THE DESIGN OF COMPRESSION

STRUCTURES FOR MINIMUM WEIGHT

by
D. J. FARRAR, M.A., [Link].S.

1. INTRODUCTION 1.2. TYPE OF STRUCTURE CONSIDERED

1.1. FUNDAMENTAL IMPORTANCE OF THE It is assumed that it is required to design a
1
PROBLEM surface carrying a certain ultimate compres-
'T'HE primary structure of an aeroplane sive end load. The surface consists of a skin
-*- usually consists basically of a set of stiffened by longitudinal stringers, and
tubular beams. The main structural box of supported at intervals by ribs or frames.
the wing or tailplane is a well-known As a limiting case this type of structure can
example: a semi-monocoque fuselage is include the two-spar wing, in which there are
another. For any given loading condition of only two " stringers " which are stabilised by
the aircraft the material in the tube is stressed the spar webs. The trend of modern design
mainly in tension, in shear, or in compression, has been towards the thick stiffened skin type
depending on its location in the tube cross of construction, the thick skin being needed
section. for torsional stiffness in the case of the wings
The aim of the designer is to make the or tailplanes and for carrying pressurising
material fulfil these three functions in the loads in the case of fuselages. In addition to
most economical manner. In tension, he is the reinforced skin cover there may also be
limited only by the quality of material avail- spar booms which carry part of the end load.
able. In shear, this is again substantially the
case, although it is well known that very light 1.3. AIMS OF THE INVESTIGATION
shear webs over great depths do not develop Of the various design conditions, the one of
as high an effective failing stress as do more over-riding importance is that the structure
sturdy webs. This property of dependence should have adequate strength. The primary
on the intensity of loading is much more purpose of this investigation therefore is to
marked in the case of the compression determine the structure which has a minimum
structure, which is liable to instability in weight when its external dimensions and its
various ways. strength are given.
The fundamental problem of the design of In general the design conditions often
such a structural element is the stabilising of involve other aspects besides strength. The
the compression surface. The ingenuity of provision of adequate stiffness may be one:
j the designer in this respect has led to various ease of manufacture, and limitation of work-
' types of structure being adopted: one may ing stresses by fatigue considerations, may be
mention the two-spar wing box, the thick others: they are not inevitable design
skins reinforced by stringers and supported conditions, whereas strength considerations
by ribs in the case of a wing, or by frames are.
in the case of a fuselage, and the sandwich A means is therefore provided for assessing
structure. It is clear that a generalised the effect of geometrical limitations which
approach cannot cover all forms of structure, may be imposed by these other requirements
and in the present case the second type only on the structural weight, so that a structure
is considered. can be chosen in which the weight increase,
Paper received March 1949.
above that necessary for strength alone
Mr. Farrar is Assistant Chief Designer, Aircraft in the minimum weight structure, is also a
Division, Bristol Aeroplane Co. Ltd. minimum.
1041

Downloaded from http:/[Link]/core. Monash University, on 11 Dec 2016 at [Link], subject to the Cambridge Core terms of use, available at
http:/[Link]/core/terms. [Link]
 D. J. FARRAR

