Reijo Kouhia Pekka Marjamäki
Example 1.1 Determineallequilibriumpathsofthe stru ture onsistingoftworigidbarsand
a linearelasti rotationalspring. Investigate alsothe stabilityof all paths.
P = λ4k/L
.b
b
b
@@
b
@@
b
P
k
L/2
L/2
Solution: Let's assumethat the barsdispla e by anangle
ϕ
,then inthe middlepin the angle willbe 2ϕ
.b
b````
```
`
b
b
@
@
b b
@
@
P
k
ϕ
2
ϕ
The total potentialenergy
Π
of the stru ture is thus:Π =
1
2
k(2ϕ)
2
− P L(1 − cos ϕ)
(1)∂Π
∂ϕ
= 4kϕ − P L sin ϕ
(2)∂
2
Π
∂ϕ
2
= 4k − P L cos ϕ.
(3)The stru turewillbeinequilibriumwhen thetotal potentialenergy attainsitsminimum,thus
the rst variationof the TPE willvanish.
δΠ =
∂Π
∂ϕ
δϕ = 0 ∀ δϕ 6= 0 ⇒
∂Π
∂ϕ
= 0
(4)⇒
ϕ =
0
primary pathP =
4k
L
ϕ
sin ϕ
sek ondary path(5)
Let us rst investigate the primary path. A point on an equilibrium path is stable, if a small
hange(disturban e) inthe equilibriumpositionwillwillin rease the valueof
Π
.Sin etherst variation is zero on an equilibrium path, then the se ond variationwill determine the hangein the TPE. Sin e now
ϕ = 0
,δ
2
Π =
∂
2
Π
∂ϕ
2
(δϕ)
2
= 0
⇒
∂
2
Π
∂ϕ
2
= 4k − P L cos ϕ = 4k − P L
⇒ P
kr
= 4
k
L
(6)The primary equilibriumpath is thusstable up to the point
(ϕ = 0, P
kr
)
. Next, the stabilityproperties of the se ondary pathis investigated.∂
2
Π
∂ϕ
2
= 4k − P L cos ϕ
(7)Insering the equation of the se ondary path
P = 4kϕ/L sin ϕ
to the equation above,gives∂
2
Π
∂ϕ
2
P
II
= 4k
1 −
ϕ
tan ϕ
> 0 ∀ϕ
(8)The se ondary path is this stable for all values of
ϕ
, ex ept the bifur ation point whereϕ =
0, P
kr
= 4k/L
, and the se ond variation ofΠ
iszero.The equilibriumpaths are shown inthe
λ − ϕ
- oordinatesystem in the gure below.0
0.2
0.4
0.6
0.8
1
1.2
-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
λ
ϕ
a linear elasti rotational spring. Investigate also the stability of all paths. Are there riti al
points onthe paths?
b
b````
```
`
b
b
@@
b
@@
b
P
k
ϕ
0
cos ϕ
0
L/2
cos ϕ
0
L/2
Solution: The total potentialenergy expression isnow
Π =
1
2
k[2(ϕ − ϕ
0
)]
2
− P L(cos ϕ
0
− cos ϕ)
∂Π
∂ϕ
= 4k(ϕ − ϕ
0
) − P L sin ϕ
∂
2
Π
∂ϕ
2
= 4k − P L cos ϕ
The stru turewillbeinequilibriumwhen thetotal potentialenergy attainsitsminimum,thus
the rst variationof the TPE willvanish.
δΠ =
∂Π
∂ϕ
δϕ = 0 ∀ δϕ
⇒
∂Π
∂ϕ
= 0
⇒ P =
4k(ϕ − ϕ
L sin ϕ
0
)
An equilibriumstate is stableif the se ondvariationof the TPE ispositive
δ
2
Π =
∂
2
Π
∂ϕ
2
(δϕ)
2
> 0
⇒
∂
2
Π
∂ϕ
2
> 0
⇒ P <
L cos ϕ
4k
.
Inserting the equilibrium equation
P = 4k(ϕ − ϕ
0
)/L sin ϕ
in the expression above, gives the ondition for stability4k
1 −
ϕ − ϕ
tan ϕ
0
> 0
⇒
ϕ − ϕ
tan ϕ
0
< 1,
whi hisvalidforallnon-negativevaluesof
ϕ
.Thusthis equilibriumpathdoesnot have riti al points.0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
-2
-1.5
-1
-0.5
0
0.5
1
1.5
2
λ
ϕ
Inthegureabove,dottedlineshowstheequilibriumpathoftheperfe tstru ture
ϕ
0
= 0
, and solid line indi atesthe stable path whenϕ
0
> 0
. The path in the negative part ofϕ
axis shown by asolid lineis a omplementary path. The load parameterλ
isdened asλ = P/P
kr
= P L/(4k).
(9)Noti e, that the omplementary path is not stable everywhere. Determine the unstable
and stable parths of the omplementary path! Note too, that this means an existen e of a
onsisting of two rigid bars (length
L/2
)and a linear elasti rotationalspring. Investigate also thestabilityof allpaths.TheperturbationloadF = ǫ4k/L
,whereǫ
isadimensionless(se ond) perturbation parameter.b
b````
```
`
b
b
@@
b
@@
b
P
k
ϕ
0
?
