Chapter 11 · Section 2

Stability of Finite-DOF Systems

Starting from finite-DOF models, we build the two main lines for solving elastic stability problems: the static method (equilibrium equations in the small-perturbed configuration) and the energy method (stationary-potential principle). Through one example — two-DOF system — we verify that both methods give the same critical load $F_{\mathrm{Pcr}} = kl/3$.

How to use

Click any item in the left Contents to jump · Click the green "Next" button at the end of each subsection · Each animation has a ▶ button: click to resume from the last breakpoint; at the end it changes to for replay.

11.2.1

Buckling degrees of freedom of a system

Degrees of Freedom
Question · How many buckling DOF?

For the same class of columns (fixed at the bottom, compressed at the top), under the three conventions axial deformation ignored, only the right column carries axial load, no pre-buckling bendinghow many buckling DOF does it have? Consider both $EI = \infty$ (rigid) and finite $EI$.

Answer: for finite $EI$ (elastic column) the deflection curve is continuous, so the buckling DOF is infinite; for $EI = \infty$ (rigid), the buckling DOF is determined by the number of hinges / elastic supports — typically a finite number like 1 or 3.

When solving a stability problem, the first question is: how many independent parameters are needed to describe the new post-buckling displacement configuration? This number is the system's buckling DOF.

The DOF depends on constraints and rigidity assumptions. For an elastic column the deflection curve needs infinitely many parameters — it is an infinite-DOF problem (discussed in §11-3); whereas in simplified engineering models the column is often taken as a rigid bar with deformation concentrated at individual hinges, giving a finite buckling DOF.

① Infinite DOF

Any cross-section of an elastic column may deflect laterally — the deflection curve $y(x)$ is continuous and needs infinitely many independent parameters; this is an infinite-DOF problem.

Anim. 11.2.1-a

② Simplified finite-DOF model

When the column itself is treated as rigid and deformation is concentrated only at elastic supports or hinges, the buckling configuration is described by finitely many independent parameters. Below, a column formed by three rigid bars connected by two elastic hinges has buckling DOF $n = 2$ or $3$, depending on the number of hinges.

Anim. 11.2.1-b
Why the finite-DOF model matters

Although real structures are usually infinite-DOF, the finite-DOF model reveals the fundamental principles of stability analysis in the simplest form — the static method's eigenvalue equation and the energy method's stationary-potential condition both have clean algebraic expressions under finite DOF, providing the foundation for understanding infinite-DOF (§11-3) and energy approximation (§11-4).

11.2.2

Static method · Basic idea

Static Method

The static method (equilibrium method) rests on the idea: near the original equilibrium, assume a small new displacement configuration and write the equilibrium equations in that new configuration; if the equations admit a non-trivial solution, the new configuration can coexist with equilibrium — the corresponding load is the critical load.

① Equilibrium in the new configuration

Take Animation 11.2.2-a's single-DOF model: a rigid column $AC$ fixed at the bottom with a horizontal support at the top ($A$'s rotational stiffness $k_\theta = 3EI/l$). After buckling, the rotation $\theta$ at $A$ is the only DOF. Writing the moment equilibrium for the rigid bar:

$$ F_{\mathrm P}\, l \sin\theta \;-\; k_\theta\, \theta \;=\; 0 $$
(11.2-1)
using $\sin\theta \approx \theta$ and substituting $k_\theta = 3EI/l$
Anim. 11.2.2-a

② Non-trivial solution of the homogeneous equation

Rearranging gives a homogeneous equation in $\theta$:

$$ \left( F_{\mathrm P}\, l \;-\; \frac{3EI}{l} \right)\theta \;=\; 0 $$
(11.2-2)
$\theta = 0$ is the trivial (straight) equilibrium; setting the coefficient $= 0$ gives the non-trivial condition

Setting the coefficient to zero: $F_{\mathrm P} l - 3EI/l = 0$, giving the critical load

$$ F_{\mathrm{Pcr}} = \dfrac{3EI}{l^{2}} $$
(11.2-3)

③ Generalization to n DOF · Characteristic equation

For a system with $n$ DOF, set up $n$ independent equilibrium equations in the new configuration, giving a homogeneous linear system in the $n$ independent parameters. The condition for a non-trivial solution is that the coefficient determinant vanishes:

$$ D \;=\; 0$$
(11-1)
Eq. (11-1) is the stability equation or characteristic equation; $n$ real roots give $n$ eigenvalues
Anim. 11.2.2-b
11.2.3

Example 11-1 · Two-DOF system (static method)

Worked Example

In the system of Animation 11.2.3-a, bars $AB$, $BC$, $CD$ are all rigid; at nodes $B$ and $C$ there are elastic supports (stiffness $k$); at the $D$ end there is an axial compression $F_{\mathrm P}$. Find the critical load and the corresponding mode shape.

