Accepted Manuscript

A review of simple formulae for elastic hoop stresses in cylindrical and spherical
pressure vessels: what can be used when
G.B. Sinclair, J.E. Helms
PII:

S0308-0161(15)00007-1

DOI:

10.1016/j.ijpvp.2015.01.006

Reference:

IPVP 3430

To appear in:

International Journal of Pressure Vessels and Piping

Received Date: 24 June 2014

Revised Date:

13 January 2015

Accepted Date: 16 January 2015

Please cite this article as: Sinclair GB, Helms JE, A review of simple formulae for elastic hoop stresses
in cylindrical and spherical pressure vessels: what can be used when, International Journal of Pressure
Vessels and Piping (2015), doi: 10.1016/j.ijpvp.2015.01.006.
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to
our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please
note that during the production process errors may be discovered which could affect the content, and all
legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT

A review of simple formulae for elastic hoop stresses in cylindrical and spherical
pressure vessels: what can be used when

RI
PT

G.B. Sinclaira*
J.E. Helmsb
Department of Mechanical Engineering, Louisiana State University, Baton Rouge, LA 70803,
U.S.A. E-mails: a sinclair@me.lsu.edu, b helms@me.lsu.edu

M
AN
U

SC

Abstract
Classical simple formulae for elastic hoop stresses in cylindrical and spherical pressure
vessels continue to be used in structural analysis today because they facilitate design procedures.
Traditionally such formulae are only applied to thin-walled pressure vessels under internal
pressure. There do exist, however, some variations of these formulae that remain simple yet
permit wider use. Here, by reviewing various underlying rationales for simple hoop stress
formulae, we make a determination of when and how well different formulae apply. For the
formulae that do apply to thicker vessels than usually recognized, we give companion results for
external pressure.

TE
D

Keywords: Hoop stresses, pressure vessels, design formulae

AC
C

EP

_______________
To receive correspondence. Telephone: (225)767-5786
Home e-mail: glennsinclair@cox.net

ACCEPTED MANUSCRIPT
1

1. Introduction
Distinguished by the subscript c, the classical formulae for the elastic hoop stress, ,

c =

RI
PT

produced by an internal gauge pressure p acting within thin-walled pressure vessels have
pr
pr
, c =
,
t
2t

(1)

pR
,
t

TE
D

B =

M
AN
U

SC

for cylindrical and spherical vessels, respectively. In eqns (1), r and t are corresponding inner
radii and thicknesses.
Possibly the earliest development of an expression like the first of eqns (1) is that
obtained experimentally by Mariotte circa 1670, [1]. Mariotte tested closed thin-walled
cylindrical tanks by connecting them to standpipes placed on top of them. He found that the
height of the water in the standpipes when cylinders burst was proportional to the cylinders wall
thickness and inversely proportional to its radius. That is, in effect, p t/r at rupture. Thus
Mariotte experimentally confirmed the essential elements of the first of eqns (1).
It is not clear to us when the explicit expressions for the hoop stresses in eqns (1) were
developed. An alternative form for the first of eqns (1) was provided by Barlow circa 1830, [2].
Identified with the subscript B, this has
(2)

AC
C

EP

for cylinders, where R = r + t is the external radius. Otherwise we are unaware of just who
developed eqns (1) and when. It would seem likely, however, that this occurred before Lams
derivation of the more complex formulae for hoop stresses in thick-walled cylindrical and
spherical vessels. This derivation was reported in [3] in 1852.
The classical formula for the hoop stress in cylinders, the first of eqns (1), has been given
in strength of materials texts, and occasionally in statics texts, for over a century. Examples of
the former are [4,5]. An example of the latter is [6]. For more than 50 years this formula has
been given in introductory elasticity books [7,8], texts on shell theory [9], design books [10], and
engineering handbooks [11].
This formula continues to be cited in modern mechanics of materials texts right up to the
present time. In chronological order, examples are [12-19].
It is also given in current
introductory solid mechanics books [20,21] and machine design books [22-25], as well as
current/recent handbooks [26-28]. At this time it would appear to be fair to say that the classical
formula for the hoop stress in cylinders under internal pressure has gained long-standing and
widespread acceptance.
Similarly the classical formula for the hoop stress in spheres, the second of eqns (1), has
been given in texts of yore [4], [6,7], [9], [29]. It too continues to be given in current/recent texts

ACCEPTED MANUSCRIPT
2

RI
PT

including all of the modern mechanics of materials texts previously cited, as well as an
introductory solid mechanics text and some handbooks, [20], [26, 27]. It is not usually given in
machine design books. Nonetheless it would seem reasonable to view the classical formula for
the hoop stress in spheres under internal pressure as having also gained long-standing and
widespread acceptance.
The formulae of eqns (1) only apply away from any discontinuities that cause stress
concentrations, and provided pressure vessel walls are sufficiently thin. The most commonly
accepted range of sufficiently small wall thicknesses is
(3)

SC

t r 1 10.