1.4. LIMITATIONS AND DIFFICULTIES of the ribs or frames is completely dictated

In a subject with so broad a scope it is by practical considerations, one need go no
inevitable that any solution involving a farther.
reasonable amount of labour is not a precise In a structure of fair size it is generally
and complete one: in fact it may be assumed possible to vary the rib or frame spacing.
to be a design condition of the structure that The second stage of the problem is then to
the labour involved in design should not be vary the rib or frame spacing so that the total
prohibitive. Some reasonable limits to the weight (ribs + skin + stringers) is a minimum.
range of the problem explored, and perhaps
to the accuracy of the answer desired, must 2. CHARACTERISTICS OF THE SKIN-
therefore be set.
STRINGER COMBINATION
To illustrate this point, one may note that
it is necessary to include many forms of 2.1. CONDITIONS ASSUMED
stringers so as to pick the best type for the
particular structure—typical shapes being Z- It is assumed that the number of stringers
section, top hat, bulb angle, and so on. A is sufficiently large for the instability stresses
certain amount of experience is necessary in to be the same as jn an infinitely wide panel
order to choose a limited range of sections (more than three or four stringers usually
which must be subjected to theoretical ensures that this is so). The ribs are assumed
analysis. Even when a basic type such as to supply simple support so that the com-
the Z-section has been chosen, there are bination of skin and stringers acts as a
several shape parameters which can be widely pin-ended strut of length L.
varied.
When a single shape of stringer has been NOTATION
chosen for analysis, the situation is still far P—compressive end load carried per
from favourable on the theoretical side. The inch width of skin-stringer combina-
various modes of instability are often quite tion.
heavily coupled so that analysis is difficult L = r i b or frame spacing.
and depends on a large number of para- T = thickness of skin which has the same
meters : the failing stress may be appreciably cross sectional area as the skin-
different from the instability stress because of stringer combination.
initial eccentricity or lateral air loading, and E= compression Young's modulus of
so on.
skin-stringer material.
In view of these difficulties one may feel E T = tangent modulus of skin-stringer
that it is not possible to find an optimum material.
structure. However, the nature of an K=radius of gyration of skin-stringer
optimum is such that the weight is a combination.
mathematical minimum with respect to the / = m e a n stress realised by skin and
design parameters: it follows therefore that stringers at failure. (Note: f=P/T).
if our methods of analysis provide only a
reasonable approximation to the truth, the d= flange width of stringer.
weight will still be very close to the mathe- h = depth of stringer.
matical minimum achieved by the ideal b—stringer spacing.
structure. ts = stringer wall thickness.
t = skin thickness.
As=cross sectional area of stringer.
1.5. DIVISION OF THE PROBLEM
/. = 3.62 E{tj b)\
Let us suppose that the compression skin- fb — initial local buckling stress.
stringer combination is required to carry an fL = secondary local buckling stress in
end load P per unit width, and that the ribs stringer.
or frames are uniformly spaced at a distance fT = secondary torsional buckling stress
L apart along the length of the structure. in stringer.
If we fix P and L, and assume that the D = depth of rib.
ribs or frames possess a certain minimum TB= thickness of an ideal plate rib which
stiffness, we must first determine the lightest has the same weight as the actual
skin-stringer combination which can carry the rib.
loading P over the bay length L. This is the W = weight of skin-stringer combination
first stage in the analysis: and if the spacing per unit area.
1042

Downloaded from http:/[Link]/core. Monash University, on 11 Dec 2016 at [Link], subject to the Cambridge Core terms of use, available at
http:/[Link]/core/terms. [Link]
 THE DESIGN OF COMPRESSION STRUCTURES FOR MINIMUM WEIGHT

2.2. GENERAL CASE(1) At this stage the analysis becomes more

Consider a skin-stringer combination which detailed since a variety of types must be con-
is equivalent in area to a skin of thickness T. sidered and the individual instability stresses
Let the radius of gyration of the combination must be explored for each type over a range
be K. of geometrical parameters. Once the value
Then if the mean compressive stress in the of A0 has been determined, families of
skin and stringers is /, optimum structures can immediately be
produced for all P and all L, so that the
P = fT . . . (1) actual design process is extremely rapid once
Euler instability of the combination is the general investigation is completed. As a
given by specific example here we consider the case of
Z-section stringers, which are widely used.
f = ^ . • (2)

where ET is the tangent Young's modulus of 2.3. INSTABILITY CHARACTERISTICS OF Z-

the material. Since the skin-stringer com- SECTION STRINGERS
bination can be regarded as an assembly of The Z-section stringer-skin combination
interconnected plates, it will also develop can develop several separate types of
instability of a local plate-buckling type. The instability, which may be coupled to a greater
mean stress at which this local instability or less degree.
occurs is given by (a) Skin buckling (or initial buckling). This
f = AET(T/Kf . . (3) generally involves waving of the skin between
where A is a constant depending on the stringers in a half-wavelength comparable
geometry of the cross section. There may be with the stringer pitch. There will also be a
several modes of local buckling; we will certain amount of waving of the stringer web
assume that the one to which A is appro- and lateral displacement of the free flange.
priate will involve distortions of the section For some proportions these latter may
sufficiently large to induce premature flexural become larger than the skin displacements,
instability and failure. and the mode becomes more torsional or
It is easily demonstrated that the lightest local in nature (see (b) and (c)).
structure is that in which flexural instability (b) Local instability. A secondary short-
and local instability coincide. If we impose wavelength buckling may take place in which
this condition we find that the stringer web and flange are displaced out
of their own planes in a half-wavelength
f=«*A*(^y. . (4) comparable with the stringer depth. There
will be smaller associated movements of the
Now the coefficient A depends on the type skin and lateral displacements of the stringer
of skin-stringer combination chosen, and for free flange.
the optimum structure will have a maximum
value. If we call this maximum value A0, (c) Torsional instability. The stringer
then the highest stress which could ever be rotates as a solid body about a longitudinal
achieved is given by axis in the plane of the skin, with associated
smaller displacements of the skin normal to
u=*Aa^y . (5) its plane and distortions of the stringer cross-
section. The half-wavelength is usually of
The implication of this result is an the order of three times the stringer pitch.
important one: it is that the maximum stress (d) Flexural instability. Simple strut
which could possibly be achieved is dictated instability of the skin-stringer combination in
entirely by the value of the structure loading a direction normal to the plane of the skin.
coefficient P/L, which is the fundamental There may be small associated twisting of the
quantity dictating the weight of the skin- stringers. The half-wavelength is generally
stringer combination. It is thus economic to equal to the frame spacing.
make P as large as possible by concentrating
all the compression material in the sheet- (e) Inter-rivet buckling. Buckling of the
stringer cover, and almost eliminating spar skin as a short strut between rivets: this can
booms. be avoided by using a sufficiently close rivet
In seeking to determine the best type of pitching along the stringer.
skin-stringer combination, we must determine (f) Wrinkling. A mode of instability
the best value of A which each can achieve. similar to inter-rivet buckling, but analogous
1043