F
cos ϕ
0
L/2
cos ϕ
0
L/2
Solution: The total potentialenergy of the stru ture
Π
isΠ(ϕ; ϕ
0
, ǫ) =
1
2
k[2(ϕ − ϕ
0
)]
2
− P L(cos ϕ
0
− cos ϕ) −
1
2
F L(sin ϕ
0
− sin ϕ)
(10) A ne essary ondition of an equilibrium state is the stationarity of the TPE, thus the rstvariation of the total potentialenergy must vanish
δΠ =
dΠ
dϕ
δϕ =
4k(ϕ − ϕ
0
) − P L sin ϕ +
1
2
F L cos ϕ
δϕ = 0
∀ δϕ 6= 0
(11)An equilibriumpath is thusdened by
P = 4
k
L
ϕ − ϕ
0
+
1
2
ǫ cos ϕ
sin ϕ
(12)This equationdetermines aunique path with respe t to
ϕ
if the perturbationparameters does not satisfy the onditionǫ = 2ϕ
0
. In su h a ase the stru ture is a straight bar of lengthL
at the unloaded stateP = 0
.Let usexamine this spe ial ase rst.Case
ǫ = 2ϕ
0
The equilibriumequation isnowdΠ
dϕ
= (ϕ − ϕ
0
) − P L sin ϕ + 4kϕ
0
cos ϕ
(13)= 4kϕ − P L sin ϕ + 4kϕ
0
(cos ϕ − 1) = 0
(14)and the two solutionsare
ϕ = 0
primary pathP
I
,
(15)P = 4
k
L
ϕ + ϕ
0
(cos ϕ − 1)
sin ϕ
se ondary pathP
II
(16)An equilibriumstate is stableif the se ondvariationof
Π
:δ
2
Π =
d
2
Π
dϕ
2
(δϕ)
2
is positive.Let usrst examinestability of the primary path, i.e.when
ϕ = 0
,thusδ
2
Π|
P
=
d
2
Π
dϕ
2
P
(δϕ)
2
= (4k − P L)(δϕ)
2
(18)The primary path is thus stable when
P < 4k/L
and unstable whenP > 4k/L
, and the riti al load is thusP
cr
= 4k/L
. Let us examine wheather the riti al point is a symmetri or asymmetri bifur ationpoint.The expression of the third variation of the TPE isδ
3
Π|
P
=
d
3
Π
dϕ
3
P
(δϕ)
3
(19) whered
3
Π
dϕ
3
= −P L sin ϕ − 4kϕ
0
cos ϕ
(20) Atthe riti alpoint the valueof the third derivatve of the TPE is ond
3
Π
dϕ
3
kr
= −4kϕ
0
6= 0
(21)thus the riti al point is an asymmetri bifur ation point. The equilibrium path is drawn in
gure 1.
Case
ǫ 6= 2ϕ
0
Let us examine stability of the equilibrium path, dened in (12). The se ond variation of the TPEδ
2
Π =
d
2
Π
dϕ
2
(δϕ)
2
(22)
is obtained from the expression of the rst variation (11). An equilibrium state is stable if
the se ond variationof the TPE ispositivefor allkinemati allyadmissiblevariations
δϕ
, thus in this single degree of freedom example it is su ient to investigate the sign of the se ondderivativeof the TPE
d
2
Π
dϕ
2
= 4k − P L cos ϕ − 2kǫ sin ϕ
(23) Let's insert the expression of the equilibriumpath (12) in the expression above,givesd
2
Π
dϕ
2
= 4k
sin ϕ − (ϕ − ϕ
0
) cos ϕ −
1
2
ǫ
sin ϕ
(24)Letus examine the ases
ǫ > 2ϕ
0
andǫ < 2ϕ
0
separately.In the ase
ǫ > 2ϕ
0
, the stru ture is below the horizonal line dened by the supports before applying the ompressive load, thus the stru ture will ontinue to displa e below thesupportline, thus
ϕ < 0
.Let usdeneǫ = 2ϕ
0
+ ¯ǫ
, and the expression (24) givesd
2
Π
dϕ
2
= 4k
sin ϕ − ϕ cos ϕ − ϕ
0
(1 − cos ϕ) − ¯ǫ
Sin e now
ϕ < 0
and both the nominator and denominator are negative, thusδ
Π
is always
positive, i.e.the path is stable when
ǫ > 2ϕ
0
.The ase
ǫ < 2ϕ
0
is more interesting. Nowϕ > 0
and the denominator of the expression (24) isalwayspositivebut thenominator an have zeropoints.These roots anbesolved fromthe trans endental equation
sin ϕ − (ϕ − ϕ
0
) cos ϕ −
1
2
ǫ = 0.
(26)Sin eanalyti alsolutionisimpossible,let'strytheasymptoti analysisassumingthattheangles
ϕ
andϕ
0
are small, thussin ϕ ≈ ϕ −
1
6
ϕ
3
,
cos ϕ ≈ 1 −
1
2
ϕ
2
,
and the expression (26) willhas a form
1
3
ϕ
3
−
1
2
ϕ
0
ϕ
2
+ (ϕ
0
−
1
2
ǫ) = 0
(27)Thethirdorderpolynomialabove anhavebothnegativeandpositivevaluesforpositivevalues
of
ϕ
. To show that,let ussrt al ulate the minumum pointϕ
2
− ϕ
0
ϕ = 0
=⇒
ϕ = ϕ
0
.
(28)The minimum value of the fun tion dened in (27) (kun
ϕ > 0
) and the ondition for the nagativitywe get an inequality (let's deneǫ = ηϕ
0
)−
1
3
ϕ
2
0
+ 1 −
1
2
η < 0
=⇒
η > 2 −
1
3
ϕ
2
0
Takingthe ondition
ǫ < 2ϕ
0
intoa ountwe'llget a onditionfor the perturbationparameterǫ = ηϕ
0
:2 > η > 2 −
1
3
ϕ
2
0
i.e.2ϕ
0
> ǫ > (2 −
1
3
ϕ
2
0
)ϕ
0
forthe existen e ofa limitpoint ontheequilibriumpath. Inthe followinggure,some
equilib-rium paths are shown for some values of the perturbation parameter
ǫ
To sum up, the equilibrium pathsof this stru ture an have•
atrivialequilibriumpathand anasymmetri bifur ationpointifǫ = 2ϕ
0
.These ondary path isdened inequation (16).•
A stable equilibriumpath without riti al pointsifǫ > 2ϕ
0
or ifǫ
/ (2 −
1
3
ϕ
2
0
)ϕ
0
.•
An equilibriumpath has a limitpointif(2 −
1
3
ϕ
2
0
0.5
1
1.5
2
-2
-1.5
-1
-0.5
0
0.5
1
1.5
2
λ
ϕ
ǫ = 2ϕ
0
ǫ = 1.99ϕ
0
ǫ = 1.8ϕ
0
ǫ = 2.2ϕ
0
The stru ture in the problem1 isan idealized olumnhaving a onstant bending stiness
EI
. Determine the spring oe ientk
and how the riti al load will dier from the exa t beam solution.Solution: The spring onstant
k
an be determined either by•
lettingthe bifur ationloads tobeequal forboth models,•
tomake the displa ementsat the middleequal under pointload at the middle,•
tomake the displa ementsat the middleequal under uniformload. Dee tionunder a point load isδ
P
p
=
1
48
F L
3
EI
and for a unform load
δ
p
q
=
5
384
qL
4
EI
.