M
AN
U

That is, that the thickness be at least one order of magnitude smaller than the inner radius:
requirements of this nature are common in engineering when defining relatively small
dimensions (see, e.g., [12]). On occasion, further justification for the range in inequality (3) is
offered by noting that it leads to no more than about a 5% deviation from the maximum hoop
stresses, [13],[18]. Though less frequent than inequality (3), there are some other upper limits
for applying eqns (1) given in the literature. Examples are t/r 1/20 in [24], t/r 1/5 in [22],
[27]. When upper limits are exceeded, presumably formulae for thick-walled vessels are to be
used.
These Lam formulae for the maximum hoop stress, max , produced by internal pressure

max =

p ( R2 + r 2 )
t (R + r)

TE
D

within thick-walled pressure vessels have

, max =

p ( R 3 + 2r 3 )

2t ( R 2 + Rr + r 2 )

(4)

AC
C

EP

for cylindrical and spherical vessels, respectively. Derivations of the formulae in eqns (4) may
be found in [30]: therein it is also explicitly pointed out that in the limit as t 0 the stresses in
eqns (4) recover their counterparts in eqns (1). Clearly, therefore, there has to be some
sufficiently small range of thicknesses such that eqns (1) suffice. Within this range, then, the
relative simplicity of eqns (1) facilitates design by enabling the ready determination of design t
given allowable and specified p and r. This simplicity is a key reason for the continued use
of formulae like eqns (1) today. A detailed and precise explanation of the role of these formulae
in the design of pressure vessels can be found in the ASME code [31], or the exposition of this
code in Megyesy [32].
On the other hand, as t in eqns (4), max p, p 2, respectively, in marked
contrast to eqns (1) as t which has 0 for both cylinders and spheres. Hence there has
to be some upper limit or limits on t/r for eqns (1) to provide reasonable respective estimates of
eqns (4). One objective of the present review is to check inequality (3) in this role and, if
needed, to augment it so that upper limits are set in a clear and consistent way.

ACCEPTED MANUSCRIPT
3
While eqns (1) contain by far the most common simple expressions for hoop stresses in
the literature, there are also some variations in these expressions themselves. One such is the
already noted expression of Barlow for cylinders, eqn (2). Another for cylinders under internal
pressure has
pr
,
t

RI
PT

S =

(5)

pr
,
2t

M
AN
U

R =

SC

where r = r + t/2 is the mean radius. As far as we can discern, this formula was first given in
Shigleys Mechanical Engineering Design circa 1972, thus the subscript S. It continues to be
given in the current version of Shigley [23]. It is also provided in [24], [32-34]. A further
alternative for spheres under internal pressure has
(6)

the analogue of eqn (5) in effect. As far as we can discern, this formula is only given in Roarks
Formulas for Stress and Strain [28], thus the subscript R.
Two other simple formulae are furnished in the ASME code [31]. Distinguished by the
subscript d for design code, these have
(7)

TE
D

d = c + 0.6 p, d = c + 0.1 p,

AC
C

EP

for cylinders and spheres under internal pressure, respectively. Precise ranges of application for
these formulae are set out in [31] (t/r 0.500, 0.356, respectively).
While it seems certain that the various alternatives of eqn (5), eqn (6) and eqns (7) would
not have been put forward unless they offered some advantages over their counterparts in eqns
(1), it is not obvious from any of the references listed here just exactly what these advantages
are. Accordingly, as this review continues we seek to try and clarify the relative merits of eqns
(5) (7) versus eqns (1).
All of the foregoing applies to internal pressure. With external pressure, buckling is
possible. Consequently most discussions of classical formulae in the literature preclude their use
with external pressure because of a sense that, in the limited thickness ranges in which formulae
like eqns (1) could be physically appropriate, buckling occurs. There are instances in the
literature, however, which, recognizing that buckling may not always so dominate, suggest that
then eqns (1) apply with a sign change. That is, the hoop stress with external pressure, e , is
simply given by

e = - ,

(8)