Downloaded from http:/[Link]/core. Monash University, on 11 Dec 2016 at [Link], subject to the Cambridge Core terms of use, available at
http:/[Link]/core/terms. [Link]
 D. J. FARRAR

1 O We will therefore consider, separately, two

cases. In the first, primary local buckling
causes flexural failure. In the second, skin
16 buckling develops and flexural failure is
precipitated by secondary local buckling. It
1-4 will be assumed that inter-rivet buckling and
1-4 wrinkling are avoided by suitable disposition
of the riveting.
1-2

1-2 3. Z-SECTION STRINGERS—INITIAL

fb LOCAL BUCKLING CAUSING
fo t
FAILURE' 2 '
10
3.1. INITIAL LOCAL BUCKLING
The local buckling mode which is the first
08 to develop is a mixture of modes (a), ip) and
(c) of Section 2.3, the predominant type of
buckling being dictated by the geometry of
0-6 the particular skin-stringer combination used.
The stress at which this initial buckling
occurs has been determined theoretically by
0-4 J. H. Argyris' 3 ' for a wide range of skin-
stringer combinations, allowance being made
for all the inter-actions between the various
0 8 modes of distortion. Typical results are
0-2
shown in Fig. 1 in which the non-dimensional
0 6 buckling stress (fblfo) is plotted against AJbt
04 for various values of tjt. (Note: f0 is the
0
0-4 0-8 1-2 1-6 2() buckling stress of the skin if pin-edged along
As the stringers, and fb the actual initial buckling
bt stress.)
Fig. 1. The upper portions of the curves corres-
Initial bucklinjI stress ()f flat p inels wit h Z-section pond to a skin-buckling-cum-stringer local
stringer s (d/h = 0.3). type of instability, while the lower portions
of the curves correspond to a stringer
torsional-cum-lateral type of instability over
to wrinkling of a sandwich structure, in which a longer wavelength. The change of slope in
the skin develops short-wavelength buckling the curves takes place when these two types
as an elastically supported strut. For all of initial buckling occur at the same stress.
practical skin-stringer combinations it can be
avoided by keeping the line of attachment
very close to the stringer web. 3.2. FLEXURAL INSTABILITY
Considering a stringer associated with a
2.4. FAILURE OF Z-SECTION STRINGERS pitch b of skin, the whole cross section is
When the skin-stringer combination fully effective until initial local buckling
approaches its Euler instability stress, occurs.
development of instabilities (a), (b), (c), (e) In general the stringer will not develop
or ( /) will so reduce the flexural stiffness as pure flexural instability: there will also be a
to cause premature collapse. certain amount of stringer twisting, the
If the Euler instability stress is reasonably analysis of which is far from simple.
remote, instability (a) (skin buckling) will not Fortunately the type of design which this
precipitate failure, and the structure will carry analysis will show to be most efficient is one
increased load, with the skin buckled, until in which flexural-torsional coupling is small,
failure occurs by the onset of instability (b), and we will therefore assume that pure
(c), (e) or (/). In general an excessive margin flexural instability occurs. If we consider
of flexural stiffness is needed to prevent d/h=0.3, the second moment of area of the
failure due to any of these latter four modes. skin-stringer combination is
1044

Downloaded from http:/[Link]/core. Monash University, on 11 Dec 2016 at [Link], subject to the Cambridge Core terms of use, available at
http:/[Link]/core/terms. [Link]
 THE DESIGN OF COMPRESSION STRUCTURES FOR MINIMUM WEIGHT

