Академический Документы
Профессиональный Документы
Культура Документы
Stability Of
Structures:
Basic Concepts
171
Lecture 17: STABILITY OF STRUCTURES: BASIC CONCEPTS 172
TABLE OF CONTENTS
Page
17.1. Introduction 173
17.2. Testing Stability 173
17.2.1. Stability of Static Equilibrium . . . . . . . . . . . . 173
17.2.2. Stability of Dynamic Equilibrium . . . . . . . . . . 174
17.3. Static Stability Loss 175
17.3.1. Buckling Or Snapping? . . . . . . . . . . . . . . 175
17.3.2. Response Diagrams . . . . . . . . . . . . . . . 175
17.3.3. Stability Models . . . . . . . . . . . . . . . . 176
17.3.4. Stability Equations Derivation . . . . . . . . . . . 177
17.4. Exact Versus Linearized Stability Analysis 177
17.4.1. Example 1: The HCR Column: Geometrically Exact Analysis . 178
17.4.2. Example 1: The HCR Column: LPB Analysis . . . . . . 179
17.4.3. Example 2: The PCR Column: Geometrically Exact Analysis . 1710
17.4.4. Example 2: The PCR Column - LPB Analysis . . . . . . 1712
17.5. Discrete Stability Analysis As Eigenproblem 1713
17.5.1. Example 3: A Cantilevered Two-Strut Column . . . . . . 1713
17.5.2. Example 4: A Pinned-Pinned Three-Strut Column . . . . . 1715
172
173 17.2 TESTING STABILITY
17.1. Introduction
The term stability has both informal and formal meanings. As regards the former, the American Heritage
Dictionary lists the following three. 1. Resistance to sudden change, dislodgment, or overthrow. 2a.
Constancy of character or purpose: tenacity; steadfastness. 2b. Reliability; dependability. Related
verb: to stabilize. Related adjective: stable. Opposite terms: stability loss, instability, to destabilize,
unstable.
The formal meaning is found in engineering and sciences, concerning stability of systems.* Broadly
speaking, structural stability can be defined as the power to recover equilibrium. It is an essential
requirement for all structures. Jennings provides the following historical sketch:
Masonry structures generally become more stable with increasing dead weight. However
when iron and steel became available in quantity, elastic buckling due to loss of stability of
slender members appeared as a particular hazard.
The stability of a mechanical system, and of structures in particular, can be tested (experimentally
or analytically) by observing how it reacts when external disturbances are applied. Here we have to
distinguish between statics and dynamics.
For simplicity we will assume that the structure under study is elastic, since memory effects such as
plasticity or creep introduce additional complications such as historical dependence, which are beyond
the scope of the course. The applied forces are characterized by a loading parameter , also called a
load factor. Setting = 0 means that the structure is unloaded, at which it takes up an equilibrium
configuration C0 = (0) called the undeformed state. Furthermore, assume that this state is stable in
the sense defined below.
As is varied from 0 the structure deforms and assumes equilibrium configurations C(). These are
assumed to be (I) continuously dependend on and (II) stable for sufficienttly smaller values of .
How is stability tested? Freeze at a specific value, say d where d connotes deformed. The
associated equilibrium configuration is Cd = C(d ). Apply a perturbation to Cd , and remove it. What
sort of perturbation? An action that may disturb the state, for example a tiny load or a small imposed
deflection. More precise restrictions on such perturbations are described later. For now we will
generically call admissible perturbations those that are allowed in the application of the test.
The perturbation triggers subsequent motion of the system. Three possible outcomes are sketched in
Figure 17.1.
S For all admissible perturbations, the structure either returns to the tested configuration Cd or
executes bounded oscillations about it. The equilibrium at d is called stable.
* A system is a functionally related group of components forming or regarded as a collective entity. This definition uses
component as a generic term that embodies element or part, which connote simplicity, as well as subsystem, which
connotes complexity. In this course we shall be concerned about mechanical systems governed by Newtonian mechanics,
with focus on structures.