ACCEPTED MANUSCRIPT
4

M
AN
U

SC

RI
PT

where is as in eqns (1) or eqn (5) for cylinders, eqns (1) or eqn (6) for spheres. This view of
the hoop stress produced by external pressure on cylinders is stated in [15], [28], [35,36] for the
first of eqns (1) and [32] for eqn (5) . For spheres, it is stated in [35, 36] for the second of eqns
(1), and [28] for eqn (6). Here, for instances in which buckling does not preclude their use, we
also seek to review the accuracy of eqns (1), eqn (5) and eqn (6) in conjunction with eqn (8)
when pressure is applied externally.
In the remainder of this review, we begin (Sect. 2) with a recap of the traditional statics
derivations underlying eqns (1), followed by systematic improvements afforded by synthetic
division. Shigleys formula eqn (5) is a natural outcome. The resulting formulae for hoop
stresses are lower bounds. Thus next (Sect. 3) we develop upper bounds. Barlows formula (2)
is one outcome of this exercise. Thereafter (Sect. 4) we combine lower and upper bounds. This
leads to design formulae, including those of ASME in eqns (7). Then (Sect. 5) we examine what
formulae apply with external pressure. In light of this review, we conclude (Sect. 6) by
summarizing what simple formulae can be applied when.

pr
= c ,
t

TE
D

2. Lower bounds
A straightforward way to obtain lower bounds is to determine average hoop stress values
using equilibrium: being averages, these hoop stresses have to be lower bounds for hoop stress
maxima.
To this end, first we view the free-body diagram in Fig. 1 as being for half of a cross
section of a cylindrical pressure vessel under internal pressure. Then balancing forces per unit
out-of-plane length gives

(9)

EP

where is the average cylindrical hoop stress and c is as in the first of eqns (1). Accordingly,
as is well recognized, this c is an average value and so a lower bound.

AC
C

Second we view Fig. 1 as being for half of a cross section of a spherical pressure vessel.
Then, as shown in [12], balancing forces on an inner circle with those on an outer annulus gives

pr 2
,
2rt

(10)

where is now the average spherical hoop stress.

As noted in [30], the lower bound in eqn (9) coincides with the dominant term in the first
of eqns (4) as t 0. This lower bound can therefore be improved by including the next t terms
from eqns (4) as t 0.

ACCEPTED MANUSCRIPT
5
To obtain such an improved estimate, we use synthetic division rather than binomial
expansions because then there is no ambiguity as to the sign of the entire remaining terms. Thus
for cylinders from the first of eqns (4) and eqn (9),

t
t2
+
.
2r 2r ( 2r + t )

(11)

RI
PT

max = 1+

Hence the first-order correction results in

pr
= S,
t

(12)

SC

wherein is the consequent improved lower bound for cylindrical hoop stresses, r continues as

M
AN
U

the average radius, and S is as in eqn (5). That is still a lower bound is confirmed by the
positive sign of the last term in the brackets in eqn (11). This improved hoop stress estimate is
explicitly observed to be so in Shigley, [23].
To gain an appreciation of the improvement afforded by , we compare errors with eqn

AC
C

EP

TE
D

(12) with those of the first of eqns (1). If we adopt the conventional range of application of
inequality (3), eqn (1) underestimates cylindrical hoop stress maxima on averaging by
integrating over this range by 2.5%, and at most by 5.0%. The latter is a confirmation for
cylinders of the observations to this effect in [13], [18]. In contrast, for the same range, eqn (12)
underestimates hoop stress maxima on average by 0.08% and at most by 0.23%: clearly
markedly more accurate. Moreover, suppose instead we choose to be somewhat more cautious
because errors with lower bounds are not conservative and limit underestimates to 1%. Then the
first of eqns (1) is only applicable for 0 < t/r 1/50: eqn (12), on the other hand, is applicable
for 0 < t/r 2/9, or eleven times the range for eqns (1). Tradition and familiarity aside,
therefore, there would seem to be little if any reason to use the first of eqns (1) for cylindrical
hoop stresses instead of eqn (12).
Proceeding similarly for spheres, the second of eqns (4) and eqn (10) have

max

t 2 ( 2r + t )
t
.
= 1+ + 2
2
2
2r 2r ( 3r + 3rt + t )

(13)

Now the first-order correction results in

pr
= c,
2t

(14)

ACCEPTED MANUSCRIPT
6
wherein is now the improved lower bound for spherical hoop stresses, and c is as in the
second of eqns (1). That is still a lower bound is confirmed by the remainder in eqn (13).
The fact that this c realizes an improved estimate of max over that of the average value is

RI
PT

generally recognized in the literature (a good specific example of this being the case is given in
[12]). In fact, of eqn (14) is about an order of magnitude more accurate than of eqn (10)
in the conventional range of inequality (3).
If, as previously, we adopt the range of inequality (3), the c of eqn (14) underestimates
spherical hoop stress maxima on average by 0.21% and at most by 0.63%. These are
approximately an order of magnitude lower than corresponding percentages for c for

M
AN
U

either c or for cylinders.

