the fatigue of vessels, rather than to reduction of the lives indicated in Fig. 19, bearing in
provide a rational explanation of the phenomenon. An mind the unforeseen operational variations that can occur
interesting fact immediately apparent is that the ASME in plant, could well bring designs into a danger region.
design code curve for carbon and low alloy steels more or Offsetting these effects is the influence of temperature.
less provides a lower limit for all the experimental results. For a fair range of temperature above ambient and below
Fig. 17, however, does not reveal all the implications of the creep range, the fatigue life would be expected to be
the above remarks. Present practice allows higher design better than at ambient, but this is not certain for alloys
stresses for high yield material with an overriding factor without experimentation, see Fig. 16. It has been sug-
on the ultimate tensile strength; the permissible design gested that the initial overpressure test experienced by
strength being Proof Stress/l.5 or UTS/2.35, whichever most pressure vessels may have some bearing on the subse-
is the smaller. The actual factor will depend upon the quent fatigue behaviour. Taylor (29) showed that the
particular code in question. Fig. 19 is a replot of Fig. 17 on effect was not very significant, at least for the number of
the basis of Semi-stress range/DesignStress wherethe design cycles and stress levels used in his tests. There are obvious
stress has been taken as the lesser of the above conditions limits to any such procedure and, in addition, side effects
as obtained from Table 5. The idea is not new (3), but is such as change in ductility would need to be considered.
Proc Instn Mech Engrs 1967-68 Downloaded from pme.sagepub.com at City University Library on April 25, 2016 Val 182 Pt 1 No 31
674 J. SPENCE AND W. B. CARLSON

3.25
MODELS FULL SIZE

3 00 ?4 A 2 0 1

2 75
B.W.RA. I
A
MnCr Mo V
Supertough
Mild steel

m i l d steel t e s t s
Drumheads
Present H MnCrMoV
results
fB M i l d steel
2 50 B Mild steel
Notched
} rotating
bending
(Drumhead)

Ln

225 Belgian
k-
m
z
2
2
.a
w
w
200

z
4
oz 1.75
Ln
m
W
(L
I-
m
2
W
I50
l.n

I25
.ommon p e r m
,tress l e v e l

I00 r
nun if B.W.R.A.
tee1

0 75

0 50
lo3 104 lo5 I06
NUMBER OF C Y C L E S

Fig. 19, Code based comparison of pressure vessel fatigue results

The conclusions reached from the above remarks rein- probably introduced a t the edge of the reinforcement,
force those made by Wellinger (30) after testing small- see reference (8).
scale specimens. He summarized as follows: 'The design The stress level in a nozzle in a cylinder can be satisfac-
safety factor (German) must be corrected and adjusted torily approximated from the theories available for nozzles
to suit steels with high yield strengths'. Similarly Payvar in spheres but a closer approximation may be obtained for
and Vasarhelyi (31) after limited testing on small scale flush nozzles by using Fig. 18 of this paper.
specimens of ten steels concluded: 'Increased caution is There is approximate numerical agreement between the
advisable in the use of low alloy high strength and high fatigue performance of the full size vessels with nozzles
strength heat-treated steels'. and rotating bending tests on U-notch specimens. Agree-
ment between model tests on drumheads a'nd the rotating
CONCLUSIONS bending specimens is not good, probably due to change
On the basis of both static and dynamic tests there is little .in shape of the model drumheads.
to choose between the nozzle design of vessels D1, D2, The present results are in agreement with international
D3 and D4. Nozzle D4 with 100 per cent reinforcement in fatigue results, all of which indicate that on a straight
the form of an external plate showed no marked improve- comparison of normal workshop fabricated vessels low
ment over the others. In fact a region of high stress is dloy steels show little improvement over mild steel.
Proc Instn Mech Engrs 1967-68 Downloaded from pme.sagepub.com at City University Library on April 25, 2016 Vol I82 Pt I NO 31
A STUDY OF NOZZLES IN PRESSURE VESSELS UNDER PRESSURE FATIGUE LOADING 675

