Professional Documents
Culture Documents
ABSTRACT
NOMENCLATURE
D i C y l i n d e r i n s i d e diameter.
hi Internal head height.
K~ Shape factor (BS 4994).
Po-2% Pressure at a m a x i m u m strain o f 0.2 %.
INTRODUCTION
SPECIMEN GEOMETRY
Table 1 gives the nominal properties of the three geometries, designated B, D and E
in order of decreasing knuckle radius. All geometries have hi/D ~= 0.25 and are
assumed to be of uniform thickness with t/D~ = 0.01. Also shown are the reference
numbers of the corresponding specimen vessels. Geometry B has the crown radius
equal to the cylinder diameter and thus also formed part of the group covered in
Part 1 of this paper. Geometry D has the dimensions of the 'ellipsoidar domed end
specified in the British Standard for pressure vessel details, BS 5276,4 and is also very
similar to the Korbbogen type end of the German code for GRP vessels. 5 It is, in
fact, an approximation to a semi-ellipsoidal shape of the same head height.
Geometry E has the knuckle radius equal to 10 ~oof the cylinder diameter and can be
regarded as complementary to geometry C of Part 1 of this paper, which geometry
had the same knuckle radius but h l / D i = 0"19, R~/Di = 1"0.
TORISPHER1CAL ENDS FOR GRP PRESSURE VESSELS 33
TABLE 1
NOMINAL PROPERTIES OF END GEOMETRII~ B, D AND E
1 2 3 4 5 6 7 8
Geometry Vessel ri/Di Rl/Dl ~o Properties from BS 4994
reference reference
letter numbers Ks Effective Po.2 ~ P.l~
SCF (kPa I (kPa)
B 3 and 4 0.188 1.0 22-5 ° 1.80 1.98 104 818
D 11 ° 0.146 0.8 32-8 ° 2.25 2.48 46 364
12 83 655
E 13 and 14 0.1 0.708 41.1 o 2.25 2.48 83 655
" This vessel has only five layers of 450 8/m 2 CSM. All others have nine layers.
All geometries have hl/D ~= 0"25.
Also shown in Table 1 are the shape factors (Ks) from BS 4994 for each geometry
and the effective stress concentration factors (SCF) these represent when eqn. (17) of
the code is compared with the corresponding equation for the unit load in a cylinder.
Geometry D just qualifies for the same shape factor as geometry E (the higher one of
the two specified at this head height) because it has ri/D i less than 0.15. It is
interesting to note that the effective SCF from the German code for this geometry is
1.5* for the knuckle and 0.9 for the crown area, i.e. the sphere may be made thinner
than the torus region. The value of 1.5" is even less than the BS 4994 shape factor for
geometry B. Table 1 also gives values of P0.2 ~. and Pu, calculated according to the
BS4994 procedure using the properties specified in the code ( X = 12.7kN/mm
width per kg/m 2 of glass; u = 2 0 0 N / m m width per kg/m 2 of glass) and the
nominal construction of each vessel (vessel 11: five layers of 450 g/m 2 chopped
strand mat (CSM); all other vessels: nine layers of 450 g/m 2 CSM).
THEORETICAL INVESTIGATION
BOSOR 4, 6 the computer program described in Part 1 which is able to carry out a
stress analysis of axisymmetric shells of revolution assuming linear elastic material
properties and allowing for changes of shape, was again used to give theoretical
predictions of the behaviour of these dished ends. Stress and strain distributions at a
relatively high pressure, obtained assuming a uniform thickness of 10 mm and a
value of Young's modulus of 7.0 GPa, are shown for geometries D and E in Figs 1
and 2, respectively. They are directly comparable with the corresponding
distributions for geometry B shown in Fig. 4 of Part 1. It can be seen that the
intensification of stress and strain in the torus is significantly greater for geometry E.
* In the tentative update of A D - M e r k b l a t t N I o f February, 1976, this value has been increased to 1.75.
34 D. C. BARTON, P. D. SODEN, S. S. GILL
80 I I I ~ l 1
Sphere _ L~Torus~ _ _ C ),L:nJer
60--
~.0 - -
:' 20
~ 0 1 l ~ 1 t
~\
-20 t "~-J
~\ iiII Inside surfoce | . .
- - - - - outside surfoceJMer=dl°nal
-A,O
. . . . Inside surfoce I .
