Академический Документы
Профессиональный Документы
Культура Документы
A R T I C L E I N F O A B S T R A C T
Article history: An exact non-linear formulation of the equilibrium of elastic prismatic rods subjected to compression and
Received 14 April 2008 planar bending is presented, electing as primary displacement variable the cross-section rotations and
Received in revised form 29 September 2008 taking into account the axis extensibility. Such a formulation proves to be sufficiently general to encom-
Accepted 12 November 2008
pass any boundary condition. The evaluation of critical loads for the five classical Euler buckling cases
Keywords: is pursued, allowing for the assessment of the axis extensibility effect. From the quantitative viewpoint,
Buckling it is seen that such an influence is negligible for very slender bars, but it dramatically increases as the
Extensible rod slenderness ratio decreases. From the qualitative viewpoint, its effect is that there are not infinite critical
Elastic instability loads, as foreseen by the classical inextensible theory. The method of multiple (spatial) scales is used
Post-buckling to survey the post-buckling regime for the five classical Euler buckling cases, with remarkable success,
Method of multiple scales
since very small deviations were observed with respect to results obtained via numerical integration of
the exact equation of equilibrium, even when loads much higher than the critical ones were considered.
Although known beforehand that such classical Euler buckling cases are imperfection insensitive, the
effect of load offsets were also looked at, thus showing that the formulation is sufficiently general to
accommodate this sort of analysis.
© 2008 Elsevier Ltd. All rights reserved.
0020-7462/$ - see front matter © 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijnonlinmec.2008.11.005
C.E.N. Mazzilli / International Journal of Non-Linear Mechanics 44 (2009) 200 -- 208 201
w – w0
P P
M M
R R S
u
+u z ϕ
P
w – w0 w
S S
R x, u R z
S x+u x
z, w
M0 P’
M0
P P
Fig. 1. (a) Prismatic elastic rod under bending and compression; (b) Bernoulli–Euler beam kinematics.
2. Non-linear equilibrium equation where, without loss of generality, it was assumed in the last of (6)
that ū0 = 0 and w̄0 = 0. Hence, combining (3) and (5), as well as (4)
The prismatic beam-column of Fig. 1(a), with length , cross- and (6)
section area A and moment of inertia I, made of an elastic material
of Young's modulus E, is considered. It is subjected to an initial axial EA(¯ − 1) = −P cos + R sin , (7)
compression P. It may be the case that end-bending moments come
into play, as result of constraint conditions and/or load offsets. In −EI = M − R[( + ū ) − (x + ū)] − P(w̄ − w̄)
the general case, to restore equilibrium with respect to moments, it
may happen that end transversal forces R also appear. = M0 + R(x + ū) + P w̄. (8)
Fig. 1(b) introduces the notation and refers to the Bernoulli–Euler
After derivation with respect to x and taking (1) into account, (8) is
kinematics, which is characterised by the following well-known re-
rewritten as
lationships:
evaluated as
The exact non-linear Eq. (12) can be approximated up to the order
N = EA(¯ − 1), (3) 3 , where 0 < >1, by
M = −EI . (4) d2
2
+ 1 + 2 2 + 3 3 = 0 , (14)
d
Considering the applied end loads, the normal force and the bending
moment can also be written as p(1 − p) p(1 − 4p)
0 = − , 2 = − ,
2
N = −P cos + R sin , (5)
p(1 − p + 2 p) p(1 − 4p + 42 p)
M = M − R[( + ū ) − (x + ū)] − P(w̄ − w̄) 1 = , 3 = − . (15)
6
= M0 + R(x + ū) + P w̄, (6)
Notice that the non-homogeneous term and the coefficient of the
1
Filipisch and Rosales [16] consider other statements for Hooke's law, depend-
quadratic term are scaled in (14) as of the order , since is null or
ing on which stress (engineering, second Piola–Kirchhoff, Cauchy) and strain (linear, at least small compared to the unity in the five classic Euler buckling
Green, Almansi, Hencky) definitions are used. cases, which are the main concern of this study.
