• Introductory example

- cylindrical plate buckling

• Deriving the Equation of loss of

stability
• Some classical cases
- simply supported rectangular plate under
one-side compression
- simply supported rectangular plate under
in-plane bending and compression
- shear buckling of a rectangular plate

• FEA linear buckling example

• FEA post-buckling example
Chapter 8. Buckling of Plate Elements
(textbook Chai & Sung)
Lecturer
Djebar Baroudi, Dr. Eng.
Civil Engineering Department
Aalto University version
26.3.2023
 Examples of common structures

Many structures or structural elements have plate

elements or substructures that under certain loading
conditions can buckle

Plate buckling
 This course textbook Must read classics
Reading
assignment e-book

This week
Chapter 8. Buckling of Plate
Elements

Next week
Chapter 9. Buckling of
Thin Cylindrical
Elements
Homework #5
This is practically the Bryan’s form
Weeks #3-4 – Lectures series

Using
1) energy and
2) differential approaches Readings: Chapter 8. Buckling of Plate
Elements
Chai & Sung
 Plate shear
Week #4 – Lecture series buckling

• Introductory example
➢ Cylindrical plate buckling
• Deriving the Equation of loss of stability
• Some classical cases
➢ simply supported rectangular plate
under one-side compression
➢ simply supported rectangular plate
under in-plane bending and
compression
➢ Shear buckling of a rectangular
plate
• FEA linear buckling example
• FEA post-buckling example

Plate buckling
Stability challenge

What is the mechanisms

Buckling modes of a sheet paper involved in such buckling?
Djebar’s room h. 299 – 2/2/2023
Recall of two important points missed from last
lectures
… in the work of initial
stresses during buckling
Ref. Bazant
 Strain increments between initial pre-
buckled and post-buckled state

Notation by Novozhilov

Strain energy

Notation from our course

e = Linear part of strains
= quadratic part of strains
= full strains = finite strain

Our course notation:

 Plate buckling
(ala-laipan lommahdus)

N.B. Plate buckling starts

Clamped cantilever
(ala-laipan lommahdus)

N.B. Here, we did elastic post-

buckling analysis as it is the
topic of this course.
Plate buckling The next analysis step is to add
Waking-up ... for me too possibility for yield of the material
(GMNA) postuckling analys.
Recall about GNA and GMNA
a
t
e
r
i
a
l

• Here, we do Geometrically Non-linear

Elastic post-buckling Analysis (GNA) as it
is the topic of this course

• The next analysis step to do is to add

possibility for material non-linearities,
yield or other types of failure of the
Material. Such postuckling analys is
known as GMNA .
Waking-up Example from my research work
experiment

lateral deflection

modelling

axial shortening
Stiffened plates By using stiffeners, the buckling resistance of a plate can be
considerably increased.
• Local / global buckling
• Design for simultaneous local ang global buckling lead to
economic design
• For some structures it may be required the local buckling to
occur first; so what will be the needed minimal rigidities of the
stiffeners to achieve that?
Energy principles to estimate buckling load ... example coming soon....

before solving that, let's start from the begining

2D versus 1-D:
Plate model
versus beam-
model for
torsional
buckling
 Pure torsional buckling and plate buckling

The primary mechanism the rotational mode of the thin-walled beam open cross section is be
the result of plate buckling.
Pure torsional buckling and plate buckling
2D versus 1-D: Plate model versus beam-model for torsional buckling
Beam model

Will be derived on next slide

(good approximation of exact solution by Timoshenko)

buckling stress in pure torsional

buckling for an angle cross-section
STA TS
Shear buckling of thin plates
Levyn leikkauslommahdus

2021 - Rhône (Lyon)

Photo: D. Baroudi
 Curvatures: Plane stress constitutive law

Also called
warping
 Strains:
Transversal
pressure

Curvatures:

Also called
warping
 Put p = 0 for buckling problems
equilibrium Transversal
pressure

In the following, we will derive such equations

via energy principles
• Without using the free-body diagram (no need to open the Pandora box, see previous slide)
• We obtain the linear buckling equations (p = 0)
• And consequently, a direct mean for finding best approximations for buckling stresses and modes
(Orthotropic plate)

(Isotropic plate)
Initially flat paper sheet
buckles under in-plane
pushing

Cylindrical plate
buckling
Analogous to Euler column
with unit width a
Introductory example
Cylindrical plate buckling