A. Jennings, Structures: From Theory to Practice, Taylor and Francis, London, 2004, Chapter 7.
173
Lecture 17: STABILITY OF STRUCTURES: BASIC CONCEPTS 174
U If for at least one admissible perturbation the structure moves to (decays to, or oscillates about)
another configuration, or takes off in an unbounded motion, the equilibrium is unstable.
N The transition from S to N occurs at a value = cr called the critical load factor. The
configuration Ccr = C(cr ) at the critical load factor is said to be in neutral equilibrium. The
quantitative determination of this transition is a key objective of stability analysis.
The foregoing classification leaves several gaps and details unanswered.
First, speaking about moving or returning introduces time into the picture. Indeed the concept of
stability is necessarily dynamic in nature There is a before: prior to applying the perturbation, and an
after: what happens upon removing it. Many practical methods to assess critical loads, however, factor
out the time dimension as long as certain conditions are verified. Those are know as static criteria.
Second, the concept of perturbation as small imposed change is imprecise. How small is a tiny
load or a slight deflection? The idea is made more mathematically precise later when we introduce
linearized stability, also called stability in the small. This is a natural consequence of assuming
infinitesimal configuration changes.
For example, Bazant and Cedolin in Stability of Structures: Elastic, Inelastic, Fracture and Damage Theories,Dover, 2003,
comment on p. 144: Failure of structures is a dynamical process, and so it is obviouly more realistic to approach buckling
and instability from a dynamical point of view.
E.g., conservative loading: applied loads derive from a potential.
174
175 17.3 STATIC STABILITY LOSS
To make such problems mathematically tractable it is common to restrict the kind of motions in such a
way that a bounded reference motion can be readily defined. For example a bounded periodic motion
of a oscillating structure. Departures under parametric changes are studied. Transition to unbounded
or unpredictable motion is taken as a sign of instability.
An important application of this concept are vibrations of structures that interact with an external or
internal fluid flow: bridges, buildings, airplanes, fluid pipes. The steady speed of the flow may be
taken as parameter. At a certain airflow or liquid-flow speed, increasing oscillations may be triggered:
this is called flutter. Or a non-oscillatory unbounded motion happens: this is called divergence. A
famous example of flutter in a civil structure was the collapse of the newly opened Tacoma-Narrows
suspension bridge near Seattle in 1940 under a moderate wind speed of about 40 mph.
Modeling and analysis of dynamic instability is covered in other courses, primarily at the graduate
level because it requires fancier math tools and heavier use of complex calculus. In this course we
consider only static stability. Moreover, the class of problems, methods and examples will be severely
constrained so as to fit into three lectures.
As just noted, we restrict attention to stability of static equilibrium. Two assumptions are introduced:
Linear Elasticity. The structural material is, and remains, linearly elastic. Displacements and rota-
tions, however, are not necessarily small.
Conservative Loading. The applied loads are conservative, that is, derivable from a potential. For ex-
ample, gravity and hydrostatic loads are conservative. On the other hand, aerodynamic and propulsion
loads (wind gusts on a bridge, rocket thrust, etc) are often nonconservative.
The main reason for the second restriction is that loss of stability under nonconservative loads is
inherently dynamic in nature, and thus lies beyond the scope of the course.
17.3.1. Buckling Or Snapping?
Under the foregoing restrictions, two types of instabilities may occur:
Bifurcation. Structural engineers use the more familiar name buckling for this one. The structure
reaches a bifurcation point, at which two or more equilibrium paths intersect. What happens after the
bifurcation point is traversed is called post-buckling behavior.
Snapping. Structural engineers use the term snap-through or snap buckling for this one. The structure
reaches a limit point at which the load reaches a maximum. What happens after the limit point is
traversed is called post-snapping behavior.
Bifurcation points and limit points are instances of critical points. The importance of critical points in
static stability analysis stems from the following property:
Transition from stability to instability can only occur at critical points
Reaching a critical point may lead to immediate destruction (collapse) of the structure. This depends
on its post-buckling or post-snapping behavior, and nature of the material. For some scenarios the
knowledge of such behavior is important since immediate collapse may lead to loss of life. On the
other hand, there are some configurations where the structure keeps resisting significant or even
increasing loads after traversing a critical point. Such load-sustaining designs are obviously
preferable from a safety standpoint.