When vessels are designed to codes where the design Techniques de 1’Industrie des Fabrications Metalliques,
stress is proportional to the yield point but a fatigue Bruxelles. Part I, 1964, Part 11, 1965.
(13) LIBERTINY, G. Z. ‘Short-life fatigue under combined
analysis is not required, the expected fatigue life of an stresses’,Jl Srrain Analysis 1967 2 (No. l), 91.
alloy vessel may be significantly lower than that of its (14) MANSON,S. S. ‘Fatigue: A complex subject-some simple
mild steel counterpart under the same service conditions, approximations’, Experi. Mech. 1965 5 (No. 7), 19A.
unless special precautions are taken. (15) LANGER, B. F. ‘Design of pressure vessels for low-cycle
fatigue’, Am. SOC.Mech. Engrs Paper 61-WA-18.
All the experimental fatigue results reported herein lie (16)RUIZ, C. and BICKELL,M. B. ‘Recent trends in pressure
above the design curve for fatigue taken from Section I11 vessel design’, Commonwealth Welding Conference at
of the A.S.M.E. code for nuclear vessels. B.W.R.A. 1965.
(17)ANON. ‘Experimental and analytical determinations of the
ACKNOWLEDGEMENTS stress systems in a welded pressure vessel’, British Weld-
ing Research Association, London 1951.
The authors wish to thank the management of Babcock (IS)WATERS, E. O., HARDENBERGH, D. E., TAYLOR, C. E., LIND,
and Wilcox (Operations) Ltd for permission to publish N. C., SCHWEIKER,J. W., WILLIAMS,F. S. G. and
this paper. The work of many colleagues in the Babcock AULER, E. P. ‘Welding Research Council Bulletin’
and Wilcox Research Station is gratefully acknowledged. 1959 (June, No. 51).
(19) MERSHON,J. K. ‘P.V.R.C. research on reinforcements of
openings in pressure vessels’, ibid. 1962 (May, No. 77).
APPENDIX (20) HARDENBERGH, D. E., ZAMRIK,S. Y. and EDMONDSON, A. J.
REFERENCES ‘Experimental investigation of stress in nozzles in
W. B. and MCKEAN,J. D.
(I) CARLSON, ‘Cylindrical pressure cylindrical pressure vessels’, ibid. 1963 (July, No. 89).
vessels: stress systems in plain cylindrical shells and in (21) WELLS, A. R., LANE,P. H. R. and ROSE, R. T. ‘Stress
plain and pierced drumheads’, Proc. Instn mech. Engrs analysis of nozzles in cylindrical pressure vessels’, Symp.
1955 169,269. Pressure vessel research towards better design, Proc. Instn
(2) SCHOESSOW, R. and BROOKS, T. E. ‘Analysis of experimental Mech. Engrs 1961 (Paper 2), 17.
data regarding certain design features of pressure (22) ROSE, R. T. ‘New design method for pressure vessel
vessels’, Trans. Am. SOL.mech. Engrs 1950 72, 367. nozzles’, Engineer, Lond. 1962 (20th July).
(3) WELLINGER, K. and KUSSMAUL, K. ‘Strength properties of (23) ROSE,R. T. and THOMSON, J. M. T. ‘Calculated stress con-
steels subject to overelastic stresses’, Mitteilungen der centration factors for nozzles in spherical vessels’, Symp.
V.G.B. Heft 92, 1964 (October). Pressure vessel research towards better design, Proc. Instn
(4) KOOISTRA,L. F. ‘Effect of plastic fatigue on pressure mech. Engrs 1961 (Paper l), 1.
(24) LECKIE, F.A.and PENNY, R. K. ‘Stress concentration factors ‘
vessel materials and design’, Welding Jl Res. suppL 1957
36 (3), 120. at nozzle intersections in pressure vessels’, W.R.C.
(5) KOOISTRA, L. F., LANGE, E. A. and PICKETT, A. G. ‘Full Bulletin No. 90 1963.
size pressure vessel testing and its application to design’, (25) O’CONNELL,J. M. and CHUBB,E.J. ‘An improved method
Am. SOC. Mech. Engrs Paper 63-WA-293. of calculating stresses at the intersection of a cylindrical
(6) KOOISTRA,L. F. and LEMCOE, M. M. ‘Low cycle fatigue nozzle and a spherical vessel’, Proc. Instn mech. Engrs
research on full scale pressure vessels’, Welding JZ Res. 1963-64 178 (Pt 3J), 224.
suppl. 1962 41 (2), 297. (26) ERIGEN, A. C.,NAGHDI,A. K. and THIEL, C. C. ‘State of
(7)WELTER,G. and DUBUC, J. ‘Fatigue resistance of simulated stress in a circular cylindrical shell with a circular hole’,
nozzles in model pressure vessels of T-1 st#, ibid., 368. W.R.C. Bulletin No. 102,1965.
(8) LANE, P. H. R. and ROSE,R. T. ‘Comparative performance (27) SIGNBS,E. G. et d. ‘Investigation of the factors affecting the
of pressure vessel nozzles under pulsating pressure’, fatigue strength of welded high strength steels’, Brit.
Symp. Pressure vessel research towards better design, Proc. Weld. yl 1967 (March) 14 (No. 3), 108.
Instn mech. Engrs 1961 (Paper 4), 67. (28) LECKIE, F. A. and PAYNE, D. J. ‘Some observations on the
(9) TAYLOR, T. E. Private communication. Also ‘Low cycle design of spherical pressure vessels with flush cylindrical
fatigue behaviour of low alloy steel pressure vessels’, Brit. nozzles’, Proc. Instn mech. Engrs 1965 180 (Pt l), 66.
Weld. Jl 1967 14 (12), 641. (29) TAYLOR, T. E. ‘Effect of test pressure on the fatigue perfor-
(10)SOETE,W. et al. ‘Pressure vessel research in Belgium’, mance of mild steel cylindrical pressure vessels containing
Symp. Pressure vessel research towards better design, Proc. nozzles’, Brit. Weld.JI 1967 (March) 14 (No. 3), 117.
Instn mech. Engrs 1961 (Paper 5), 82. (30) WBLLINGER, K. ‘The significance of high yield steels in
(11) HIERMAN, J. and STOCKMAN, G. ‘Essais sur tubulures boiler construction’, Mitt. der V.G.B.Heft 91 Aug. 1964.
inclin6es’, Revue de la Soudure No. 1 1964 (January), 47. (31) PAWAR,K. and VASARHELYI, D. D. ‘Fatigue tests of struc-
(12) STOCKMAN, G. ‘Le dimensionnement du rtinforcement des tural steels and weldments in notched specimens’, Weld.
ouvertures’, Centre de Recherches SaentXques et Res. Suppl. 1966 (April), 161.