You and I will derive these equation very

easily

Analogous to Euler column

with unit width a
Cylindrical-type buckling of plate slab
This stability equation will be derived

By analogy with Euler-buckling:

Usually, slenderness is defined

as the L/h compare to column
L/r, where r – gyration radius

= ‘1 / slenderness’

What is beautiful is that we now

discovered this complex looking formula
just by understanding the analogy with
Euler column buckling. Such analogy is
not visible and remains hidden to
‘lookers’ not having the needed
background in structural mechanics
Plate buckling: Euler column buckling:

Effect of boundary
conditions
Buckling coefficient
Lommahduskerroin

The buckling coefficient K depends on = ‘1 / slenderness’

• Loading
• boundary conditions and
• the aspect ratio length to width
(for rectangular plates)
• shape of the plate
 ( loss of )

From where these equations came?

Can be loading
also edge shear
Pre-buckled intial (membrane = in-
and
configuration plane) stresses

Huom.! Here, this is not the Laplace

Post-buckled
operator, it is just a difference = increment
configuration
At buckling, these u and v do not change and are approximately the same u0 and v0 as in the prebuckled
state (when deriving linear buckling equations we do not go far from the critical equilibrium point)
Buckling of plate

Quadratic part of strain increment (at

buckling)

At buckling, the derivatives of displacement components of the

primary equilibrium are small (can be ignored)
Recall column buckling where u,x was ignored since it doesn’t change during
buckling (in the neighborhood of critical point)
 Torsion modes
Bending modes
Bending rigidity
 Initial (membrane) stress
resultants satisfy
equilibrium equation in
pre-buckled state

Refer to you course: change in u

Plates and Shells
 Initial (membrane) stress
resultants satisfy
equilibrium equation in
pre-buckled state

Non-linear (quadratic) part of strain increment:

Refer to you course:
Plates and Shells

Linear part of strain increment:

Initial membrane stresses work
with their conjugate NL-strain
increments (they are work-
conjugate)
 This part provides
the mechanical
boundary
conditions

Equations
of loss of
stability:
The membrane
stresses have to
equilibrium
equations in the
pre-buckled state
“I think that you should be more explicit here in step two.”
Solving stability problems
We use intensively this approach

1) using energy method: stationarity provides buckling load and modes

Two methods beside the

experimental approach

2) Solving the differential equation of loss of stability (= linear eigenvalue problem)

Energy principle to estimate buckling
loads with examples
1.

2. 2-D versus 1-D: Plate model

versus beam-model for
torsional buckling

3.
Engineers in the class:
What is the current practice in structural design?

Remember, for steel reinforced concrete, we prefer that

the still yields first far before crashing of compressed
concrete to achieve a ductile failure of the member.
Otherwise, if the design is such that the reinforcement
fails at the same time when the compressed concrete
fail, the member failure will be practically brittle.
Energy principles to estimate buckling load

loading plane
Energy principles to estimate buckling load
Trial bucling modes

loading plane
Energy principles to estimate buckling load

Provides handles to structural

design of the stiffener
Provides handles to structural design of the stiffener
Less accurate ... but enough
good
 % Work increment of initial stresses (applied N0x along the boundaries x=0
% and x=L
% Energy method to approximate buckling load %--------------------------------
% for a stiffened thn plate with in-plane compressive Nox along one side delta_W(L, b, w0, N0x) = 0.5 * int( int(N0x * d1x_w * d1x_w, x, [0 L]) , y,
% the Poisson expansion is not restrained by the supports. [0 b] )
% ---------------------------------------------------- %% Note that here Nox is negative
% Author: Baroudi D. 2021 % ------------------------------------
% --------------------------------------------------- % Total increment of potential energy
syms x y % -----------------------------------
syms delta_P delta_W delta_Pi = delta_U(L, b, D, w0) + delta_W(L, b, w0, N0x) + delta_U_beam(L, b,
D, w0)
syms w w0 delta_Pi = simplify( delta_Pi)
syms L b D EI nu % Equations of neutral equilibrium
syms N0x % ------------------------------------
% --------------------------------------- delta_Pi_w = simplify ( diff(delta_Pi, w0) )
% Displacement approximation (you can use better approximations) % ----------------------------------------------
% ---------------------------------- texti = 'Remember Nox is now negative : = - Nox_ref'
%% w(x, y, w0, L) = w0 / ((L^2) * (b^2)) * x * ( L - x) * y * (b - y) %
less good than the trigonometric
w(x, y, w0, L) = w0 * sin(pi* x /L) * sin(pi* y /b) % best one

