On the Applicability of the Southwell Plot

to Plastic Buckling

by Josef Singer

ABSTRACT--A completely forgotten paper by C.T. Wang on man's 1910 short columns,' to Gerard's 1947 Republic
the extension of the Southwe~l plot to inelastic buckling of Aviation Corporation short columns (detailed in Ref.
columns is revived, rederived and amplified. A theoretical 11), and then applied the Southwell plot, as modified by
justification is presented for the application of the Southwell Lundquist ~ to Horsfall and Sandorff's 1946 short
method to plastic buckling of columns made of a strain-
column, '3 and showed that the test data lay closely to
hardening material, showing that it predicts the double (or
reduced) modulus buckling load, Typical experimental verifica- straight lines a n d that the Southwell plot predicted
tions are recapitulated. This puts practical applications of the buckling loads close to the double-modulus theory
Southwell method in the plastic region on firmer ground. buckling load Pr. It is astonishing that this important
1948 paper has been completely forgotten, and is not
mentioned in any of the reviews or scores of papers on
Introduction applications of the Southwell plot published in the 40
The Southwell plot, originally proposed in 1932 by years since.* Even Massey's 1963 paper, ' ' in which a
SouthwelP for determination of the theoretical buckling modification of the Southwell plot for lateral instability
load P~ of a perfect column from experiments on real of I-beams is developed and shown to be approximately
imperfect columns, has been widely used and extended to applicable also in the inelastic region, does not mentio0
other structures (see for example Refs. 2-5). The South- Wang's work. Nor is it mentioned in Newman's 1973
well method, and its extensions, are based on small paper 9 which successfully employs the Southwell plot to
deflection theory and assume elastic material behavior. predict plastic buckling of pressurized tubular columns,
Hence its application to plastic buckling appears to be though without justification; or in Sobel's 1983 paper '~
ruled out a priori. SouthweU' already observed that when which applies the Southwell method to plastic buckling
he applied the method to the classical strut tests of von loads of elbows.
K~rm~n,' the results were excellent for slender struts, but It seems therefore appropriate to remind the solid-
failed, as he expected, for the 'medium' and 'thick' struts mechanics community of Wang's important theoretical
since practically every observation there related to deflec- justification of the use of Southwell's method in the
tions for which the material had ceased to be elastic. plastic region, which puts the later (and future) papers on
Tuckerman, 7 when reviewing extensions of the Southwell firmer ground. A slightly different derivation (obtained
plot, stated clearly that independentlY by the author, essentially similar to Wang's,
but the arguments have been carried a step further) will
None of these methods can be expected to give be presented.
straight-line graphs for high loads at which
plastic yielding has become appreciable, and for
which the theory is no longer applicable.
Analysis
If one reflects and reconsiders the derivation of the
However, the beauty and simplicity of the Southwell Southwell method, one realizes that a certain class of
method, and in particular its usefulness in 'smoothing' plastic-buckling problems may also theoretically qualify.
experimental data in parametric studies, by removing In other words, if buckling occurs well into the plastic
most of the imperfection effects (geometric imperfections, region and the measurements are practically all taken
load eccentricities and minor variations in stress distribu- when the dominant part of the structure is plastic, and in
tion) related to the specific specimens, as pointed out addition the stress-strain behavior is essentially linear
already in 1934 by Bridget et al., ~ has tempted the strain hardening, the conditions resemble elastic buckling
investigators to apply the method even well into the and the Southwell plot should yield good and consistent
plastic range. And surprisingly sometimes it worked results. It will, however, not predict the classical elastic
extremely well (see for example Refs. 9 and 10). buckling load of the perfect structure, but its plastic
In 1948, Wang" showed that the Southwell method can buckling load, based on the linear strain-hardening
be extended into the inelastic region, provided that Er, material properties.
the Engesser-von K,qma~ double modulus, is approximately A simply supported column is reconsidered and its
constant. Wang applied the SouthwO1 plot to von Kar- buckling in the plastic regime is analyzed. The shape of
the stress-strain curve of the material is assumed to be
such that it can be approximately replaced by two straight
Josef Singer (SEM Member), Sherman Fairchild Distinguished Scholar,
California Institute of Technology, 1987-88, is Shirley L. Tark Professor lines (see Fig. 1)--one elastic (up to the yield stress ey)
of Aircraft Structures, Department of Aeronautical Engineering, Tech-
nion, Israel Institute of Technology, Haifa, Israel 32000. *The author only discovered it by chance after having finished a similar
Original manuscript submitted: January 19, 1988. Final manuscript derivation independently and having applied the method successfully, and
received: October 25, 1988. with great joy, to all von Kdrmdn's short columns.