① DOF · New displacement configuration

At buckling, the new configuration is fully determined by the two vertical displacements $y_1, y_2$ of hinges $B, C$ — DOF $n = 2$. The elastic-support reactions are $k y_1$ and $k y_2$.

Anim. 11.2.3-a

② Free-body equilibrium

Overall moment equilibrium about $A$ ($\sum M_A = 0$) gives the vertical reaction at $D$:

$$ F_{yD} \;=\; \tfrac{1}{3}\,(k y_1 + 2\, k y_2) $$
(11.2-4)

Taking moments to the right of hinges $B$ and $C$ ($\sum M_B = 0,\ \sum M_C = 0$):

$$ \left\{\begin{aligned} F_{\mathrm P}\, y_1 \;+\; k y_2\, l \;-\; F_{yD}\cdot 2l &= 0 \\[2pt] F_{\mathrm P}\, y_2 \;-\; F_{yD}\, l &= 0 \end{aligned}\right. $$
(11.2-5)

③ Reduce to a homogeneous system

Eliminating $F_{yD}$ gives a homogeneous system in $y_1, y_2$:

$$ \left\{\begin{aligned} (3 F_{\mathrm P} - 2kl)\, y_1 \;-\; kl\, y_2 &= 0 \\[2pt] -\,kl\, y_1 \;+\; (3 F_{\mathrm P} - 2kl)\, y_2 &= 0 \end{aligned}\right. $$
(11.2-6)
$y_1 = y_2 = 0$ is the trivial equilibrium; buckling requires a non-trivial solution

④ Stability equation · Eigenvalues

Non-trivial solution requires the coefficient determinant to vanish:

$$ D \;=\; \begin{vmatrix} 3F_{\mathrm P} - 2kl & -\, kl \\[2pt] -\, kl & 3F_{\mathrm P} - 2kl \end{vmatrix} \;=\; 0 $$
(11.2-7)

Expanding: $3F_{\mathrm P}^2 - 4klF_{\mathrm P} + (kl)^2 = 0$, giving two eigenvalues:

$$ F_{\mathrm{P1}} \;=\; \dfrac{kl}{3} \quad,\quad F_{\mathrm{P2}} \;=\; kl $$
(11.2-8)

⑤ Take the smaller · Critical load

$$ F_{\mathrm{Pcr}} \;=\; \min\{F_{\mathrm{P1}},\, F_{\mathrm{P2}}\} \;=\; \boxed{\;\dfrac{kl}{3}\;} $$
(11.2-9)
The smaller of the two eigenvalues is the actual critical load
11.2.4

Buckling mode shapes · Symmetric & antisymmetric

Mode Shapes
Question · How to find the critical load when $EI \ne \infty$?

In this example the bars are treated as rigid ($EI = \infty$), giving 2 DOF and 2 eigenvalues. If the bars are actually elastic with finite $EI$, the DOF becomes infinite — how to find the critical load?

Answer: the finite-DOF method no longer applies; one has to extend to infinite DOF — use the deflection ODE $EIy''=-M$, apply the displacement boundary conditions to get a homogeneous system, and the determinant $= 0$ gives a transcendental stability equation (see §11-3). Number of eigenvalues = number of DOF; the smallest is still taken as the critical load.

The two eigenvalues $F_{\mathrm{P1}} = kl/3$ and $F_{\mathrm{P2}} = kl$ of Ex. 11-1 correspond to two independent buckling mode shapes. Back-substitution into the homogeneous system yields the ratio $y_2 / y_1$ at the two hinges:

Antisymmetric mode $y_2/y_1 = -1$ · $F_{\mathrm P} = kl/3$

Anim. 11.2.4-a
  • The two hinges move in opposite directions — $B$ up, $C$ down (or vice versa);
  • corresponds to the smaller eigenvalue $F_{\mathrm{P1}} = kl/3$, the real critical load;
  • geometrically antisymmetric about the mid-axis.