d1x_w(x, y, w0, L, b) = simplify( diff(w, x) )

d2xy_w(x, y, w0, L, b) = simplify( diff(d1x_w, y) )
d2x_w(x, y, w0, L, b) = simplify( diff(d1x_w, x) )

d1y_w(x, y, w0, L, b) = simplify( diff(w, y) )

d2yx_w(x, y, w0, L, b) = simplify( diff(d1y_w, x) )
d2y_w(x, y, w0, L, b) = simplify( diff(d1y_w, y) )

% Strain energy (plate alone)

delta_U(L, b, D, w0) = 0.5* D * int( int(d2x_w * d2x_w, x, [0 L]) , y, [0 b]) +
...
0.5* D * int( int(d2y_w * d2y_w, x, [0 L]) , y, [0 b]) +
...
nu * D * int( int(d2x_w * d2y_w, x, [0 L]) , y, [0 b]) +
...
(1 - nu)* D * int( int(d2xy_w * d2xy_w, x, [0 L]) , y, [0 b])

% Strain energy (stiffner beam alone)

d2x_w_beam = d2x_w(x, b/2, w0, L, b)
%% delta_U_beam(L, b, D, w0) = 0.5* EI * int( int(d2x_w_beam * d2x_w_beam, x,
[0 L]) , y, [0 b])
delta_U_beam(L, b, D, w0) = 0.5* EI * int( d2x_w_beam * d2x_w_beam, x, [0 L] )
2-D versus 1-D: Plate model versus beam-model for torsional buckling

Task: use stationary total potential energy

principle and estimate buckling load
2-D versus 1-D: Plate model versus beam-model for torsional buckling 1(3)
Task: use sationary pot. energy principle and estimate buckling load
Trial:

Not so good approximation

Linear one is better, see next slide

Energy-approximation:

This is a good analytical approximation:

Buckling stress for this plate

 2-D versus 1-D: Plate model versus beam-model for torsional buckling 2(3)
Task: use sationary pot. energy principle and estimate buckling load Trial:

This is a good analytical approximation:

This a better result than with the

quadratic approximation in y. model
2-D versus 1-D: Plate model versus beam-model for torsional buckling 3(3)
Task: use sationary pot. energy principle and estimate buckling load Trial:

This a better result than

with the quadratic
approximation in y.
There is a reason.

This is a good analytical approximation:

Reation between plate bucklng and pure torsional buckling of thin plates

Why we obtained a very good approximation?

A better approximation
in y-direction

Without
stiffener

Contribution of
the stiffener
% Energy method to approximate buckling load
% for a stiffened thn plate with at y=b (free end) in-plane % Strain energy (plate alone)
compressive Nox along one side delta_U(L, b, D, w0) = 0.5* D * int( int(d2x_w * d2x_w, x, [0
% the Poisson expansion is not restrained by the supports. L]) , y, [0 b]) + ...
% ---------------------------------------------------- 0.5* D * int( int(d2y_w * d2y_w, x, [0
% Author: Baroudi D. 2021 The Matlab code that produced L]) , y, [0 b]) + ...
the previous result
% --------------------------------------------------- nu * D * int( int(d2x_w * d2y_w, x, [0
clear all L]) , y, [0 b]) + ...
(1 - nu)* D * int( int(d2xy_w * d2xy_w, x, [0
syms x y L]) , y, [0 b])
syms delta_P delta_W
% Strain energy (stiffner beam alone)
syms w w0 ystiff = b;
syms L b D EI nu d2x_w_beam = d2x_w(x, ystiff, w0, L, b)
syms N0x delta_U_beam(L, b, D, w0) = 0.5* EI * int( d2x_w_beam *
syms n m d2x_w_beam, x, [0 L] )

% --------------------------------------- % Work increment of initial stresses (applied N0x along the