Proc Instn Mech Engrs 1967158 Vol182 Pt 1 No 31


Downloaded from pme.sagepub.com at City University Library on April 25, 2016
676

Communications
Mr C. M. White, B.Sc., C.Eng., M.1.Mech.E.-The nozzles in cylindrical vessels according to the following
warning of the potentially reduced fatigue life of low alloy equation :
steel vessels compared with mild steel vessels designed to [1+ 1.77(d/D)do/T
the same code is indeed timely. It highlights the necessity + ( d / D ) 2 / d s p [l+(T/D)/%‘/sISl
]
of checking against this mode of failure and the possible
need to reduce stress concentrations by even more atten- K = Max. of )
1+(d/D)Z/(s,/S)dqS
tion to the design of details. The somewhat increased either 1 [1 . 6 7 v ’ ~ d ~ T0565(d/D)]
+
numbers of cycles to failure, Table 3, and more particularly I x [l+(T/D)/dsISl
the transference of the point of failure, Fig. 12, for vessel 0.67d~d~+0.565(d/D)/(s/s)
D4 when compared to the other alloy steel vessels give
some indication that fatigue life can be increased by giving where d is the internal diameter of branch pipe; D the
further attention to the local shape and disposition of internal diameter of main pipe; r = 4 2 ; R = 012;
nozzle reinforcements. t the branch pipe wall thickness; T the main vessel wall
The experimental results back up this warning in a thickness; s = d/2t, the nominal stress in branch pipe;
forceful manner, but it seems unfortunate that some shade S = D/2T, the nominal stress in main vessel (cylindrical);
of doubt must be cast upon the absolute value of these and S = D / 4 T is the nominal stress in main vessel
results because the alloy steel vessels were so constructed (spherical).
that there was mild steel, or welds between mild steel and Calculated results were compared to a variety of nozzle
alloy steel at or near the points of maximum stress results and these were assessed in an ASME paper at their
concentration. Winter meeting (November 1967). A series of curves from
The collected vessel fatigue results, Fig. 17, provide a the above equation are shown for a given set of parameters,
valuable justificationfor using the ASME I11 criteria as the Fig. 20. Perhaps the authors would like to comment on the
basis of a fatigue analysis. They also emphasize the con- above.
servatism of the B.S. 1515 criteria for determining Another point is that it would be very interesting to
whether fatigue analysis is required, and underline the hear more’%bout the ‘unpublished German work’, men-
overdesign which could result if the B.S. 1515 criteria tioned by the authors. Were these SCFs based on measure-
were to be used as a basis for determining vessel scantlings ments at the inside surface of the vessel or were they

----
-particularly for those sections of vessels which do not extrapolations from measurements taken from the outside
contain openings, supports or other major discontinuities. surface? If the latter is the case, what order of accuracy
are we to expect from these data ?
20 I { I
Mr E.Hay, M.Sc., B.Sc., C.Eng., M.1.Mech.E.-The D II T A
s/S=l.O
authors point out in the discussion section of their paper
that the Leckie-Penny values they quote are for nozzles in I /r- Inn I
spherical vessels. The recommended use of such results, by
approximation, to nozzles in cylindrical vessels is however
rather disturbing even though this approach has been
fostered by B.S. 3915. When dealing with stress concentra-
tions due to pressure alone, the above assumption may be
reasonably acceptable. However, when a nozzle in a cylin-
drical vesselisbeing analysed for fatigueperformance under
actual operating conditions, the effects of external loads, as
well as the thermal stresses in the system, may well make
the errors that we are condoning tip the balance for a failure
as opposed to a no-failure system. An improved method of
stress concentration assessment was put forward in 1967 dlo
by Professor Lind of Waterloo University for flush Fk. 20
Proc Inrrn Mech Engrs 1967-68 Voli82 P I i No 3i
Downloaded from pme.sagepub.com at City University Library on April 25, 2016
A STUDY O F NOZZLES IN PRESSURE VESSELS UNDER PRESSURE FATIGUE LOADING 677

Finally, a point related to the results put forward by correlates to within f10 per cent a large number of
the authors. These are very useful only as applied to the measured stress concentration factors. Here p is the
parameters tested. Would it not be possible to interrelate parameter r / R d r T which indicates the specific size of
the area replacement factor (say 0.6,0*8,0.9 and 1.0) to the opening in the vessel. It may be noted that the expression
fatigue life or the SCF of the nozzle ? This approach, in gives the SCF in terms of a single group and may therefore
future publications, may give a better guide to the designer be drawn as a single line on a sheet of graph paper. The
for any of his new applications. expression may be written in the form

Mr K. D. Clare, M.Sc., C.Eng., M.1.Mech.E.-In the SCF = 2.5 [p0’408(T/t)0’408]


discussion section of their paper the authors have pre- The Leckie and Payne(34) curves for stress concentration
sented a well reasoned argument in favour of the generally at flush nozzles in spheres may also be plotted on a log-log
accepted approach that the stress levels at nozzles in scale in terms of a single group p2( T/t)llz,Fig. 21, and a
cylinders can be approximated from theory for nozzles in single line may be drawn through all the points plotted
spheres. This is probably somewhat pessimistic for a fairly which cover values of p from 0-4 to 4.0 and (Tit) from
wide range of practical nozzles and it is worth considering 1.0 to 4.0. The expression given by this line which again
the influence of the different dimensional parameters on correlates all the results to within f10 per cent is
the nozzle stress levels.
For the cylinder/cylinder intersection Money (40) has SCF = 3.3 [pov56(T/f)o’14]
shown that an empirical expression,
From these last two empirical expressions it is possible
log SCF = 0.2042 log p z ( T/t)’+O.3979 to obtain an idea of how the two parameters p and ( T / t )

F*. 22
Proc Instn Mech Engrs 196768 Downloaded from pme.sagepub.com at City University Library on April 25, 2016 Vol I82 Pt I No 31
678 COMMUNICATIONS