Symmetric mode $y_2/y_1 = 1$ · $F_{\mathrm P} = kl$

Anim. 11.2.4-b
  • The two hinges move in the same direction, same amount — $B$ and $C$ both sag;
  • corresponds to the larger eigenvalue $F_{\mathrm{P2}} = kl$ — only possible when the antisymmetric mode is restrained;
  • geometrically symmetric about the mid-axis.
Basic features of multi-DOF buckling
  1. An $n$-DOF system has $n$ eigenvalues and $n$ possible buckling mode shapes;
  2. For symmetric structures under symmetric loads, the buckling mode is necessarily symmetric or antisymmetric — exploit this to simplify analysis;
  3. The real critical load corresponds to the smallest eigenvalue — only its associated mode actually occurs.
11.2.5

Energy method · Stationary-potential principle

Energy Method

The energy method starts from the stationary-potential principle: a system is in equilibrium if and only if, for any admissible displacement / deformation, the total potential $E_{\mathrm P}$ is stationary, i.e. the first variation vanishes:

$$ \delta E_{\mathrm P} \;=\; 0$$
(11-2)

The animation below shows the stationary-potential principle intuitively — a perturbed system remains in equilibrium only if the first variation of its potential is zero:

Anim. 11.2.5-a

The sign of the second variation $\delta^{2} E_{\mathrm P}$ near the stationary point classifies the stability. Intuitively, the three equilibria of a ball on a surface capture this:

● Stable

Bowl: $\delta^{2} E_{\mathrm P} > 0$

Potential is at a local minimum. Any small perturbation raises the potential and the system returns.

Anim. 11.2.5-b
◆ Critical (neutral)

Flat: $\delta^{2} E_{\mathrm P} = 0$

Potential does not change to second order — the ball can rest anywhere; boundary between stable and unstable.

Anim. 11.2.5-c
▲ Unstable

Dome: $\delta^{2} E_{\mathrm P} < 0$

Potential is at a local maximum. Any small perturbation lowers the potential; the system accelerates away (no animation here, text only).

Total potential of a deformable system · Two parts

For a deformable system, the total potential consists of strain energy $U$ and load potential $U_{\mathrm P}$:

$$ E_{\mathrm P} \;=\; U \;+\; U_{\mathrm P}$$
(11-3)
$U$ = internal-force (strain) energy; $U_{\mathrm P}$ = external-load potential (typically taken as zero before buckling)

Two equivalent ways to assess stability via the potential: (i) stationary potential $\delta E_{\mathrm P} = 0$ with a non-trivial displacement — i.e. an eigenvalue problem; (ii) second variation vanishing $\delta^{2} E_{\mathrm P} = 0$ — the critical-state criterion. Both are equivalent; the first is used in practice.

11.2.6

Energy method · Single-DOF derivation

Single-DOF Derivation

Using the single-DOF rigid column of the figure, we re-derive $F_{\mathrm{Pcr}}$ by the energy method and compare with the static-method result. Before buckling the total energy is zero; we look at the energy change due to rotation $\theta$ at $A$.

Strain energy

$U = \tfrac{1}{2}\,k_\theta\, \theta^{2}$

Strain energy of the elastic hinge at $A$ rotating by $\theta$:

$$ U \;=\; \tfrac{1}{2}\, k_\theta\, \theta^{2} \;=\; \tfrac{1}{2}\cdot \tfrac{3EI}{l}\cdot \theta^{2} $$
(11.2-10)
$k_\theta = 3EI/l$ is the rotational stiffness at $A$
Anim. 11.2.6-a
Load potential

$U_{\mathrm P} = -\, F_{\mathrm P}\, \Delta_{yB}$

The loading point $B$ drops vertically by $\Delta_{yB}$; the load does positive work, so the potential is negative:

$$ \Delta_{yB} \;=\; l\,(1 - \cos\theta) \;\approx\; \tfrac{l\,\theta^{2}}{2} $$
(11.2-11)
using the small-angle approximation $1 - \cos\theta \approx \theta^{2}/2$
$$ U_{\mathrm P} \;=\; -\, F_{\mathrm P}\cdot \tfrac{l\,\theta^{2}}{2} $$
(11.2-12)
Anim. 11.2.6-b

