7142 The simple plastic bending of beams

Thepurebending of anunsymmetricalcross-sectionisconsidered. As noted by

Brown,13 the principal‘zero stress axes’for plastic bending do not in general coincide
withtheelasticprincipalaxes,nor are theynecessarilyorthogonal.Thepresent
Paper discusses the bending of an unequal angle section, and some numerical results
are given; a simple test is reported which confirms the theoretical findings.

Introduction
The theory of bending of unsymmetrical cross-sections appearsalways to
have given some difficulty. Thesolutionforthe symmetrical cross-section
was slow in being produced, and even Navier,l who was the first to give a
modern statement of the elastic problem, was not aware of the notion of
principal axes of bending. Thus he gave wrong expressions for the bending
of a rectangular cross-section about an axis not parallel to one of its sides, and
it fell to Saint-Venanta to discuss fully thequestion of principalsecond
moments of area. An account of the development of this subject is given by
Timoshenko3; see also S t r a ~ band,
, ~ for more detailed information, Todhunter
and Pear~on.~
2. Saint-Venant extended his analysis to cover non-linear behaviour of the
material but confinedhis work in this connexion to symmetrical cross-sections.
The elastic/perfectly plastic material is a special case of Saint-Venant’s more
general material,and the plastic bending problemwas considered separatelyby
Ewing.6Ewing again discussed only the rectangularsectionbent about a
principal axis, and,indeed, none of themodernstandardtextsonplastic
theory (Prager and H ~ d g eBaker
, ~ et Neal,’ Phillips,lo Beedle,l’ Massonnet
and Savela) does more than at most mention the unsymmetrical problem.
3. Brown13 seems to be the first to have recorded the general features of
plastic unsymmetrical bending,and heestablishes the concept of the ‘centroidal
locus’. He gives no specific solutions,butnotes that the principal axesof
elastic and plastic bendingneed not coincide (this is a property alsoof sections
havingonly one axis of symmetry), and that the principal axes of plastic
bending are not necessaray orthogonal.
Simple bending theory
4. The usual assumptions of simple bending theory will be made. Applied
to the elastic problem of pure bending of a general cross-section, Fig. 1, these
assumptions lead to the notion of a neutral axis ofbending,whichpasses
through the centre of gravity of the cross-section. If this neutral axis is G x
in Fig. 1, then the elastic stressat any point in thecross-section is proportional
to the ordinatey .
5 . If now the neutral axisis to be parallel to the axis of the applied bending
moment, then the bending moment about the axis G y perpendicular to G x
Written discussion closes 31 January, 1969, for publication after April, 1969.
* Lecturer, Engineering Department, University of Cambridge.
751
Downloaded by [ University of Sussex] on [14/09/16]. Copyright © ICE Publishing, all rights reserved.
 HEYMAN

Zero stress

Fig. 1 Fig. 2

must be zero. This condition'leads immediately to the equation, valid for

elastic bending.
J xydA = 0 . . . . . . * (1)
where d A is an element of cross-sectional area, and the integration is carried
out over the whole cross-section. Equation 1 characterizes the principal axes
of elastic bending; there are, in general, two solutions to the equation, and,
since it is symmetrical in X and y , the directions of the two principal axes will
be orthogonal.
6. Elastic bending can be discussed only in terms of the principal axes;
a bending moment applied about any other axis must first be resolved into
components about the principal axes, and the results superimposed.
7. In discussing the fully plastic bendingof a cross-section it will be assumed
thatthe materialiselastic/perfectly plastic, andthat elasticstrains are
infinitesimal. The yield stresses in tension and compression will be taken to
be the same (uo). Thus, in the state of fully plastic bending, the material of
the entirecross-section is at theyield limit (ro, either in tension or compression.
8. The condition that there is no net thrust across the cross-section leads
at once to theresult that the areayielding in tension must equalthe area yield-
ing in compression; in Fig. 2, the zero stress axis (so-called to distinguish it
from the elastic neutral axis) is an equal area axis. The centre of tension T
is the centre of area of AT, and similarly C is the centre of Ac, but the zero
stress axis will not, in general, pass through the centre of area of the whole
cross-section.
9. A principal zero stress axis must be parallel to the axis of the applied
bending moment, exactly as for the elastic problem. If the x-axis coincides
with the zero stress axis in Fig. 2, then the bending moment about the y-axis
must be zero. Formally, this condition leads to the equation
J*T X& = J*C xdA . . . . . . . (2)
where the integrations 'are carried out over the two separate areas, in tension
and compression.
10. Thus the line CT in Fig. 2 must be perpendicular to a principal zero
stress axis, and there are, in general, two possible solutions of equation 2.
That is, there will be a 'strong' and 'weak' principal axis of plastic bending,
corresponding to themajor and minorprincipal axesof elastic bending.
Since equation 2 is not symmetrical in X and y , the two principal zero stress
axes are not necessarily orthogonal.
752
Downloaded by [ University of Sussex] on [14/09/16]. Copyright © ICE Publishing, all rights reserved.
 T H E S I M P L E P L A S T I C B E N D I N G OF B E A M S