affect the stress levels at nozzles. In particular the expres- by some of the pressure vessel fabricators, e.g. the root
sions indicate the different influence of the shell to nozzle face should not be greater than 7% in to conform to the
thickness ratio in the two cases. It is clear that more optimum preparation for efficient welding methods instead
advantage may be gained by thickening the nozzle in a of Q in as shown in Fig. 3 of the paper.
cylinder than by thickening the nozzle in a sphere. Fig. 10 of the paper indicates the point of fracture
This advantageous influence of nozzle thickening on initiation and it would appear that the failure did not
cylinders is such that it should not be overlooked merely originate at this very substantial artificial notch.
for the sake of having a single rule for both spheres and I n the conclusions, the authors state that on the basis of
cylinders. This is perhaps brought out more clearly by both static and dynamic tests, there is little to choose
Fig. 22, which is similar to the authors’ Fig. 18, but between the nozzle design of vessels Dl, D2, D3 and D4,
compares the Leckie and Payne results with the Money and they also state that certain special precautions should
expression. be taken when vessels are designed to Codes where the
It is interesting to note that if a similar exercise is carried design stress is proportional to the yield point.
out comparing limit pressures of spherical and cylindrical But the trend in nozzle detail design at present is to
vessels with flush nozzles it is again apparent that increas- be more sophisticated and rational rather than the area-
ing nozzle thickness increases strength more rapidly in for-area arbitrary reinforcement method allowed for in
cylindrical vessels than in spherical vessels. B.S. 1113:1958 and the latest draft of this Code, i.e.
Users Specifications, may require a fatigue analysis check
Mr D. K. Common (London)-In my view more to be allowed for in the detailed examination of the stress
should be made of the differences in behaviour between system on the lines given in ASME III or B.S.3915
the three types of nozzles tested-M1, D4 and the Group (which is mandatory for Nuclear Vessels), and the set-on
D1, D2, D3-the latter arc all sufficiently similar in nozzle is the latest fashion with varying reasons given in the
design and appearance to be considered as one type. Specifications for the necessity of its use. Figs 23 and 24
Considering the behaviour of these three types in the show typical examples.
context of the question ‘How does one design a nozzle for Document I.S.O. 66/30384 states:
optimum fatigue performance in an alloy steel vessel?’ (a) Area replacement methods. are not adequate for
various aspects mentioned by the authors assume some- advanced pressure vessels.
what different significance. For instance there can hardly (6) B.S. 3915 protruding nozzles offer economies in
be any doubt that nozzle D4 failed as the result of over- material.
compensation (this is confirmed incidentally by the (c) Simplified versions of B.S. 3915 design charts for
circumferential nature of the fracture-in every other case protruding nozzles are proposed for the 13.0. Boiler
this was radial). Even in spite of this, D4 gave a slightly and Pressure Vessel Codes, enabling rapid determina-
higher fatigue life than Group D1, D2, D3 and one cannot tion of economic reinforcement configurations.
help feeling that a significantly higher fatigue life would
have been achieved had less (or perhaps nil) compensation
been provided with a mild steel set-through nozzle of this
type. Some extension of the work to confirm this would
appear to me to be highly desirable to identify quite
unequivocally what the optimum design practice would
be.
I should also like to comment on the fact that the stress
concentration observed in nozzle M1 was significantly in
excess of theoretical predictions and also higher than that
observed on the alloy steel vessels. Surely this confirms
what many of us believe, i.e. that observed values of SCF
are largely a function of macroscopic material behaviour
under local yielding conditions/strain gauge size and
location and that ‘abnormally’ large values measured on a
structure should be considered more as evidence of
healthy stress redistribution taking place rather than as
potential points of weakness.

Mr H. Crowther (Renfrew, Scotland)-The authors


state the current Code was B.S. 1113:1951 when the
fifteen nozzles were designed and it would appear that in
Fig. 3 a detail for set-in nozzle is shown which would not
be recommended by this Code. B.S. 1113:1958 shows, in
Fig. 35b, a set-in nozzle which has been successfully used Fig. 24
Proc Instn Mech Erigrs 1967-68 Vo1182 Pt 1 No 31
Downloaded from pme.sagepub.com at City University Library on April 25, 2016
A STUDY O F NOZZLES IN PRESSURE VESSELS UNDER PRESSURE FATIGUE LOADING 679

In a fatigue test recently carried out on a set-in small of the component is important if maximum advantage is
bore nozzle array the joint required the following fatigue to be gained from such methods.
cycles before failure occurred. Theoretical analysts have generally adopted the Tresca
200 000 cycles at 0 -design pressure- O+ 200 000 criterion for first yield, e.g. limit analysis by Dr Gill (32)
cycles at 0 - 1.5 x design pressure -0 +approximately and shakedown analysis by Professor Leckie (33). In the
100 000 cycles at 0-3 x design pressure-0. Even then correspondence to reference (34), Mr Spence showed that
the failure did not originate at the artificial notch in the calculated values of shakedown factors are higher when
bore of the nozzle but at the crotch corner, i.e. the the von Mises instead of the Tresca criterion is used in
recognized highest stressed point. Professor Leckie’s theory. These higher values have since
It is also most interesting to compare the authors’ con- been supported by experiments (35) which immediately
clusions against some of the earlier statements on preferred implies that the use of Tresca would result in unnecessarily
nozzle designs, e.g. ‘The set-in protruding thick nozzle conservative designs. The authors, in their paper, have
was found to have the longest life’. Paper No. 5 Sympo- implied that the von Mises criterion is also conservative.
sium on Pressure Vessels research towards better design. In view of this, some of my own observations may be of
Proc. Imtn Mech. Engineers 1961. interest.
It can be shown that the minimum stress concentration During the investigations into shakedown behaviour of
for a hole in a cylinder may be obtained when the hole is cylinder to sphere attachments (35) similar observations
an ellipse with axes ratio 1:2 and the minor axis of the were noted to those of the authors, is. the onset of non-
elliptical opening is perpendicular to the hoop direction. linearity of the strain/pressure graphs usually appeared to
Accordingly, elliptical shaped nozzles might be stresswise be slightly higher than that required by the von Mises
ideal and desirable for a cylindrical pressure vessel. But yield criterion. It was considered, however, that such small
the manufacture of such a nozzle presents an economic volumes of material are involved with first yield that the
problem, since there has to be a change of cross-section surrounding elastic material retains control of the region
from an ellipse to a circle for the supply pipe connection. during the immediate pressure increase. Thus, so far as can
However, such a nozzle has been invented by Hiltscher, be determined from an experimental plot, strain behaviour
(Swedish) Patent Application No. 9134/63 and may be appears to remain linear. When making comparisons of
in the form of an unsymmetrical push out in the vessel this type it is, of course, important to use the correct value
wall or a separately formed unsymmetrical nozzle to be of uni-axial yield stress for the materials. In this case
welded into the elliptical hole in the vessel wall. Other tensile specimens were machined from material taken from
methods may be used such as machining a thick wall vessel the ends of the nozzle forgings and from the shell plate
to achieve the same stress pattern. cut-outs after the basic manufacturing procedures had
‘The set-on branch might sometimes offer advantages been completed, i.e. machining, rolling, etc. The specimen
in fabrication, provided the plate had been checked very material was also subjected to the same heat treatment as
carefully for laminations. Also the protruding nozzle the completed vessel.
constituted a highly efficient reinforcement.’ (Paper No. 2 The experimentally determined shakedown pressures
Symposium 1961 in the replies to the discussion.) were almost twice the von Mises first yield pressure and
‘The findings after investigations for control rod nozzles approximately 2.2 times the Tresca first yield pressure.
in nuclear pressure vessels showed protruding type as Since shakedown would not be expected beyond twice
final choice.’ (Penetrations Reinforcements studies in 1963 first yield pressure the Tresca criterion is clearly too
by The Levivier Company, France). conservative. Evidence suggests that the Von Mises
Studies on composite vessels of the multilayer types criterion is, at least, approximately correct and conse-
all indicate that manufacturers are using set-through quently I would like to urge the theoretical analysts to
type nozzles as there appears at present to be no practical follow experimental practice and use Von Mises rather
alternative. than Tresca.
I welcome the trend in the paper wherein the designer
is encouraged to observe more fully the principle that
every design in all its parts must be judged individually, Mr J. F. Poynor, B.Sc.(Eng.), C.Eng., F.1.Mech.E.-
and more particularly where alloy steels are to be used. Fig. 18 is of particular interest to those designing nozzles
This paper should help to discourage attempts to intro- in cylindrical vessels. Use of the cylinder/sphere theory is
duce into the conventional Boiler Codes simplification of sometimes limited to r / R = 3 for cylindrical vessels, and I
calculations which tend to arbitrarily bias the designer’s would be interested to know how many of the experi-
judgement towards a solution which may not be the best on mental points in Fig. 18 came within the range of, say,
purely engineering grounds. r / R = 0.75 to 1.0 which is relevant to both the design of
vessels and piping systems. Would the authors have any
Mr E. Procter, C.Eng., M.1.Mech.E.-Future designs reservations in the use of Fig. 18 for these large diameter
of pressure vessel attachments, operating below the creep cylinder junctions irrespective of the value of (r/R)dm.
range, are likely to be based on shakedown and/or limit I would like to develop the authors’ views concerning
analysis methods. The determination of first yield pressure fatigue implications of the area replacement method for
Proc Instn Mech Engrs 196748 Vo1182 Pt 1 No 31
Downloaded from pme.sagepub.com at City University Library on April 25, 2016
680 COMMUNICATIONS