. . . . Outside surfocel Cwcumferentl°
-60 .______J__ l 1 J L 1
100 200 300 6,00 500 600 7IX) 800 ~)0
Meridionol disfnnce from pole mm
(a)
I.~ 1 [ T 1 f 1 F T - -
1"2-- ~ -
o.~
-0"2 - - \ \
-O.t.
I \,,
' ~
,'r
/ ~lnsidesurf,~-e 1
I
(b)
Fig. l. Predictions of BOSOR 4 for geometry D (uniform thickness) at P = 1200 k Pa. (a) Distribution
of surface stress. (b) Distribution of surface strain.
TORISPHERICAL ENDS FOR GRP P R E S S U R E VESSELS 35
t I I I I I I |
60
g. -
T
20 --
0 J I I "\.,\I~. i/,'~1 I
-20 -
I~ /.l--Inside surfoce | . --
(a)
1"6
I T i T 1 _-7 1 1
1.4
~ _ _ Sphere _ . _ Cylinder_ -.~
1-2
1-0
0'8
0'2
-0"2
-0"~,
-0'6
-0.8
0 100 200 300 /~0 500 600 700 800 900
Heridiono( disfonce from pole mm
(b)
This is confirmed by the relative values of SCF and SNCF (strain concentration
factor) defined exactly as in Part 1 and shown for geometries B, D and E in Table 2.
Values are specified at zero pressure (corresponding to no change of geometry) at a
pressure to give a maximum strain of 0.2 ~ and at 1500 kPa (close to the estimated
cylinder burst pressure). It can be seen that, although BS 4994 requires the same
TABLE 2
STRESS AND STRAIN CONCENTRATIONFACTORSFROM BOSOR 4
(higher) shape factor for geometry D as for geometry E, SCF's and SNCF's from
BOSOR 4 for geometry D are very much closer to those of geometry B. Thus,
whereas for geometries B and E the SCF's at zero pressure are very similar to the
corresponding shape factors shown in Table 1, for geometry D the SCF is
significantly less than Ks. This apparent anomaly in the Standard is the result of the
tabular method of presenting shape factors.
Again, both SCF and SNCF decrease with pressure as the end deforms into a
stronger shape. At P0.2 ~ the reduction in SCF and SNCF is greatest for geometry B
and least for geometry E. In proportional terms, this is also true at P = 1500 kPa,
suggesting that the effects of changes of shape decrease with decreasing crown radius
for ends of the same head height.
Table 2 also shows values of SCF and SNCF for a semi-ellipsoidal end of the same
head height and wall thickness. It can be seen that both SCF and SNCF for this
shape at zero pressure are much less than the corresponding values for any of the
three torispherical ends, including geometry D which is an approximation to an
ellipse. This would seem to validate the lower shape factor in BS 4994 for a semi-
ellipsoidal end (Ks = 1-45 for h J D i = 0-25). However, as the pressure is increased,
the change of geometry effect is not as great for this shape and, at 1500 kPa, the SCF
and SNCF values are only slightly less than those of geometry B, the 'strongest' of
the torispherical ends at this pressure.
TEST SPECIMENS
The results for specimens 11-14 only are reported here, those for vessels 3 and 4,
corresponding to geometry B, having been covered in Part 1. As before, the vessels
were manufactured by Plastics Design and Engineering Ltd using the same
TORISPHERICAL ENDS FOR GRP PRESSURE VESSELS 37
materials and methods of construction (see Part 1). The reinforcement was again
nine layers of 450 g/m 2 CSM except in the case of vessel 11 where only five layers
were used. All vessels were post-cured for several hours at 50 °C in a polyethylene
tent.
TEST PROCEDURES
Vessels
The test rig and instrumentation were as described in Part 1. The hand pump was,
however, replaced by a pneumatically :operated one (Charles Madden & Co. Ltd,
Airhydro Power Unit Type J.S.) in order to obtain a steadier rate of pressurisation.
The first loading cycle (up to a maximum strain of approximately 0.2 %) to
determine the 'elastic' behaviour was carried out in the normal manner for each
vessel. Pressure was applied in five equal increments up to the predetermined
maximum for the cycle and subsequently released in a similar manner. At each
increment, the valves were closed for a dwell period of 10 min.