202 C.E.N. Mazzilli / International Journal of Non-Linear Mechanics 44 (2009) 200 -- 208
I II III IV V
P P P P P
M = M0 M = M0
x
R
M0 M0 M0 M0
P P P P P
3. Linearised equilibrium equation: classic Euler buckling cases Critical loads are evaluated from (22), provided the solution (21)
is non-trivial. For the inextensible bar, an infinite number of Euler
Fig. 2 refers to the five classic Euler buckling cases [3,4]. To buckling loads pE are found, the smallest of them being
determine the corresponding critical loads for the perfect systems
2
(no load offsets here) it suffices to consider the linearised form Case I k = ⇒ pE = ,
of (14): 2 4
Case II k = ⇒ pE = 2 ,
d2
+ 1 = 0 . (16)
d
2 k
Case III = ⇒ pE = 4 2 ,
2
Notice that in Cases (I)–(IV) =0 ⇒ 0 =0, because no transversal
force is needed at the beam ends to secure equilibrium. In Case V, Case IV k = ⇒ pE = 2 ,
equilibrium requires that Case V k = 4.493 ⇒ pE = 20.19. (23)
d
(0) Should the axis stretching be considered, 1 would play the same
R EI (0) d d p(1 − p)
= = ≈ ⇒ (0) ≈ . (17) role as (k)2 . By the way, in Cases I–IV, it is seen that 1 = p(1 − p)/ ,
P ū p(1 − p) d since = 0. Even in Case V, 1 =p(1−p)/ still holds for the unbuckled
P 1 +
solution, up to the critical state. Hence, the smallest critical loads
Yet, from the first of (15) and (17), it is found that for Case V considering stretching are
d √ 1 − 1 − 2
0 ≈ − (0). (18) Case I 1 = ⇒ pcr = ,
d 2 2
In the classical inextensible rod solution, (16) is replaced by √ 1 − 1 − 4 2
Case II 1 = ⇒ pcr = ,
2
d2
2
+ (k)2 = 0 , (19) √ 1 − 1 − 16 2
d Case III 1 = 2 ⇒ pcr = ,
2
where
√ 1 − 1 − 4 2
Case IV 1 = ⇒ pcr = ,
p 2
k = (20)
√ 1 − 1 − 80.76
√ Case V 1 = 4.493 ⇒ pcr = . (24)
is an approximation for 1 . 2
√
Solution to (16)—or equally (19), if 1 is replaced by k—is A synthetic relationship between the critical loads, considering
√ √ axis stretching (pcr ) or not (pE ), can be proposed:
= 0 + C1 sin( 1 ) + C2 cos( 1 ), (21)
1 1 − 1 − 4pE
pcr = . (25)
where C1 and C2 are real constants to be determined from the bound- 2
ary and/or symmetry conditions:
Thus, the actual critical load pcr is always larger than the Euler
2
d buckling load pE . Notice that = I/A = (r/ )2 —where r is the gy-
Case I ⇒ (0) = 0 and (1) = 0,
d ration radius of the cross section—is inversely proportional to the
square of the slenderness ratio. It is seen that for very slender bars,
d d
Case II ⇒ (0) = 0 and (1) = 0, pcr ≈ pE . Yet, as the slenderness ratio decreases, and consequently
d d pE increases, tending to 14 , the critical load tends to twice the Euler
1 buckling load, as it can be seen in Fig. 3.
Case III ⇒ (0) = 0 and = 0,
2 It is also clear from (25) that there is not such a thing as a critical
load pcr when pE > 14 . Taking Case II as an example, it means that
Case IV ⇒ (0) = 0 and (1) = 0,
there is not a critical load for a bar with < 2 r, that is to say that the
d distance between two consecutive inflection points of the elastica
Case V ⇒ (0) = 0 and (1) = 0. (22)
d cannot be smaller than 2 r. Hence, there is a limited number of
C.E.N. Mazzilli / International Journal of Non-Linear Mechanics 44 (2009) 200 -- 208 203
Notice that the term on the right-hand side of (31) leads to an un-
bounded solution for 2 , contrary to what it is expected from the
proposed asymptotic expansion (26), according to which any new
term added should be a small correction to the accumulated ex-
pansion value. Therefore, to enforce that the expansion (26) be uni-
formly convergent, the right-hand side of (31), which is said to be a
secular term, must be eliminated, leading to the so-called solvability
condition D1 1 = 0 ⇒ A = A(2 , 3 , . . .). It is further noticed that the
homogeneous solution for 2 may be considered as already included
in (30) and can thus be disregarded in what follows.