0.6336/+ 0.37M, 3 F- 0-5 0« 07 075 OS

h hL (6)
l.6hts + bt
The flexural instability stress is therefore
, Jt2ET ,, 0.633bt +0.37 hts
JE= -yr- *hJ7, x (7)
u " (iMts+bty
We equate this to the initial local buckling
stress
h=hlf0x3.62EI(tjbf . (8)
The load per inch run is then given by
l.6ht.
'=*(>+ *£*) (9)
where / = / E = / B .
Combining equations (7), (8), (9) in the
same manner as equations (1), (2), (3) we find

'-W^ (10)
where F is a function of AJbt and t,/t,
corresponding to the quantity ^A* of
equation (4); it is a measure of the structural
efficiency of the skin-stringer combination.

3.3. OPTIMUM STRUCTURE

F = is
The quantity ( f J pjr) Plotted
against AJbt and tjt in Fig. 2.
It is seen that an optimum value of AJbt
and tjt exists, at which for a given P, ET
and L the stress realised will be a maximum.
For this optimum design with AJbt= 1.5 and
tjt=1.05 we can write _ Fig.2.
Contours of / y—— for Z-section stringers where
j = 0.95^f • • (ID rL,T
It is also noticeable that a ridge of high initial buckling coincides with failure.
realised stress exists for the family of designs
where "the two types of local buckling occur A useful feature now becomes apparent:
simultaneously; if for any reason the if for any reason it is not possible to use the
minimum-weight design cannot be used, it is minimum-weight design, Fig. 3 shows the
economic to use designs of this family. (The amount of weight penalty incurred. For
general principle seems to emerge that the example, if practical considerations demand
most efficient designs are those in which
failure occurs simultaneously in all possible that b ( p p - ) cannot be less than 2.0, the
buckling modes.)
weight of the structure will inevitably be
3.4. DESIGN CHARTS 21 per cent, greater than the lightest possible.
If we consider only the more efficient If b ( ~ \ could be reduced to 1.5, the
designs, in which the two modes of local
buckling occur at the same stress, the results increase would only be 9 per cent.
can be presented in the form of Fig. 3. Thus the efficiency of the structure from a
From these curves the stress realised by a weight point of view can be related quanti-
given type of structure, together with the tatively to geometrical limitations imposed by
stringer pitch and depth, and the skin and consideration of ease of manufacture, stiff-
stringer thicknesses, may be found at once ness, ease of maintenance and so forth, and
for any basic design conditions P, ET and L. with such a quantitative relationship a
1045

Downloaded from http:/[Link]/core. Monash University, on 11 Dec 2016 at [Link], subject to the Cambridge Core terms of use, available at
http:/[Link]/core/terms. [Link]
 D. J. FARRAR

16
y
:V
t
•B^_
1-2

10

0-8
L
" ^
0-6

L V PL
0-4

0-2

PL3
Fig. 3.
Design chart for optimum Z-section stringers where initial buckling coincides with failure.

compromise is more readily found in which more efficient structures, is to provide a

allowance is made for all the various design slightly greater margin of flexural stress than
conditions. is allowed for local buckling, the actual
margin depending on the standard of
3.5. EFFECT OF INITIAL ECCENTRICITY straightness of the stringers which can be
achieved.
In practice the full theoretical value of It should be noted that even when initial
eccentricity has been allowed for, the
F( =
/A/PF~ rs n o t ac
hieved: experimental theoretical optimum design is also the
results indicate that about 90 per cent, is an practical optimum. However, in an actual
average realised value. This reduction is due design it is convenient to eliminate errors in
to the effect of initial eccentricity of the the estimated failing stress (due to approxi-
structure, and (unlike the case of a simple mations in the present theory, to the use of
strut) it occurs even at low design stresses. clad skins, or to initial eccentricity) by one
Suppose that the Euler stress and skin or two panel tests.
buckling stress of a particular skin-stringer Although the present analysis was intended
combination are both 30,000 lb./in. 2 . When only as a guide to determining the best skin-
an actual compressive stress of 27,000 lb./in. 2 stringer combination, in practice it has pro-
is reached, it is likely that the bow due to vided a quite reliable estimate of the strength
initial eccentricity will give an additional of panels, agreeing closely with test results.
stress (due to bending) of 3,000 lb./in. 2 in
the skin: thus the total stress in the skin will 4. Z-SECTION S T R I N G E R S -
be 30,000 lb./in. 2 , the skin will buckle and SECONDARY LOCAL BUCKLING
induce premature flexural failure. CAUSING FAILURE' 4 )
A simple method of allowing for these
effects is to design for a theoretical failing 4.1. GENERAL
load per inch P (assuming no eccentricity) In this class of design, the skin buckles at
somewhat greater than that actually required. some stress below the failing stress of the
A more refined method, which gives slightly panel. Strictly speaking the mode of buckling
1046

