You are on page 1of 14

Int. J. Mech. Sci. Vol. 25, No. 9 10, pp.

623 636, 1983


Printed in Great Britain.

0020-7403/83 $3.00+.00
PergamonPress Ltd.

BENDING COLLAPSE OF RECTANGULAR


AND SQUARE SECTION TUBES
DU~AN KECMAN'~
Faculty of Mechanical Engineering, University of Belgrade, 27 Marta 80, l l000 Belgrade
Yugoslavia
(Recieved 23 November 1982; in revised from 28 March 1983)

Summary--The bending collapse behaviour of rectangular and square section tubes is studied
theoretically and experimentally. A limit analysis technique was employed and a set of formulae
relating the hinge moment and associated angle of rotation was derived. The theoretical
predictions were verified by comparison with 56 quasi-static bending tests on 27 different sections
having aspect ratios (a/b) varying between 3.0 and 0.33 and with width to wall thickness ratios
(a/t) between 128 and 9.14. Very good agreement was found between the theoretical predictions
and experimental results for the whole range of sections.

a(mm)
ae (mm)
b(mm)
E(N/mm 2)
h(mm)
M(Nmm)
Mm(Nmm)
Mp(Nmm)
rap(N)
r(mm)
t (mm)

w(J)
v(-)
p (rad)
0 (tad, o)
ae,(N/mm 2)
a~(N/mm 2)
op(N/mm 2)
ae~(N/mm 2)

~(-)

NOTATION
section width, i.e. width of the compression flange;
"effective" width of the compression flange;
section depth;
Young's modulus;
half of the hinge length;
hinge moment;
maximum hinge moment (at 0 = 0);
fully plastic moment of a section;
fully plastic moment of a unit width of the section wall;
rolling radius of the wall deformation;
wall thickness;
energy absorbed at a hinge;
Poisson's ratio;
half of the hinge rotation angle;
hinge rotation angle (radians in all of the formulae);
critical buckling stress of the compression flange;
edge stress of the compression flange;
yield stress of the material in uniaxial tension;
maximum flow stress of the material in uniaxial tension--nominal value (ap~ = maximum tensile
force/original area of the specimen cross-section, see Figs. 8-13).
3.141592...
I. INTRODUCTION

Rectangular and square section tubes are widely used in buses, coaches, special purpose
vehicles, roll-over and falling object protective structures and other areas subject to safety
requirements. Under impact such tubes usually buckle locally and hinges may develop and
permit rotations more than 30 before the structure starts intruding into the survival space.
A study of the collapse behaviour is therefore important for both the design and analysis
of weight-efficient safety structures.
Methods already exist [1--4] that allow for varying hinge moments in computer analyses
of the collapse behaviour of complete structures. However, these rely on the experimental
input data, while the current paper enables prediction of the hinge properties. A detailed
discussion can be found in Ref. [5], while only a summary of the results will be given here.
Only uniaxial bending collapse will be considered, since it is the most common in practice
[6-8] and it has been demonstrated [3, 4] that biaxial collapse may be analysed on the basis
of uniaxial hinge properties.
2. MAXIMUM BENDING STRENGTH OF A RECTANGULAR SECTION TUBE
Rhodes and Harvey have examined 19] the maximum bending strength of a lipped channel section with
lips in tension and a full-width flange in compression. The critical stress trc, of the compression flange was
derived in a form:
a,.,

12(l~v~5.23+0.16Z

tConsultant to Cranfield Impact Center, Cranfield, England.


623

(1)

624

D. KECMAN

where E, v, u, h and t are respectively, Young's modulus, Poisson's ratio, flange width, section depth and
wall thickness. The effect of lips on the flange stability is negligible
If a,r is less than the yield stress ap of the material, the flange buckles and an "effective" flange width
(a.) is introduced in the following simplified form:

a e =

(l

07
.

'r+0.3

(2)

tTe

where a,. is the stress at the edge of the flange and a e = a if a,r >~%. A further simplifying assumption, which
gave a good agreement with experiments, was that the section starts collapsing when the edge stress a e equals
the yield stress %. The actual maximum moment (/14,,) depends on the total stress distribution when the
compression flange yields, with formulae given in [9].
Dawson and Walker developed a similar method[10], but assumed that the flange of the lipped channel
has an initial imperfection with an amplitude fitting the lower bound experimental data by Winter[Ill and
Chilver[12]. The calculation of the maximum moment shows that a,,, of a simply supported plate (lower than
equation (1) should be at least 2.5 times higher than % in order to produce full plasticity in the compression
flange.
Closing of the section on the tension side has little effect on the stability of the compression flange, hence
formulae in [9 10] could also be used for uniaxial bending of a rectangular section tube. A comparative study
of the two methods indicated [5] that the maximum moment (M,,) of a rectangular or square section tube
can be calculated using equations (1) and (2) with a,, = ap and:
if a,r <~ap
3~;+2

2a+b+a,.