SC

cylindrical hoop stresses. Accordingly it is not consistent to give a single range of applicable
thicknesses for both classical formulae. If instead we adopt a limit on nonconservative errors of
1%, then c of eqn (14) is applicable for 0 < t r 1 8. This range is quite distinct from that for

pR
= B,
t

(15)

EP

+ =

TE
D

3. Upper bounds
There are quite a number of pressure vessels that have thicknesses outside of the
applicable ranges given for the improved lower bounds in the preceding section (see, e.g., the
cylinder in [27] that has t/r = 0.9). With a view to avoiding the nonconservative errors that limit
the applicability of lower bound estimates of hoop stresses, here we obtain upper bounds instead.
For cylindrical pressure vessels, recognizing that (R2 + r2)/ (R+ r) R in the first of eqns
(4), we have max pR t . Hence for an upper bound for cylindrical hoop stresses

wherein B continues to be from Barlows formula (2). Now the correct limit results for t ,

AC
C

namely p. In terms of the errors produced by estimating corresponding hoop stress maxima
using eqn (15), these are now conservative for all thickness. However, these errors can have
significant magnitudes (up to 21%).
For spherical pressure vessels, recognizing that (R3 + 2r3)/(R2 + Rr + r2) R in the
second of eqns (4), we have max pR 2t . Hence for an upper bound for spherical hoop
stresses

+ =

pR
.
2t

(16)

ACCEPTED MANUSCRIPT
7
Again the correct limit results for t , namely p/2, and errors produced by estimating
corresponding maxima with eqn (16) are conservative for all thickness. Again, too, these errors
can have significant magnitudes (up to 40%).
The upper bounds in eqn (15) and eqn (16) both admit to being expressed as
+ = + p 2 , where is as in eqn (12) for cylinders, eqn (14) for spheres. This relationship

SC

RI
PT

between upper and lower bounds raises the possibility that by adjusting the coefficient of p in the
foregoing expression we may be able to obtain hoop stress estimates that are superior to
estimates from either upper bounds or lower bounds by themselves. We explore this possibility
next.

M
AN
U

4. Design formulae
Motivated by the preceding observation, we consider a class of design estimates of hoop
stress maxima given by

d = + p,

(17)

wherein continues to be as in eqn (12) and eqn (14) for cylinders and spheres, respectively.

TE
D

In eqn (17), is a nondimensional parameter with a range of 0 . When = 0, just the

improved lower-bound estimates of hoop stress maxima are recovered: when = , the upperbound estimates.
To consider other values, we introduce an error measure. In line with the common
practice of designing against yield, we choose our error measure, e, to be the relative error
between the maximum shear stress calculated with d , d , and that calculated with max , max .

AC
C

EP

This choice is readily implemented. However, it is not critical. If instead the von Mises stress is
used, maximum errors for cylindrical vessels are reduced by less than 4% of their values, while
errors for spherical vessels are unchanged. Furthermore, if as earlier the relative error in hoop
stress maxima is chosen, while error measure magnitudes are altered, the ordering of the
performance of the various hoop stress estimates entertained remains the same.
With this choice, the percentage error is given by
e = ( d - max ) max x 100 ( % ) ,

(18)

where d = ( d + p ) , max = ( max + p ) . These maximum shear stresses occur on the inner
walls of pressure vessels whereon p acts.
For cylinders, from the first of eqns (4), eqn (12) and eqn (17), eqn (18) has
e 0 for t r 4 (1- 2 ) . That is, errors are conservative in this range of thicknesses. Within
this conservative thickness range, the maximum error occurs when de/dt = 0. Thus

ACCEPTED MANUSCRIPT
8

emax =

t2
x 100 ( % ) ,
4r ( r + t )

(19)

e 0 for t r

) (1- 2 ).

1- 3 2 + 3 -1

Within this conservative thickness range, the

maximum error is
6rt 2 ( r + t ) + t 4
9r 2 ( r + t )

x 100 ( % ) ,

(20)

SC

emax =

RI
PT

when t/r = 2 .
For spheres, from the second of eqns (4), eqn (14) and eqn (17), eqn (18) has

M
AN
U

when t/r = 6 + 4 - 2.
Turning to specific choices for , we first note the choice effectively made by the ASME
Boiler and Pressure Vessel Code, [31]. Now and henceforth reserving the subscript d to
distinguish these design code estimates of hoop stress maxima, from [31]

d = + 0.1 p,

(21)

wherein continues to be as in eqn (12) for cylinders and eqn (14) for spheres. The d of eqn
(21) are consistent with the earlier reporting of these estimates using c in eqn (7). This