175
Lecture 17: STABILITY OF STRUCTURES: BASIC CONCEPTS 176
Figure 17.2. Graphical representation of static equilibrium paths and their critical points: (a)
a response path with no critical points; (b) multiple response paths showing occurrence of two
critical points types: bifurcation point (B) and limit point (L).
To illustrate the occurrence of static instability as well as critical points we will often display load-
deflection response diagrams. This is a plot of equilibrium configurations taken by a structure as a
load, or load parameter, is gradually and continuously varied. The load (or load parameter) is plotted
along the vertical axis while a judiciously chosen representative deflection is plotted along the horizontal
axes. A common convention is to take zero deflection at zero load. This defines the reference state,
labeled as point R in such plots.
A continuous set of equilibrium configurations forms an equilibrium path. Such paths are illustrated
in Figure 17.2. The plot in Figure 17.2(a) shows a response path with no critical points. On the other
hand, the plot in Figure 17.2(b) depicts the occurrence of two critical points: one bifurcation and one
limit point. Those points are labeled as B and L, respectively.
Students should be familiar with this visualization technique if they have done tension or torsion mechanical tests.
176
177 17.4 EXACT VERSUS LINEARIZED STABILITY ANALYSIS
Stability equations for discrete models may be constructed using various devices. For lumped parameter
models one may resort to either perturbed equilibrium equations built via FBDs, or to energy methods.
For FEM models only energy methods are practical. All techniques eventually lead to matrix stability
equations that take the form of an algebraic eigenproblem.
Both continuous and lumped-parameter discrete models for column problems are presented later in
this Lecture, and in the following two Lectures.
As noted above, we will use only the equilibrium method to set up stability equations. Its key feature
is that FBDs must take the perturbed configuration into account. For certain simple problems it is
possible to establish the equations using the exact geometry of the deflected structure. Such analyses
will be called geometrically exact. A couple of examples follow to illustrate exact versus linearized
results.
* Often this term is restricted to the determinantal form of the stability eigenproblem. This is the equation whose roots give
the eigenvalues, which can be interpreted physically as critical loads, or as critical load parameters.
177
Lecture 17: STABILITY OF STRUCTURES: BASIC CONCEPTS 178
L Bifurcation
rigid point
MB = k L cr
k Equilibrium
;;;
path of straight
B (untilted) column
B
P Stable
Unstable
Figure 17.3. Geometrically exact analysis of the hinged cantilevered rigid (HRC) column: (a) untilted
column, (b) tilted column, (c) equilibrium paths intersecting at bifurcation point.
Pr e f
k = Pr e f d = Pr e f L sin k sin = 0. (17.1)
L
The equation on the right hand side has the two solutions
k
= 0 for any , = . (17.2)
Pr e f L sin
These pertain to the untilted ( = 0) and tilted ( = 0) equilibrium paths, respectively. Since
lim(/ sin ) 1 as 0, the two paths intersect when
k k
cr = , or Pcr = . (17.3)
Pr e f L L
The two paths are plotted in Figure 17.3(c). The intersection (17.3) characterizes a bifurcation point B.
The diagram shows four branches emanating from B. Three are stable (full lines) and one is unstable
178
179 17.4 EXACT VERSUS LINEARIZED STABILITY ANALYSIS
(dashed line). Note that the applied load may rise beyond the critical Pcr = cr Pr e f = k/L by moving
to a tilted configuration. It is not difficult to show that the maximum load occurs if 180 , for
which P ; this is a consequence of the assumption that the column is rigid (and that it may fully
rotate by that amount about the hinge).
This is the stability equation. Since the product of two numbers is zero, at least one must be zero:
= 0 or k Pr e f /L = 0, or both. The solution = 0 reproduces the untilted configuration. For
buckling to occur, we must have = 0. If so, the expression in parenthesis must vanish, which requires
= cr = k/(Pr e f L). This reproduces the critical load given in (17.3).
Note that this analysis yields only the critical load. It does not provides any information on post-
buckling behavior. If this is necessary, a more comprehensive analysis, such as done in the previous
subsection, is required.