M m = a 0. t - b '

3(a + b)

(3a)

if a,r >~ 3ap (fully plastic moment Mp)


M,, = Mp = ~7p . t [a(b - t) + 0.5(b - 2t) 21

(3b)

and, finally, if % < a,~ < 3o-p, a moment:


M"e = a p , t . b ,

(3c)

a+-~

is defined first, and the maximum strength becomes, from linear interpolation:
Mm=M' o+(M.-M;)

~7 c r - -

2ap

(7 p

"

(3d)

3. A C T U A L BENDING COLLAPSE M E C H A N I S M
The hinge collapse modes observed in actual accidents [6], static and dynamic laboratory tests on standard
and non-standard sections [5, 13], bus rings [14] and complete structures [7, 8, 15] proved to be very repeatable.
A typical hinge mechanism produced by bending a square section (38 x 38 x 1.6 ram) tube as a 1 m long
cantilever is shown in Fig. l(a). The mechanism is symmetric with respect to the lateral (Fig. lb) and longitudinal
(Fig. lc) middle planes.
A dense pattern of 2.54 mm dia. circles was applied electrochemically to all four walls at the hinge area (Fig.
la). The test was stopped after 9.5, 19 and 29.8 of hinge rotation (0) in order to measure the deformations
recorded by the pattern.
The hinge mechanism occurs between the two plastically undeformed beam segments with clearly defined
stationary wall bending lines G H N K G and E F M L E in Fig 2(a). A detailed analysis of the hinge mechanism and
the deformed pattern revealed the following main points:
(a) Most of the plastic deformation is concentrated along the stationary yield lines EF, GH, EB, GB, F(',
HC, BC, BA, C J, GK, NH, EL and F M (Figs. 2 and 3 and Table 1).
TABLE I. DEFORMATION OF CHARACTERISTIC CIRCLES 1N FIGS. ] (a) AND

0
(deg)
9.5

Circle
Numbert
d]*(div)
d2(div)

1
51
35

50 55
50 50

3
10

11

12

-~50
-48

55
30

55
47

54
48

53
47

2 5 0 50
~ 50 48

52 52
50 49

19

dB(div)
d2(div)

-~50
~30

~50
-50

57
50

-~50
-~40

>75
<20

60
35

58
35

54
45

=50
~50

50
45

53
49

53
47

29.8

dl(div)
d2(div)

-~50
-~20

~-50
-~50

59
50

-~50 Frac= 4 0 ture

>75
<25

>75
<25

54
45

~50
-50

50
45

51
47

52
48

tCircle numbers correspond to those in Fig. 3.


+tUndeformed circle: 50 x 50 divisions (div); d I and d2 are the longer and shorter ellipse
diameters.

Bending collapse of rectangular and square section tubes

(a)

(b)

(c)
FIG. 1. A typical hinge collapse mechanism (a) with cross (b) and longitudinal (c) sections.

625

626

D. KECMAN

FIG. 4. Collapse modes in various rectangular and square sections.