E x p e r i m e n t a l M e c h a n i c s * 205
 with slope E (Young's modulus), and one plastic (from wo = Wo, (9)
ay onwards) with slope E, (the tangent modulus).
For a perfect column, the equation of equilibrium when and eq (8) can be rearranged as
a > ~ry is now, according to the Engesser-K~rm~n double
(or reduced) modulus theory (see for example Ref. 15, b =/'. (~) - Wo (10)
pp. 12-13),
The inverse slol~e of the plot of -~- versus iS, the South-
d2Y (EI~ + Etlz) = - P y (1)
dx 2 well plot, yields therefore the buckling load Pr in the
inelastic range of the corresponding perfect column,
If one writes according to the Engesser-K~mhn reduced modulus
theory. Note that P , is the bifurcation critical load in the
E~I = E l , + EtI~ (2) inelastic range, and that as in the elastic range the South-
well plot yields the Euler load, which is the elastic bi-
or furcation load, it yields P~ in the plastic range. One
EL E,/2 should remember that the reduced modulus theory assumes
E r - - + - (2A) strain reversal. If bifurcation occurs at A in Fig. 1 and the
I I
column begins to bend, on the concave side the compres-
equation (I) becomes sive stresses increase due to bending at a rate proportional
to Et (along A B in Fig. 1), whereas on the convex side
Ej y~ + Py = 0 (3) there will be a reduction in the longitudinal stresses due
to the bending stresses, which only relieves the elastic
For an imperfect column, with initial geometric im- portion of the strain and hence is proportional to E
perfections yo(X), and if y ( x ) is the additional deflection (along A C in Fig. 1).
(see Fig. 2), eq (3) is replaced by It is important to note that the Engesser-K~rm~n
reduced modulus theory is the generalization for the in-
E~I y= + P( y + yo) = 0 (4) elastic case of the reasoning used in elastic bifurcation
theory (see Ref. 16). The Southwell method for plastic
Following the usual derivation of the Southwell plot buckling should therefore predict the reduced modulus (or
by assuming double modulus) critical load P~ for the corresponding
perfect column, though imperfect columns usually buckle
9 inTrxx nearer to the tangent modulus critical load Pt, as
yo(x) = ~ Wo. sm t - T - ) pointed out by Shanley" and generally accepted today
n=l
(5) (see for example Ref. 15 or 18). One may note that also
and by using a Shanley model, plastic buckling according to
the Engesser-K~rm~n reduced modulus theory implies a
n Trx
y(x) = ~ w. sin (--~---) neutral equilibrium postbuckling path (see Ref. 19), as in
n=l
the case of the elastic Euler column.
In the above analysis linear strain hardening has been
leads to stipulated to make the derivation a rigorous generalization
n2Pr of the Southwell's analysis. This restriction can, however,
y(x) = ~ Wo. [ 1 ]-' sin (n__~) (6)
n=l P be relaxed and the derivation applies also to the more
general case of a strain-hardening material, whose stress-
where
P~ = /'bT~r2___._:_. (7)
L=

the critical load according to the Engesser-K~m~n reduced

(or double) modulus theory.
As the critical load is approached, when P is a fairly
Fig. 2--Deflection of an imperfect
considerable fraction of P,, the central addition deflection, L
column
= = = w, = - (8)

P
As P - - P,, the imperfection component that represents
the buckling mode is the one that is primarily magnified.
Hence one can write

A B
~.,oill ~..,l IIIIIIII
Fig. 3--Southwell plot
o~ Et I loading
! fibIrS) Of van K&rm,4n's test
Fig. I--Idealized uniaxial I,K i ! i/~,~,ivi~,,d,,,l_l

? ''.
results for short
stress-strain diagram: linear 9 odirlg

strain-hardening
~, I IXXX~ "~/h'hYV.~liI I columns (from Ref. 11)
~ I I I I I
41~., D t f l e l t I o . ~ in ram.