For vessels 11, 13 and 14, it was decided that the intermediate cycles used
previously--whereby the specimen was loaded cyclically two or three more times to
progressively increasing maximum pressures before loading to failure--should be
omitted. This seemed reasonable because tests carried out on four-point bend
specimens both with and without these cycles had not produced any significant
variation in material strength properties, v For vessel 12, however, acoustic emission
studies were carried out by Dunegan Endevco Ltd and, at their request, two
additional cycles up to approximately 50 ~ and 75 ~ of the burst pressure were
included. For all four vessels the final pressurisation was carried out in a slow, steady
manner (without the 10-min dwell periods) to give failure in approximately 4h.
Because the rate of loading was so slow, it was possible to halt the pump
momentarily to allow a scan of the transducers to be taken without significant
changes in the pressure occurring in the 30 s that this required. This method had the
advantage that many more scans of the transducer readings could be taken during
the final loading without unduly prolonging the period of the test and also it was
more likely that a scan would be obtained immediately prior to failure.
After burst, each vessel was cut in half as before and measurements of wall
thickness taken, along the two instrumented meridians. Again, four samples were
taken from the cylinder at equidistant positions around the circumference and a
total of eight from the torisphere at various meridional and circumferential
38 D . C . BARTON, P. D. SODEN, S. S. GILL
positions. Burn-off tests were conducted on these to determine their glass content. In
addition, three samples were taken from each vessel (one from the cylinder, one each
from the knuckle and crown regions of the torisphere) in order to obtain
measurements of the thickness of the resin-rich layers present on the inside and
outside surfaces of the vessel wall material. The samples were cold-mounted in plastic
before being ground on water-lubricated abrasive paper and polished on diamond
paste impregnated cloth wheels. On viewing the samples prepared in this way
through a low-power microscope (magnification of the order of 10 × ), the bundles of
fibres in the main part of the specimen and the fibre-free resin-rich surface layers
could be clearly distinguished. Using a scale calibrated in micrometres and projected
onto the focal plane of the microscope, the thickness of each surface layer was
measured at two points per sample. This was relatively straightforward for the inside
surface because the gel coat here was generally smooth and parallel-sided. The final
(outside) layer, however, was usually very much more uneven and more judgement
was required to estimate the thickness.
RESULTS
Vessels
(1) Measured and derived properties: The measured wall thickness profiles for all
four vessels along both the strain-gauged and comparative meridians are shown in
Fig. 3. Mean thicknesses along the strain-gauge meridian in the torisphere and in the
cylinder (excluding the tapered hub) of each vessel are shown in column 3 of Table 4.
The mean glass content results from the burn-off samples are shown in columns 4 to
6. Columns 7 and 8 give the mean thickness of each resin-rich layer from the two
samples taken in the case of the torisphere and the single sample for the cylinder.
In Part 1, individual panel results were used to derive representative properties for
the corresponding vessel. There are doubts about the validity of this method, partly
because the percentage glass content of the vessels was invariably significantly lower
than those of the panels and partly because of the effect the resin-rich layers were
found to have, particularly on the apparent stiffness of the laminate in bending. It
was thought that these differences were likely to be more important than the dfects of
using different batches of nominally similar materials and of minor variations in
laminating and curing conditions that were supposed to be accounted for by the
testing of 'representative' panels. It was therefore decided to use the results from all
fourteen panels tested to date in an attempt to derive more realistic properties for
TORISPHERICAL ENDS FOR G R P PRESSURE VESSELS 39
o~o~o~
~.,~ ~ ~,r-
I-
<
II
<
~..~ -....
4:1.
2z. ~ I - ] I 1 I i I I --t 24 I - I I r I I I I I -I
20~_ Vessel 11 (strain gauged meridian) 20~-- Vessel 13 (strain gouged mer,dion)
16 Sphere Torus l Cylinder ~[ 16~- .... ".
.. ..., ...,, ..
12[:':'"" " """. . . . . •......
0 i........; ................... o I I I J li Ji J I t
o loo zoo 30o 40o soo 6oo ?oo 8oo too 0 100 200 300 400 S00 600 700 800 900
24. l - I I I I I I I I -,I 21~1- I I I I I 1 I I -I
20 ~- Vess~ 11(comporntive meridiomi -~ 20 L- Vessel 13 (comp~'xitive meridian)
,,,,p -. ....... -<,,,- '~1~
, F ......... - ...........................
."". • , - . t
1
.... ...... . ...... ........ ...° sp.er. ,,~,-'-~,~"~
e.F:_
~ ~2 "'"................