Downloaded from http:/[Link]/core. Monash University, on 11 Dec 2016 at [Link], subject to the Cambridge Core terms of use, available at
http:/[Link]/core/terms. [Link]
 THE DESIGN OF COMPRESSION STRUCTURES FOR MINIMUM WEIGHT

involves movement of the stringer as well as (c) Stringer torsional instability stress
the skin, and is in fact a complex one as The simple analysis developed by
mentioned in Section 3.1. H. L. Cox(5) gives for the edge stress at
We now require to investigate the post- which stringer torsional instability
buckled behaviour, and so to arrange the occurs
geometry that the flexural instability stress GJ
is coincident with the critical stress for one U= ^2-£ME. . (H)
or more secondary local buckling modes. where
These secondary local buckling modes may Stringer polar moment of
be of the nature of stringer local instability inertia J' =0.633fc3fs
or stringer torsional instability, modified by Stringer St. Venant torsion
restraint from the skin. constant J = 0.533/tfs3
A rigorous solution to this problem is not Bending torsional con-
yet in sight. Accordingly, approximate stant H -Olh'dX
methods will be used; they are justifiable in Skin support stiffness K—EF/b
the first place because, as has already been Shear modulus G =0.385E
remarked, the weight of an optimum structure It should be noted that at edge stresses
designed by such methods is very close to the greater than about three times the skin
true ideal minimum. A second justification buckling stress, the stringer rotation in
appears at the end of the analysis; it is that the torsional instability mode is in the
the optimum structures determined here same direction for all stringers, owing
happen to be of such a type that the to the effect of the skin buckles. We
approximations used seem to be very close also note that the potential energy of the
to the truth. We can make two general skin and lateral distortion of the stringer
observations regarding efficient structures: — have been neglected; these are valid
(a) Coupling of modes reduces the lower approximations in the present case.
instability stress and raises the higher,
thus leading to an inefficient structure. (d) Flexural instability stress
(b) The most efficient structure is that in We assume that after the skin buckles
which every type of instability which it has a constant
could cause failure occurs simul-
taneously. ^ A 7 E R A G E of0.3.
"/EDGE
By letting failure occur at more than about For a stringer and a pitch of skin the
three times the skin buckling stress we tangent second moment of area is then
achieve designs in which the stringers are . 0.l9bt + 0A7htSJSt „„
relatively sturdy, and in which the coupling
between skin buckling and stringer local ^=03bt+l.6ht. h%
- • (15)
distortion is negligible. The load per inch at flexural instability
We now attempt to satisfy (b), namely that is then
the local, torsional and flexural instability x2ET (0.\9bt + 0.47hts\
P= h% (16)
stresses shall be equal, and assuming that bU V 0.3bt+l.6hts )
failure occurs at more than three times the
skin buckling stress. (e) Actual load per inch
The expression for the load carried is

4.2. INSTABILITY STRESSES bP = hBaJl.6hts + bt(o.3+0.7y^-)) (17)

Notation as in Section 3.1. It is assumed
thatd/A=0.3. 4.3. OPTIMUM DESIGNS
Using equations (12) - (17) we impose the
(a) Skin buckling stress condition that flexural, torsional and local
f„ = Mf. x3.62E r (f/fc) 2 . . (12) buckling should occur at the same edge
(b) Stringer local instability stress stress. Owing to the form of equation (14)
it is not possible in this case to evolve a
The following approximation is used simple solution of the form
for the edge stress at which stringer local
instability occurs
fL = 3.62ET(tJhy . . (13) H PE T
- constant,
1047

Downloaded from http:/[Link]/core. Monash University, on 11 Dec 2016 at [Link], subject to the Cambridge Core terms of use, available at
http:/[Link]/core/terms. [Link]
 D. J. FARRAR

01 0-2 0-3 h 0-4 0-5 0-6 0-7

b
Fig. 4.
Design chart for optimum Z-section stringers where secondary buckling causes failure.
(Light alloy material.)