AC
C

EP

TE
D

common choice of = 0.1 for both cylinders and spheres realizes low conservative errors in
hoop stress maxima over quite wide thickness ranges. These conservative error ranges extend up
to t/r = 0.500 for cylinders and t/r = 0.356 for spheres: these thickness limits are precisely the
ones prescribed in [31], so that there the code restricts the use of eqn (21) to thicknesses that only
result in conservative errors. Within these thickness ranges, the average value of the error
measure e is 0.5% and the maximum value is 0.8% at t/r = 0.20 for cylinders, while the average
value is 0.8% and the maximum value is 1.2% at t/r = 0.14 for spheres.
When designing pressure vessels in accordance with code, clearly one has to comply
completely with the applicable code (in the U.S., [31]). Nonetheless, during preliminary design,
wider use of simple formulae can facilitate exploring a range of configurations before settling on
one to subject to the full requirements of the governing code. With this in mind, we seek to
develop formulae for preliminary design that extend over yet larger thickness ranges by
entertaining larger, but still conservative, percentage errors of about five percent. We note that
because such formulae are more conservative than their code counterparts, any design
thicknesses determined with them has to be in compliance with the thickness ranges permitted by
code formulae.

ACCEPTED MANUSCRIPT
9
To this end we merely select = , its mid-range value. Then from eqn (12) and eqn
(17), we have, for a preliminary design estimate of cylindrical hoop stress maxima distinguished
from its code counterpart by the subscript D,
p(R + r )
,
2t

(22)

RI
PT

D =

D =

SC

for 0 t/r 2.0. The average error with eqn (22) is 2.5% while the maximum error is 4.2% at t/r
= 0.50. Likewise from eqn (14) and eqn (17), we have, for a preliminary design estimate of
spherical hoop stress maxima,
pr
=R,
2t

(23)

M
AN
U

for 0 t/r 1.3. The average error with eqn (23) is 3.6% while the maximum error is 6.0% at t/r
= 0.35. As indicated in eqn (23), this design estimate is the same as the formula for spherical
vessels given in eqn (6) and furnished in Roark [28].
In Fig. 2, a graphical comparison of the errors in the maximum shear stress, e of eqn (18),
is shown for d of eqn (21) and D of eqn (22) and eqn (23). We reiterate that [31] prohibits the
use of d outside the ranges given previously: above the upper limits of these ranges, this code

TE
D

uses Lam formulae and so has no error. What Fig. 2 shows, however, is that, if as earlier we
admitted a 1% nonconservative error in preliminary design, increases in the applicable ranges of
d are modest (where d intersect the horizontal dashed lines).
Shown for comparison in Fig. 2 are of eqn (12) and eqn (14) ( = S , c for
cylinders, spheres respectively), as well as + of eqn (15) and eqn (16) ( + = B for cylinders ) .

EP

For these curves, to be consistent we continue to use e of eqn (18) but now with d replaced by
the relevant hoop stress estimate. Consequently the magnitudes of maximum errors for + are

AC
C

different from those stated in Sect. 3 because of the use of the relative error in the maximum
shear stress instead of that in the maximum hoop stress. What Fig. 2 clearly illustrates is that, as
expected, either of the design estimates perform better than or + for both cylinders and
spheres.
Also for comparison, e of eqn (18) for c of eqns (1) is included in Fig. 2(a) for
cylinders, as a broken line. This comparison shows the markedly higher errors that occur with
the traditional classical formula for cylinders.
5. External pressure
With external pressure instead of internal, buckling is the key phenomenon to be guarded
against in pressure vessels. Marks [27] furnishes a simplified approach to meet this objective,

ACCEPTED MANUSCRIPT
10

SC

RI
PT

while the ASME code [31] provides complete procedures for designing against buckling.
Nonetheless, when vessels become thick enough, stresses also need to be considered. Marks
notes that there exist instances of yielding before buckling, and ASME code recognizes the
possibility that allowable stresses instead of buckling criteria may set design limits when
t r > 1/ 4. As thickness ranges for conservative errors for the simple formulae in eqns (15),
(16), (21-23) all extend past t/r = 1/4, here we consider corresponding formulae for external
pressure.
The development of such formulae follows immediately from superposition. For
example, for a cross section of a cylinder (Fig. 3), the combined application of a uniform
pressure applied externally and internally leads to an all-round uniform applied pressure that
induces a hydrostatic state of uniform, normal, compressive stresses of magnitude equal to the
applied pressure. This hydrostatic state includes the hoop stress. Hence for hoop stresses under
external pressure, e , we simply have

M
AN
U

e = - ( + pe ) ,

(24)