Bending collapse of rectangular and square section tubes

627

(b) Travelling yield lines of rolling deformation were also observed along GA, AE, KA and LA, while the
distance BA changes during hinge rotation (point A in Fig. 2b corresponds to the current tip of the "bulge"--A~,
A 2, A 3 etc.).
(c) The corner angle along KL and M N remains almost constant (Fig. 2b);
(d) The deformation of the tension flange includes travelling yield lines N K and M L (dotted in Fig. 3c),
approximately corresponding to the intersection of the undeformed flange planes and the current cylindrical
surface 1D in Fig. 2(b), (see also Fig. 1);
(e) In-plane deformation of the walls was also observed, particularly along the rolling lines GA, EA, BA, HC,
CF and CJ (the original circles in Fig. l(a) deformed into "droplet" shapes 6 and 7 in Fig. 3 and Table 1). This
deformation is essential to maintain hinge kinematics;
(f) The rolling radius " r " varies slightly along GA measured by cutting at right angles to GA ) and its estimated
value was approximately proportional to the hinge length, i.e. r = (0.03-0.05)' KL. Larger values correspond to
thicker tubes of comparable sections, but the test programme did not indicate a strong effect of either wall
thickness or the initial corner radius (this conclusion is in contrast with [16], where r = 0.5 t). The radius reduces
during collapse, so that rolling becomes more and more "difficult", until the mechanism "jams" and starts a new
phase of collapse;
(g) The rolling radius along KA, LA, M J and NJ varies from infinity at K, L, M and N to a value close to
" r " in (f) above. Deformation pattern along these lines is very diffused.
Four distinct phases in the hinge development were identified:
(1) Initial phase, with a "bulge" in the webs (At in Fig. 2b) starting approx. 0.2 .Kg below the compression
flange, but without apparent rolling deformation (this phase spreads over hinge angles between zero and 5-10 ,
with larger values in relatively thicker walls).
(2) Second phase, with rolling as described above, and the cross-section shape as in Fig. 2(b) (A2). The angle
0 is between 5-10 and 25 to 35.
(3) Third phase, initiated by "jamming" of the rolling deformations, with the cross-section taking the shape
B3A3TD in Fig+ 2(b) and with new bending lines A T and T K a n d their symmetric counterparts.
(4) Fourth phase, initiated by contact between the two buckled halves of the compression flange (Fig. 1). This
totally jams the original hinge and starts an adjacent secondary one with new bending lines formed at 0.5 KL
away from the section GHNK (Fig. 2), the webs buckling "inwards" and the compression flange "out". There
is no clear-cut transition from one phase to another, so one can say that intermediate, hybrid forms are also
present.
The collapse mechanisms in rectangular sections (Fig. 4) displayed the same general pattern, with a hinge
length KL approximately equal to the shorter of the section sides.
4. THEORETICAL BENDING COLLAPSE MECHANISM
The theoretical model of an actual hinge collapse mechanism is based on the second phase of collapse which
includes the hinge angles normally allowed and displayed in vehicle safety structures. The experiments also show
that there is a smooth transition in the hinge moment (M)-rotation (0) curves between the first and second and
second and third phases of collapse.
The theoretical mechanism is based on the assumptions that the walls of a section deform along the
concentrated yield lines only, that the walls are incompressible and inextensible and that structural continuity
is maintained in the two characteristic sections shown in Fig. l(b, c). It was already mentioned that some in-plane
deformation is essential for the hinge kinematics and indeed Wierzbicki and Abramowicz have shown recently
[17] that 1/3 of the energy absorbed by axially collapsing thin-walled columns can be attributed to this source.

//

a)

r"

i c3
\

b)
FIG. 2. Hinge mechanism (a) at various stages of development (b).
MS Wol. 25, No. 9--10--B

628

D. K I:X'MAN

-2.&

',,

A?

%-~'--.f
I /"'

b)

/[
/I

'\d

;i

11o

~/' .~

c)

FIG. 3. Characteristic circles for measuring the deformation pattern.

A similar conclusion may apply to the bending collapse, but the experimental evidence shown later in the paper
proved that the approximation used can also produce a very useable set of formulae.
Wall continuity is considered in two main sections only, since an overall continuity would result in a very
complex analysis.
The first theoretical model in Fig. 2(a) resulted in a set of formulae with two unknown parameters: hinge
length KL = 2h and rolling radius r (a possible simplification was published in [18], but this result is now
improved). Attempts to derive the u n k n o w n parameters from the conditions of m i n i m u m m o m e n t or energy and
wall continuity were not successful, so a new model in Fig. 5 was introduced.
The rolling radius is assumed very small in comparison with other hinge dimensions and points A, B, C and
J are on a straight line.
Introducing an angle p = 0/2, where 0 is the angle of hinge rotation, coordinates of point B in Fig. 5(b) are:
x~ = h: Yo = b cos p - w.'b :sin p (2h ~ h sin p); ZB= 0.

a)

-0-/2 - ~

KL

bl
FIG. 5. The theoretical model (a) and its longitudinal section (b).

(4)

Bending collapse of rectangular and square section tubes

629

Continuity in the middle section implies:


b = z A + x / Y a + ZA, with YA = Yn.

(5)

z A = b - sin z p - h sin p + x/~ sin p . (2h - b sin p ) . cos p.