Forthe spring-bar system the orresponding dee tions are
δ
P
j
=
1
8
F L
2
k
and δ
j
q
=
1
16
qL
3
k
.
Letδ
P
j
= δ
P
p
⇒ k
P
=
6EI
L
δ
j
q
= δ
q
p
⇒ k
q
=
12EI
L
The riti al load of the spring-bar system is thus
P
cr
=
4k
L
⇒ P
P
kr
=
24EI
L
2
and P
q
cr
=
48EI
L
2
.
One additionalway to ompute
k
is to make the bending strain energies equal under a uniform load.rigidbarsand elasti springs.Investigatealsothestabilityofthe equilibriumpaths.Investigate
espe ially ases
k
1
= k
2
jak
1
= 5k
2
. What kindof real stru tures these models imitate?Solution: The total potentialenergy expression is
Π = U + V
U =
1
2
k
1
L
2
sin
2
ϕ + k
2
u
2
+
1
2
k
2
[u − 2L(1 − cos ϕ)]
2
V = −P u
(29)The equilibriumpaths an beobtained from the stationarity ondition of the TPE:
δΠ =
∂Π
∂ϕ
δϕ +
∂Π
∂u
δu = 0
(30)Sin e the variationsof the displa ement
u
and rotationϕ
are arbitrary, the equilibrium paths are obtained from equations∂Π
∂ϕ
= k
1
L
2
sin ϕ cos ϕ + k
2
[u − 2L(1 − cos ϕ)](−2L sin ϕ) = 0
∂Π
∂u
= 2k
2
u + k
2
[u − 2L(1 − cos ϕ)] − P = 0
(31)After some manipulationswe get
sin ϕ[k
1
L
2
cos ϕ − 2Lk
2
u + 4k
2
L
2
(1 − cos ϕ)] = 0
(32)u =
P
3k
2
+
2
sin ϕ = 0
tai(k
1
− 4k
2
)L
2
cos ϕ + 4k
2
L
2
− 2k
2
Lu = 0
(34) Ifequation (33) is put into equation(34) and denek
2
= k
jak
1
= αk
, we get⇒
P = kL 4 + (
3
2
α − 4) cos ϕ
(35)
whi histhe proje tion ofthese ondary pathontothe
(ϕ, P )
-plane.A ordinglyfromequation (33) we getcos ϕ = 1 +
P
2kL
−
3
2
u
L
,
whi his substituted into(35)
⇒
P =
kL
4 − α
h
2α + (8 − 3α)
L
u
i
,
whi hdes ribes the proje tion of the se ondary path onto the
(u, P )
-plane. The primary paths are dened as
ϕ =
0
u =
P
3k
and the se ondary paths
P = [4 + (
3
2
α − 4) cos ϕ]kL
P =
kL
4 − α
h
2α + (8 − 3α)
L
u
i
Let's investigatethe ases
α = 1
jaα = 5
.α = 1 ⇒
P = (4 −
5
2
cos ϕ)kL
P =
1
3
kL
2 + 5
u
L
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
P
kL
ϕ
0
0.5
1
1.5
2
2.5
3
3.5
4
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
P
kL
u/L
0.333*(2+5*x)
3*x
We noti e, that displa ements are in reasing more rapidly on the se ondary path than
in the primary path. However, the load an still be in reased over the riti al value at the
bifur ation point.
(P
kr
=
3
2
kL
, thus the se ondary path is stable. In ompressed thin plates su h kind of behaviour an be obtained. The strong stability of the se ondary paths an beutilizedalso indesign for some ases.
α = 5 ⇒
P = (4 +
7
2
cos ϕ)kL
P = −kL
10 − 7
L
u
0
1
2
3
4
5
6
7
8
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
P
kL
ϕ
0
1
2
3
4
5
6
7
8
0
0.5
1
1.5
2
2.5
3
P
kL
u/L
-10+7*x
3*x
In this ase the bifur ation load is mu h higher than in the preious one. However, the
se ondary equilibrium path is now unstable. Shells, espe ially exhibit su h kind of unstable
behaviourafter bifur ation.Ifthe post-bu klingregimeis unstable, su hstru tures are
imper-fe tionsensitive,whi hmeansthatthe riti alloadofanimperfe tstru tureismu hlowerthan
the theoreti al bifur ationload. Imperfe tions are due to e entri ities, geometri aldeviations
et .
Example 1.5 Investigate the ee t of imperfe tions in the previous example. Draw the
im-perfe tion sensitivity diagram for the ase
k
1
= 5k
2
.Solution: Let's determine the riti al load asa fun tion of
ϕ
the imperfe tion amplitudeϕ
0
. NowU =
1
2
k
1
L
2
(sin ϕ − sin ϕ
0
)
2
+ k
2
u
2
+
1
2
[u − 2L(cos ϕ
0
− cos ϕ)]
2
and∂Π
∂ϕ
= k
1
L
2
(sin ϕ − sin ϕ
0
) cos ϕ + k
2
[u − 2L(cos ϕ
0
− cos ϕ)](−2L sin ϕ) = 0
∂Π
∂u
= 2k
2
u + k
2
[u − 2L(cos ϕ
0
− cos ϕ)] − P = 0
Solving
u
fromthe equation above and substitute it intothe equationbelow, givesu =
k
1
2k
2
sin ϕ − sin ϕ
0
tan ϕ
L − 2L(cos ϕ − cos ϕ
0
)
P = 3k
2
u − 2k
2
L((cos ϕ − cos ϕ
0
) =
3k
1
2
sin ϕ − sin ϕ
0
tan ϕ
L − 8k
2
L((cos ϕ − cos ϕ
0
)
Figure2: The maximum load
λ
max
asa fun tion of the imperfe tionamplitudeϕ
0
0
1
2
3
4
5
6
7
8
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
P
kL
ϕ
ǫ
0
= 0
0.003
0.03
0.1
ation pointson the paths? (
EI = EA = ∞
)c
a
a
c
c c
P k k L L LSolution: Let's determine the displa ementsby using the following gure
c!!
