THE ANTICLASTIC CURVATURE OF

RECTANGULAR BEAMS AND PLATES

by

D. G. ASHWELL, M.A.

1. STATEMENT OF THE PROBLEM each other, but if d is small compared with

b, i.e. if the beam becomes a flat plate, it is
The pure bending of a rectangular beam found experimentally that there is no appre-
about an axis parallel to a principal axis of ciable distortion of the cross-section, and in
its cross-section is well understood if its depth such cases it is usually assumed that no anti-
d is comparable in magnitude with its breadth clastic curvature takes place.
b. The usual theory, noticing the geometrical The behaviour of blades, 1or flat plates, is
fact that at any cross-section the longitudinal discussed
strains of filaments of the beam are pro- out that by G. F. C. Searle' '. Searle points
portional to the distances of the filaments filament with small a length of longitudinal
from some neutral axis parallel to the axis of curvature r about the axis ofTprincipal
a net tension and a radius of
bend-
bending, postulates that the longitudinal
stresses will vary in the same way. The ing experiences a resultant force acting
absence of a resultant longitudinal force towards its centre of curvature. Such a
length, s, of filament is shown in Fig. 2. The
means that the neutral axis passes through resultant
the centre of gravity of the cross-section, and when B isforce towards O is 2Tsin£0, or
infinitesimal, TO. But 6=sjr so
if e is the longitudinal strain at a distance y
from the neutral axis, p is the corresponding
stress, R is the radius of curvature of the
unstrained filaments and E is Young's
modulus for the material of the beam, it is
found that
e=ylR=p/E
giving
1/R=ely.
These strains e cause strains of -<re at
right angles to them, so that the system of
strains e, implies a similar system of transverse
strains -<re, and these imply a curvature of
the beam at right angles to the original,
principal, curvature, given by
\IR'=o-ely=o-IR.
This curvature, <r/R, is the transverse or
" anticlastic " curvature and gives rise to the
sort of distortion shown in Fig. 1. This anti-
clastic curvature can be shown to exist in
beams for which b and d are comparable with
Paper received February 1947.
Mr. Ashwell is a Fellow of Trinity College,
Cambridge. Fig. 1.
708
 ANTICLASTIC CURVATURE OF RECTANGULAR BEAMS AND PLATES

/'=moment of inertia of distorted

cross-section of beam
k=foundation modulus for beam
on elastic foundation
M=bending moment in beam
p=longitudinal stress
r=radius of curvature of any
filament of beam
R=radius of curvature of neutral
(unstrained) filaments
R' = R/<r
s=length of filament
T=tension in filament
x=co-ordinate measured parallel
to axis of bending
Fig. 2. X=b/2-x
y=co-ordinate perpendicular to
that the radial force becomes Tsjr or T\r x—co-ordinate, in plane of
per unit length of filament. Searle shows typical cross-section of beam
.that these radial forces tend to reduce the a- Vf3(l-^)1
anticlastic curvature, and estimates2 that it
will be nearly neutralised when b J{Rd) is A=maximum "dip" of arc of
greater than about six. small curvature
Case(2) investigates the question on the 8/=small element of breadth. {See
same lines as Searle, and also concludes that Section 2)
the value of b2/{Rd) decides the mode of &e defined in Section 2. {See
distortion, a conclusion that is confirmed Fig. 3)
later in this paper. Both Searle and Case 9=angle subtended at axis of
state that flat plates may be expected to curvature by filament of
remain substantially flat except near the length s
edges, and Case further expects that the cr=Poisson's ratio
change from bending in which anticlastic <t>=MR/EI. {See Section 5)
curvature occurs to bending in which it is very
nearly neutralised will appear as a definite
discontinuity, rather than as a gradual 2. DERIVATION OF THE FUNDA-
change. MENTAL EQUATION
It is the object of this paper to determine Figure 3 shows the distorted cross-section
the nature of the distortions of the cross- of a bent beam. The neutral axis, or axis
sections of rectangular beams of any dimen- of zero longitudinal strain, has been taken as
sions, and to show how the transition occurs the jc-axis. Consider the elemental strip
between the two types of bending. pqrs, of which the sides pq and sr are each
perpendicular to CL, the centre-line of the
NOTATION
B, C defined in equations (5)
A'JB'.C'jy, constants of integration. {See
equation (3))
b—breadth of beam, i.e. dimension
parallel to axis of bending
d=depth of beam or thickness of
plate
e—longitudinal strain
E=Young's modulus
E'=E/(l-<r2)
F,f functions defined at the end of
Section (5)
/=moment of inertia of undis-
torted cross-section of beam Fig. 3.
709
 D. G. ASHWELL