~o I I I I II II I I
0 100 200 300 400 500 600 700 800 9QO 0 100 200 3100 Z~O 500 600 700 800 gO0
2~,~_ I I I I I I I I -1 I 21.1_ I I I I I I I I -t
~20 I: ,'e.,, 12 ..-. _ 10 E V i i l l l 1/* Istrom gouged ll~idieml
16 k (strain gouged meridian) .." "'.. .t '~ t 6 +.,.~ ......,.+--.~.,
ol- I i i I II il I i -I o I 1 I I II II I I
0 100 200 300 zOO 500 600 730 800 I 0 100 200 300 zOO 500 600 ?00 800 ~00
2¢ i- I I I I I I I I -I 2z'l- I 1 I 1 I I I I J
20 ~- vessel 12 2 0 ~'- Vlisll 1¢, (comporotivi.,meridion) -~
16 V (compofotive meridian) .-"" "
~ 7 I-~.-'"............ .... _..,........."" "., .--°,. 162F[--- ..........'" "'..... ............. ~.
, ~- Spi,e~, ±To, us _7"'""~7i~i',:"'".-:i 15~____ Sphere _~OCUSL_ Cylinder _--I
o
'P I I I I ;1 T I I I
0 100 200 300 /,00
600 700 500 800 0 100 200 300 ~0 500 600 700 800 900
Meridlonol distance from pole mm Meridionot distance from pole mm
2 3 4 5 6 7 8 9 I0 11 12
1 Derived properties
Vessel Region Mean measured properties
Thickness Weight Percentage Specific Thickness of Young's modulus Ultimatestress
ref. No. of (Gea) (gea)
vessel on strain- of glass thickness resin-rich Z
gauged glass contentby (ram per layers ~e
meridian (kg/m 2) weight squaremetre (ram)
(mm) (%) of glass) Inside Outside Tension Bend Tension Bend
4~
42 D. C. BARTON, P. D. SODEN, S. S. GILL
these four vessels, taking into account both the differences in percentage glass
content and the effect of the resin-rich surface layers. The exact procedure whereby
this was done is described in detail in reference 8, but, basically, difference in glass
content were accounted for using approximate empirical relationships between
measured panel properties and the corresponding glass content, whilst the
correction for the resin-rich layers is based on simple laminate theory using the
measured surface layer thickness and an assumed modulus (3.5 GPa). The values of
laminate modulus and UTS derived in this way in tension and in bend are shown in
columns 9 to 12 of Table 4. From these it is a straightforward matter to calculate
values of extensibility and UTUS, when required, by multiplying by the specific
thickness shown in column 6.
(2) Comparison of experimental and theoretical dejbrmations: As previously,
BOSOR 4 was run for each vessel using the measured thickness distribution along
the strain-gauged meridian and pressure increments corresponding to both the
'elastic' cycle and the final loading to failure. This time, however, a different modulus
was used for each vessel, the value being the derived bend modulus for the torisphere
shown in column 10 of Table 4. The value of Poisson's ratio was also increased from
the value of 0.3 used in Part 1 to 0.34, the latter being the overall mean value from a
total of forty separate results from strain-gauged specimens (both tension and four-
point bend) tested as part of the complete programme of work.
In Part 1 it was shown that good agreement was obtained in the general pattern of
experimental and theoretical results at strain levels below the limit of
proportionality (which occurs at approximately 0 . 3 ~ strain for CSM GRP)
although, even at these low pressures, peak deflections and strains were usually
underestimated by BOSOR 4. As the stress-strain curve for this material is nearly
linear right up to failure, one might expect BOSOR 4 to give reasonable predictions
at much higher strain levels. In Part 1, data was not presented for high strains but the
discrepancies were of a similar order of magnitude to those at low pressures. 7
Experimental and theoretical deformations are now presented for vessel 13 at
pressures up to a maximum just below the actual burst pressure. Vessel 13 was
selected because (a) a scan of transducer readings was available just before failure
and (b) no large-scale tilting of the cylinder took place even at high pressures. This
latter, which occurred for most vessels, often made measured radial displacements
difficult to correlate with BOSOR 4 predictions.
The distribution of meridional surface strain is shown in Fig. 4, that of
circumferential strain in Fig. 5 and that of radial deflection in Fig. 6. In each case,
experimental and theoretical values are plotted at an initial pressure of 211 kPa
followed by four equal increments of 200 kPa up to a maximum of 1011 kPa.