206 9 June 1989

strain curve can be approximated by linear portions at case, the Southwell predictions are within +_5 percent of
the points of buckling. or, and the predicted trso=thwfor the corresponding perfect
column is always above or equal to the experimental
Experimental Verification and Conclusions buckling stress of the imperfect column, + 0 percent to
+ 9 percent above.
For experimental verification, the extended SouthweU
method was applied by Wang to the short columns of
von K~rm~n ~ of HorsfaU and Sandorff, ~3 and to Gerard's
Extension to Other Classes of Columns
tests described in Ref. 11. The application to von K~'rn~n's The concept employed in the derivation of the South-
columns has been reconfirmed independently by the well method in the plastic region for the simply supported
author with practically identical results. Figure 3 (re- column can be extended to other classes of columns
produced from Ref. 11) shows the application to von whose instability is essentially a linear bifurcation
Karman's short columns. Similar excellent fits are shown phenomenon. For then replacement of E by the reduced
in Ref. 11 for the remaining two short columns of modulus Er in the differential equation governing the
yon Kfirm~m and for those of Gerard and that of Hors- behavior of the column with an initial geometric imperfec-
fall and Sandorff. Table 1 (reproduced from Ref. 11) tion leads to the Southwell plot in the plastic region, for
shows that the extended inelastic Southwell predictions a linear (piecewise linear) strain-hardening material, in the
from the tests are indeed very close to the buckling stress same manner as it does in the simple case, originally
calculated with the Engesser-K~rm~n double (or reduced) considered by W a n g / ' derived in the section above.
modulus theory for perfect columns, try. Except in one As an example consider a column on an elastic founda-
tion (Fig. 4), considered in the elastic region by Donnell
in 1938. 2 Here, as before, yo(x) is the initial geometric
P imperfection and y ( x ) is the additional deflection, and
is the spring constant per unit length (with dimensions
force divided by length squared) usually called the modulus
of foundation.
Fig. 4--An elastically supported Differentiating eq (4) twice and adding the elastic
imperfect column support B one obtains
E r l y ..... + Y + P ( Y , x x + Y . . . . ) = 0 (11)

Substituting y and Yo from eq (5) in eq (11) yields, as in

TABLE 1--A COMPARISON BETWEEN THE ULTIMATE STRESS, THEORETICAL BUCKLING STRESSES, AND THE
BUCKLING STRESS ESTIMATED BY SOUTHWELL'S METHOD FOR SHORT COLUMNS

Ot Or ~rSouthw
Effective Calculated Buckling Stress aso,,h~ Buckling Stress oso,,h~ ~e~p
Slenderness Tangent Modulus Double Modulus Estimated ~,=p Ultimate
Column No. Ratio, ( L / e ) , Theory Theory ar from Test Stress

von K,~rm~n* Kg per sq cm Kg per sq cm Kg per sq cm Kg per sq cm

7a 88.1 2,400 2 690 1.033 2,780 1.007 2,760
7b 88.0 2,400 2 690 1.033 2,780 1.035 2,685
8 82.O 2,600 2 900 0.945 2,740 1.000 2,740
9a 73.1 2,960 3 050 1.000 3,050 1.007 3,030
9b 73.1 2,960 3 050 1.018 3,105 1.083 2,866
10a 58.6 3,100 3 150 1.029 3,240 1.017 3,185
10b 58.6 3,100 3 150 0.994 3,130 1.016 3,080
11 53.6 3,120 3 175 1.030 3,270 1.033 3,165
12a 48.2 3,130 3 210 0.969 3,110 1.010 3,080
12b 48.2 3,130 3 210 0.950 3,050 1.030 2,960
13 47.3 3,140 3 215 0.964 3,100 1.013 3,060
14b 38.2 3,160 3 320 1.048 3,480 1.048 3,320
15a 28.8 3,220 3 560 1.039 3,700 1.090 3,395
16 24.8 3,290 4 100 0.951 3,900 1.003 3,890
17 22.0 3,450 Approx. 0.978 4,500 1.039 4,330
4,600

Gerardt Ib per sq in. Ib per sq in. Ib per sq in. Ib per sq in.