!!
!!
!!
!!
a
a
b
b
b
b
b
b
b
b
bb c
ϕ
0
ϕ
1
!!
ϕ
12
ϕ
2
ϕ
3
v
1
v
2
sin ϕ
0
=
v
1
L
, sin ϕ
12
=
v
2
− v
1
L
, sin ϕ
3
=
v
2
L
In additionϕ
1
= ϕ
0
− ϕ
12
= arcsin
v
1
L
− arcsin
v
2
− v
1
L
ϕ
2
= ϕ
3
+ ϕ
12
= arcsin
v
2
L
+ arcsin
v
2
− v
1
L
Assuming small rotationswe an approximate
arcsin x ≈ x + 1/6 x
3
, and the rotationsat the
springs are
ϕ
1
= δ
1
+
1
6
δ
3
1
−
δ
2
− δ
1
+
1
6
(δ
2
− δ
1
)
2
= 2δ
1
− δ
2
+
1
3
δ
3
1
−
1
6
δ
3
2
+
1
2
δ
1
δ
2
2
−
1
2
δ
2
1
δ
2
ϕ
2
= δ
2
+
1
6
δ
3
2
+
δ
2
− δ
1
+
1
6
(δ
2
− δ
1
)
2
= 2δ
2
− δ
1
+
1
3
δ
3
2
−
1
6
δ
3
1
−
1
2
δ
1
δ
2
2
+
1
2
δ
2
1
δ
2
Dee tion underthe load
∆ = L
3 −
q
1 − δ
2
1
−
p1 − (δ
1
− δ
2
)
2
−
q
1 − δ
2
2
≈ L
δ
2
1
+ δ
2
2
− δ
1
δ
2
+
1
4
δ
4
1
+
1
4
δ
4
2
−
1
2
δ
3
1
δ
2
+
3
4
δ
2
1
δ
2
2
−
1
2
δ
1
δ
3
2
Π = Π(δ
1
, δ
2
) =
1
2
kϕ
2
1
(δ
1
, δ
2
) +
1
2
kϕ
2
2
(δ
1
, δ
2
) − P ∆(δ
1
, δ
2
)
Equilibriumpaths are determined from the stationarity ondition
δΠ = 0
,whi hgives∂Π
∂δ
1
= kϕ
1
∂ϕ
1
∂δ
1
+ kϕ
2
∂ϕ
2
∂δ
1
− P
∂∆
∂δ
1
= 0
(36)∂Π
∂δ
2
= kϕ
1
∂ϕ
1
∂δ
2
+ kϕ
2
∂ϕ
2
∂δ
2
− P
∂∆
∂δ
2
= 0
(37) in whi h∂ϕ
1
∂δ
1
= 2 + δ
2
1
+
1
2
δ
2
2
− δ
1
δ
2
∂ϕ
2
∂δ
1
= −1 −
1
2
δ
2
1
−
1
2
δ
2
2
+ δ
1
δ
2
∂∆
∂δ
1
= L
2δ
1
− δ
2
+ δ
1
3
−
3
2
δ
2
1
δ
2
+
3
2
δ
1
δ
2
2
−
1
2
δ
3
2
∂ϕ
1
∂δ
2
= −1 −
1
2
δ
2
1
−
1
2
δ
2
2
+ δ
1
δ
2
∂ϕ
2
∂δ
2
= 2 + δ
2
2
− δ
1
δ
2
+
1
2
δ
2
1
∂∆
∂δ
2
= L
2δ
2
− δ
1
+ δ
2
3
−
1
2
δ
3
1
+
3
2
δ
2
1
δ
2
−
3
2
δ
1
δ
2
2
Equations (36) and (37) are satised, if
ϕ
1
= ϕ
2
= ∆ = 0
i.e.δ
1
= δ
2
= 0
. Let's investigate stabilityof this primarypath. The se ond variation of the total potentialenergy isδ
2
Π =
∂
2
Π
∂δ
2
1
[δ(δ
1
)]
2
+ 2
∂
2
Π
∂δ
1
∂δ
2
δ(δ
1
)δ(δ
2
) +
∂
2
Π
∂δ
2
2
[δ(δ
2
)]
2
= (δ(δ
1
) δ(δ
2
))
∂
2
Π
∂δ
2
1
∂
2
Π
∂δ
1
∂δ
2
∂
2
Π
∂δ
1
∂δ
2
∂
2
Π
∂δ
2
1
δ(δ
1
)
δ(δ
2
)
!
,
in whi h the matrix
K = [∂
2
Π/∂δ
i
∂δ
j
]
is the stability matrix. The path is stable, if and only ifδ
2
Π > 0
. This is true if the stability matrix
K
is positive denite, whi h means that all its eigenvalues are positive.On the primary path
δ
1
= δ
2
= 0
the elements of the stabilitymatrix are∂
2
Π
∂δ
2
1
= kϕ
1
∂
2
ϕ
1
∂δ
2
1
+ k
∂ϕ
1
∂δ
1
2
+ kϕ
2
∂
2
ϕ
2
∂δ
2
1
+ k
∂ϕ
2
∂δ
1
2
− P
∂
2
∆
∂δ
2
1
(38)∂
2
Π
∂δ
2
1
= 5k − 2P L
∂
2
Π
∂δ
1
∂δ
2
= −4k + P L
∂
2
Π
∂δ
2
2
= 5k − 2P L
Denoting
P = λk/L
and solving the eigenvalues ofK
. The path is stable, whenK
is positive denite, i.e. allitseigenvalues are positiveK ¯
x = ω ¯
x ⇒ (K − ωI)¯x = 0 ⇒ det(K − ωI) = 0
det
5 − 2λ − ω
λ − 4
λ − 4
5 − 2λ − ω
!