(6)

From (4) and (5) follows:

Continuity in the longitudinal direction is established for a fibre originally parallel to the beam axis and passing
through the current point A in Fig. 5. This fibre is at a distance A D = b - ZA from the edge KL. Before rotation
the coordinates of the intersection of this fibre and the line G K are (O, b - z A, O) and after rotation
((b - ZA) sin p, (b -- ZA) COSp, O). The distance between this rotated point and point A, with coordinates (h, YA,
zA), must be equal to " h " if the walls are inextensible and if the bending radii are assumed to be very small.
Hence:
h 2 = [h - (b - ZA) sin p]2 + [YA -- (b -- zA) cos p]2 + z~.

(7)

Substituting (4)-(6) into (7) it can be shown that continuity conditions (5) and (7) are met simultaneously for
any ratio a / b and any 0 when:
2h = a , or 2h = b .

(8)

The correct value is the one that yields lowest energy absorbtion, hence it is concluded that:
2h=aifa~<b,

or2h=bifa~>b,

(9)

which is in excellent agreement with experimental evidence (Fig. 4).


These continuity conditions may explain why the static and dynamic collapse mechanisms are so repeatable
and virtually the same. It means that the current analysis may also be applied to the dynamic collapse, but with
allowance for the strain rate effect on the material properties.
Jamming o f the theoretical collapse mechanism starts when the two buckled halves of the compression flange
contact each other. The appropriate angle of jamming is (from Fig. 5(b):
O s = 2 a r c sin - -

(lO)

5. N O M I N A L E N E R G Y ABSORBED D U R I N G H I N G E ROTATION
From the assumptions made, all the energy is absorbed by plastic deformation along the yield lines of the
theoretical collapse mechanism.
Rotation of the hinge through an angle 0 = 2p creates a relative rotation along the flange walls E F and G H
(Fig. 5b):
a=g-~/2-p-fl=g/2-p-arcsin

1-~sinp

(11)

hence the energy becomes:


Wl=g"er+an=2"mp'a

rG2-p-arcsin

1-~sinp

(12)

where m 0 is an average fully plastic moment of a unit width of the wall, a and b are section width and depth
and h is known from (9).
Similarly:

n-2arcsin

W2=WBc=mp'a

l-~sinp

(13)

The length o f the yield line A B = z A is known from (6) and the angle of relative rotation (r~ - 2fl ) can be found
from (11), (angle fl is equal to the last term in (11)). Thus,
W3=WAn+Cs=2me(bsin2p-hsinP+x/bsinp(2h-bsinp).cosp)x

rr - 2 arc sin \ 1 - b sin p ) ) .

(14)

This "corner" deformation is treated somewhat differently by Abramowitz[19]. The angle o f relative rotation
along BGis rr/2 and B G = h, so that
W 4 = WBa+ Be +cn +cr = 4 " me" h 2 = 2 . mp. h n .

(15)

Relative rotation (r/) along G K can be seen in Fig. 6. Plane to is normal to G K and contains point A (Fig. 5a).
Point A ' is the normal projection o f A into the K G D plane (also plane x y ) and A o is obtained by rotating co

63(1

D. KECMAN

Y~/

Ao

FIG. 6. The angle "tl" of rotation along the line GK.

about its intersection m~,. with the x y plane. Coordinates of point A" are:
h tan p + b . c o s p - xfb s i n p ( 2 h Z b s i n p )
. . . .
1 + tan2p
- "

.!'4.

(16)

and:
x A = y , , tan p.

(17)

From Fig. 6 one can now derive the angle q and then:

Ws

W~;s,+Ez}n,~.,F M

h--x.4")'+O"A'

4 mp-barctan

Y~)-

(18)

where y~ is known from (4). The area swept by the yield line GA is equal to the triangle ABG in Fig. 5a and
the rolling radius r is reasonably constant along GA, so that allowing for bending and flattening of the walls
during rolling, we obtain:

42mphzA
h
r 2 =4mprZ'4"

W= W(i4+Ae+CW+CF=

(19)

The rolling radius decreases during hinge rotation and an approximate variation within the observed range:

is suggested to fit the experimental results.


The observed variation of the rolling radius along KA was such that a hyperbolic approximation:
rKA =

KA

(21)

was used, where lk is the distance from K along KA and r is given by (20).
A linear variation of the rolled length between K and A (from zero to ZA) is assumed:
lk
l~ = KA zA

122)

and the energy absorbed along KA becomes:


KA

KA

W~A = o 2me KAdl k = 2rap


0

KA zA KA-r dl~ -

3r

(23)

or:
zA

W7 = WKA+ LA+ NJ~ MJ = ~ rnp r~ x / ~ + y~ + z]

(24)

with all parameters known.