It should be noted that if the loss of stability is by snap buckling, it cannot be obtained by LPB. The
main reason is that deformations prior to buckling are essential in the determination of limit points,
whereas the first LPB assumption listed above explicitly neglects those.
179
Lecture 17: STABILITY OF STRUCTURES: BASIC CONCEPTS 1710
; ;
(a) (b)
P = Pref vA = L sin
;; ;;
C P = Pref
A A
C' A' uA = L (1 cos )
k
L
rigid L
;;B
P
B
Figure 17.4. Geometrically exact analysis of a propped cantilevered rigid (PRC) column with
extensional spring remaining horizontal: (a) untilted column, (b) tilted column.
(a) (b)
vA = L sin
P = Pref
A
A' Bifurcation
FA = k vA point Equilibrium path
Equilibrium of tilted column
path of straight
(untilted) column
L cos
cr = k L/Pref
B 90 +90
Stable
P Unstable
Figure 17.5. Geometrically exact analysis of a propped cantilevered rigid (PRC) column with
extensional spring remaining horizontal: (a) FBD for tilted equilibrium, (b) equilibrium paths intersecting
at bifurcation point.
Consider next the configuration pictured in Figure 17.4(a); this differs from the HRC column in the type
of stabilizing spring. A rigid strut of length L is hinged at B and supports a vertical load P = Pr e f at
end A. The load remains vertical as the column tilts. The column is propped by an extensional spring
of stiffness k attached to A. This configuration will be called a propped cantilevered rigid column; or
PCR column for short. The only DOF is the tilt angle .
For the geometrically exact analysis is it important to know what happens to the spring as the column
tilts. One possible assumption is that it remains horizontal, as depicted in Figure 17.4(b). If so the
FBD in the tilted configuration will be as shown in Figure 17.5(a). Taking moments about B yields
1710
1711 17.4 EXACT VERSUS LINEARIZED STABILITY ANALYSIS
; ;
(a) P = Pref (b)
L
vA = L sin
; ;
P = Pref
A A
A' uA = L (1 cos )
C C
k
L
rigid L
;; B
P
B
Figure 17.6. Geometrically exact analysis of a propped cantilevered rigid (PRC) column with wall-attached
extensional spring: (a) untilted column, (b) tilted column.
= angle A'CA
vA = L sin
positive CW P = Pref
C A
A' uA = L (1 cos )
FA = k ds
ds : tilting
spring
elongation
B
P
Figure 17.7. Geometrically exact analysis of a propped cantilevered rigid (PRC) column with wall-attached
extensional spring: (a) FBD for tilted equilibrium, (b) equilibrium paths intersecting at bifurcation point.
4 2
3
= 1/4 1.5
= 1/2
2 1
1 0.5
60 40 20 20 40 60 80
60 40 20 20 40
angle (deg) 0.5 angle (deg)
1
1
1
1
0.75 0.8
0.5 =1
0.6 =5
0.25
0.4
60 40 20 20 40 60 80
0.25
angle (deg) 0.2
0.5
angle (deg)
0.75
60 40 20 20 40 60 80
1
Figure 17.8. Geometrically exact analysis of a propped cantilevered rigid (PRC) column with wall-attached
extensional spring: vs./ response diagrams for four values of (defined in Figure 17.6(a)), with Pr e f = k = L = 1.
1711
Lecture 17: STABILITY OF STRUCTURES: BASIC CONCEPTS 1712
These yield the vertical and tilted equilibrium paths, respectively, plotted in Figure 17.5(b) on the
versus plane. The two paths intersect at = k L/Pr e f , which is a bifurcation point. Consequently
the critical load parameter is cr = k L/Pr e f and the critical load is Pcr = cr Pr e f = k L. Of the 4
branches that emanate from the bifurcation point B, only one (the = 0 path for < cr is stable.
Once B is reached the tilted column supports only a decreasing load P, which vanishes at = 90 .
Consequently this configuration is poor from the standpoint of post-buckling safety. three are unstable.