(3) Summary ofpredictedandmeasured results ( Table 5):Table 5isequivalentto
Table 4 of Part 1 and all results are calculated in a similar manner. Values of SCF
and SNCF (columns 2 and 3) were calculated from BOSOR 4 stress and strain
predictions for the measured wall thickness using eqns (6) and (7) of Part 1 but
taking t as the mean measured thickness in the torisphere. This latter deviates from
1.2
I l T F i l" Bosor ~,
• E xperlmentol
1"0-- - - -
0"8
0-6
Cylinder
~, 0.1.
0.2
100 200 300 400 500 600 ?00 800 900 1000
Heridionol distun£e from pole mm
(o)
o.8 I I I I l I Bosor
0.7 _
• Experilntal
I
Sphere , Torus _ , _ Cylinder
0.6 ~ - : _ _ =
• \
0.5 • --
\
". • ,
~ -
~ 0.2
i - -
0-1 --
o I I I I I ~
-0'1 -- --
-0"2 -- - -
-o.3 I I I I I I i I I
0 100 200 300 /.00 500 600 700 800 900 1000
Meridionol dstonce from pole m m
(b)
Fig. 4. Distribution ofmcridional strain t'orvessel 13 at P = 211, 411, 611, 811 and I011 kPa. (a) Inside
surface strain. (b) Outside surface strain.
0.8 I I ~ I I I T T I _
Sphere - I - Torus -1 - Cyhnder
0"6
0"~.
0"2
-0"2
-0"~.
- 0 "6
100 200 300 ~00 500
J
600
J
700
~
•
800
Bosor I.
Experimentul
900 1000
Meridionol distance from pole mm
(o)
O. 8 r - ~ - - - - - - 1 - - - T--T~T--T T T T T l
,,u
Sphere .~ _ CyI,.def _ _~
0.6--
0,~,--
0.2
~ 0
-0.2
-O.L,
Bosor 4
-0'6 ~ l ~ l ~ l J j J
100 200 300 ~,00 SO0 600 700 BOO 900 1000
Moridionol distance from pole mm
(b)
Fig. 5. Distribution ofcirfumferential strain for vessel 1 3 a t P = 2 1 1 , 4 1 1 , 6 1 1 , 8 1 1 and 1011kPa.(a)
Inside surface strain. (b) Outside surface strain.
TORISPHERICALENDSFOR G R P PRESSUREVESSELS 45
S'O!
4..0
3"0
i 2"0
~ 1.o
¢s o
Ls 1 ~.
-1 ' 0
-2.0
-3"0[ I 1 I I l I I I
-1000 -800 -600 -400 -200 200 /,00 600 800 1000
Meridional distance fro• pole mm
Fig. 6. Distribution of radial deflection for vessel 13 at P--21 l, 411, 61 l, 811 and 1011kPa.
Part 1, where the nominal thickness of the dished end was used, but seems more
realistic in that the mean measured thickness was always significantly higher than
the nominal value.
P0.2y. from BS 4994 (column 4) was calculated as in Part 1 using the extensibility
in bend (derived in the improved manner) and the mean measured weight of glass in
the torisphere. The corresponding value from BOSOR 4 (column 5) is defined as
that pressure at which the maximum predicted strain in the torisphere is equal to
0-2 %. A similar definition applies to the experimental value (column 6). Failure
pressures from BOSOR 4 are defined as the pressure at which the maximum
predicted stress is equal to the derived UTS in tension (Table 4, column 11) for the
cylinder, or in bend (Table 4, column 12) for the torisphere. Burst pressures from
BS4994 use corresponding UTUS values and the mean weight of glass in the
cylinder and torisphere, respectively.
DISCUSSION
were less. However, it is thought unlikely that strength results for the latter in bend
are accurate due to the large deflections occurring for these thin specimens near
failure. All values of ultimate strain were greater in bend.
Vessel tests
(1) Thickness distributions (Fig. 3): As before, there is reasonable consistency
between the two meridians measured in each case. Each vessel is thickened to some
extent in the torus region and is generally thicker in the torisphere than the cylinder.
Vessels 11 and 12, in particular, show a large increase in thickness centred on a
region just on the knuckle side of the sphere/torus junction. This is presumably due
to a build-up of overlaps of successive glass mat layers which occurs during
manufacture. Of the four, vessel 13 is probably the most uniform, particularly along
the strain-gauged meridian.