alloy steel vessels. This method of nozzle design has been properties required to complete the correlation. The upper
used successfully for many years, and continues to be used and lower curves in Fig. 25 are predictions of fatigue life
with the increased design stresses now permitted by most on this basis, for the extremes of material tested by the
Codes. In addition to the high yield materials tested by the authors. The curve for Fortiweld material has been
authors, quenched and tempered materials are becoming partially interpolated. Experimental values plotted are
of increasing interest. taken from the authors’ paper and reference (9).
The cumulative result of increased code stresses and Fig. 25 shows that fatigue life seems to be independent
the use of high yield material is design stresses much even of the other sacred design criterion, ultimate strength.
higher than those used several years ago for carbon steels. The small improvement which Manson expects from a
Local stresses will also increase, resulting in the fatigue material with an increased ultimate strength is not
implications shown in the paper, and more attention must reflected in the experimental results. There are exceptions
be given to the SCFs permitted at nozzles if fatigue is to which may reveal the cause. The three ground and
be assessed. This requires the designer to know the SCF he polished examples clearly stand out from their fellows
is permitting, but use of the area replacement method does even on a logarithmic plot. The three examples incorporat-
not itself give a known SCF; the method can result in ing a ground radiused crotch may also have benefited.
SCFs ranging up to about 4.0. The SCF should therefore This kind of behaviour is well known in the field of
be known, in addition to meeting area requirements, if welded structures and has been attributed (27) to the
fatigue could be a mode of failure. It is evident that this presence of very small sharp defects in regions of high
procedure would be simplified by the use of a nozzle strain. These defects produce local strains of sufficient
reinforcement method which is based on a known SCF magnitude to reduce all materials to the same level of
constant for all nozzles satisfying the method. fatigue life/strain range. (Fig. 25 shows that this happens
at 2 per cent strain range.) The authors doubt whether this
Mr G. T. Gray, B.Sc. (Glasgow).-Full size workshop mechanism applies here but it is known that fine fatigue
fabricated vessels are not often included in cyclic pressure cracks can be generated prematurely in the surface layer
testing programmes. The present excellent contribution of as-received plate even in the absence of welds (e.g.
to the literature is all the more useful because careful reference (36)). It is possible then that removal of this
experimental work has been supplemented by a valuable layer, including polishing to avoid further notches, may
review of available data. One hopes that this paper may allow the material to develop its maximum fatigue
squash any lingering notions that fatigue strength is strength.
related to yield strength. Anyone doubting the significance of surface effects in
A comparison of the full scale results with the experience such thick sections should refer to Wellinger (reference
of Manson (14) may be interesting. He found a simple (37) which is the same as (30)), who reports ‘fine net-like
relationship between cycles to failure and total strain surface cracks’ on a boiler drum wall, which were propa-
range for small uniaxial specimens cycled between con- gated under low stresses in a short service period.
stant strain limits. Ultimate strength and reduction of area Polishing is hardly a practical solution to the problem, but,
at fracture in a uni-axial tensile test were the only material in any case, surface layer effects may develop in service.