% Displacement approximation (you can use better boundaries x=0
approximations) % and x=L
% ---------------------------------- %--------------------------------
%% w(x, y, w0, L) = w0 / ((L^2) * (b^2)) * x * ( L - x) * delta_W(L, b, w0, N0x) = 0.5 * int( int(N0x * d1x_w * d1x_w, x,
y * (b - y) % less good than the trigonometric [0 L]) , y, [0 b] )
%% w(x, y, w0, L) = w0 * sin(pi* x /L) * sin(pi* y /b) % %% Note that here Nox is negative
best one % ------------------------------------
w(x, y, w0, L) = w0 * sin(pi* x /L) * (y*y) / (L*L) % Total increment of potential energy
% -----------------------------------
or w(x, y, w0, L) = w0 * (y / L) * sin(pi* x /L)
d1x_w(x, y, w0, L, b) = simplify( diff(w, x) ) delta_Pi = delta_U(L, b, D, w0) + delta_W(L, b, w0, N0x) +
d2xy_w(x, y, w0, L, b) = simplify( diff(d1x_w, y) ) delta_U_beam(L, b, D, w0);
d2x_w(x, y, w0, L, b) = simplify( diff(d1x_w, x) ) delta_Pi = simplify( delta_Pi)
% Equations of neutral equilibrium
% ------------------------------------
d1y_w(x, y, w0, L, b) = simplify( diff(w, y) ) delta_Pi_w = simplify ( diff(delta_Pi, w0) )
d2yx_w(x, y, w0, L, b) = simplify( diff(d1y_w, x) ) % ----------------------------------------------
d2y_w(x, y, w0, L, b) = simplify( diff(d1y_w, y) ) texti = 'Remember Nox is now negative : = - Nox_ref'
 Some classical
analytical solutions
of
the
partial differential
equations of buckling
 Can move
simplifies

Can move

We look for a solution in this form

 N.B. Stability loss Criteria: here
it is asked from the trial to
fulfill the eigenvalue problem

Buckling stress-resultant [N/m]

Buckling occurs when:

n = 1 gives the smallest

critical (the buckling) stress-
resultant
 Buckling occurs when:
n = 1 gives the smallest
critical (the buckling) stress-
resultant
Buckling stress-resultant [N/m]

Buckling coefficient

take the closest integer m

Buckling stress-
resultant [N/m] Let’s make a graph to see how many buckles
(half-waves) we have depending on a/b
 Buckling coefficient
Buckling stress-resultant [N/m]

Buckling coefficient
m number of half-waves

Buckling stress-resultant [N/m]

Buckling stress-
resultant [N/m]
(alin verhokäyrä)

Let’s make a graph to see how

many buckles (m half-waves) we
have depending on a/b

At the limit kc = 4
 Linear Buckling Analysis
Buckling coefficient
Lommahduskerroin

Buckling stress-resultant [N/m]

Can be written in
this canonical form
Buckling coefficients forsome boundary conditions and axial edge load
Reaction stress
resultant:
Buckling stress reduces to 3/4
No restraints:
Bi-axial loading – the idea We just present the idea. If time remains, we may provide an example using
virtual work principle to construct the buckling interaction diagrams for non-
proportional bi-axial loading like we did with columns

Buckling interaction diagram

Biaxial compression – buckling interaction An example using stationarity of the total potential energy
to construct the buckling interaction diagrams for non-
diagram proportional bi-axial compression
For combined biaxial compression and tension, in principle, we obtain similarly as above the interaction
diagrams. However, we probably should take mode trials having hyperbolic functions or similar for tension. It

Buckling interaction diagram came out trough further analysis that the critical mode for equal biaxial tension and compression is two-half
sine waves in the compression direction and half a sine wave in tension…
Biaxial loading – buckling interaction diagram

Combined
compression
and tension

The diagram is analytically

solved by energy method using
K = 8.36
the modes displayed of the (Abaqus)
figure 7

-3/7
Biaxial loading – buckling interaction diagram . . . continues

Thanks to Dr. J. Jaaranen

for this Abaqus
simulation.
Combined biaxial
compression and tension -
buckling interaction diagram

K = 8.36 = -1
(Abaqus)
Reference:
K = 8.36 ; s = -1
(Abaqus)
 Local buckling of plates

A
P t
b
L
b

A A-A
Example from exam 11.4.2019 Solution
Mathematica script
Buckling of plates beyond the proportional limit
local buckling of plates
 local buckling of plates

concept of effective-width in local buckling resistance of plates

concept of effective-width in local buckling resistance of plates
concept of effective-width in local buckling resistance of plates
Engineering practice

(all experiments: simply

supported plates)

concept of effective-width in local buckling resistance of plates

 This problem can be found in Timoshenko's textbook

Boundary conditions:
• simplification: insulated panel a x b simply supported at all 4 edges
1. This is an approximation leading to a lower-bound for the critical buckling load
2. In reality, the upper and lower edge-connections of the web-plate to the flanges correspond to those of
rotational springs due to their rotational rigidity while the web buckles. This type of boundary condition
is not impossible to address even theoretically.