cross-section, at their points of intersection point from some neutral axis, and since there
with it. As it will be shown that the slope of can be no resultant transverse force, this
CL is everywhere small, the distance between neutral axis will be the horizontal centre-line
pq and sr, measured along CL, will be called of the cross-section of the transverse beam.
&c. Now as 8x becomes infinitesimal, pqrs This means that the transverse stresses due
becomes a rectangle, so that its centre of to the change in the "initial curvature" of
gravity will be on CL. the transverse beam add up, at any section,
Consider the stress system defined by the to a pure couple about this neutral axis. The
expression p/y=E/R. changes in the longitudinal stresses produced
It will be seen later that there is a second by these transverse stresses are o- times the
stress system to be superimposed on this, transverse stresses producing them, so that
but for the moment attention will be confined the changes in longitudinal stresses over the
to this one. The longitudinal stresses over face of such an element as pqrs in Fig. 3, due
the face of pqrs are proportional to then; to a change in the curvature of the transverse
distances from the neutral axis, and so the beam, add up to a pure couple.
average stress over the element will be the In other words, the change in curvature of
stress at its centre of gravity. Thus the the transverse beam does not affect the
resultant force on pqrs will be T= expression for the net force on the element
(Ey/R)d&x, where y is now the distance-of pqrs, and the effective pressure on the longi-
the centre-line from the neutral axis. T will tudinal strip of which the element 2is a cross-
be a tension when y is positive. section, is still given by Eyd&x/R per unit
Thus the longitudinal strip of which pqrs length longitudinally.
is a cross-section, suffers a net tension if its Thus the transverse beam, in effect, is
centre of gravity lies above the neutral axis, subjected to2 a non-uniformly distributed load
and a net compression if its centre of gravity of EydSl/R per unit length transversely, a
lies below the neutral axis. But this strip is load which is proportional to y, the distance
bent to a radius R, (measured to its un- of the centre-line from the x-axis. So the
strainedfilaments),about the axis of principal transverse beam evidently behaves as a beam
•bending, and by the pressure analogy for on an elastic foundation.
circumferential tensions and compressions Such beams are shown by the usual theory
{see Section 1) this has the effect of a pressure to have the equation
of T/R or EydSx/R* per unit length
longitudinally, tending to deflect it towards
the neutral axis. Thus, as Searle shows, EI^+ky=0 . . (1)
there is a tendency for the anticlastic
curvature to be neutralised. where E and / have their usual significance,
Consider a slice of the beam whose and ky is the load per unit length of the
dimension parallel to the axis of the beam is beam. This equation, however, depends on
8/. This slice will be called the " transverse equating d'y/dx2 to the curvature of the
beam," and its dimensions are evidently beam, an approximation which demands that
length b, breadth §/, depth d. Owing to the (dy/dxf shall be everywhere negligible com-
tendency of the principal beam to bend anti- pared with unity, so that this condition must
clastically, this transverse beam has an be fulfilled before the equation may be
"initial curvature" of <r/R, which cannot used. If anticlastic curvature took place
take place fully on account of the foregoing unhindered it would be quite possible for
considerations, so that the curvature of the dy/dx to be large, but there is experimental
transverse beam is in general less than a-/R. evidence that the distortion of the cross-
This change in curvature of the transverse sections of all types of beams is never great,
beam produces transverse strains which so that in practice, dy/dx is always small and
imply additional longitudinal strains in the equation (1) may safely be used. This is
principal beam. But no longitudinal strains confirmed later.
can occur other than those given by e=y/R, Therefore
so the implied additional longitudinal strains
appear as changes in the longitudinal stresses.
The usual theory for the transverse beam is
assumed, i.e. that the transverse stress at any
point is proportional to the distance of the where E'=E/(l-cr2) (by the usual theory for
710
 ANTICLASTIC CURVATURE OF RECTANGULAR BEAMS AND PLATES

beams such as the transverse beam, in which lateral contraction and extension
cannot take place),
I=U.d3l\2 and k=EdUjR2
E d38ld*y EdU
ie
' l - o - 2 12 dx" + R2 y
which reduces to
dy 12(1-0 . . . . (2 )
dx* d2R2 y K>

This equation gives the shape of the centre-line of the distorted cross-section.
Its complete solution is
y = cosh ax (A' sin ax+B' cos ax) + sinh ax(C sin ax + D' cos ajc)

where 4«*= - ^ - o r a = ^ [ - ^ J . , . (3)

and A', B', C", and D' are constants to be determined.