TE
D

wherein pe is the applied external pressure and is the hoop stress under an internal pressure of
magnitude pe. That is, all we have to do to obtain hoop stress formulae for external pressure is
take the corresponding result for internal pressure, exchange p for pe, add pe, then change the
sign. Moreover, this result holds not only for cylinders but also for spheres, and for that matter,
for any pressure vessel. This is because, when any pressure vessel is subjected to an all-round
uniform applied pressure, the same hydrostatic stress state is induced.1
Accordingly, the corresponding upper bound formulae for external pressure are
e + = - ( + + pe ) , where + is as in eqn (15) for cylinders and eqn (16) for spheres. Here what is

AC
C

EP

being bounded is the magnitude of the maximum compressive hoop stress, and these stresses
continue to be bounded for all thicknesses. However, these formulae can be expected to be
precluded from applying for smaller thicknesses because buckling dominates. Corresponding
maximum errors for e of eqn (18) continue to be as for eqn (15) and eqn (16). This is because
the maximum shear stresses are the same for external pressure as they are for internal pressure.
Further, design code estimates of hoop stress maxima with external pressure are, from
eqns (21), (24),

ed = - ( +1.1 p ) ,

(25)

where is as in eqn (12) for cylinders and eqn (14) for spheres, with respective thickness
ranges of up to t/r = 0.500 and t/r = 0.356 so as to maintain conservative errors. Lower ends of

Equation (24) also holds for hoop stress maxima in thick-walled vessels.

ACCEPTED MANUSCRIPT
11

eD = - pe + , eD = - pe + ,
t 4
2t 4

RI
PT

these ranges are set by when buckling dominates response (t/r 1/4). The formulae from eqn
(25) are not currently included in the ASME code [31].
Likewise preliminary design estimates of hoop stress maxima with external pressure are,
from eqns (22-24),
(26)

SC

for cylinders and spheres, respectively, with thickness ranges up to t/r = 2.0 and t/r = 1.3 so as to
maintain conservative errors. Lower ends of these ranges are set by when buckling dominates.
To close this consideration of external pressure, we review the performance of eqn (8)
instead of eqn (24). When, as suggested in [15], [28], [35] and [36], eqn (8) is applied to c of

M
AN
U

eqns (1) for cylinders in the usual range of inequality (3), the maximum value of e of eqn (18) is
13%. Even if external pressures were light enough so as not to induce buckling, it is difficult to
see such a large nonconservative error being acceptable. If instead we were to limit such
nonconservative errors to 1%, then the range of applicability is t r 1 150, a range of
thicknesses within which buckling would almost certainly preclude application of c of eqns (1)
with eqn (8) for anything but the lightest of external pressures.
For spheres, performance is not significantly different. When, as suggested in [35, 36],
eqn (8) is applied to c of eqns (1) for spheres in the range of inequality (3), then the

TE
D

corresponding maximum value of e from eqn (18) is 17%. If such nonconservative errors were
to be limited to 1%, the range of applicability is t r 1 200.
While the situation is improved somewhat by using eqn (8) with S of eqn (5), [32], for
cylinders and with R of (6), [28], for spheres, it is still not good. Corresponding maximum

AC
C

EP

errors within inequality (3) are both 9%, while corresponding ranges with limited 1% errors are
both t/r 1/100. All told, the performance of eqn (8) for external pressure in conjunction with
any of the suggested simple formulae is less than satisfactory.
6. Concluding remarks
Throughout the summary of formulae for hoop stresses that follows, estimates are first
given for cylindrical pressure vessels then for spherical pressure vessels. In these formulae for
, p continues as the internal gauge pressure, t as a vessels wall thickness, and r, r and R as its

inner, mean and outer radius. The errors reported are relative percentage errors in maximum
shear stresses.
Consistent classical formulae for hoop stresses are

S =

pr
pr
,c =
.
t
2t

(27)

ACCEPTED MANUSCRIPT
12

pR
pR
,+ =
.
t
2t

M
AN
U

B =

SC

RI
PT

The first of these is apparently due to Shigley circa 1972 and should be used instead of the
traditional expression, pr/t, because it is far more accurate. The second has been in existence for
more than 120 years [4], and hence has the subscript c for truly classical. These formulae are
lower bounds on corresponding hoop stress maxima. Hence errors are nonconservative.
Because they are so, here we limit their magnitude to 1%. Then applicable respective ranges are
0 < t r 0.25 and 0 < t r 0.14. 2 While these precise ranges are a consequence of the limiting
error chosen as well as the error measure, with any consistent basis applicable ranges for
cylinders are uniformly markedly more extensive than those for spheres.
Upper bounds for hoop stresses are
(28)

The first of these is given in Barlow [2]. Errors are conservative for all thickness but can be
quite large: up to 12.5% and 17.6%, respectively.
Hoop stress estimates permitted by the ASME code [31] are

d =

pr
pr
+ 0.1 p, d =
+ 0.1 p.
t
2t

(29)