Fig. 3 Fig. 4

11. Brown13 pointed outthatthepoints C and T traceout a skew-

symmetric ‘centroidal locus’ as random equal area axes are chosen, and he
gave a graphical construction for locating the principal zero stress axes from
this centroidal locus.

The unequal angle

12. As an illustration of these ideas, suppose that an unequal angle section
is used as a cantilever beam withone leg vertical and the otherhorizontal, and
is loaded at its tip, Fig. 3. The elastic principal axes are sketched in Fig. 3,
and elastic deflexions must be computed by resolving the load into these two
directions. On superimposing the resulting deflexions, it is found that the tip
of the cantilever moves both horizontally and vertically under the action of a
purely vertical load.
13. Although the calculation has not yet been made, it will be appreciated
that the principal zero stress axes for the plastic bending of the unequal angle
are not parallel to the legs. Thus the same cantilever beam of Fig. 3, loaded
as before by a vertical load, will be bent about an axis which is not parallel to
a principal zero stress axis. At collapse of the beam, therefore, the tip will
again move horizontally as well as vertically although the direction of motion
at collapse will be different from the direction under the action of an elastic
load.
14. An idealizedsection will be used for the numericalcomputation of
principal zero stress axes. In Fig. 4 the legs have lengths 2a and 26 and the
same thickness r ; the dimension t will be considered to be small compared with
a and b (this will introduce small errorsif centre-line dimensions of real angles
are substituted into the formulae derived below).
15. The area of the whole cross-section is 2(a+ b)r; thus the zero stress
axis will divide the cross-section into two equal parts, each of area (a b)t. +
In Fig. 4 the ‘strong’ principal axisis shown, making an angle al(whose value
is to be determined) with the direction of Ox. Geometrically, the centre of
tension T and the centre of compression C, Fig. 5, could be found; the line
C T , which is perpendicular to the zero stress axis, would then give the value
of al. In fact, it is simpler to compute directly the bending moments M , and
My about theX and y axes, and tocombine theseto give the required condition.
753
Downloaded by [ University of Sussex] on [14/09/16]. Copyright © ICE Publishing, all rights reserved.
 HEYMAN

Fig. 5

16. The following expressions will be found for the values of the bending
moments:
M , = (a a + 2ab - ba)too
M y = -2b2t00
. . . . . * (3)
The condition that the total bending moment acts about an axis parallel to
the zero stress axis is that
M, sin ctl + M , cos al=O . . . . . . (4)
that is,
2b2
tan a1 = a2+2ab-b2
* . ' ' * * (5)
17. Table 1 gives somenumericalresults:

Table 1

! a, in. 1 b, in. 1 tan al ! al 1 Elastic

18. The elastic values are book' values taken from the section tables; the
inclinations of the principal axes computed for the idealized section of Fig. 4
are very slightly different from these book values. It will be seen that there
is only a small angular difference between the elastic principal axis and the
corresponding strong principal zero stress axis.
19. The full plastic moment may be calculated from the expression
MP = M , cos a1- M y sinal = M , sec al . . . (6)
20. The 'weak' principalzerostress axis cuts both flanges. In Fig. 6 its
location is specified by the parameter z, whose value is to be determined; the
754
Downloaded by [ University of Sussex] on [14/09/16]. Copyright © ICE Publishing, all rights reserved.
 T H ES I M P L EP L A S T I CB E N D I N G O F BEAMS
axis has been drawn as an equal area axis, cutting the section into two equal
halves of area ( u + b ) t . The inclination of the axis is given by
.
21. A second expression for tan may be found by writing the condition
that the axis of the applied bending momentis parallel to the zero stressaxis.
The bending moments about the X and y axes have values
+ +
M+ = {(U' - 2ab - ba) 2 ( ~ b)z - z'}t~o
My = (2ba-za)t,o . . } (8)
and the condition that there is no bending momentabout anaxis perpendicular
to the zero stress axis is that
M y sina, = M, COS . . . . . * (9)
Thus
tan al =
( 2 - 2ab - P ) 2 ( ++
~ b)Z - z2
. . . . (10)
2ba - za
22. Thesimultaneoussolution of equations 7 and 10 results in a cubic
equation in z havingonerealandtwoimaginaryroots.Fromthereal
solution the numerical results of Table 2 may be obtained:

Table 2

Equal angle (a = b)
6 x 3 X ~ 14.4
X lb.
4~3xfxll.OIb. .
: 1 I 1 1
0%8
0.799
10.051
0326
45"
14.6"
28.4"

23. Thedirection of theprincipal weak zerostress axisis somewhat

different from that of the corresponding elastic principal axis for both the
unsymmetricalsections.Further, the strong and weak axes are not ortho-
gonal,butintersect atabout 79" forbothsections.(Theresults forthe
6 X 3 X 3 angle will be in error due to the very small inclinationof the weak zero
stress axis to the leg; the dimension t has been taken as small, so that the
various portions of the angle in Fig. 6 could be approximated by rectangles.)
24. The full plastic moment about the weak axis is given by
MP = M y sec az , . . . . . (11)
where M y is given by the second of equations 8.
25. The general case of plastic bending about any axis (not a principal zero
stress axis) can nowbe discussed with referenceto the yield locus for the cross-
section. Taking axes M,, M y , Fig. 7, equations 8 may be plotted to give the
curve AB; the same equations, with signs reversed (bending in the opposite
sense) give the curve A'B'. At the point B the value of z (Fig. 6 ) is zero, and
for the portions A'B and A B ' of the yield locus in Fig. 7 the zero stress axis
lies wholly within the longer leg of the angle section. Equations 3 plot as the
point A in Fig. 7. The general shape of Fig. 7 will be preserved for any angle
section ; the particular proportionsof the figure correspond toa ratioa / b = 413.
756
Downloaded by [ University of Sussex] on [14/09/16]. Copyright © ICE Publishing, all rights reserved.
 HEYMAN

JfY

.M*

Fig. 7 Fig. 8

26. The generalpoint P in Fig. 7 representscombinations of bending

moments M , and M ywhich will just producefull plasticity of the cross-section.
In accordance with the normalityrule of plasticity, thecorresponding in-
clination of the axis of bending (a2in Fig. 6), is given by the outward normal
to the yield locus as shown. The position of the weak principal axis is marked
in Fig. 7, for which the axesof the applied bending momentand of the resulting
deformation coincide. The points A and A' are cusps of the yield locus, and
the usual indeterminacy of the direction of deformation at these points results;
this corresponds to the fact that the expressions forM , and M y , equations 3,
are independent of the angle ctl in Fig. 4. Indeed, a1 can vary between 90"
and - tan-l (a - 6)/2bat the two cusps, and the direction of deformation for
bending about the strong principal axis will depend strongly on very slight
variations in the ratioM,IM,.

Experimental result
27. In Table 2, the difference in inclination of the plastic zero stress axis
and the corresponding elastic principal axis, 18.1" as compared with 28.4" for
the 4 X 3 unequal angle, is large enough to be easily observed experimentally.
28. A length of nominal I + in. X I + in. X in. mild steel angle was machined
to the average dimensions of Fig. 8 ; it will be seen that this is a scale model
of the 4 in. X 3 in. X f in. unequal angle. The specimen was arranged as a
cantilever by profile welding one end to a base plate which could be clamped
to a virtually rigid support. Two small holes were drilled in the legsof the
angle, through which a wire loop was passed supporting a loading hanger,
Fig. 9: the elrective length of the cantilever was 10.0 in. from the root to the
loading point.
29. The base plate could be rotated in a vertical plane before clamping, so
that the inclination a to the horizontal of the leg in Fig. 9 could be set to any
desired value. In practice, the angle 6 was measured by means of an accurate
spirit level to well within 0.1"; the computed valueof a was probably accurate
to f0.5".
30. Deflexions at the tip of the cantileverwere measured by threedial
gauges actuated remotely by fine wires attached to the specimen. Readings on
the gauges were estimated to 0.0001 in., and were reproducible to +04005 in.
756
Downloaded by [ University of Sussex] on [14/09/16]. Copyright © ICE Publishing, all rights reserved.
 T H E S I M P L E P L A S T I CB E N D I N G O F BEAMS