(2) Measured and derived properties (Table 4): The mean thicknesses in
column 3 of Table 4 confirm the trend suggested by the thickness plots, all vessels
being, on average, significantly thicker in the torisphere than in the cylinder. It can
be seen that it is, in fact, the five-layer specimen (vessel 11) that comes nearest in the
dished end to the nominal thickness of 10 mm. The measured weight of glass per unit
surface area (column 4) was also greater in the torisphere in each case, but the
percentage glass contents (column 5) and the specific thicknesses (column 6) both
indicate that the proportion of glass to resin was less. As usual, percentage glass
contents are considerably less than for the corresponding flat panels. The
thicknesses of the resin-rich layer on the inside surface (column 7) were roughly
twice those on the outside (column 8). Both thicknesses tended to be higher than
those measured for the flat panels, the overall mean values for which were 0.8 mm
and 0.3 mm, respectively (see reference 8).
The derived properties shown in columns 9 to 12 of the Table accurately reflect the
major trends observed in the results from the flat panel tests. Values of modulus in
bend (column 10) are lower than those in tension (column 9) and are particularly low
for vessel 11. UTS values in bend (column 12) are higher than in tension
(column 11).
(3) Comparison of experimental and theoretical deformations (Figs 4, 5 and 6):
Considering first the strain distributions shown in Figs 4 and 5, it can be seen that the
overall pattern of experimental results is well predicted by BOSOR 4, even at the
highest pressure. The main discrepancy occurs near the pole where the strain gauge
readings suggest a high degree of bending, which is theoretically not present. The
maximum measured strain is, as predicted, that on the inside surface in a meridional
direction at a position near the centre of the torus region. At the lowest pressure
loading (211 kPa), which corresponds to a maximum strain of about 0.2 %, the
measured strains in the torus are slightly less than predicted. In contrast, for the
maximum pressure (1011 k Pa) they are generally greater. This is due to the gradual
reduction of the tangent modulus of CSM GRP which occurs at strains above
48 D . C . BARTON, P. D. SODEN, S. S. GILL
properties and measured weights of glass are much greater than those calculated
using the nominal values shown in Table 1, Despite the greater weight of glass in the
torisphere compared with that in the cylinder and the greater strength in bend,
failure pressures from BS 4994 in each case are still much lower for the dished end
than for the cylinder. Values of Po.2% and Puit predicted by BOSOR 4 for the
torisphere are greater than those from BS 4994, particularly for vessels 11 and 12 of
geometry D design. This latter is mainly due to the discrepancy in the shape factor
from BS 4994 for this geometry. Values of Punt in the cylinder are lower from
BOSOR 4 because the program takes account of the stress-raising effect of local
thickness variations in the meridional direction. The net results of these differences is
that the BOSOR 4 failure pressure in the torisphere is only lower than the
corresponding value in the cylinder for vessels 13 and 14--i.e. BOSOR 4 predicts a
cylinder failure for vessels 11 and 12. However, although the maximum stress was in
the cylinder for these vessels, the maximum strain predicted by BOSOR 4 was still in
the torus.
As can be seen from the.actual burst pressures given in the Table, vessels 11 and 12
did, in fact, fail in the cylinder. The mode of failure in each case was very similar to
that described for vessels 1,2 and 3 in Part 1. Both experimental failure pressures are
less than predicted by either BS 4994 or BOSOR 4. As discussed in Part 1, this is to
be expected because of the simplifications in the theoretical model and the use of
mean, rather than minimum, properties and glass contents.
Vessels 13 and 14 both failed in the knuckle of the dished end. The appearance of
the fracture was similar to that described for vessels 4, 5 and 6 in Part 1. The main
crack through the gel coat on the inside surface for vessel 13 was at a meridional
distance of about 580 mm from the pole, very close to the position of maximum
theoretical (and experimental) stress and strain for the measured thickness
variation, as shown in Fig. 7. The meridional distance for the corresponding crack in
vessel 14 was about 610 mm--i.e, much closer to the tangent line. This region is still
highly stressed on the inside surface although theoretically the position of maximum
stress and strain is as for vessel 13. Far more circumferential surface resin cracks
could be seen on the inside surface of the knuckle for vessel 14 and they extended
further around the circumference. This, together with the high burst pressure
recorded in this case (column 12), suggests that more non-catastrophic damage was
allowed to accumulate before the vessel ultimately failed. More energy also seemed
to be released at failure for this vessel since a complete segment was blown out from
the main body of the specimen, as shown in Fig. 8.