TE
D

In effect, these are the consistent classical formulae with a constant correction term added. Both
formulae are conservative provided 0 < t/r 0.500 and 0 < t/r 0.356, respectively:
corresponding maximum errors are 0.8% and 1.2%.
Possible hoop stress estimates for preliminary design are
p(R + r )
pr
,R =
.
2t
2t

EP

D =

(30)

AC
C

These are simply the average of respective lower and upper bounds. The second formula for
spheres is given in Roark [28]. Both formulae are conservative for broader ranges than those for
eqn (9), namely 0 < t/r 2.0 and 0 < t/r 1.3, respectively: corresponding maximum errors are
4.2% and 6.0%. Because these hoop stress estimates are more conservative than hoop stresses in
the ASME code, any preliminary design of thicknesses using them within the given ranges is
guaranteed to be in compliance with [31].
For external pressure pe, replace p with pe in the chosen formula for internal pressure, add
pe, and change the sign. That is,
2

By way of comparison, the applicable range with this error limit for the traditional expression for cylinders is but
0 < t r 0.02.

ACCEPTED MANUSCRIPT
13

e = - ( + pe ) ,

(31)

p = pe

RI
PT

where e is the hoop stress with external instead of internal pressure. Then upper limits on
applicable thickness ranges remain the same, but lower limits are replaced by those set by
buckling. Thus classical formulae can be expected not to apply, but other formulae can. When
applicable, errors are the same as for internal pressure.

M
AN
U

SC

Acknowledgment
We are grateful for comments from Dr. S.F. Hoysan of Stress Engineering Services Inc.,
Houston.

References

[1]

Mariotte E. Oeuvres de Mr. Mariotte de lAcadmie Royal des Sciences. Leiden, The
Netherlands: Piere Vander; 1717.

[3]

TE
D

[2] Barlow P. A treatise on the strength of materials, 6th ed., pp. 117-120. London, England:
Lockwood & Co.; 1867.
Lam MG. Leons sur la thorie mathmatique de lelasticit des corps solides. Paris,
France: Bachelier; 1852.

EP

[4] Bovey HT. Theory of structures and strength of materials, pp. 586, 591. New York, NY:
John Wiley & Sons; 1893.

AC
C

[5] Timoshenko S. Strength of materials, Part I elementary theory and problems (in Russian,
page numbers and publisher that follow refer to the English ed. of 1940), pp. 9, 31. New
York, NY: D. Van Nostrand Co.; 1908.
[6]

Lamb H. Statics, including hydrostatics and the elements of the theory of elasticity, pp.
330, 331. England: Cambridge University Press; 1912.

[7]

Timoshenko S, Lessells JM. Applied elasticity, p. 19. Pittsburgh, PA: Westinghouse

Technical Night School Press, 1925.

[8] Southwell RV. An introduction to the theory of elasticity for engineers and physicists, pp.
146, 148. England: Oxford at the Clarendon Press; 1936.

ACCEPTED MANUSCRIPT
14
[9] Timoshenko S. Theory of plates and shells, pp. 398, 407. New York, NY: McGraw-Hill
Inc.; 1940.

RI
PT

[10] Shigley JE. Mechanical engineering design, p. 43. New York, NY: McGraw-Hill Inc.;
1963.
[11] Baumeister T, Marks LS, eds. Mechanical engineers handbook, 6th ed., p. 5-63. New
York, NY: McGraw-Hill Inc.; 1958.

SC

[12] Gere JM, Timoshenko SP. Mechanics of materials, 4th ed., pp. 549-559. Boston, MA:
PWS Publishing; 1997.

M
AN
U

[13] Riley WF, Sturges LD, Morris DH. Mechanics of materials, 5th ed., pp. 238-241. New
York, NY: John Wiley & Sons; 1999.
[14] Craig Jr, RR. Mechanics of materials, 2nd ed., p. 586. New York, NY: John Wiley &
Son; 2000.
[15] Vable M. Mechanics of materials, pp. 239, 242-24. New York, NY: Oxford University
Press; 2002.

TE
D

[16] Patnaik SN, Hopkins DA. Strength of materials: A unified theory, pp. 465, 467. Oxford,
England: Elsevier; 2004.
[17] Spiegel L, Limbrunner GF. Applied statics and strength of materials, 4th ed., pp. 573-576.
Upper Saddle River, NJ: Prentice Hall; 2004

EP

[18] Hibbeler RC. Mechanics of materials, 6th ed., pp. 421-423. Upper Saddle River, NJ:
Prentice Hall; 2005.