Stationary potential → critical load

Total potential:

$$ E_{\mathrm P} \;=\; U + U_{\mathrm P} \;=\; \tfrac{1}{2}\left( \tfrac{3EI}{l} - F_{\mathrm P}\, l \right)\theta^{2} $$
(11.2-13)

Stationary condition $\mathrm d E_{\mathrm P}/\mathrm d \theta = 0$:

$$ \left( \tfrac{3EI}{l} - F_{\mathrm P}\, l \right)\theta \;=\; 0 $$
(11.2-14)
$\theta = 0$ is trivial; setting the coefficient $= 0$ gives the non-trivial condition
$$ F_{\mathrm{Pcr}} \;=\; \dfrac{3EI}{l^{2}} $$
(11.2-15)
Identical to the static-method result in 11.2.2 ✓
Equivalence of the two methods

The stationary condition $\mathrm d E_{\mathrm P}/\mathrm d \theta = 0$ and the equilibrium equation $F_{\mathrm P} l - k_\theta \theta = 0$ are essentially the same homogeneous equation — because equilibrium is stationary potential. The second variation $\mathrm d^{2} E_{\mathrm P}/\mathrm d \theta^{2} = 3EI/l - F_{\mathrm P} l$ vanishes exactly at $F_{\mathrm P} = F_{\mathrm{Pcr}}$, giving the critical state.

11.2.7

Energy method · Generalization to n DOF

Multi-DOF Generalization

For an $n$-DOF system, let $a_1, a_2, \ldots, a_n$ be $n$ independent generalized coordinates. The total potential after buckling is a function of these coordinates:

$$ E_{\mathrm P} \;=\; E_{\mathrm P}\!\left( a_{1},\, a_{2},\, \ldots,\, a_{n} \right) $$
(11.2-16)

① Stationary-potential condition

From $\delta E_{\mathrm P} = 0$ and the arbitrariness of $\delta a_1, \ldots, \delta a_n$, we require all first-order partial derivatives of $E_{\mathrm P}$ with respect to $a_i$ to vanish:

$$ \dfrac{\partial E_{\mathrm P}}{\partial a_{i}} \;=\; 0 \qquad (i = 1, 2, \ldots, n)$$
(11-5)

② Homogeneous linear system

Since $E_{\mathrm P}$ is a quadratic form in $a_i$, its first derivatives are linear in $a_i$, so (11-5) is a homogeneous linear algebraic system in $a_1, \ldots, a_n$. $a_1 = \cdots = a_n = 0$ is the original equilibrium.

③ Stability equation · Eigenvalues

Buckling requires a non-trivial solution, so the coefficient determinant must vanish:

$$ D \;=\; 0 $$
(11.2-17)
same form as Eq. (11-1) from the static method

$D = 0$ gives $n$ eigenvalues — $n$ possible critical loads. The smallest is the actual critical load, and the corresponding $a_i$ ratios give the buckling mode.

Static and energy methods meet at the same answer

Whether starting from equilibrium in a perturbed configuration (static method) or from stationary potential with respect to generalized coordinates (energy method), we reach the same homogeneous system, the same characteristic equation $D = 0$, and the same eigenvalues. This is because equilibrium is stationary potential — the two methods are mathematically equivalent.

11.2.8

Example 11-2 · by the energy method

Worked Example · Energy

Re-solve Ex. 11-1 (see Animation 11.2.8-a) by the energy method and compare with the static-method result $F_{\mathrm{Pcr}} = kl/3$.

① Geometry and animation

The buckling configuration is still described by two parameters $y_1, y_2$. The rigid bars store no energy; all strain energy is in the elastic supports at $B$ and $C$.

Anim. 11.2.8-a

② Strain energy U

Sum of strain energies in the two elastic supports:

$$ U \;=\; \tfrac{k}{2}\,\bigl( y_{1}^{2} + y_{2}^{2} \bigr) $$
(11.2-18)

③ Load potential · Horizontal displacement at D

Using (11-4) approximation $l(1-\cos\theta) \approx l\theta^{2}/2$ and summing over the rigid segments gives the horizontal displacement at $D$:

$$ \Delta_{xD} \;=\; \tfrac{1}{l}\,\bigl( y_{1}^{2} - y_{1} y_{2} + y_{2}^{2} \bigr) $$
(11.2-19)
$$ U_{\mathrm P} \;=\; -\, F_{\mathrm P}\, \Delta_{xD} \;=\; -\, \tfrac{F_{\mathrm P}}{l}\,\bigl( y_{1}^{2} - y_{1} y_{2} + y_{2}^{2} \bigr) $$
(11.2-20)