as for the unbuckled structure. Instead, is about 1.20, which is greater than for
the equations lead to families of solu- unbuckled designs; in fact as far as failure
tions which are shown in Fig. 4 for light by elastic instability is concerned the buckled
alloy (EV = 107). It is notable that the structure is superior to the unbuckled one.
realised mean stress / again depends on the The limitation on the buckled structure
structure loading coefficient P/L, and the occurs when the edge stress begins to exceed
mean stress realised by the optimum that permitted by present materials, which it
structure varies very nearly as s/(P/L) as in does for 2values of P/L exceeding about
the case of the unbuckled designs. If we 100 lb./in. . For values of P/L above this,
adopt this as a standard of comparison we the unbuckled type of structure will tend to
become more efficient.
see that the value of F ) realised
\'N'PET 4.4. RELATIVE MERIT OF BUCKLING AND
1-2 NON-BUCKLING DESIGNS
In Fig. 5 the structural efficiency, as
II
measured by / J - ^ - , of the designs is plotted
'fit
10 against the ratio of the skin buckling stress
to the edge stress at failure. It is seen that
0-9 the skin should either not be allowed to
buckle at all, or should buckle at a com-
paratively low stress, if good structural
0-8 efficiency is desired.
0 0-2 0-4 0-6 0-8 10 1-2
RATIO OF SKIN BUCKLING STRESS TO FAILING
In Fig. 6 the best possible results using
EDGE STRESS. Z-section stringers are given for current light
Fig. 5. alloy material. It is assumed that the
Effect of skin buckling on theoretical stress realised optimum design is used, and the mean stress
by optimum Z-stringer-skin combinations. thus achieved is plotted against the value of
1048

Downloaded from http:/[Link]/core. Monash University, on 11 Dec 2016 at [Link], subject to the Cambridge Core terms of use, available at
http:/[Link]/core/terms. [Link]
 THE DESIGN OF COMPRESSION STRUCTURES FOR MINIMUM WEIGHT

P/L. It is seen that the working stress, and where F is a constant for the optimum design.
hence the structural weight, is dictated The skin-stringer combination is therefore
entirely by the value of P / L which is used, equivalent in weight to a skin of thickness T
and typical values, of this quantity for where
various aircraft components are shown.
T=P//
In practice, requirements of torsional
stiffness may often cause the unbuckled
=Wf ••'(«>
design to be used. If a certain minimum skin If the density of the material is p, then the
thickness is required for stiffness, reference weight per unit area of skin-stringer surface
to Figs. 3 and 4 gives at once the stress is
realised by the best unbuckled design and the I PL
best buckled one, from which the choice of W-
•V p-T • (19)
F v E
the better type of structure can be made for
the particular case concerned. We see therefore that W is a minimum
when the value of \JET/ p is a maximum. (It
is noteworthy that an analysis of tubular
5. OTHER TYPES OF STRINGER struts'6* by Pugsley also yields this result.)
5.1. BASIS OF COMPARISON Qn this basis Fig. 7 has been prepared to
We have now established that the show the relative efficiency of optimum com-
structural efficiency of a skin-stringer com- pression skin-stringer combination in various
bination can be measured by the constant F, current materials. (It should be noted that
for a given P and L the geometry of the
which is / . hrp^and which has a definite structure will be different for each material,
\PET being adjusted to retain the optimum design
maximum value for any given type of for that particular material. It is assumed
stringer. The results of some similar that such optimum designs realise F = 1.15.)
calculations and tests for stringer sections
other than Z are given below. 6.2. COMPARISON OF MATERIALS
Type Theoretical Realised
best value value The results show that present aluminium
alloys are the most efficient at high values of
the loading coefficient P/L; at lower values
Z-section, primary buckling of P/L, magnesium alloys are more efficient;
causing failure 0.95 0.88 at lower values still, wood is efficient, and for
Z-section, secondary buck- very low values of P/L (appropriate to model
ling causing failure ... 1.20 1.14
Hat section, primary aeroplanes) balsa wood is indicated as the
buckling causing failure 0.96 0.89 best structural material.
Y-section, primary buckling The potentially most efficient material is
causing failure 1.25 1.15 that with the largest value of s/E/p and
5.2. STRUCTURAL EFFICIENCY OF VARIOUS
TYPES OF STRINGER (OPTIMUM
DESIGNS)
We see that of the range explored, the
Y-section stringer is the most efficient,
although for values of P/L less than 100 the
buckled skin and Z-section is as good arid
in practice is more robust. Design charts
similar to Fig. 3 can easily be plotted for
each shape of stringer.