The relative rotation along K N and L M is p = 0/2 and along KL and M H the angle ~ = arc tan (z~/y~) (Fig.
5b) should be used. Hence:
W8 = WKN + LM+ KL + M N =

2mp(ap + 2h arc tan (ZA]]'\YA//

(25)

Although the mechanism in Fig. 5(a) does not resemble well the true deformation of the tension flange (Fig. lb
and c), it was concluded [5] that equation (25) should provide a reasonable approximation.
The nominal value of the energy absorbed at a hinge now becomes:

W(O) = L Wi(O)
I

(26)

Bending collapse of rectangular and square section tubes

631

valid for 0 ~< 0 ~< Os, where 0 s is known from (10). Equation (26) is difficult to differentiate, so that a hinge
moment M ( O ) at any angle (0) is calculated numerically from:
M(O)

w(o + 40) - w(o)

A0

(27)

where A0 represents a small, but finite increment of the hinge angle 0.


6. SELECTION OF THE M A T E R I A L PROPERTIES
The theory assumes that the material is ductile enough not to fracture during hinge collapse, which is true
for most mild steels and some aluminium alloys. However, proper selection of material properties is not simple
because:
(a) Wall bending occurs under plane strain conditions and even in ideally elastic-plastic materials there may
be a 15% difference in the effective yield stress, depending on whether the Tresca or Von Mises yield criterion
is used.
(b) In actual materials the yield stress is not necessarily equal in tension and compression (it is quite common
that the latter is 10-15~o higher).
(c) Reversal of plastic deformation is present in parts of the section subject to rolling or fully plastic
deformation prior to bending, so that a Bauschinger effect may occur. However, the theoretical model in Fig.
5(a) follows the path of plastic deformation along the yield lines and does not obscure strain reversal.
(d) Anisotropy due to seam welding and/or rolling of tubes is always present. Anisotropy normal to the wall
(resistance to thinning) is particularly interesting, since it deforms the VonMises ellipse. Experiments on
specimens taken from some of the tubes indicated [5] that the effective yield stress is 9.7% higher than the value
in uniaxial tension (i.e. it is "between" the Tresca and VonMises criteria).
(e) Plastic deformation along the yield lines is quite excessive, so that strain hardening must be allowed for.
It is also present during tube manufacture, so that the yield stress at corners or seam welds of standard tubes
is higher than in the middle parts of section sides.
(f) The effect of strain rate on material properties must also be allowed for, but fortunately the same
theoretical model can be used.
(g) Presence of the axial load in the section walls has an effect on the fully plastic moment developed at a
particular location.
After a detailed consideration of these effects and allowing for the fact that the theoretical mechanism does
not reflect the first phase o f hinge collapse, it was decided[5] to use the data which are commonly available for
mild steel, i.e.:
ap in equation (3a~l) represents the yield stress in uniaxial tension of a speciment parallel to the beam axis
and cut away from corners and seam welds. For design purposes ap would correspond to the minimum
guaranteed yield stress.
~p~ is the maximum nominal flow stress of the material in uniaxial tension and this value should be used when
predicting the collapse performance of the hinge. Hence:
O-pu . t 2

mp=

(28)

To illustrate the second point, hinge moment (M) - rotation (0) curves of a tube with a = 100 mm, b = 103 mm
and t = 0.81 mm are shown in Fig. 7. The stress-strain curves 1, 2 and 3 correspond to specimens taken from
the appropriate locations on the section. The experimental curve is chain-dotted, curve A is theoretical with
apu = 253 N/mm 2 and B corresponds to ap = 161 N/mm 2.
All curves in Fig. 7 start from M(0) = M m because 0 corresponds to plastic deformation only, i.e. elastic
curvature before and during collapse is eliminated as explained in Section 8.
7. THE FINAL F O R M OF THE THEORETICAL H I N G E M O M E N T A N D
THE E N E R G Y ABSORBED
The basic (nominal) moment (M) - rotation (0) curves are calculated from the previous formulae, but the
model does not cover adequately the first phase of collapse, particularly in relatively thick tubes. However, the
maximum moment Mm at 0 = 0 is known and we also have the nominal M - 0 curve between 0 = 0 and 0 = 0j.
A simple and smooth transition between the two is obtained by drawing a tangent to the curve, passing through
point (Mm, 0)--dotted line in Fig. 7. The contact point T and the appropriate transition angle Or are found
approximately while calculating discrete points on the nominal M - - O curve (if Or > 0 s then Or = 0~).
The difference between the tangent line and the nominal M - 0 curve need not be great (except for small
0), even when Or shifts beyond 10 to 20 degrees.
For hinge angles beyond 0j the following empirical formula is suggested:
M(O) = M(0j) + 1.4[M,, - M(Oj) ] (0 - Oj); for 0 > 0j.