Now the " natural shape " of the transverse beam is such that it has an " initial
curvature" of <r/R, and this will still be true at its ends, where x= ±b/2, so that
(0 d2y/dx2=<r/R at x= ± 6 / 2 .
Also, the shear in the transverse beam is proportional to d3yjdx3, and this will
be zero at the ends, i.e.
(ii) d3yldx3=0 at *= ± 6 / 2 .
Finally,
(Hi) by symmetry, y is an even function of JC.
These conditions enable A' B', C and D' to be determined, and the equation
of the centre-line is found to be
Ra2 sinh (aft/2)cos (afr/2)-cosh(afr/2)sin(afr/2)
•y— ; - r : ; COSD. &X COS &X ~p
o- smhaft + sinaft
sinh (ab 12) cos (ab / 2) + cosh(afr /2) sin (ab/ 2) .
T •t i . f sinii &.x sin Q..X.
sinh ab + sin ab
Now «*=\/[3(l-«r?)]/(d J RX so that i?a 2 /«•= V[3(l-o- a )]/(do-), giving
V[3(l -<r2)] v _ sinh (afr/2)cos (afr/2)-cosh(a6/2)sin(afr/2)
rf o- ^ sinh ab + sin 06
sinh (a6/2)cos (a6/2)+cosh(a6/2)sin(a6/2)
sinh ax sin ax
sinh a6 + sin ab
(4)
An examination of this expression shows th(at if <r is considered a constant, the
relationship between y/d and xjb depends only on the factor ab and may be written
y/d=Bcosh ab (x/b)cos ab (x/b) + Csinh ab (x/b) sin ab (x/b),
<r sinh (ab 12)cos(afr / 2) - cosh(afe/ 2) sin (a6 / 2)
where B= 2
V[3(l-o- )] sinh ab+sin a6 (5)
o- sinh (a6 / 2) cos (ab / 2) + cosh (ab / 2) sin (afr / 2)
and C = ^[3(1-a- 2 )] sinh ab + sin a6
These expressions were41
first given by Lamb (3) , and have recently received
experimental confirmation* .
711
 D. G. ASHWELL

-0-061
Fig. 4. Fig. 5.

3. EXPRESSION OF RESULTS
It may be seen that
«Z>=bil [3 (1 - a-2)] / >/(Rd)= i/ [3 (1 2)]b/^(Rd),
so that the relationship between y/d and xjb depends only on o- and the non-
dimensional factor b2/(Rd), and for a given value of o-, it is this factor which
determines completely the type of distortion experienced by the cross-section.
Therefore by plotting y/d against x/b for various values of b2/(Rd) pictures may
be obtained showing the modes of distortion of the cross-sections of all types of
rectangular beams for all degrees of bending. A number of such curves have been
plotted in Figs. 4 and 5 for <r=l/3. The broken lines in Fig. 5 show portions of
the curve for b*/(Rd)=4Q0 with y/d magnified by 10 and 1,000 in order to show
its shape more clearly.
4. BEAMS FOR WHICH b*/(Rd) IS VERY LARGE
From Figs. 4 and 5 it appears that if ba/(Rd) is large, the deflection of the
cross-section is substantially confined to the edges of the beam, the deflections near
the centre being quite small in comparison. Also, the similarity of the curves for
b2/(Rd)=l00 and 400 for values of x/b beyond the node next nearest the edge of
the beam suggests that as b2/(Rd) becomes large, the deflection near the edge of
the beam may tend to some constant shape.
Now as b2/ (Rd) becomes large, ab becomes large also,a so that
(0 sinh (ab/2) c± cosh (ab/2) c~ \ g»»' ,
(it) sinh ab czd \ e*,
(Hi) near the edges, cosh ax ^ sinh ax c=: \ e*,
and (iv) sin ab becomes negligible compared with sinh ab.
Using these relations we obtain from equation (4),
\/[3(l-o- a )] y_ \ e*!* [cos (ab / 2) - sin (ab / 2)H e" cos ax
<r d~ i i«"* +
i ii
. jg° '/ [cos(a&/2) + sin(afe/2)H
+ ^
{x b )
= ie" - ^ [cos(ab/2)cosax-sm(ab/2)cosax + cos(ab/2)sitiax+siD.(ab/2)siaax]
=ie-'a'l2-x)[cosa(bl2-x)-sma(b!2-x)l
Now let X—b/2-x(j.e. measure x from the edge of the beam inwards, and
call it X), then,
x
3' • 2 ^ 3 ( 1 - ^ ) ]-' e~" [cos aX - sin aX]
2 2
(6)
where «Z= ^[3 (1 - or )] X/ \/(Rd)= ^[3 (1 - a- )] [b/ \/(Rd)] X/b.
111
 ANTICLASXIC CURVATURE OF RECTANGULAR BEAMS AND PLATES