6. CHOICE OF MATERIALS
6.1. EXTENSION OF THE THEORY
Consider a structure having a bay length
L and sustaining a compressive load per inch 100 ISO 200 2S0 100

width P. We have demonstrated that

/= IP%
=Wf Fig. 6.
Theoretical stress realised by optimum Z-stringers
and skin in aluminium alloy.
1049

Downloaded from http:/[Link]/core. Monash University, on 11 Dec 2016 at [Link], subject to the Cambridge Core terms of use, available at
http:/[Link]/core/terms. [Link]
 D. J. FARRAR

JU STEEL • • ^ ^ - ^

8/
40 ^^\o^
6
I0 W 7 ^ ^
ypL sa
30 7 / ^s~
*0^
LB. IN.
*
3rt
C\J ;

10

j
50 100 150 200 250
- £ LB./IN.'

Fig. 7.
Relative weight of optimum skin-stringer combination in current materials.

magnesium alloys promise very well in this 7. CONSIDERATIONS AFFECTING

respect, since a 100 per cent, increase in proof RIB SPACING*7)
stress would make them more efficient in
general application than any increased 7.1. OBJECT OF ANALYSIS
strength aluminium alloy. A further pro- Since the stress achieved by the skin-
mising development is an aluminium alloy stringer combination varies as V (P/L), a light
with its Young's modulus increased by combination for a wing is achieved by placing
30 per cent.: comparison with best present the ribs close together. When this is done,
results is made in Fig. 8, where the effect of however, the weight of ribs may be consider-
increased alloy strength is also shown. able, and the structure of minimum weight
in fact may be associated with a wider rib
spacing, the increase of weight of the skin
30 and stringers being more than offset by the
"l^-u
*>--'/ saving oh ribs.
Theoretical investigation of this problem is
_5J
made difficult by the complex nature of the
20 loads on a rib, and consequent ignorance of
1. HICHER STRENGTH
io6w ALUMINIUM ALLOY.
rib weights.
VPL 2. ALUMINIUM ALLOY WITH
YOUNC'S MODULUS
LB. IN. INCREASED 30'/: 7.2. DETERMINATION OF RIB WEIGHT
10
STREI VCTH INCR EASED The design of rib is a detail matter and the
SO'/: weight of a rib cannot be predicted in
practice from pure theoretical considerations.
Accordingly a more or less empirical
so iso 250 approach must be adopted.
LB/lN- 2 It can be demonstrated that in order to
carry air load the strength of the rib must be
Fig. 8. directly proportional to the rib spacing,
Relative weight of optimum skin-stringer whereas to prevent general instability the
combination in possible future materials. stiffness of the rib must be inversely
1050

Downloaded from http:/[Link]/core. Monash University, on 11 Dec 2016 at [Link], subject to the Cambridge Core terms of use, available at
http:/[Link]/core/terms. [Link]
 THE DESIGN OF COMPRESSION STRUCTURES FOR MINIMUM WEIGHT

proportional to the rib spacing. Since some

parts of the rib will be designed by one con- $
sideration, some by the other, and some by
minimum practical sizes, it appears likely
that the weight per rib will not vary very 0-3

greatly with the rib spacing. ••*

Figure 9 has been prepared on this basis
for some weighed bracing ribs in wings, So,
tailplanes and fins on various aircraft. The '\y''
low scatter of the values indicates that the
ignoring of the effect of rib spacing does u. 0 1
not introduce any serious error. In default <*%^^
of better information we will then assume
15
that the weight of a rib does not depend on
the rib spacing within practical limits. 10 20 30 40 SO
DEPTH OF RIB INS.

7.3. DETERMINATION OF BEST RIB SPACING Fig. 9.

Typical weights of bracing ribs.
Consider unit width of compression
structure carrying a load per inch P at a mean
stress /. The ribs are a distance L apart, Weight of ribs per unit skin surface area
while each rib has a depth D and is
equivalent in weight to a plate of thickness
TR. We will assume for the present that all -»(<¥)" [(£)'• (A)1]
stresses are within the elastic range. 7.4. PRACTICAL APPLICATION
Then
The foregoing analysis sho>vs that the
t-T^iIPE weight of ribs should be equal to about half
the weight of the surfaces designed by com-
and the equivalent thickness T of the skin- pression. The analysis assumes that
stringer combination is __
r=p/f=Xj V
§L 005

F PL
When we include the weight of ribs, the total TOTAL

equivalent thickness becomes \ V 004

(T • DT
* V
skin.