= 0
ω
1
= 1 − λ ja ω
2
= 9 − 3λ
The zero points of the eigenvalues o ur when
λ
have valuesλ = 1
andλ = 3
. The primary pathis stablewhenλ < 1
. Theeigenmodedes ofK
i.e.the bu klingmodesof the stru tureare obtained fromλ = 1 ⇒
3
−3
−3
3
! (
δ
1
δ
2
)
=
(
0
0
)
⇒ δ
1
= δ
2
The bu klingmode orrespondingto the riti al load
P
kr
= k/L
is:c
a
```
a
`````
`` c
c c
λ = 3 ⇒
−1
1
1
−1
! (
δ
1
δ
2
)
=
(
0
0
)
⇒ δ
1
= −δ
2
And the bu lingmode orrespondingto the riti alload
P
kr
= 3k/L
is:c
aa
aa
aa
aa
aaa a
c
c c
Let's nally investigate the post-bifur ation paths after the bran hing point at
λ = 1
. Substituting displa ementsδ
1
= δ
2
= δ ⇒ ϕ
1
= ϕ
2
intothe equation of equilibrium(36).∂Π
∂δ
1
= 0 ⇒ k δ +
1
6
δ
3
2 +
1
2
δ
2
− 1 − P L δ −
1
2
δ
3
= 0
⇒
δ
k 1 +
1
6
δ
2
1 +
1
2
δ
2
− P L 1 +
1
2
δ
2
= 0
⇒
δ = 0 tai P
II
= 1 +
1
6
δ
2
k
L
(39)The same orresponding to the higher bifur ationload:
λ = 3
kohdalla(δ
1
= −δ
2
= δ ⇒ ϕ
2
=
−ϕ
1
)∂Π
∂δ
1
= 0 ⇒ k 3δ +
3
2
δ
3
2 +
5
2
δ
2
− (−1 − 2δ
2
)
− P L 3δ +
9
2
δ
3
= 0
⇒ δ
3k 1 +
1
2
δ
2
3 +
9
2
δ
2
− P L 3 +
9
2
δ
2
= 0
⇒ δ = 0 tai P
III
= 1 +
1
2
δ
2
3k
L
(40)0
0.5
1
1.5
2
2.5
3
3.5
4
-0.4
-0.2
0
0.2
0.4
λ
δ
P
II
P
III
Ifwewant toinvestigatestability properties ofthe paths
P
II
andP
III
wehaveto substi-tute the equationsof the paths(39) and (40) intothe expression of the se ondvariationoftheTPE (38). Forpath
P
II
it isvalid(δ
1
= δ
2
)∂
2
Π
∂δ
2
1
P
II
= k
δ +
1
6
δ
3
δ + k
2 +
1
2
δ
2
2
+ k
δ +
1
6
δ
3
· 0 + k(−1)
2
− k
1 +
1
6
δ
2
2 +
3
2
δ
2
= k
3 +
7
6
δ
2
+
1
6
δ
4
> 0 ∀ δ
Forthe path
P
III
(δ
1
= −δ
2
, ϕ
1
= −ϕ
2
)∂
2
Π
∂δ
2
1
P
III
= k
3δ +
3
2
δ
3
3δ + k
2 +
5
2
δ
2
2
− k
3δ +
3
2
δ
3
(−2δ) + k(−1 − 2δ
2
)
2
−3k
1 +
1
2
δ
2
2 +
15
2
δ
2
= k
−1 +
7
2
δ
2
+
3
2
δ
4
< 0,
when
δ
issu iently small.Therefore the pathP
II
isstable andthe pathP
III
isunstable near the bifur ation point.translational springs. Investigate alsostability of the paths. It is assumed that the point C is
not moving horizontally.
A C B
?
v
θ
-
a
Da
-?
P
6
?
L
Solution: The total potentialenergy expression is
Π(v, θ) =
1
2
k∆
2
A
+
1
2
k∆
2
B
− P ∆
D
= k(v
2
+ a
2
sin
2
θ) − P [v + L(1 − cos θ)]
= ka
2
(u
2
+ sin
2
θ) − P
u +
1
α
(1 − cos θ)
,
in whi h
v = au
anda = αL
.By deningP = λka
, weget the formΠ
ka
2
= ˜
Π = u
2
+ sin
2
θ − λ
u +
1
α
(1 − cos θ)
The equilibriumequations are
∂ ˜
Π
∂u
= 2u − λ = 0 ⇒ u =
λ
2
∂ ˜
Π
∂θ
= 2 sin θ cos θ − λ
1
α
sin θ = 0
= sin θ
2 cos θ −
α
1
λ
= 0
⇒
(
sin θ = 0 ⇒ θ = 0
λ = 2α cos θ
Thusthe primarypath isdened as
u = λ/2
jaθ = 0
.Stabilityof the primarypath∂
2
Π
˜
∂u
2
= 2,
∂
2
Π
˜
∂u∂θ
= 0,
∂
2
Π
˜
∂θ
2
= 2 cos 2θ −
λ
α
cos θ
Let's substitute the expressions of the primary path
u = λ/2
andθ = 0
, into the stability matrix:K =
"
2
0
0 2 −
α
λ
#
The riti alvalue ofthe load parameter
λ
kr
anbeobtained fromthe onditiondet(K) = 0 ⇒
λ
kr
= 2αka = 2α
2
kL
.On the se ondary equilibriumpath
u = λ/2
jaλ = 2α cos θ
:K =
"
2
0
0 2 cos 2θ − 2 cos
2
θ
#
= 2
"
1
0
0 − sin
2
θ
#
The se ondary path isunstable sin e
K
2
2 = ∂
2
Π/∂θ
˜
2
= − sin
2
θ ≤ 0 ∀ θ
.0
0.5
1
1.5
2
-0.4
-0.2
0
0.2
0.4
λ/α
θ
Example 2.1 Derive the expression of urvature for a plane beam using (a) the Lagrangian
and (b) the Eulerianapproa h.