INCLINATION OF ANGLE, d
Fig. 9 Fig. 10

31. Preliminary elastic tests were made for different values of the inclina-
tion a ;six typical results are shown in Fig. 10. The observed inclination of the
elastic principal axis would appear tobe about 28.9",which agrees well enough
with the book value of 28.4",Table 2. Had the cantilever been tested exactly
at this angle, no elastic lateral deflexion would have been observed.
32. The specimen was then clamped at an inclination a = 26.4",and loaded
in steps of 20 Ib to 60 lb, and thereafter in 2 Ib increments. The load/lateral
deflexion curve of Fig. 11 was obtained. It willbe seen that the direction of
lateral deflexion reversed, and that thedeflexions eventually became very large.
The test was stopped when the vertical deflexion of the cantilever tip exceeded
1 in.

Lwd W.Ib f

- 50 - 40 - 30 - 20 - 10 0
LATERAL DEFLEXION S. IN. X 10.)

Fig. 1 1
767
Downloaded by [ University of Sussex] on [14/09/16]. Copyright © ICE Publishing, all rights reserved.
 HEYMAN

. ' VERTICAL
DEFLEXION
0 IN. X IO-'

2w c 0
0

L
- 60 - 50 -40 - 30 - 20 - 10 0 10
LATERAL DEFLEXION 6 ,IN. X I0 -3

Fig. 12

33. The reversal of direction of lateral deflexion is, of course, a consequence

of the fact that the test inclination of 26.4" lay between that corresponding to
the elastic principal axis and the predicted inclinationof the plastic principal
zero stress axis. After an initial elastic deflexion in one direction, the onset
of yielding compelled the tip of the cantilever to start moving in the opposite
direction.
34. Theratio M J M , inthe test was tan 26.4", or 0.496. Using centre
line dimensions (2a=0.935 in., 26=0.685 in.), equations 8 may be solved for
this case to give z=O.222 in., and equation 7 then gives tan a2 =0.378, or
a, = 20.7". Thus the axis of bending is predicted to lie at an angle with the
horizontal of 26.4 - 20.7 = 5.7". The experimental data are replottedin Fig. 12
using axes of vertical and lateral deflexion;the line of slope tan-' 5.7" = 0.100
is shown, and itwill be seen that theexperimental points lieclosely on this line
over a considerable range.

References
C. L.M. H.
1. NAVIER Rdsumt des lefons donndes a I'Ecole des Ponts et Chausse'es.
2nd Edition, Paris, 1833.
2. NAVIER C . L.M. H. Rdsume'des lefons donndes a I'Ecole des Ponts et Chaussdes.
3rd Edition, Paris, 1864, with notes and appendices by Barre de Saint-Venant.
3. TIMOSHBNKO S. P. History of strength of materials. McGraw-Hill, 1953.
4. STRAUB H. A history of civil engineering. London, 1960.
768
Downloaded by [ University of Sussex] on [14/09/16]. Copyright © ICE Publishing, all rights reserved.
 T H ES I M P L EP L A S T I CB E N D I N G O F BEAMS
I. and PEARSON
5. TODHUNTER K. A history of thetheory of elasticity. 3 Vols.,
Cambridge, 1886-1893.
6. EWINGJ. A. Thestrength of materials. Cambridge,1899.
7. PRAGER W. andHODGEP. G . Theory ofperfectlyplastic solids. London, 1951.
8. BAKERJ. F.,HORNEM. R. and HEYMAN J. Thesteel skeleton, Vol. 2: plastic
behaviouranddesign. Cambridge,1956.
9. NEAL B. G . Theplasticmethods ofstructural analysis. London, 1956.
10. PHILLIPS A. Introduction toplasticity. New York, 1956.
11. BEEDLE L. S. Plasticdesign of steel frames. New York, 1958.
12. MASSONNET, CH. and SAVEM. Calculplastique des constructions. Bruxelles,
1961.
13. BROWNE. H. Plasticasymmetricalbending of beams. Znt. J. mech. Sci. 9,
77-82.

759
Downloaded by [ University of Sussex] on [14/09/16]. Copyright © ICE Publishing, all rights reserved.