Therefore if yjd is plotted against aX, or Stresses (if), as shown earlier, form a system
preferably against [b/ y/(Rd)] X/b, the curve of stresses similar (but reduced in the ratio
gives the shape of the edges of any bent beam 1: o-) to that arising in the transverse beam
for which b2/(Rd) is large (say greater than due to the change in curvature from o-/R to
about 100). This has been done in Fig. 6. something less than <r/R.
From this it will be seen that as b2/(Rd) Thus the contribution per unit length
increases for a beam, or a set of beams, the transversely of this system of stresses, to the
maximum deflection occurs at the edges, and bending moment in the principal beam, will
tends to about O.ld (actually, 0.102d if be <r times the bending moment per unit
<r=l/3). length longitudinally in the transverse beam,
which, in the usual way, will be given by

5. RELATIONSHIP BETWEEN CURVA- Ed3

M
TURE AND BENDING MOMENT '= ( l - ^ ) l 2 [ C h a n g e 0 f 1/2? ]
'
per unit length longitudinally
The orthodox theory states that when anti- Ed3
clastic curvature is allowed to take place
d-o- 2 ) 12 IR dx2} '
M=EI/R,
and when it is completely neutralised or the contribution to the principal bending
moment will be <TM' per unit length trans-
M=(EIIR)/(l-<r2) versely, giving a total contribution for the
but the theory gives no clue as to how this whole cross-section of
change occurs. This may now be investigated. +6/2
It has been shown earlier that the stresses •Ed3 <r dY]
contributing to the bending moment consist of M, dx.
(l-o- 2 )12. JR. dx2
(/) the system given by -6/2

p/y=E/R, Thus the total bending moment in the

this being the stress system which would principal beam will be given by
exist if the anticlastic curvature were +6/2
allowed to take place unhindered, and (if) a
set of stresses superimposed on (0, due to the
change in curvature of the transverse beam -6/2
from <r/R to something less than <r/R. or if / is the usual moment of inertia for the
Stresses (0 are dealt with by the usual undistorted beam,
theory and the moment resulting from them +6/2
is given by EI Ed{ <r2Ebds
M=-z+ir)y>dx+
M^EI'/R, 12R (1 -a- 2 )
-6/2
where / ' is the moment of inertia of the dis- +6/2
torted section about the neutral axis. •Ed3
dx.
12(1- )}dx*
OI2 -6/2

OIO
This can be evaluated, and expressed as
i ooe
M=(EllR)<p
/ where $ is a function of <zb or b2l(Rd) given
006 by
/ 0-04 1 . 3 , , ,, 2^30-
/ 002
T
1- + 2ab U*b)- ab*/(l~<r2) F(<xb)
f1 O y/J
where
s0 4 3
° i fx\ 0*"~ ^ O 1 f(ab)=2(B2 + C2) (sinha&+sina&) +
0
°y -0-02 + (B2 - C2 + 2 BC) cosh <xb sin «b +
00* + (B2 - C2-2 BQ sirih <zb cos <zb +
Fig. 6. + 2(B2-C2)ab
713
 D. G. ASHWELL

and F (aft)=(B + Q sinh (aft /2) cos (aft /2) -

- (fi-Qcosh(aft/2)sin(aft/2). 12-5
Again, the only variable is aft or tf/iRd)
(for constant <r), and in Fig. 7 is plotted per-

fJ
10

centage increase in <p against2 ft2/(iW)(for

<r = 1 / 3), f being unity whenft / (Rd)=0. It M
is seen that as b2l(Rd) becomes large, <j> Zui
tends to a value which can 2be shown to be §2' UJ K
Q. U
1.125. This equals 1/(1 -a- ), which is the
value given by the usual theory for beams in O
which the anticlastic curvature is quite IOO 200
neutralised. TQ
Fig. 7.
Therefore
6. COMPARISON OF RESULTS WITH
USUAL THEORY FOR b2j{Rd) /v\ 2<r ft2 1 ft2 ...
SMALL W ^ = 6 / 2 3 8/W = 3 6 / W ( l f < r = 1 / 3 ) -
=