Solution: The dieren e between the Lagrangian and Eulerian approa hes is the meaning of
theindependentvariable
x
.IntheLagrangianapproa hthe oordinateisatta hedtoamaterial point.The dispa ement atpointx
is the displa ementof the pointa upyingthe positionx
at the initialundeformed onguration. In the Eulerianapproa h the oordinateis referringonlytoa spatial point
x
.da = dx
-x, u
?
y, v
ϕ
da
dv
u
u + du
Lagrange: It is seen fromthe gure
sin ϕ =
dv
da
=
dv
dx
= v
′
⇒ ϕ = arcsin v
′
sin e the urvature is
κ = 1/R = ϕ
′
we get1
R
= ϕ
′
=
v
′′
p1 − (v
′
)
2
,
d arcsin x
dx
=
1
√
1 − x
2
da
-x, u
?
y, v
ϕ
da
dv
dx
Euler: From the gure
tan ϕ =
dv
dx
= v
′
⇒ ϕ = arctan v
′
we obtainfor the urvature
κ =
1
R
=
∂ϕ
∂a
not
∂ϕ
∂x
!!
=
1
1 + v
′2
∂
∂a
dv
dx
where
d arctan x
dx
=
1
1 + x
2
=
1
1 + v
′2
d
2
v
dx
2
1
√
1 + v
′2
(where da = dx
√
1 + v
′2
)
=
v
′′
(1 + v
′2
)
√
1 + v
′2
=
v
′′
(1 + v
′2
)
3/2
Note! When the higherorder terms are negle ted we get the same result forboth approa hes:
Example 2.2 Determine
P
cr
startingfrom the dierentialequation.e e
e e
2 EI EI L/2 L/2 -x P Solution: In part 1 1v
(4)
1
+ k
2
v
1
′′
= 0
, wherek
2
= P/2EI
.(see problem ??) 2v
(4)
2
= 0
BC : v
1
−
L
2
= v
1
′
−
L
2
= v
2
L
2
= v
2
′
L
2
= 0
v
1
(0) = v
2
(0)
v
′
1
(0) = v
′
2
(0)
M
1
(0) = M
2
(0)
Q
1
(0) = Q
2
(0) + P v
2
′
(0)
P
XX
X
z
M
1
-Q
1
XXXX
XXXX
XXXX
P
Q
2
M
2
Solutionsfor the homogenious equationsare
v
1
= C
1
sin kx + C
2
cos kx + C
3
x + C
4
v
′
1
= C
1
k cos kx − C
2
k sin kx + C
3
v
′′
1
= −C
1
k
2
sin kx − C
2
k
2
cos kx
v
′′′
1
= −C
1
k
3
cos kx + C
2
k
3
sin kx
v
2
= C
5
x
3
+ C
6
x
2
+ C
7
x + C
8
v
′
2
= 3C
5
x
2
+ 2C
6
x + C
7
v
′′
2
= 6C
5
x + 2C
6
v
′′′
2
= 6C
5
Takingthe boundary onditions intoa ount
Q
1
(0) = Q
2
(0) + P v
2
′
(0)
−2EIv
′′′
1
(0) = −EIv
2
′′′
(0) + P v
2
′
(0)
2C
1
k
3
= −6C
5
+ 2k
2
C
7
C
5
= −
1
3
k
3
C
1
+
1
3
k
2
C
7
M
1
(0) = M
2
(0)
−2EIv
′′
1
(0) = −EIv
2
′′
(0)
2C
2
k
2
= −2C
6
⇒ C
6
= −k
2
C
2
v
′
1
(0) = v
2
′
(0)
C
1
k + C
3
= C
7
⇒ C
5
= −
1
3
k
3
C
1
+
1
3
k
2
(C
1
k + C
3
) =
1
3
k
2
C
3
v
1
(0) = v
2
(0)
C
2
+ C
4
= C
8
v
1
−
L
2
= 0 ⇒ C
4
= C
1
sin
kL
2
− C
2
cos
kL
2
+ C
3
L
2
v
′
1
−
L
2
= 0 ⇒ C
3
= −k(C
1
cos
kL
2
+ C
2
sin
kL
2
)
v
2
L
2
= 0 ⇒
1
3
k
2
C
3
L
2
3
− k
2
C
2
L
2
2
+ (C
1
k + C
3
)
L
2
+ C
2
+ C
4
= 0
⇒
kL
2
− kL cos
kL
2
1 +
1
24
(kL)
2
+ sin
kL
2
C
1
+
1 −
1
4
(kL)
2
− cos
kL
2
− kL sin
kL
2
1 +
1
24
(kL)
2
C
2
= 0
v
′
2
L
2
= 0 ⇒ k
2
C
3
L
2
2
− 2k
2
C
2
L
2
+ C
1
k + C
3
= 0
⇒
1 −
1 +
1
4
(kL)
2
cos
kL
2
kC
1
+
−kL −
1 +
1
4
(kL)
2
sin
kL
2
C
2
= 0
⇒
"
a(kL) b(kL)
c(kL) d(kL)
# (
C
1
C
2
)
=
(
0
0
)