AC
C

[19] Nash WA, Potter MC. Strength of materials, 6th ed., pp. 49-51. New York, NY: McGrawHill Inc.; 2014.
[20] Budynas RG. Advanced strength and applied stress analysis, 2nd ed., pp. 162-164. New
York, NY: McGraw-Hill Inc.; 1999.
[21] Shames IH, Pitarresi JM. Introduction to solid mechanics, 3rd ed., pp. 13, 14. Upper Saddle
River, NJ: Prentice Hall; 2000.

ACCEPTED MANUSCRIPT
15
[22] Collins JA, Busby HR, Staab GH. Mechanical design of machine elements and machines,
2nd ed., pp. 385, 386. Hoboken, NJ: John Wiley & Sons; 2009.

RI
PT

[23] Budynas RG, Nisbett JK. Shigleys mechanical engineering design, 9th ed., p. 114. New
York, NY: McGraw-Hill Inc.; 2010.
[24] Schmid SR, Hamrock BJ, Jacobson BO. Fundamentals of machine elements, 3rd ed., pp.
246, 247. Boca Raton, FL: CRC Press; 2013.

SC

[25] Norton RL. Machine design an integrated approach, 5th ed., p. 236. Upper Saddle River,
NJ: Prentice Hall; 2014.

[27]

M
AN
U

[26] Kreith F, Goswami DY, eds. The CRC handbook of mechanical engineering, 2nd ed., pp. 188, 1-89. Boca Raton, FL: CRC Press; 2005.
Avallone EA, Baumeister III T, Sadegh AM, eds. Marks standard handbook for
mechanical engineers, 11th ed., pp. 5-45, 5-46. New York, NY: McGraw-Hill Inc.; 2007.

[28] Young WC, Budynas RG, Sadegh AM. Roarks formulas for stress and strain, 8th ed., pp.
575, 602, 608, 609, 614. New York, NY: McGraw-Hill Inc; 2012.

TE
D

[29] Timoshenko S. Strength of materials, Part II advanced theory and problems (in Russian,
page numbers and publisher that follow refer to the English ed. of 1940), p. 160. New
York, NY: D. Van Nostrand Co.; 1910.

EP

[30] Sokolnikoff IS. Mathematical theory of elasticity, 2nd ed., pp. 300, 302, 345. New York,
NY: McGraw-Hill Inc.; 1956.

AC
C

[31] ASME. The 2010 ASME boiler and pressure vessel code: Section VIII, Division I
p. 19, 22., New York, NY: American Society of Mechanical Engineers; 2010.
[32] Megyesy EF. Pressure vessel handbook, 14th ed., p. 14. Oklahoma City, OK: PV
Publishing; 2008.
[33] Ugural AC, Fenster SK. Advanced strength and applied elasticity, pp. 247, 248. New
York, NY: Elsevier North-Holland Publishing Co.; 1975.
[34] Bickford WB. Advanced mechanics of materials, p. 276. Menlo Park, CA: AddisonWesley; 1998.

ACCEPTED MANUSCRIPT
16
[35] Bednar HH. Pressure vessel design handbook, 2nd ed., pp. 48, 49, 57. New York, NY:
Van Nostrand Reinhold Co.; 1986.

AC
C

EP

TE
D

M
AN
U

SC

RI
PT

[36] Bedford A, Fowler W, Liechti KM. Statics and mechanics of materials, pp. 524, 526.
Upper Saddle River, NJ: Prentice Hall; 2003.

ACCEPTED MANUSCRIPT

Figure captions

Fig. 1. Cross section of half a pressure vessel including pressure exerting fluid.

AC
C

EP

TE
D

M
AN
U

SC

Fig. 3. Superposition of cylindrical pressure vessel configurations.

RI
PT

Fig. 2 Comparison of percentage errors for hoop stress formulae: (a) cylindrical pressure
vessels, (b) spherical pressure vessels.

AC
C

EP

TE
D

M
AN
U

SC

RI
PT

ACCEPTED MANUSCRIPT

AC
C

EP

TE
D

M
AN
U

SC

RI
PT

ACCEPTED MANUSCRIPT

AC
C

EP

TE
D

M
AN
U

SC

RI
PT

ACCEPTED MANUSCRIPT

AC
C

EP

TE
D

M
AN
U

SC

RI
PT

ACCEPTED MANUSCRIPT

ACCEPTED MANUSCRIPT

Highlights
Consistent derivation of lower and upper bounds for hoop stresses
Justifies ranges of application for hoop stress formulae

RI
PT

Furnishes the cylindrical analogue of the improved formula for spheres in Roark

AC
C

EP

TE
D

M
AN
U

SC

Provides correct formulae for external pressure, with applicable upper limits