It is useful to compare the results obtained (y/d)x=bi2 is now tabulated against b2/(Rd)
with those given by the usual theory for small for the two theories as follows:—
values of b2/(Rd). This is of interest
because the usual theory is the basis of a (y\ Author's / y \ Usual
standard method of determining Poisson's KRd) \dhx=bi2 theory \d)x=bl2theory
ratio, and the beams used have a very small
value of b2l(Rd). 0.25 0.00693 0.00694
The deflection of the beam at its edges is 1 0.0272 0.0278
taken as the criterion, and the results obtained
by the present theory are compared with 4 0.0857 0.111
those obtained by assuming that anticlastic 16 0.104 0.444
curvature takes place unhindered.
(0 Author's Theory 25 0.102 0.694
The deflection at the end of the beam is
given by It is seen that agreement is very close for
b2/(Rd) less than about one, but beyond this,
(y/d)x=b/2=B cosh (aft/2) cos (aft/2) + the value given by the usual theory becomes
+ C sinh (aft / 2) sin (aft / 2) considerably larger.
which reduces to
, .- 0.102 (sinh aft sin aft)
(W<0*=6/2= sinh aft + sin aft
(for <r=l/3).
(ii) Usual Theory
Anticlastic curvature takes place so that
the curvature of the transverse beam is
everywhere <r/R. If A is the maximum
"dip" as shown in Fig. 8,
(b/2fc^A2R/(r.
Thus, A = (<r/8)(ft2//?),
Now the origin lies on the neutral axis, at
the centre of gravity of the distorted section,
and this, for an arc of small angle, is at a
distance A/3 from the lowest point. So the
maximum deflection at the ends of the trans-
verse beam is 2 A/3=(2/3)o-ft2/(8U).
714
 ANTICLASTIC CURVATURE OF RECTANGULAR BEAMS AND PLATES

7. CONCLUSIONS Consider a beam or plate made of duralu-

min, for which E is, say, 107 lb./in.22 and the
It is seen from the graphs and Table that elastic limit occurs at 30,000 lb./in. From
(0 the present theory agrees closely with the formula p/y=E/R, if such a beam is
the orthodox theory for small values of bent to a radius of less than 167d the elastic
b2/(Rd) (say b2/(Rd) less than about limit is exceeded, so that this value of R is
one), i.e. for beams in which either d is assumed to be the smallest possible. For
not small compared with b, or R is very steel, this figure is about twice as great.
large, We have
(ii) for large values of b2j(Rd) the distorted yjd=B cosh ax cos ax + C sinh ax sin ax,
cross-section changes its shape consider-
ably, becoming undulating, the undula- giving, when x = b / 2, dy j dx = ad x F (a b)
tions decreasing in size, towards the where F (ab) is given in Section 5.
centre of the beam. Now if R = l67d, this can be expressed,
(iii) as2 b2/(Rd) becomes very large (say dy s/(Rd) , p . M </(Rd) ab . ,
b /(Rd) greater than about 100), the
beam is substantially flat except near the
edges where the maximum deflections which depends only on ab, i.e. on b2j(Rd).
tend to the value 0.102J. Thus dy I dx for these conditions, (i.e. x=b/2
and R = l67d), can be plotted against
b2/(Rd), and if this is done, it is found that
8. ASSUMPTIONS dyjdx has a maximum value of about 20.022
It is assumed that: — and tends to about 0.020 as b /(Rd)
(/) the material has the usual characteristics becomes very large.
Now the approximation
as to elasticity, homogeneity, and so
forth. El(diy/dxi)-^ky=0
(ii) the effect of shear in the transverse beam asdepends on neglecting a term in (dyjdx)2,
being small compared with unity, so that
is negligible. in using this expression something of the
(iii) dyjdx remains small compared with order of (0.02)2, i.e. about 4 in 10,000, is
unity, otherwise the equation being neglected.
El(diy/dxi)+ky=0
is not valid, see Section 2, equation (1). REFERENCES
It is worth investigating (iii). Considering
a beam of given dimensions and varying its 1. SEARLE, G. F. C. (1908). Experimental Elas-
radius of principal curvature, its value of ticity, Cambridge, 1908, pp. 40-58.
b2j(Rd) will change, and its modes of dis-
tortion will be those shown in2 Figs. 4, 5 and 2. CASE, J. (1938). Strength of Materials, Arnold,
3rd edition, 1938, pp. 546-554.
6, for the various values of b /(Rd) and the
corresponding values of R. The greatest 3. LAMB, H. (1891). On the Flexure of a Flat
slope occurs at the edges of the beam, and Elastic Spring. Philosophical Magazine, March
increases as b2l(Rd) increases. Hence the 1891.
greatest slope of the centre-line occurs where 4. ASHWELL, D. G. and GREENWOOD, E. D. The
b2j(Rd) is as large as possible, or R is as Pure Bending of Rectangular Plates. Engin-
small as possible. eering, 21st and 28th July 1950.