det = 0 ⇒ kL ≈ 7.55 ⇒ P
kr
= 114
EI
L
2
Example 2.3 Determine
P
cr
startingfrom the dierentialequation.(α = 2, β = 1
)@@
@@
@@
@@
α
EIβ
EA EI EA L/2 L/2 -x PSolution: Sin e
β = 1 ⇒ (EA)
1
= (EA)
2
⇒ P
1
= P
2
= P/2
.Part 1:
v
(4)
1
+ k
1
2
v
1
′′
= 0
k
1
2
=
P/2
2EI
=
P
4EI
2:v
(4)
2
− k
2
2
v
2
′′
= 0
k
2
2
=
P/2
EI
=
P
2EI
k
2
2
= 2k
2
1
⇒ k
2
=
√
2k
1
BC : v
1
−
L
2
= v
′
1
−
L
2
= v
2
L
2
= v
2
′
L
2
= 0
v
1
(0) = v
2
(0)
v
′
1
(0) = v
2
′
(0)
M
1
(0) = M
2
(0)
Q
1
(0) = Q
2
(0) + P v
2
′
(0)
P
2
XX
X
z
M
1
-Q
1
XXXX
XXXX
XXXX
P
Q
2
M
2
P
2
XX
X
z
Solutionsfor the homogenious dierentialequations are
v
1
= C
1
sin kx + C
2
cos kx + C
3
x + C
4
v
1
′
= C
1
k cos kx − C
2
k sin kx + C
3
v
′′
1
= −C
1
k
2
sin kx − C
2
k
2
cos kx
v
′′′
1
= −C
1
k
3
cos kx + C
2
k
3
sin kx
v
2
= C
5
sinh kx + C
6
cosh kx + C
7
x + C
8
v
′
2
= C
5
k cosh kx − C
6
k sinh kx + C
7
v
′′
2
= −C
5
k
2
sinh kx − C
6
k
2
cosh kx
v
′′′
2
= −C
5
k
3
cosh kx + C
6
k
3
sinh kx
Takingthe boundary onditions intoa ount
Q
1
(0) = Q
2
(0) + P v
′
2
(0)
−2EIv
′′′
1
(0) = −EIv
′′′
2
(0) + P v
1
′
(0)
2C
1
k
3
1
= −C
5
k
2
3
+ 4k
1
2
(C
1
k
1
+ C
3
)
C
5
= −
1
k
3
2
2k
3
1
C
1
+ 4k
2
1
C
3
=
1
√
2
C
1
+
2
√
2k
1
C
3
M
1
(0) = M
2
(0)
−2EIv
′′
1
(0) = −EIv
′′
2
(0)
2C
2
k
2
1
= −C
6
k
2
2
⇒ C
6
= −2
k
1
k
2
2
C
2
= −C
2
v
′
1
(0) = v
′
2
(0)
C
1
k
1
+ C
3
= C
5
k
2
+ C
7
⇒ C
7
= C
1
k
1
+ C
3
− C
5
k
2
= −C
3
v
1
(0) = v
2
(0)
C
2
+ C
4
= C
6
+ C
8
⇒ C
8
= C
2
+ C
4
− C
6
= 2C
2
+ C
4
v
1
−
L
2
= 0 ⇒ C
4
= C
1
sin
k
1
L
2
− C
2
cos
k
1
L
2
+ C
3
L
2
v
′
1
−
L
2
= 0 ⇒ C
3
= −k
1
C
1
cos
k
1
L
2
+ C
2
sin
k
2
L
2
⇒ C
4
=
sin
k
1
L
2
−
k
1
L
2
cos
k
1
L
2
C
1
−
cos
k
1
L
2
+
k
1
L
2
sin
k
1
L
2
C
2
v
2
L
2
=
1
√
2
C
1
+
2
√
2k
1
C
3
sinh
k
2
L
2
− C
2
cosh
k
2
L
2
− C
3
L
2
+ 2C
2
+ C
4
= 0
⇒
1
√
2
−
2
√
2
cos
k
1
L
2
sinh
k
2
L
2
+ sin
k
1
L
2
C
1
−
√
2
2
sin
k
1
L
2
sinh
k
2
L
2
− cosh
k
2
L
2
+ 2 − cos
k
1
L
2
C
2
= 0
v
′
2
L
2
=
1
√
2
C
1
−
2
√
2
C
1
cos
k
1
L
2
−
2
√
2
C
2
sin
k
1
L
2
k
2
cosh
k
2
L
2
−C
2
k
2
sinh
k
2
L
2
+ k
1
C
1
cos
k
1
L
2
+ k
1
C
2
sin
k
1
L
2
= 0
⇒
k
√
2
2
cosh
k
2
L
2
−
2k
2
√
2
cos
k
1
L
2
cosh
k
2
L
2
+ k
1
cos
k
1
L
2
C
1
+
k
1
sin
k
1
L
2
− k
2
sinh
k
2
L
2
−
2k
2
√
2
sin
k
1
L
2
cosh
k
2
L
2
C
2
= 0
⇒
"
a(kL) b(kL)
c(kL) d(kL)
# (
C
1
C
2
)
=
(
0
0
)
The riti al load isobtained from the equation det[℄
= ad − bc = 0
⇒ k
1
L ≈ 10, 637 ⇒ P
kr
= 4k
1
2
EI
L
2
= 452, 6
EI
L
2
Ifthe dire tionof the load is reversed, the result is
247, 5EI/L
2
tensible beam and small dee tions. Solve the equations and determine the eigenmodes and
show that the eigenmodes are orthogonal.