(29)

In this manner, the complete rotation range is divided in segments 0 < 0 ~< Or, Or < 0 <~ Oj and 0 > 0j. The final
expression for the energy absorbed are:

for 0 < 0 ~<Or: W ( O ) = O . 5 [ M , , ( 2 - f f ~ r ) + f f r M ( O r ) ] . O


for Or < 0 <~ Os: W(O) = 0.5[M m+ M(Or) ] Or + W(O) -- W(0r)

(30)
(31)

632

D. KECMAN

M(Nm)I
Mm'e t
Mm,t

1680
1400

%(N/mm2)~

1200
1000
8O0
600
400
2O0

. . . . . . . . .

.....

eT

~b

2'0

io

~b

~p = 161
o(oT

FIG. 7. Theoretical (solid) and experimental (chain-dotted) curves of section 18 (Table 2).
and, finally, if Og< 0 ~<0end, where 0no is the angle for which the experimental verification exists (Table 2):
W(O) = 0.5 [m m+ m ( 0 r ) ] Or + W(Oj) - W(Or)
+ 0.512M(0j) + 1.4(M,, - M(Oj))" (0 - Og)]" (0 - Os).

(32)

All quantities M(O) and W(O) in (30)-(32) are calculated using the basic theory.
8. E X P E R I M E N T A L VERIFICATION OF THE THEORY
The theoretical predictions were checked against experimental results from 56 quasi-static cantilever bending
tests on beams with section dimensions and material properties shown in Table 2. In 13 cases, 3 tubes per section
were tested and the general repeatability of results remained within ___5~o of the average values, with somewhat
larger scatter in section 12. The wall thickness, measured at the middle of all sides varied within + 3'?/o of the
nominal value, with some exceptions of + 5~o. Sections with a/t ratios greater than 50 were hand-made, while
all others were purchased off-the-shelf, unmodified standard seam welded mild steel tubes. All cantilevers were
1 to 1.2m long.
Comparative tests showed [5] that pure bending produced virtually identical results as those obtained with
I-1.2 m long cantilevers. The shear effect on collapse properties is therefore considered negligible, unless, of
course, the beams are so short (in relative terms) that a different collapse mode is produced. The test procedure
was therefore simpler than the one in Ref. [20], but provision of a rigid fixture at the root of the cantilever was
not simple. Some beams were therefore cast into concrete blocks to enable a firm but non-crushable support for
the section.
Load increments were applied quasi-statically up to the onset of collapse and further deformation was
produced in increments of displacements. When calculating the hinge moment-rotation curves allowance was
made for the geometry of the beam deformation and load application. The onset of collapse was assumed at
the point of maximum moment, so that a relatively small (0-3 ) elastic-plastic curvature was eliminated from
the total rotation (in the computer analysis of a complete structure the elastic deformations are covered by the
elastic beam stiffness matrices). Stresses due to shear and small axial forces were negligible.
Some of the comparative theoretical (solid) and experimental curves are shown in Figs. 8-13, illustrating the
range of sections to which the formulae are applicable. Bending took place about axes shown in the small section
drawings and the quasi-static nominal stress-engineering strain curves of the tube material are also shown.
The energy absorbed, proportional to the area under the M - - O diagram is a particularly useful parameter
indicating the agreement between the theory and experiments. Results of all sections with theoretical (Wth) and
experimental (Wexp) energies are therefore summarised in Table 2, showing good agreement over the complete
range of section dimensions.
The accuracy of the theoretical prediction can be regarded as very satisfactory, particularly in the light of
production tolerances of materials, tubes and complete structures, degree to which the structures are idealised
by finite-element models and the reserve of strength and energy absorbing capacity that is usually required.
9. PRACTICAL APPLICATION
Almost all rectangular and square section tubes used in structural design have a/b and a/t ratios within the
range which was experimentally verified. The method presented has therefore been used extensively for both
design and analysis purposes, mainly in combination with the CRASH-D program for collapse analysis of
complete structures. Good prediction was achieved in both static and dynamic tests on bus rings [14] and a full
scale test on a complete bus body [15]. A program for section optimization from the weight, cost and material
points of view has also been developed [21].