__ - __
- -I-,

v MN-CR-Mo-V

1-1
L

10’
CYCLES TO FAILURE

Fig. 25
Proc Instn Mech Engrs 1967-68 Downloaded from pme.sagepub.com at City University Library on April 25, 2016 Vo1182 Pt 1 No 31
A STUDY OF NOZZLES IN PRESSURE VESSELS UNDER PRESSURE FATIGUE LOADING 681

Fig. 26

Wellinger also draws attention to the problem of cracking early in the life of their vessels with failure frequently
in the protective magnetite layer which forms on boiler occurring after twice the cycles to crack initiation. The
drums in service. Reference (38) also describes fatigue present authors conclude that failure occurred quickly
failures due to cracking of surface oxide films. after crack initiation, perhaps they would care to comment
To conclude, there seems to be no valid reason for upon this.
increasing the allowable fatigue strength for a material The curves of Fig. 18 relating SCF to r / R v ‘ m a r e very
which has an increased yield or ultimate strength. The irregular. If a cross plot of t/T against SCF for a fixed
surface layer of the component seems to be responsible. value of r / R d W is made it will be found to have a very
Other ‘fringe’ effects such as temperature, corrosion and irregular shape so that interpolation between the curves
cycling speed have not been considered and may be is difficult. If the SCF is plotted against (T/T)~T/Ra
equally significant. much better distribution is achieved, the vast majority of
points falling within 10 per cent of the mean line, Fig. 26.
The predicted SCF for Spence’s vessels by this method
Mr H. A. Money, C.Eng., M.1.Mech.E.-Reference are 2.85 for vessels Dl-D4 and very similar for M1. The
(5) of the paper refers to work carried out by Pickett and measured stress from M1 is outside the scatter band but as
Grigory (39) which contradicts the present authors’ work neither of the other two methods of predicting SCF
on two points. First, Pickett and Grigory show that local mentioned in the paper suggests a value as high as this, the
deviations from the round, of about 2 per cent, can increase result remains an enigma.
membrane stresses by up to 100 per cent whereas the The ability to predict SCF is important in fatigue work
present authors state that up to 1 per cent out of roundness as the calculation of expected life usually gravitates to
produces negligible increase in stress. Does the act of locating the maximum stress in the structure on an S-N
welding on a nozzle locally distort a vessel and contribute curve and reading off the life. As most of the vessels of the
to the scatter of maximum stress found between nominally type considered in this paper are designed for high
identical attachments ? temperatures, I hope that Dr Carlson will consider
Second, Pickett and Grigory found fatigue cracks quite extending his test work to vessels at high temperature.

Proc Instn Mech Ewrs 1967-68 Vo1182 Pt I No 31


Downloaded from pme.sagepub.com at City University Library on April 25, 2016
682

Authors’ Reply
M r J. Spence and D r W. B. Carlson-In reply to r I R d v T < 1.1 and the qualitative trends and magni-
Mr White’s comments on the shade of doubt cast by tudes were quite similar. Comparisons of Professor Lind’s
having mild steel nozzles in alloy vessels, we would point work with any other source such as Fig. 18 or Figs 22 and
out that it seems likely that in vessels D1 and D3, and 26, do not lead to any positive conclusions for r / R d m
possibly in D2 as well, the mild steel would have little < 1.5. At higher values, however, Lind’s results tend to be
effect due to its physical position relative to the maximum higher than any of the others shown by Table 6 .
stress position. In addition the small difference in life I n the German unpublished work the stresses were
between alloy and mild steel anyway makes it seem calculated using Fig. 18 as pointed out in the paper. The
unimportant. The basic idea of having a more ductile accuracy is relatively unimportant for the present purpose.
material in the region of high stress is good (28) but is It is the similarity between the fatigue behaviour of mild
hardly fulfilled by using mild steel nozzles; it may, steel and the Cu-Ni steel on Fig. 17 and their separation
however, be exploited in a better way in advanced future in Fig. 19 that is the important point. We cannot agree
designs. Mr White emphasizes the conservatism of the with Mr Hay that the results (fatigue) are only useful as
B.S. 1515 criterion for determining whether a fatigue applied to the parameters tested and we are not convinced
analysis is required. For all cases that we have checked one as to the usefulness of the area replacement factor. It is
might even say it is grossly over-conservative. necessary to specify the distribution of the reinforcement
Mr Hay refers to Professor Lind’s excellent engineering carefully. The lee-way allowed by putting arbitrary limits
approach to the problem of nozzle in cylinder stressing to the reinforcement can give rise to large variations in
which can be found in reference (41). This work was not SCF and fatigue life.
available when the paper was drafted and, as was pointed Mr Clare has produced two expressions for the SCFs at
out, events are changing rapidly. The business of using nozzles in spheres and nozzles in cylinders which will be
nozzle in sphere results to predict nozzle in cylinder very useful to designers and has brought out well that
stresses has, of course, to be treated with care. In the more advantage is to be gained by thickening the nozzle in
paper only pressure loading was being considered. The a cylinder than by thickening the nozzle in a sphere. This
analysis of cylinder/cylinder connections under external was also apparent in Fig. 18 and brought out in the dis-
loads is complicated, and under realistic thermal loading cussion under ‘Stress levels’ but the paper urges care with
results are as yet non-existent. Lind’s formulae and Fig. 20 large r / R ratios since it is known that the maximum stress
(taken from reference (41)) are also for pressure loading. may not be at the crotch. Unfortunately most experimental
Unfortunately the bulk of Fig. 20 refers to very unlikely results (and Lind’s theory) even at large r / R refer to the
nozzle configurations, e.g. with s/S = 1.0, T / t = l/(d/D), crotch. It would be interesting to know if any of the
i.e. for dlD = 0.2, T / t = 5.0 and smaller d / D refer to results covered by Mr Money’s survey referred to a
higher T / t . This is the unfortunate thing about some position other than the crotch. Mr Clare’s Fig. 22 showing
presentations; they do not easily show the physical limits the two expressions for SCF is interesting although surely
of the problem. Similar remarks apply to the other two the Leckie-Penny results for T / t = 1.0 have been mis-
figures in reference (41).One with s/S = 2.0 is even more placed from their original publication (24). A comparison
restricted and one with sjS = 0-5 is better. of the empirical results in Fig. 18 with Fig. 22 for
A detailed comparison of Lind’s results with Eringen’s cylinder/cylinder connections shows remarkably good
latest work has been made (42) for r/R -$ and < agreement.
While we agree with Mr Common that the failure in
Table 6. Stress concentration factors nozzle D4 points to over-compensation, we think it
unlikely that any nozzle of like proportions having a
degree of compensation intermediate between D4 and D1,
Geometry I Source D2 and D3 would show greatly enhanced fatigue resist-
ance. With nozzles of these proportions it is now known
that the fatigue performance characteristic is likely to be
similar in the range of area-for-area compensation from
100 per cent down to 40 per cent, assuming that the com-
pensation is principally in the nozzle. This conclusion
would hold for soundly constructed nozzles, whether
Proc Insrn Mech Engrs 1967-68 Downloaded from pme.sagepub.com at City University Library on April 25, 2016 Yo1 182 Pt I No 31
A STUDY OF NOZZLES I N PRESSURE VESSELS UNDER PRESSURE FATIGUE LOADING 683