Trial solution:
 n – number of half-wave s in x-direction Example:
m – number of half-wave s in y-direction

N.B. Stability loss Criteria has

now the form:
 – number of half-wave s in y-direction
Buckling stress

Lommahdus kerroin
k - Buckling coefficient depends of
• ration a/b,
• boundary conditions and
• loading case
 Lommahduskerroin

Buckling stress
Buckling of simply supported thin rectangular
plate in pure bending

Buckling stress

Ref. Timoshenko & Gere

NB. About 6 times higher buckling load (buckling strength) than

for compresses rectangular plate
Length to width ratio Reference:
• Similar tables are provided our-days for practical design in standards related to structural design of metallic
structures
• For instance, the standard EN-1993-1-5, Table 6 (2006), provides similar buckling coefficients k tables for
combined compression and bending of thin plates for various boundary and loading conditions

I still admire the clarity and non-ambiguity of this table from old B7
 F
Physics of the
phenomena

F
Principle stresses under edge sharing
Buckling of a long plate under shear

Trial mode
(ref: Timoshenko):
Buckling of a long plate under shear
Exact solution is available only
Trial mode
for an infinitely long strip (Brush & Almroth
(ref: Timoshenko) (1975)).
(assumes simply supported all sides):

(ref: Timoshenko)
cr

Buckling load is therefore

smallest when

Minimising with for

respect to c we obtain
the approximation:

Taking a better trial (fulfils some additional mechanical boundary conditions… not trivial to find, we obtain a sharper approximate:
 For infinitely long strip:

Exact solution is available only

for an infinitely long strip (Brush & Almroth (1975)).
For finite strips:
 Ref: Emir prof. Markku Tuomala

Parabolic fit

Approx. analytical with

truncated series

Exact solution is available only For infinitely long strip:

for an infinitely long strip (Brush & Almroth (1975)).

For finite strips:

 For infinitely long strip:

For finite strips:

However, from where comes

this PRACTICAL design
formula?
Four edges simply
supported Analytical (numerical)
theoretical
The graph is adapted from:
A parabolic fit
(given in Timoshenko)
Good to know
Linear Buckling Analysis 3D-FE model

1 ….. 70 MPa

1.1 ….. 77 MPa

thickness = 10 mm

Simply supported rectangular plate

1.5 ….. 105 MPa

2.1 ….. 147 MPa

Buckling stresses and modes

Post-Buckling Analysis

Remains in initial flat state 1st bifurcation 2nd bifurcation 3rd bifurcation
was not
observed in
simulation
(unstable mode)
Because I used
force-control
solver… one
should switch to
displacement
control to catch
the unstable
branch

Simply supported
rectangular plate Displacements are multiplied by a scale factor which varies
to make the deflections visible, especially, the first buckling
 2nd bifurcation (unstable)
Post-Buckling Analysis was not observed in simulation Voima-ohjattuna

3rd bifurcation

Unstable
2nd bifurcation
(unstable)
1st bifurcation
Stable

Remains in initial Remains in initial

flat state flat state
 Computational Post-Buckling
Analysis
How to do?
• How to give easily (lazily) initial
tiny displacement shape as a
perturbation? Usually, a combination
of Eigen-modes is used. A random
combination can be also used.

• One idea is to use equivalent tiny

combination of loads resulting in
equivalent deformations

Generating a tiny
perturbation of the primary
equilibrium configuration

Voima-ohjattuna
Enjoy for coming spring

Photo: Djebar Baroudi, 2021

Lyon, 6/4/2021

The story ends here ... in

the beginning was water …
 Appendix
&
Miscellaneous

In a bit disorder now … will be updated

Lateral-torsional buckling by Rayleigh-Ritz

Exact analytical solution:

Criticality means that the determinant

vanishes. The roots of the quadratic
polynomial in P gives the critical loads.
The smallest one is the buckling load.
A 0
= a)
B 0

b)

c)
 y- n – number of half-
wave s in y-
direction

Plate buckling
Timoshenko;
history of
strength of
materials.