1 21.7 41,900 47,500 0.945 44,900 1.000 44,880
2 21.2 43,000 49,000 0.992 48,600 1.009 48,150

Horsfall and Ib per sq in. Ib per sq in.. Ib per sq in. Ib per sq in.
Sandorfft 29.9 37,000 42,000 0.895 37,600 1.011 37,200

*Mild steel columns

t24S-T aluminum-alloy columns

Experimental Mechanics 9 207

 Donnell's derivation 2 for the elastic region, bility of Southwell's method apply also here to the more
9 w,, . p general case of a strain-hardening material whose stress-
P = ( - - ) ,.. (12) strain curve can be approximated by linear parts at the
Wn + Won points of buckling.
but here
nTr 2 L
P,. = e,z + )2 (13) Conclusions

The main conclusion is that the Southwell plot for

columns is as valid in the inelastic region, provided the
In the elastic range, where E , = E, and when Wo, =
material is strain hardening, as it is for elastic bifurcation
0, eqs (12) and (13) reduce to the classical formula for a
buckling, to which it has been restricted since conceived
simply supported column on elastic foundation (see for
by Southwell in 1932. Further theoretical justifications are
example Ref. 20, p. 98), and in turn to Euler's formula needed for reliable extension of the 'plastic validity' to
when/3 = 0.
other structures such as plates or shells. However in
Equation (12) can be written as
practice such extensions can be tentatively attempted as
they are on as firm a ground as many of the extensions in
Wn = Pr, ( ~ ) - W,o (14) the elastic region that relied only on Southwell's original
work.
which is similar to eq (10) and represents the usual South-
well plot for a column on an elastic foundation in the References
inelastic range. The inverse slope of the plot of the 1. Southwell, R.V., "'On the Analysis of Experimental Observations
experimental values of ( w , / P ) versus w , , again yields the in Problems of Elastic Stability, "" Proc. Roy. Soc. (London) Series A,
135, 601-616 (1932).
buckling load P , , in the inelastic range of the corre- 2. Donnell, L.H., "'On the Application of Southwell's Method for the
sponding perfect column on an elastic foundation, accord- Analysis of Buckling Tests," Stephar, Timoshenko 60th Anniversary
ing to the Engesser-K~rm~n reduced modulus theory. Volume, McGraw-Hill, New York, 27-38 (1938).
When /3 is small, n = 1 will yield the lowest critical 3. Horton, W.H., Cundary, F.L. and Johnson, R. W., "'The Analysis
of Experimental Data Obtained from Stability Studies on Elastic Column
load, but larger/3 will change the buckling mode. When and Plate Structures," Israel J. Tech., 5 (1-2), 104-113 (1967).
/3 >--(4r4ErI/L'), n = 2, or larger (depending on /~), 4. Spencer, H.H. and Walker, A.C., "'Critique of South'well Plots
will yield the lowest buckling load (Ref. 20, p. 100). The with Proposals for Alternative Methods, "" EXPERIMENTAL MECHANICS,
appropriate integer n, found from eq (15), 15 (8), 303-310 (1975).
5. Bushnell, D., Computerized Buckling Analysis of Shells, Martinus
n2(n + 1) 5 = - -
/3L' (15)
Nijhoff Dortrecht/Boston (1985).
6. yon Ktirmdn, Th., "'Untersuchungen "u~ber Knickfestikeit, ""
r 4 E~ I Mitteilungen uber Forschungsarbeiten auf dem Bebiet des Ingenieurwesens,
will indicate the optimum location for deflection measure- Verein Deutscher Ingenieure, Heft 81, Berlin (1910).
7. Tuckerman, L.B., "'Heterostatic Loading and Critical Astatic
ments and the corresponding w,'s whose Southwell plots Loads, "" Res. Paper RP 1163, J. Res. Nat. Bureau of Stand., 22, 1-18
will yield Pr,, which is the lowest inelastic buckling load (1939).
for this/3. 8. Bridget, F.J., Jerome, C.C. and Vosseler, A.B., "'Some New
As another example consider the work of Ariaratnam 2~ Experiments on the Buckling of Thin Wall Construction, "" Trans. ASME,
56, 569-578 (1934).