EI
L
PSolution: The total potentialenergy fun tionalis
Π(v) =
1
2
L
Z
0
EI(v
′′
)
2
− P (v
′
)
2
dx,
wherethehorizontaldee tion
∆
undertheloadP
anbedeterminedasϕ
dx + du
dv
The Euler equations are obtained from the stationarity ondition ofthe fun tional
δΠ = Π
,v
δv =
L
Z
0
(EIv
′′
δv
′′
− P v
′
δv
′
)dx = 0,
where
δv
isthevariationofthe dee tion,i.e.anarbitraryfun tionsatisfyingthehomogenious kinemati al boundary onditionsv(0) = v
′
(0) = 0
. After integration by parts we get the term
δv
as a ommonfa tor inside the integralδΠ =
L
0
EIv
′′
δv
′
−
L
Z
0
(EIv
′′
)
′
δv
′
dx −
L
0
P v
′
δv +
L
Z
0
δvdx
=
L
0
EIv
′′
δv
′
−
L
0
(EIv
′′
)
′
δv −
L
0
P v
′
δv +
L
Z
0
[(EIv
′′
)
′′
+ (P v
′
)
′
] δvdx
At the lower limit
δv(0) = δv
′
(0) = 0
and shear for e:
M = −EIv
′′
sekäQ = −(EIv
′′
)
′
,we getδΠ = −M(L)δv
′
(L) + [Q(L) − P v
′
(L)]δv(L) +
L
Z
0
[(EIv
′′
)
′′
+ (P v
′
)
′
] δvdx = 0,
sin e
δv
isarbitraryfun tion satisfyingthe boundary onditionsv(0) = v
′
(0) = 0
, thusthe
fol-lowingequationshavetobesatised
(EIv
′′
)
′′
+ (P v
′
)
′
= 0
x ∈ (0, L)
(Euleri equation)M(L)
= 0
Q(L) − P v
′
(L) = 0
)
natural boundary onditionsv(0)
= 0
v
′
(0) = 0
)
essential boundary onditionsIfthe bendingstiness
EI
and the ompressive for eP
are onstants inthe domain,we get a homogeneousdierentialequation with onstant oe ientsEIv
(4)
+ P v
′′
= 0
⇒ EIv
′′
+ P v = Cx + D, (C, D constants)
⇒ v = A sin kx + B cos kx + Cx + D, k =
r P
EI
The derivativesare
v
′
= Ak cos kx − Bk sin kx + C
v
′′
= −Ak
2
sin kx − Bk
2
cos kx
v
′′′
= −Ak
3
cos kx + Bk
3
sin kx
v(0) = 0 ⇒ B + D = 0
v
′
(0) = 0 ⇒ Ak + C = 0
v
′′
(L) = 0 ⇒ A sin kL + B cos kL = 0
−EIv
′′′
(L)−P v
′
(L) = 0 ⇒ −EI(−Ak
3
cos kL+Bk
3
sin kL)−P (Ak cos kL−Bk sin kL+C) = 0
(41)
P = λ
EI
L
2
⇒ k =
√
λ
L
It follows from equation (41)that
A = 0 ⇒ C = 0 ⇒ B cos kL = 0 ⇒ B = 0
orcos kL = 0
. IfB = 0 ⇒ v ≡ 0
ityields atrivialsolution, hen e weshould have⇒ λ
n
=
π
2
+ nπ
2
,
and the lowest bu klingload is
λ
0
=
π
2
2
⇒ P
cr
=
π
2
4
EI
L
2
The eigenmode orresponding to the eigenvalue
λ
n
isv
n
= B(cos k
n
x − 1), k
n
=
1
L
π
2
+ nπ
It was asked to give the normalized eigenmodes. For that we should dene how this
normalizationshould be done. Itis usual to use the energy norm
||v
n
||
2
E
=
L
Z
0
EI(v
′′
n
)
2
dx.
The energy orthogonalitythus means
L
Z
0
EIv
′′
n
v
m
′′
dx = 0, kun n 6= m.
Lets normalize the eigenmodes
v
n
su h, that||v
n
||
E
= E
1
, whereE
1
is the energy unit and[E
1
] =
√
Nm
.v
′′
n
= −Bk
n
2
cos k
n
x
⇒ E
1
2
= EIB
2
k
4
n
L
Z
0
cos
2
k
n
xdx
Lets hangevariablessu h, that
y = k
n
x, dx =
1
k
n
dy rajat
(
x = 0
⇒ y = 0
x = L ⇒ y =
π
2
+ nπ
⇒ E
1
2
= EIB
2
k
n
3
π
2
+nπ
Z
0
cos
2
ydy = EIB
2
k
3
n
1
2
π
2
+ nπ
⇒ B
2
=
2E
1
2
EIk
3
n
π
2
+ nπ
=
2E
2
1
EI
π
16
4
(1 + 2n)
4
⇒ B =
4
√
2L
3/2
E
1
π
2
(1 + 2n)
2
√
EI
The energy orthonormal eigenfun tionsare thus
v
n
(x) = b
n
(cos k
n
x − 1), B
n
=
4
√
2
π
2
(1 + 2n)
2
E
1
L
3/2
√
EI
L
Z
0
EIv
n
′′
v
m
′′
dx = 0, kun n 6= m.
L
Z
0
v
′′
n
v
m
′′
dx =
L
Z
0
cos
h
π
2
(1 + 2n)
x
L
i
cos
h
π
2
(1 + 2m)
x
L
i
merk. y =
π
2
x
L
, dx =
2L
π
dy
=
2L
π
L
Z
0
span as afun tion of the ompressive for e
P
for the beam shown below.e e
e e
EI L P? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?
qSolution: The dierential equationfor the beam- olumnis
v
(4)
+ k
2
v
′′
=
q
EI
, where k
2
=
P
EI
The solutionis
v = C
1
sin kx + C
2
cos kx + C
3
x + C
4
+ Ax
2
, where A =
q
2EIk
2
=
q
2P
.
Let's hoose the zero o-ordinate atthe midspan. From the boundary onditions we get
v
′
(0) = 0 ⇒ C
1
k + C
3
= 0
v
′′′
(0) = 0 ⇒ C
1
= 0 ⇒ C
3
= 0
v
′
L
2
= 0 ⇒ −C
2
k sin
kL
2
+
qL
2P
= 0
⇒ C
2
=
qL
2kP sin
kL
2
v
±
L
2
= 0 ⇒ C
2
cos
kL
2
+ C
4
+ A
L
2
4
= 0
⇒ C
4
= −
qL
2kP tan
kL
2
−
qL
2
8P
⇒ v(x) =
qL
2kP sin
kL
2
cos kx −
qL
2kP sin
kL
2
cos
kL
2
+
kL
4
sin
kL
2
+
q
2P
x
2
DenotingP = λ EI/L
2
, thuskL =
√
λ
.⇒ v(0) =
qL
2kP sin
√
2
λ
1 − cos
√
λ
2
!
−
qL
2
8P
⇒ M(x) = −EIv
′′
=
C
2
k
2
cos kx −
q
P
EI =
qL
2
λ
kL
2
cos kx
sin
kL
2
− 1
!
x = ±
L
2
M
t
=
qL
2
λ
√
λ
2 tan
√
λ
2
− 1
!
The bending moment inthe midspan