through-welded or not. The accredited area-for-area knowing the maximum stress even with codes which do
compensation in nozzles D1, D2 and D3 is of the order of not specify or require an SCF if fatigue is a possible mode
between 30 per cent and 60 per cent, depending on the of failure.
code used. That in D4 ranges from 100 per cent to 150 We would like to thank Mr Gray for his comparisons
per cent, again depending on the code used. with Manson’s work. Comparisons on a basis of reference
Like Mr Common, one of the authors believes that the (15) would yield similar results. We are also grateful to
yield properties of the material, acting in small areas of Mr Gray for his lucid comments on the factors which
high stress concentration, can have an effect on the contribute to the similarity of observed fatigue perform-
observed allegedly elastic SCF of a component. It is a ance in alloy and mild steel vessels, and particularly for his
suggestion which is hard for the stress analyst to swallow, extension of Signes’ ideas (27)on weld defects to surface
but it could undoubtedly benefit from a bit of painstaking effects in the parent material. However, caution is advised
pure research. at high strain ranges. Where the strain range is greater
Mr Crowther has summarized well present trends and than twice the yield or proof strain the similarity in fatigue
code practice. The remark in the paper about B.S. 1113- performance may not be so strong.
1951 refers particularly to nozzle D4. The point of the In answer to Mr Money, we would point out that the
conclusion on the similarity of nozzles D1, D2, D3 and effect of out-of-roundness in a cylindrical shell is not
D4 is that the simplest one may as well be used. On the simply related to the percentage out-of-roundness.
point about preferred nozzle designs we would point out Reference (I) shows typical variations with associated
that it is not too difficult to find seemingly conflicting stresses and discusses the matter in some detail. It
statements. One needs to be careful. Soete’s work was on emphasizes that to do justice to out-of-circularity a radial
small 2 in diameter nozzles in 11 in diameter vessels and gauge is required and indicates that the increase in stress
what is actually said in the conclusions is ‘The scatter is depends on the radius to thickness of the shell, on the
so large that it is difficult to conclude with certainty as to ratio of the deviation to the shell thickness and also on
the superiority of one type of nozzle over another. It whether the out-of-roundness is local or general. Neverthe-
seems probable, however, that type D (set-in protruding less, the statement in the present paper needs amplifica-
thick nozzle) is better. . .’. Fairly well known advantages tion. Measurements of out-of-circularity were made at a
are to be had if protruding nozzles can be used but no number of stations in the test vessels and the greatest
protruding nozzles were tested here. Mr Crowther’s difference between maximum and minimum diameters
remarks on elliptical opening nozzles may be of interest was not greater than 1 per cent. The associated bending
to some people. Some results of Dr Hiltscher’s work can stresses were calculated but the values local to the actual
be found in reference (43), a private report of the Swedish nozzle positions were so low that it was decided to ignore
state power board. them. The Engineering Sciences Data Unit have this
Mr Procter’s contribution is interesting and we agree particular problem in their programme of work and will be
that it would be helpful in general to use the von Mises considering it shortly.
criterion of yielding. The conservativeness of the von The crack propagation time is not known for the present
Mises criterion in the present case, Fig. 7, was not great vessels. They were not examined at intermediate times
and might be considered marginal. It is obviously depen- but only after a failure occurred. In vessels D1 to D4 no
dent on the constraint of the system. Comparison of the cracks were found in the other nozzles after the first one
onset of non-linearity with any yield criterion on small failed. In the case of the mild steel vessel, where tests
nozzles would be expected to give more conservative continued until all three nozzles failed, nozzle No. 2 failed
results than on large nozzles if onset of linearity is used as a first at 113 816 cycles. At that stage there was no sign of
basis. It is interesting that shakedown tests have also shown cracking in the other two nozzles. Nozzle No. 1 failed at
this feature. However, experimental shakedown results 201 138 cycles and there was no sign of cracking in No. 3.
need to be treated with caution in this context since norm- It was presumed that the cracks had propagated in a small
ally it is progressive plasticity that one is preventing rather number of cycles. Conceivably it is possible that crack
than reversed yielding as described by Findlay and propagation in No. 1 occupied nearly 50 per cent of the
Spence (44). Reverse yielding is difficult to observe and life and in No. 3 about 30 per cent. On a logarithmic plot
measured shakedown factors can be expected to be at the difference hardly shows. Taylor (9) quotes propagation
least marginally higher than theoretical ones for this times varying from 2 per cent to 64 per cent of the vessel
reason. There is also the complication of true material life. The length of time occupied by crack propagation is
behaviour as distinct from idealized. not easily related to vessel or stress parameters even using
Mr Poyner asked about large r/R values. Only two of the a fracture mechanics approach (45). Anyone contemplating
results in Fig. 18 refer to r/R > 0.75. A reservation in the fatigue work in the future could perhaps make crack
use of the results and indeed in the use of any empirical propagation measurements. Recent advances in non-
relationship is the one already stated about the location destructive testing make this relatively simple to do.
of the maximum stress. Perllaps it is still advisable to use Mr Money remarks that a cross-plot of Fig. 18 is very
the nozzle in sphere values for large r / R values. We en- irregular but comparisons with his results as presented
dorse Mr Poyner’s remarks regarding the necessity of herein by Mr Clare are in good agreement as is the result
Proc Insrn Mech Engrs 196768 Val I82 Pt I No 31
Downloaded from pme.sagepub.com at City University Library on April 25, 2016
684 AUTHORS’ REPLY