The fact that, of the fourteen vessels tested to date, five have failed first in the
cylinder has not made it easy to compare different geometries by considering burst
pressures alone. However, this can be done for the two geometries with r i / D ~ = O. 1 ;
namely, geometry C with h i / D i -= 0-19 and geometry E with h i / D ~ -~ 0-25. The mean
burst pressures for the two nine-layer vessels tested in each case were 990 kPa for
geometry C and 1165 kPa for geometry E. The ratio between these values is not very
50 D. C. BARTON, P. D. SODEN, S. S. GILL
80
T ] T F T T l l F
Posalon of fmlure -'~ I Pos~honof fodure
o ~
-20
\ / . . . .
-40
100 200 300 400 500 600 700 800 900 1000
Meridionol disfonce from po(e mm
(a)
1.0
0"8
Sphere l_ Cytinder
o.4_ _ ,r_l"j \-
different from the inverse ratio of the K s factors in BS 4994 which are 2.85 and 2.25,
respectively. A torisphere burst pressure was also obtained for one specimen
(vessel 4) of geometry B design (rJD i = 0.19, hi/D~ = 0"25, K~ = 1.80) and this, at
1575 kPa, also suggests that the relative values of Ks are about right as far as failure
is concerned.
CONCLUSIONS
The theoretical and experimental results from the group of vessels reported here
confirm all the major conclusions from the previous group discussed in Part 1. In
addition, the following remarks may be made.
(1) According to BS4994, both geometry D (rJD~=O.146) and geometry E
(r~/D~ = 0.1) qualify for the higher shape factor of the two specified for the head
52 D. C. BARTON, P. D. SODEN, S. S. GILL
ACKNOWLEDGEMENTS
This work was carried out in the Applied Mechanics Division of the Mechanical
Engineering Department, University of Manchester Institute of Science and
Technology. Finance was provided by the Polymer Engineering Directorate of the
Science Research Council, to whom the authors are duly grateful. The authors
would also like to thank Plastics Design & Engineering Ltd, who manufactured the
specimens, Scott Bader Ltd, who supplied the resin, and Fibreglass Ltd, who
provided the glass fibre. Mr B. D. Gray of P.D. & E. Ltd and representatives of Scott
Bader Ltd have given guidance and advice throughout. Acknowledgement is also
due to Dr D. Bushnell of the Lockheed Missiles and Space Company for the use of
the computer program BOSOR 4 and to Professor G.D. Galletly of Liverpool
University who made the program available in immediately usable form.
TORISPHERICAL ENDS FOR GRP PRF_.~URE VESSELS 53
REFERENCES
I. ANON., BS4994. Specification for Vessels and Tanks in Reinforced Plastics, British Standards
Institution, 1973.
2. HUGIiES,D. A., BARTON,D. C., SODEN,P. D. and GILL, S. S. The effect of thickness/diameter ratio on
the structural behaviour of the torispherical ends of glass reinforced plastic cylindrical pressure
vessels, North Western Branch of I.Chem.E. and I.Mech.E. Joint Symposium on Reinforced Plastic
Constructed Equipment in the Chemical Process Industry, Manchester, 1950.
3. BARTON,D. C., SODEN,P. D. and GILL, S. S. The strength and deformations of torispherical ends for
glass-reinforced plastic pressure vessels--Part l : Effect of torus radius/cylinder diameter ratio on ends
with sphere radius equal to cylinder diameter, Int..L Pres. Ves. & Piping, 9 (1951), pp. 285-318.
4. ANON., BS5276. Specification for Pressure Vessel Details (Dimensions). Part 4: Standardized
Pressure Vessels, British Standards Institution, 1977.
5. ANON., AD-Merkblatt NI. Pressure Vessels in Glass Fibre Reinforced Plastics, Vereinigung der
Technischen Uberwachungs-Vereine e.v. (VdTUV), Essen, West Germany, 1969.
6. BUSHNELL,D., Stress, stability and vibration of complex branched shells of revolution, Comp. and
3truct., 4 09?4), pp. 399--437.
7. BARTON,D. C. Ph.D. Thesis, Faculty of Technology, University of Manchester, 1951.
8. BARTON,D. C. and SODEN, P. D. An investigation of the short-term in-plane stiffness and strength
properties of chopped strand mat reinforced polyester laminates, Composites 0952), in press.