Another reasonable assumption is that the spring attachment point to the wall, called C in Figure
17.6(a), stays fixed. The distance AC is parametrized with respect to the column length as L, in
which is dimensionless. If the column tilts, the spring also tilts as pictured in Figure 17.6(b). The
geometrically exact FBD for this case is shown in Figure 17.6(b). As can be observed, it is considerably
more involved than for the stay-horizontal case. We quote only the final result for the equilibrium path
equations:
k L cos + sin
= 0 for any , = . (17.7)
Pr e f + sin
Figure 17.8 shows response plots on the versus plane for the four cases = 14 , = 12 = 1 and
= 5, drawn with Pr e f = 1, k = 1, and L = 1. Although the bifurcation points are in the same
location, the post-buckling response is no longer symmetric for . That deviation from symmetry is
most conspicuous when < 1, since if so the spring tilting has a noticeable effect if < 0. The sharp
drop of towards occurs when the spring and the tilted column are nearly aligned. As
one recovers the solution (17.2), as may be expected.
( Pr e f L k L) = 0. (17.8)
kL
cr = , or Pcr = cr Pr e f = k L . (17.9)
Pr e f
This result is independent of assumptions on how the spring wall-attachment point behaves after the
column buckles. This is to be expected since linearization filters out that information.
1712
1713 17.5 DISCRETE STABILITY ANALYSIS AS EIGENPROBLEM
(a) P (b) vA
P
k
A A'
A
v deflections
L/2 + to the right
rigid
vB
3k/2 B
L B B'
L/2
rigid
C C
Figure 17.9. Cantilevered two-strut column. (a): structure; (b): deflected shape defined by two DOF v A and v B ;
When the discrete model has n 2 DOFs, the determination of critical loads by the LPB approach
leads to a matrix algebraic eigenproblem of order n. The n eigenvalues provide the critical loads, and
the corresponding eigenvectors give the associated buckling shapes. This is illustrated through several
examples worked out below.
1713
Lecture 17: STABILITY OF STRUCTURES: BASIC CONCEPTS 1714
P
vA P
A vA
A' FA A
(c1) A' FA
FBD of Applied forces (c3)
link AB in black, spring
vB forces in blue,
reactions in red
FA +FB B B' FB vB
B FB
P P B'
(c2) vB FA +FB FBD of whole
B B' column
FBD of FA +FB
link BC C
Restoring spring P
FA +FB forces + to the left
C
FA = k vA , FB = 3kvB /2
P
Figure 17.10. Cantilevered two-strut column. (c1,c2,c3): FBDs of link AB, link BC and whole column,
respectively. FBD color convention: applied forces in black, spring foces in blue, reactions in red.
[check the figure: merging (c1) and (c2) gives (c3)], the results of the analysis should be exactly the
same, although the stability matrix equations will be different. For the ensuing derivation we pick (c1)
and (c3).
For link AB, taking moments with respect to the displaced hinge point B, positive CW, gives
M B = P(v A v B ) FA ( 12 L) = P(v A v B ) 12 Lkv A = 0 (17.10)
For the whole column ABC, taking moments with respect to the hinge C, positive CW, gives
MC = Pv A FA L FB ( 12 L) = Pv A k Lv A 32 k( 12 L)v B = 0 (17.11)
Characteristic Equation. Combining (17.10) and (17.11) in a matrix equation gives the stability
equation (also known as the characteristic equation):
P 12 k L P vA 0
= , or Av = 0. (17.12)
P kL 4kL 3 vB 0
Matrix A is called the stability matrix or characteristic matrix. System (17.12) has two kinds of
solutions. The trivial solution v = 0, or v A = v B = 0, corresponds to the undeflected (vertical)
column, which holds for any P. To find the critical loads we must have v = 0, that is, at least one of the
{v A , v B } must be nonzero. Consequently A must be singular or, equivalenet, have zero determinant.
Since the entries of A depend on P, this is an algebraic eigenproblem in P.
Critical Loads. The eigenvalues are the solution of the characteristic equation
det(A) = 0, or P 2 74 k L P + 38 k 2 L 2 = 0 (17.13)
Solving this quadratic for P gives the two critical loads:
Pcr 1 = 14 k L = 0.25 k L , Pcr 2 = 32 k L = 1.50 k L . (17.14)
1714
1715 17.5 DISCRETE STABILITY ANALYSIS AS EIGENPROBLEM
1 1
A' A'
B' 1 2/3
B'
Figure 17.11. Buckling mode shapes for the cantilevered two-strut column.