Terms of order 3 in (14) lead still to another differential equation
Fig. 3. Critical load for extensible rod as a function of the Euler buckling load. D20 3 + 1 3 = −2D0 D2 1 − 2 21 − 3 31 , (32)
The derivative of with respect to will be useful for imposing the which can be evaluated with the help of (39) and (41), the boundary
boundary conditions: conditions and the displacements at either = 1 or 0. Notice that
from (37)
d
() = − (a) sin( +
0 )
d 8 1 0 2 0 2
a = √ −1− −4 . (43)
1 2 3 0 3 3 1 1 1 1
− + (a)2 sin[2( +
0 )]
3 1 1 1 The determination of the amplitude a, as in (43), is a key step
3 that can only be achieved once has been obtained after the impo-
− 3 (a)3 sin[3( +
0 )]. (39)
32 1 sition of the boundary conditions. In Case V, there is an additional
difficulty, since a depends on , which has to be determined iter-
The centroid displacements can be evaluated once () and atively, as it will be seen in Section 9. It is worth mentioning that
d/d() are known. For the longitudinal displacement, from the the multiple-scales solution, in general, and expressions (37)–(43),
last of (1) and (10), it is found that in particular, are valid for whichever Case I–V is considered, pro-
vided the appropriate boundary conditions are used. Each one of
these cases will be examined in detail in what follows.
ū
= [(1 − p cos + p sin ) cos − 1] d. (40)
0
5. Case I: clamped-free beam-column
Using the power series approximations for cos ≈ 1 − 2 /2 and
The boundary conditions (22) for Case I are
sin ≈ − 3 /6, (40) is re-written as
1 3
(0) = (a) cos
0 + (a)3 cos 3
0 = 0, (44)
ū 1−2p 2 32 1
= −p+p d− 2 d− p 3 d, (41)
0 2 0 3 0
d
(1) = − (a) sin( +
0 )
which can be evaluated with the help of (38) and the boundary d
conditions. 3
− 3 (a)3 sin[3( +
0 )] = 0, (45)
For the transversal displacements, from (8), it is found that 32 1
w̄ w̄(1) d d ū(1) ū from which, for the first buckling mode, it is found that
0 = /2 and
= + − + 1 + − + = /2. The first normalised critical load, from (24), is pcr =0.002474,
p d 1 d
for = 0.001. For loads p > pcr , displacements can be found from
d d ū (38), (39), (41)–(43), recalling that 0 = 2 = 0. Fig. 4 displays post-
= − − + , (42)
p d 0 d buckled configurations for distinct values of p > pcr .
Fig. 4. Post-buckled configurations and maximum displacements for (a) p/pcr = 1.0105; (b) p/pcr = 1.2894; (c) p/pcr = 1.5158; (d) p/pcr = 2.5344.
C.E.N. Mazzilli / International Journal of Non-Linear Mechanics 44 (2009) 200 -- 208 205
Fig. 5. Post-buckled configurations and maximum displacements for (a) p/pcr = 1.2037 (first mode); (b) p/pcr = 5.0155 (first mode); (c) p/pcr = 5.0155 (second mode);
(d) p/pcr = 8.0249 (second mode).
6. Case II: hinged–hinged beam-column (38), (39), (41)–(43), recalling that 0 = 2 = 0. Fig. 6 displays post-
buckled configurations for distinct values of p > pcr .
The boundary conditions (22) for Case II are
8. Case IV: clamped-guided bar
d 3
(0) = −(a) sin
0 − 3 (a)3 sin 3
0 = 0, (46)
d 32 1
The boundary conditions (22) for Case IV are
d
(1) = − (a) sin( +
0 ) 1 3
d (0) = (a) cos
0 + (a)3 cos 3
0 = 0, (50)
32 1
3
− 3 (a)3 sin[3( +
0 )] = 0, (47) 3
32 1 (1) = (a) cos( +
0 ) +
1
(a)3 cos[3( +
0 )] = 0, (51)
32 1
from which, for the first buckling mode, it is found that
0 = 0 and
= . The first normalised critical load, from (24), is pcr = 0.009969, from which, for the first buckling mode, it is found that
0 = /2 and
for = 0.001. By the way, the number of critical loads is here limited = . The first normalised critical load, from (24), is pcr = 0.009969,
to five. For loads p > pcr , displacements can be found from (38), (39), for = 0.001. For loads p > pcr , displacements can be found from
(41)–(43), recalling that 0 = 2 = 0. Fig. 5 displays post-buckled (38), (39), (41)–(43), recalling that 0 = 2 = 0. Fig. 7 displays post-
configurations for distinct values of p > pcr . buckled configurations for distinct values of p > pcr .