④ Total potential · Stationary condition

$$ E_{\mathrm P} \;=\; U + U_{\mathrm P} \;=\; \tfrac{k}{2}\,(y_{1}^{2} + y_{2}^{2})\; -\; \tfrac{F_{\mathrm P}}{l}\,(y_{1}^{2} - y_{1} y_{2} + y_{2}^{2}) $$
(11.2-21)

Setting $\partial E_{\mathrm P}/\partial y_1 = 0$ and $\partial E_{\mathrm P}/\partial y_2 = 0$ and rearranging:

$$ \left\{\begin{aligned} (k l - 2 F_{\mathrm P})\, y_1 \;+\; F_{\mathrm P}\, y_2 &= 0 \\[2pt] F_{\mathrm P}\, y_1 \;+\; (k l - 2 F_{\mathrm P})\, y_2 &= 0 \end{aligned}\right. $$
(11.2-22)

⑤ Characteristic equation · Result

Vanishing of the determinant:

$$ \begin{vmatrix} kl - 2F_{\mathrm P} & F_{\mathrm P} \\[2pt] F_{\mathrm P} & kl - 2F_{\mathrm P} \end{vmatrix} \;=\; 0 $$
(11.2-23)

Expanding: $3 F_{\mathrm P}^{2} - 4 kl\, F_{\mathrm P} + (kl)^{2} = 0$ — exactly the same characteristic equation as Ex. 11-1 by the static method.

$$ F_{\mathrm{Pcr}} \;=\; \dfrac{kl}{3} \qquad \checkmark $$
(11.2-24)
Energy and static methods give the same critical load
11.2.9

Comparison · Section summary

Summary

With the finite-DOF system as vehicle, this section has developed two solution paths — the static method and the energy method — and verified via a single example (Animation 11.2.9-a, two DOF) that both yield the same characteristic equation and the same critical load, exposing their deep equivalence.

Static method Static Method

Anim. 11.2.9-a
  • Starting point: assume a new displacement near the original equilibrium and write equilibrium equations;
  • Solution condition: homogeneous-system determinant $D = 0$;
  • Physical intuition: mechanical equilibrium — applied force = elastic restoring force;
  • Great for bar / hinge models by hand.

Energy method Energy Method

Anim. 11.2.9-b
  • Starting point: write the total potential $E_{\mathrm P} = U + U_{\mathrm P}$;
  • Solution condition: stationary potential $\partial E_{\mathrm P}/\partial a_i = 0$;
  • Physical intuition: energy balance — strain energy = work done by loads;
  • Great for complex systems / approximate solutions (basis of Rayleigh / Ritz in §11-4).
AspectStatic methodEnergy method
Basic principle Equilibrium in the perturbed configuration Stationary-potential principle $\delta E_{\mathrm P} = 0$
Equation form Homogeneous equilibrium system $\partial E_{\mathrm P}/\partial a_i = 0$ — homogeneous system
Non-trivial condition Coefficient determinant $D = 0$ (same characteristic equation)
Choice of critical load The smallest of the $n$ eigenvalues
Mode shape Back-substitute eigenvalue to obtain the $a_i$ ratios
Generalization Deflection ODE (§11-3 infinite DOF) Rayleigh / Ritz approximations (§11-4)
Section summary
  1. Finite DOF: simplify an elastic column to "rigid bars + elastic hinges / supports" with a finite buckling DOF $n$;
  2. Static method for $F_{\mathrm{Pcr}}$: write a homogeneous equilibrium system in the new configuration; $D = 0$ gives the eigenvalues;
  3. Energy method for $F_{\mathrm{Pcr}}$: write $E_{\mathrm P} = U + U_{\mathrm P}$ and $\partial E_{\mathrm P}/\partial a_i = 0$ — the same characteristic equation;
  4. $n$ eigenvalues ↔ $n$ mode shapes — for symmetric structures modes must be symmetric or antisymmetric; the real critical load is the smallest eigenvalue;
  5. Next, §11-3 extends the static method to infinite DOF via the deflection ODE.