38

54

45.2

38.1

50.8

50.8

45

25.4

45

45

25,4

25.4

14

15

16

17

50.8

50.8

50.8

38.4

38. I

38.1

29.3

25.4

25.4

19

i%

2O

21

22

23

24

~5

26

27

Op

Opu

297

283

418
161
160
380
253
308
463
297
283
290
277

0.813

0.86

1.6

1.26

1.67

3.25

1.25

1,4&

1.26

1.3

329

324

340

374

313

161

161

380

253

352

463

267

308

326

297

338

463

351

284

430

245

250

418

353

360

416

390

348

250

Numbers

0.33

0.5

0.5

0.5

0.66

0.75

0.75

0.756

0.96

0.97

1.0

1.0

1.0

1.0

1.0

1.0

1.04

1.2

1.33

430
250

1.33

1.5

1.54

1.9

2.0

2.0

2.0

3.0

284

363

463

300

326

338

290

308

277

[N/~ 2]

297

[N/~n 2 ]

2.0

1.48

2.15

1.94

1.6

1.6

0.813

0.813

1.6

1.26

1.62

3.25

1.60

1.26

1.44

1.25

1.30

[ram]

NumOers are rourlded to lhe nearest i n t e g e r ,

45

38

50.8

50,8

45.2

4~

45

19

103

I00

18

25.4

45

38.4

38.4

13

44.4

44.4

99

87

38.4

38. I

25.4

29.3

28.5

19

25.4

25.4

12

103

50.8

II

50.8

104

45

I0

r,=]

{,,=,q

Number

15

Section

'i

50
264
126
226
498
130
144
75

52

260
120
219
508
114
136

81

160

138

234

185

980

368

186

423

75

147

218

130

510

482

288

298

127

135

318

145

178

675

183

75

133

105

61

[J ]

IVexp

20

142

130

235

208

953

372

194

428

76

173

216

121

512

487

291

313

166

146

340

162

174

694

180

76

135

119

65

[J]

Wth

297

307

20o

261

248

222

1364

1392

168

476

183

242

94

97

546

210

188

230

305

315

837

163

177

638

374

398

202

178

433

202

229

995

231

99

172

151

85

[J ]

Wth

674

30

677

616

378

379

163

166

398

187

236

938

240

99

175

133

81

[J ]

Wexp

indicato e~timated values.

8,~

121

I00

9~

114

67

296

282

178

196

116

I02

213

I07

i01

362

II0

45

85

76

38

[O ]

IVth

117

73

307

286

170

186

83

93

198

93

i01

368

108

45

8O

65

36

[J]

Nexp

i0

h~ parentheses

11.5

15.1

17.6

20.3

9.14

22.8

30.2

24

52.3

123

12.7

17.5

20.9

23.2

24

27,8

127

128

31.8

40.3

23.5

13.9

33.8

30.2

35.3

40.6

34.6

O{deg]

TABLE 2. THE EXPERIMENTAL AND THEORETICAL ENERGIES ABSORBED IN HINGES*

(3O0)

271

373
213

(225)

328

1749

562

282

643

109

241

381

198

8O9

765

(397)

(325)

(293)

611

(I19)

223

214

(451)
n6
(8t7)

443

205

(47t)

511

201

236

457

275

214

1263

284

(1170)**

119
274

288

203

178

102

I J!

Wth

118

208

159

98

[j]

Wexp

40

E
Eu~

(%

634

D. KECMAN

--

M(Nm

G(N/h'~)

277

~-.~.'~.~__

~o

20

(o/.)

,oo " " Z " ' 2 ~ :


t0
20
30
FIG. 8. Theoretical (solid) and experimental curves of section l.

e(o!

50,8

M(Nm)
8004

'~ ~"/ 1~

--L

600
\\\

lo

~o

30

40 e{'/.)

400

200

OL~

lo

20

3o

40

FIG. 9. Theoretical (solid) and experimental curves of section 8.


1o4

M(Nm)

Go(N/ram2)~
250.

1000

(~

800
30

600

400

200

I0

20

30

40

FIG. 10. Theoretical (solid) and experimental curves of section 10.

O( )

e( )

635

Bending collapse of rectangular and square section tubes


M(Nm} I

:1
o

"~

~o

~o

o{ )

Fir. 11. Theoretical (solid) and experimental curves of section 13.

H(Nm)
.-.~,~
3000-

~
--" --" --'"...a__ __......~..~

45.2
-171"". . . .

- H - 3.25

2500

2000-

riO0

1000

500

FIG. 12. Theoretical (solid) and experimental curves of section 23.