of 2.85 quoted for the present vessels compared with 2.9 Although the determination of the stress is an important
from Fig. 18. Fig. 27 shows cross-plots of stress concen- step in determining the life of a vessel, it seems a pity that
tration factor for fixed values of r / R d r T . Bearing in the bulk of the discussion has been occupied by what is in
mind that the lines depicting Money’s results have an a sense a subsidiary part of the paper. It is to be hoped that
associated scatter of f10 per cent the agreement is the implication of Figs 19 and 25 and the remarks on
remarkably good. Lind’s results have not been included in fatigue have been sufficiently explicit to be of use to
Fig. 27 to avoid confusion. At low rIRd/RIT they would pressure vessel designers.
be in approximate agreement (perhaps rather more than
REFERENCES
10 per cent difference) and at large r / R d z / T they would
(32) GILL,S. S. ‘Limit pressure for a flush cylindrical nozzle in a
be high. The second value in Table 6 may be easily spherical pressure vessel’, Int. J . mech. Sci. 1964 6, 105.
compared. The accuracy suggested by Professor Lind for (33) LECKIE, F. A. ‘Shakedown pressures for flush cylinder-
his own results was f 2 0 per cent. The reason for such sphere intersections’,J. mech. Engng Sci. 1965 7 (No. 4),
367.
scatter is at least in part due to the fact that r / R d r T is (34)LECKIB, F. A. and PAYNE,D. J. ‘Some observations on the
not such a useful parameter for cylinders as it is for design of spherical pressure vessels with flush cylindrical
nozzles’, Proc. Instn mech. Engrs 1965-66 180 (Pt 1)
spheres (42). Indeed even in the case of the sphere it is not (No. 20).
so good if the stresses in the nozzle are considered. (35) PROCTER, E. and FLINDERS, R. F. ‘Shakedown investigations
Fig. 26 is very useful for design particularly in the form on partial penetration welded nozzlcs in a spherical
shell’, Nuc. Engng and Design 1968 7 (No. 7).
of the equation given by Mr Clare. Yet there may be dis- (36) TETELMAN, A.S.and MCEVILY,A. J. ‘Fracture of Structural
advantages. Usually R/T is more or less fixed and the Materials’, 1967, 356 (Wiley).
nozzle thickness or opening size is adjusted to suit the (37) WELLINGER,K. Die Bedeutung von Stahlen mit Hoherer
Streckgrenze in Kesselbau. Mitt. der V.G.B. Heft 92.
design. One cannot easily see the effect of juggling with October 1964.
the opening size and the nozzle thickness particularly if (38) COFFIN,L. F. Trans. A.S.M. 1963 56,339.
(39) PICKETT,A. G. and GRIGORY, S . C. ‘Studies of the fatigue
one wishes to extrapolate to the extreme value of t / T = 0. strength of pressure vessels’, South West Research
It must not be thought that Fig. 18 for cylinder/cylinder Institute Report 1228-1-29 (Pt 1) and 1228-1-36 (Pt 2).
connections is being defended against logical improve- (40)MONEY, H. A. ‘The design of flush cylinder/cylinder
intersections to withstand pressure’ (to be published by
ments. It is merely an empirical stop gap measure which A.S.M.E.’,
still has its usefulness, as have the other empirical relations (41)LIND,N.C. “Approximate stress-concentration analysis for
pressurized branch-pipe connections’, A.S.M.E. Paper
suggested. Events are changing rapidly as pointed out in NO. 67-WA/PVP-7.
the paper and soon theoretical curves may be generally (42)RODABAUGH, E. C.,ATTERBURY, T. J., CLOUD,R. L. and
available. The following references may be consulted: WITT. F. 1. ‘Evaluation of exoerimental and theoretical
data on ridial nozzles in pressGre vessels’. Private report
(41) and (44, (44H53)- T.1.D.-24342, United States Atomic Energy Commis-
sion, 1968.
(43)HILTSCHER, R. ‘Utveckling av Forbattrad stuts for cylin-
driska tryckkari’, Swedish State Power Board Report
No. L-206 Stockholm, 1963.
(a) FINDLAY, G. E. and SPENCE,J. ‘Applying the shakedown
concept to pressure vessel design’, The Engineer, 12th
-Money (Clare) July 1968.
+ Spence and Carlson (45) GRAHAM,M. ‘A comparison of fatigue test results with
theory’, Student Project Thesis, Strathclyde University,
@ Eringen (theory t / T = O ) 1968.
(46) HANSBERRY, J. W. and JONES, N. ‘A theoretical study of the
elastic behaviour of two normally intersecting cylindrical
shells’, Brown Univ. Report No. NSF-GK. 1268/1,
April 1968.
(47) WITMAN, W. R., HOPPER, A. G. and MERSHON,J. L.
Local stresses in soherical and cvlindrical shells due to
external loadings’, W.R.C. Bull. No. 107, 1965.
(48) STONE,R. M. and HOCHSCHILD, S . ‘The effect of nozzle
spacing on the pressure stresses at the intersection of
cylindrical nozzles and shells’, Trans. A.S.M.E. 1967
Series A 89, No. 3, 360.
( ~ ~ ) - E R I N G EA.N ,C. and SUHUBI,E. S. ‘Stress distribution at
two normally intersecting shells’, NUC Structural Engng,
2, No. 3, 253.
(50) BIJLAARD, P. P., DOHRMAN, R. J. and WANG,I. C. ‘Stresses
in junction of nozzle to cylinder pressure for equal
diameter of vessel and nozzle’, NUC Engng and Deskn
1967 5,349.
(51)LEKKERKER, J. A. ‘Stress concentration around circular
holes in cylindrical shells’, Proc. App. mech. C0n.f. 1964
Munich.
(52) VANDYKE,P. ‘Stresses about a circular hole in a cylindrical
shell’, A.I.A.A. JZ 1965 3 (9).
(53) REIDELBACH, W. ‘The state of stress at the perpendicular
intersection of two right circular tubes’, Ingenieur-
Archb. 30 (9, 293.
Proc Inrtn Mech Engrs 1967-68 Downloaded from pme.sagepub.com at City University Library on April 25, 2016 Vol I82 Pt 1 No 31

You might also like