Buckling Mode Shapes. Replacing Pcr 1 into (17.14) gives the following equation to determine the
eigenvector v1 :
k L 3 3 v A1 0 v A1 1
= v1 = = c1 (17.16)
4 1 1 v B1 0 v B1 1
where c1 is an arbitrary nonzero scaling factor.
Replacing Pcr 2 into A gives the following equation to determine the eigenvector v2 :
k L 2 3 v A2 0 v A2 3/2
= v2 = = c2 (17.17)
4 4 6 v B2 0 v B2 1
where c2 is an arbitrary nonzero scaling factor.
One common nomalization condition for eigenvectors is to make their largest component equal to +1.
This is done for (17.16) and (17.17) by taking c1 = 1 or c1 = 1 (either one works), and c2 = 2/3,
respectively. The two eigenvectors normalized as per that criterion are
1 1
v1 = for Pcr 1 , v2 = for Pcr 2 . (17.18)
1 2/3
This is a slight varint of one treated in Jennings (op. cit.). Figure 17.12(a) shows a column built with
three rigid struts of equal length, pinned at the joints, and with two extensional spring of stiffness k
as lateral supports at the internal joints B and C. Determine the critical loads of this configuration by
linearization.
Figure 17.12(b) depeicts how the structure might displace after buckling. Since the structs are rigid
and cannot change length, joints A, B and C will necesasrily moved down while the springs tilt;
1715
Lecture 17: STABILITY OF STRUCTURES: BASIC CONCEPTS 1716
rigid
L/3 (1/3) k (2vB +vC )
k
B FB = k vB
B' B'
v deflections vC
rigid
L/3 + to the right
k
C FC = k vC
rigid
C' C'
Figure 17.12. Stability analysis of three-strut column propped by extensional springs: (a) reference
configuration; (b) admissible buckled shape showing realistic geometry change (struts do not change length);
(c) linearized configuration: B and C displace by small horizontal movements whereas A stays fixed.
such geometry is correct for geometrically exact analysis. The linearized version is shown in Figure
17.9(c); here B and C move horizontally by v B and v D , respectively, and A does not move vertically.
The horizontal reactions R A and R D shown in Figure 17.9(c) are obtained in terms of FB and FD by
statics on taking moments with respect to D and A, respectively.
The two FBDs used to form the stability equations are shown in Figure 17.13. We take moments with
respect to C in the left FBD and with respect to B in the right FBD:
L 2L L 2L
MC = P vC FB + RA = 0, M B = P v B FC + RD = 0. (17.19)
3 3 3 3
vB
P P
A FB = k vB
B'
vB
RA = (1/3) (2FB +FC ) vC
= (1/3) k (2vB +vC )
FB = k vB FC = k vC
B' C'
vC
RD = (1/3) (FB +2FC )
= (1/3) k (vB +2vC )
FC = k vC D
C'
P P
Figure 17.13. Free Body Diagrams for the 3-strut column of Figure 17.12.
1716
1717 17.5 DISCRETE STABILITY ANALYSIS AS EIGENPROBLEM
B 1 B 1
1 C C 1
D D
Figure 17.14. Critical loads and buckling mode shapes for the 3-strut column of Figure 17.12.
This is a quadratic polynomial in P. Setting C(P) = 0 gives the two critical loads: Pcr 1 = k L/9 and
Pcr 2 = k L/3, as roots. The corresponding eigenvectors, normalized to 1 as largest entry, are
1 1
v1 = , v2 = . (17.22)
1 1
These are plotted in Figure 17.14. Note that the mode shape corresponding to the lowest critical load,
Pcr 1 = k L/9, is antisymmetric. This is in sharp contrast to the Euler column buckling treated in
Lecture 18.
An alternative eigenvalue problem is obtained on premultiplying by the inverse of the LHS matrix in
(17.20). This gives
3P 2 1 vB vB
= , (17.23)
kL 1 2 vc vC
an eigenproblem which has the same solutions. This is the form derived by Jennings (except for scale
factors).
1717