The boundary and symmetry conditions (22) for Case III are The boundary conditions (22) for Case V are
1 3
2
(0) = (a) cos
0 + (a)3 cos 3
0 = 0, (48) 0 2 0
1
32 1 (0) = − (a)2 +
1 1 1
2
1
= (a) cos +
0 3 0 0 2 3
2 2 − + (a) + (a) cos
0
2
1 1 1 2
1 3
+ (a)3 cos 3 +
0 = 0, (49)
32 1 2 1 1 2 3 0
+ + (a)2 cos 2
0
2 3 1 1 1
from which, for the first buckling mode, it is found that
0 = /2 and
=2 . The first normalised critical load, from (24), is pcr =0.041174, 1 3
+ (a)3 cos 3
0 = 0, (52)
for = 0.001. For loads p > pcr , displacements can be found from 32 1
206 C.E.N. Mazzilli / International Journal of Non-Linear Mechanics 44 (2009) 200 -- 208
Fig. 6. Post-buckled configurations and maximum displacements for (a) p/pcr = 1.1980; (b) p/pcr = 2.3960.
Fig. 7. Post-buckled configurations and maximum displacements for (a) p/pcr = 1.2037; (b) p/pcr = 1.8056.
d imposed for all (0), d/d(0) and ū(0)/ , even when one of the first
(1) = − (a) sin( +
0 )
d two is not strictly a boundary condition, and the integration will pro-
1 2 3 0 ceed for increasing values of . Once (1), d/d(1) and ū(1)/ are
− + (a)2 sin[2( +
0 )]
3 1 1 1 determined, the boundary conditions at this section can be checked.
If they are not satisfied to a certain prescribed small tolerance, an
3
− 3 (a)3 sin[3( +
0 )] = 0, (53) iterative scheme will be required. Note that in Case V, even for the
32 1 first iteration it is necessary to impose also ū(1)/ , to have an initial
value for , both of which values must also converge along the iter-
from which, for the first buckling mode, it is found that +
0 = ,
0 ative scheme. That is why Case V poses a more complicated conver-
being a solution of (52). The first normalised critical load, from (24), gence pattern. Table 1 displays some results correlating numerical
is pcr = 0.020615, for = 0.001. It should be recalled that here is integration and multiple scales results.
non-null in the post-buckling solution, on account of an equilibrium Multiple scales results for the maximum transversal displace-
requirement. Notice that for each value of p > pcr , an iteration scheme ment, as seen in Table 1, present deviations not larger than 1.5% with
is necessary to evaluate from (17), 0 and 2 from (15),
0 from respect to the supposedly more accurate results coming out from
(52), = −
0 , a from (43), () from (38), d/d() from (39), the numerical integration of the exact Eqs. (12) and (40), for loads
ū()/ from (41), w̄()/ from (42), from (17), etc., until convergence 20% larger than the critical one for the corresponding Case I–IV. For
has been attained. Fig. 8 displays post-buckled configurations for Case V, small deviations (2%) were still observed for a load 6.72%
distinct values of p > pcr . larger than the critical one. Deviations are seen to be larger when
axial displacements are compared: up to 8.1%, for loads 20% larger
10. Numerical integration than the critical one for the corresponding Case I–IV; and 11.7% for
Case V, for a load 6.72% larger than the critical one. For higher loads,
The solution obtained by numerical integration of (12) and (40) comparison was not possible in Case V, since convergence was not
is now discussed in this section. For Cases I–IV, since = 0, these achieved in the iterative numerical integration scheme, although the
equations are uncoupled, which means that we could integrate (12) multiple-scales solution was still available.
to obtain () and then (40) to obtain ū()/ . Nevertheless, for Case
V, since =EI (0)/P (1+ ū(1)/ ) 0, these equations are coupled and 11. Effect of load offsets
must be integrated simultaneously. As for w̄()/ , it can be evaluated
from (42) in any Case. Since this is not a truly initial-value problem, It is well known that Euler buckling is imperfection insensitive,
the fourth-order Runge–Kutta method is not strictly applicable. Yet, since it corresponds to a stable symmetric bifurcation [6,7]. In other
it is possible to adapt it in such a way that initial conditions will be words, small imperfections, such as load offsets, “round off” the
C.E.N. Mazzilli / International Journal of Non-Linear Mechanics 44 (2009) 200 -- 208 207
Fig. 8. Post-buckled configurations and maximum displacements for (a) p/pcr = 1.0187; (b) p/pcr = 1.0672; (c) p/pcr = 1.1157; (d) p/pcr = 2.4254.
Table 1
Correlation between numerical integration and multiple scales results for = 0.001.
Case I II III IV V
Acknowledgements