-) SKIN +
ST RINGERSy'

Now
//
003
DTB _ 1 / TZ DT„ o \
T+ e \ /'
~ F V E~ + L
For the structure of minimum weight L must
be chosen so as to make this quantity a
a*
CO

VA /
002
Weight, lb

TOTAL.
minimum. Differentiating with respect to L
\ and equating to zero we get (assuming F and RIBS
\ SKlh + ^ ^ ^^
E constant) \ \ STRI
001

_L /JL _ QI" =0 (20) RIBS

2 F \ EL U ^
•

and
/4F2D2T„2E y (21) so 100 20 40
L= L INS. L INS.
Fig. 10.
This is the optimum rib spacing, and we find Examples of the variation of total weight
that with rib spacing.
Weight of skin and stringers per unit surface P = 10,000 lb./in. P = 2,500 Ib./in.
Z-stringers. Z-stringers.
Alum. Alloy. Alum. Alloy.
Ribs 60 in. deep. Ribs 25 in. deep.
1051

Downloaded from http:/[Link]/core. Monash University, on 11 Dec 2016 at [Link], subject to the Cambridge Core terms of use, available at
http:/[Link]/core/terms. [Link]
 D. J. FARRAR

secondary stringer stresses due to lateral design charts of the type derived here (based
loading are unimportant: this must be on P/L) permit a good approximation to the
checked in each case but is generally an final optimum structure to be quite rapidly
acceptable approximation. obtained, while continuous allowance is made
It is also noteworthy that the depth and for other practical design factors.
equivalent thickness of the ribs have as much
influence on the configuration and weight of ACKNOWLEDGMENTS
the structure as the load per inch itself.
In many cases, owing to practical con- The author's thanks are due to the Bristol
siderations, the quantities F and E will vary Aeroplane Co. Ltd. for permission to publish >
somewhat with rib spacing; in such cases the this article. Research and test work done by f
nature of their variation always causes the the Bristol Aeroplane Company have pro- i
optimum weight of ribs to be slightly less vided much information on which some of
than half that of the skin and stringers. A the diagrams have been based, but the
satisfactory and rapid approach is to consider opinions expressed are those of the author.
a range of values of L and find the total Part of the analysis has been developed
weight for each value; this has been done in and published by workers in America, at the
Fig. 10 for a typical wing structure and for a same time as it was developed independently
large tailplane as examples. in this country.
By the nature of an optimum it is
permissible to use a frame spacing slightly REFERENCES
different from the optimum one without
much weight penalty. It is advisable in 1. ZAHORSKI, ADAM. Effects of Material Distribu-
tions on Strength of Panels. JOURNAL of Aero-
practice to use a spacing slightly greater than nautical Sciences, July 1944.
that for pure minimum weight, since a
2. FARRAR, D. J. The Failing Stress of Flat Panels
simpler and more robust structure results. with Z-Section Stringers. The Bristol Aeroplane
Co. Ltd., Technical Office Report No. 23, Sept.
8. CONCLUSIONS 1945. Scientific and Technical Memorandum
No. 19/47.
No analysis of this subject can claim to be 3. ARGYRIS, J. H. The Initial Instability Stress of
comprehensive, and the present one is no Flat Panels with Z-Section Stringers. [Link].S.
exception. Simplifications of the problem Structures Data Sheet 02.01.25.
have been made, with the object of producing 4. FARRAR, D. J. 1st Addendum, On the Design
a fairly quick approximation to the optimum of Skin-Stringer-Rib Combinations for Com-
skin-stringer-rib combination; this simplifi- pression Surfaces. Addendum to The Bristol
cation is essential because of the large Aeroplane Co. Ltd. Technical Office Report
No. 39, August 1947.
number of variables involved. One must
expect more detailed methods and a certain 5. Cox, H. L. Instability of Stringers and Sheet-
amount of testing to be used when checking Stringer Combinations. A.R.C. Report No.
Struct. 773.
the overall strength and stiffness once a
structure has been fixed. It is of interest to 6. PUGSLEY, A. G. The Relative Strength and
note that the divergences in strengths from Stiffness Properties Required for Struct
Materials. A.R.C. Report No. Struct. 1113.
those found by the present simplified theory
have been found, so far, to be quite small. 7. FARRAR, D. J. On the Design of Skin-Stringer-
Rib Combinations for Compression Surfaces.
The importance of the structure loading The Bristol Aeroplane Co. Ltd. Technical
coefficient P/L has been demonstrated, and Office Report No. 39, April 1947.

1052

Downloaded from http:/[Link]/core. Monash University, on 11 Dec 2016 at [Link], subject to the Cambridge Core terms of use, available at
http:/[Link]/core/terms. [Link]