who demonstrated in 1961 analytically that the Southwell 9. Newman, J.B., "'Inelastic Column Buckling of Internally Pres-
method also applies to elastic columns with different surized Tubes, ""EXPERIMENTAL MECHANICS, 13, 265-272 (1973).
combinations of end conditions and with varying flexural 10. Sobel, L.H., "'The Southwell Method for Predicting Plastic
rigidity, In his derivations he employed the theorem of Buckling Loads for Elbows, "" J. of Press. Vessel Tech., Trans. ASME,
105, 2-8 (1983).
minimum potential energy, instead of the differential 11. Wang, C.T., "Inelastic Column Theories and an Analysis of
equation, as the conditions for equilibrium. For a linear Experimental Observations, '" J. Aero. Sci., 15, 283-292 (May 1948).
strain-hardening material (see Fig. 1), the reduced modulus 12. Lundquist, E.E., "'Generalized Analysis of Experimental Obsetwa-
Er is not a function of the stress but a constant for any tions in Problems of Elastic Stability, '" NACA TN 658 (1938).
13. Horsfall, IV. and Sandorff, P., "'Strain Distribution During
specific cross-sectional shape. Hence in the plastic region Column Failure, "' Lockheed Aircraft Corporation, Burbank, California,
the strain energy is obtained, in this case, simply by Rep. No. 5728 (April 1946).
replacing E with Er. The theorem of minimum potential 14. Massey, C., "Southwell Plot Applied to Lateral Instability o[
energy can therefore be used also for a linear strain- Beams, ""Engineer, 218 (5666), 320 (Aug. 1964).
15. Bleich, F., Buckling Strength of Metal Structures, McGraw-Hill,
hardening material, and Ariaratnam's extensive results for New York (1952).
columns with different end conditions and varying flexural 16, von Kdrfadn, Th., "'Discussion on the Previous Paper" (Shanley
rigidity as .well as plane frameworks, apply with E re- Ref. 17), J. Aero. Sci,, 14 (5), 267-268 (May 1947).
placed by the reduced modulus Er. 17. Shanley, F.R., "'Inelastic Column Theory, '" J. Aero. Sci., 14 (5),
261-267 (May 1947).
One should, however, remember that systems involving 18. Kollbrunner, C.F. and Meister, M., Knicken, Biegedrillknicken,
plastic flow are essentially nonconservative. In a bifurca- Kippen, 2nd Ed. Springer-Verlag, Berlin (1961).
tion problem, the energy required to bring a structure to 19. Chilver, A.H. and Britvec, S.J., "'The Plastic Buckling of
an adjacent buckled state will depend upon the path of Aluminum Columns," in Aluminum in Structural Engineering, Proc.
Instit. of Struct. Eng. and AluminumFed. Symp., London, 47-58 (1963).
transition, if the material is loading into the plastic region 20. Timoshenko, S.P. and Gere, J.M., Theory of Elastic Stability,
(e.g., Ref. 22), which could cast some doubt on the McGraw-Hill, New York (1961).
employment of the theorem of minimum potential energy. 21. Ariaratnam, S.T., "'The Southwell Method for Predicting Critical
Hill 23'2' has shown, however, that as long as the in- Loads of Elastic Structures, "' Quart. J. Mech. and Appl. Math,, 14,
137-153 (1961).
finitesimal path is reasonably direct, the variation in 22. Bushnell, D., "Plate Buckling" in Pressure Vessels and Piping:
infinitesimal energy dissipation from one path to another Design Technology 1982, a Decade of Progress, ed. S.Y. Zamrik and
consists of higher order terms only. Hence though not D. Dietrich, Book No. G00213, ASME, 47-117 (1982).
completely rigorous, the employment of the theorem of 23. Hill, R., "'A General Theory of Uniqueness and Stability, in
Elastic/Plastic Solids," J. Mech. Phys. Solids, 6, 236-249 (1958).
minimum potential energy is justified here as a very close 24. Hill, R., "'Some Basic Principles in the Mechanics of Solids
approximation. Furthermore, the results of the applica- Without a Natural Time, "' J. Mech. Phys. Solids, 7, 209-225 (1959).

208 9 June 1989