10. C O N C L U S I O N S

The method presented enables a relatively simple, fast and economic prediction of the
hinge moment-rotation curves and energy absorbed during uniaxial bending collapse of
rectangular and square section tubes. The input data are section dimensions and the yield
and maximum flow stresses of the material. Hinge rotation angles can exceed 30-50 .
Experimental verification covered sections with a/b ratios from 0.33 to 3.0 and a/t ratios
from 9.14-128. The theory is applicable to dynamic collapse, but with the allowance for
the strain rate effect on material properties. The method can and was applied to both
design and analysis of weight-efficient safety structures.

636

D. KECMAN

M<.,I
~1

,o04
f

~oI~i'~

LL

JJ

t .

. . . . . . .

I0

20

30

e ('I.)

100

0
lO

2o

30

40 O( )

FIG. 13. Theoretical (solid) and experimental curves of section 26.

Acknowledgements--This study was completed in 1979 while the author worked with the Structural Design
Group (now Cranfield Impact Centre), Cranfield, England. Aid from my supervisor Mr G. H. Tidbury and other
colleagues is gratefully acknowledged.
REFERENCES
1. J. C. M1L~, The determination of collapse load and energy absorbing properties of thin walled beam
structures using matrix methods of analysis. Int. J. Mech. Sci. 18 (1976).
2. D. K~CMAN, CRASH-D user manual, structural design group, Cranfield (1980).
3. I. K. McIvoR., A. S. WINEMANand H. C. WANG, Plastic collapse of general frames. Int. J. Solids Structures
13 (1977).
4. R.J. HAYDUK, R. G. THOMSON,G. WITTLIN and M. P. KAMAT,Nonlinear structural crash dynamics analyses.
SAE paper 790588 (1979).
5. D. KECMAN, Bending collapse of rectangular section tubes in relation to the bus roll over problem. PhD
Thesis, Cranfield Institute of Technology (1979).
6. STRUCTURALDESIGN GROUP, Investigation of the PSV roll over accidents 1976-1979. Rep. to Department
of Transport, London, (1979).
7. J. C. MILESand G. A. WAR.DILL,Analysis and design of a fire engine safety cab using finite element methods.
L Mech. E. paper 18 (1976).
8. R. N. HARDY and D. KECMAN, Analysis, redesign and testing of a tractor safety cab. JUMV Masinstvo, 3/80,
Belgrade (1980), (in Serbocroatian).
9. J. RHODES and J. M. HARVEY, Design of thin wailed beams. Conf. Exp. Stress Ana(~,sis in Design, Cambridge
(1970).
10. R. J. DAWSON and A. C. WALKER., A proposed method for the design of thin walled beams which buckle
locally. The Struct. Engineer 50(2), 1972.
11. G. WINTER, Strength of thin steel compression flanges. Trans ASCE 112, AISI (1970).
12. A. H. CHILVER., The stability and strength of thin walled steel struts. The Engineer (1953).
13. STRUCTURALDESiGn GROUP, Bending collapse of car components. Rep. to the Ford Motor Company (1980).
14. STRUCTURALDESIGN GROUP, Investigation of the collapse behaviour of bus rings. Rep. to the Department
of Transport, London (1977).
15. D. KECMANand G. H. TIDBUR.Y,Theoretical prediction of the complete collapse behaviour of a coach subject
to the proposed standard roll over test. XIX Int. Congr. FISITA, Melbourne (1982).
16. N. SEKI and Y. SUNAM1,Energy absorbtion of thin walled box beam subject to combination of bending with
axial compression. XVII Int Congr. FISITA, Budapest 0978).
17. T. WIER.ZBICKIand W. ABRAMOWICZ,On the crushing mechanics of thin walled structures. MIT Rep. 82-4.
Cambridge, Mass. (1982).
18. G. H. TIDaURY and D. KECMAN, Investigation into the behaviour of hinges produced by bending collapse
of vehicle structural components. XVII Int. Congress FISITA, Budapest (1978).
19. W. ABR.AMOWICZ, Mechanics of the crushing process of thin shell structures. Int. Conf. of Bus Experts,
Budapest (1980).
20. I. K. MclvOR, W. J. ANDERSON and BIJAK-ZOCHOWSKI, An experimental study of the large deformation of
plastic hinges. Int. J. Solids Structures 13 (1977).
21. D. KEC~:AN, Program "WEST" for optimization of rectangular and square section tubes from the safety
point of view (SAE Paper 811312). Int. Conf. Vehicle Struct. Mech., Detroit (1981).

You might also like