• When the exact solution of the governing differential equation is not available,
it may yield a good approximate solution.
•
• The method has the inherent ability to converge to the exact solution if a large
number of terms are taken to represent the unknown field (say, deflection
shape).
•
• The numerical scheme of the solution technique is quite simple in view of the
advancement in numerical methods and computational fields.
•
• The method is suitable to be used in advanced techniques like FEM, which can
take up irregularities in geometry, material, boundary condition, etc.
Kashi Nath Saha, Mech. Engg. Dept., Jadavpur Univ., Kol-32 12/05/10 04:04
Strain energy U: Incrementof work done= F ⋅ dr
energy in spring = work done = (mr) ⋅ (r dt) = d( 2 mr ⋅ r ) = dT
1
1 1 = incrementof kineticenergyT
U = k∆2 = P∆
2 2
Conservation of energy:
work done = energy stored
Conservation of energy for conservative systems Kinetic energy T:
E = total energy = T + U = constant 1
T = mr ⋅ r
2
Kashi Nath Saha, Mech. Engg. Dept., Jadavpur Univ., Kol-32 12/05/10 04:04
Energy methods – Example
Work-energy principles have many
uses, but one of the most useful is
to derive the equations of motion.
Kashi Nath Saha, Mech. Engg. Dept., Jadavpur Univ., Kol-32 12/05/10 04:04
Kashi Nath Saha, Mech. Engg. Dept., Jadavpur Univ., Kol-32 12/05/10 04:04
Type of boundary conditions:
1. Geometric boundary condition – they refer to the kinematic conditions,
like, deflection, slope, curve, etc., of the boundary.
2. Forced boundary condition – refers to the forced conditions of the
boundary.
Type of functions:
1. eigen-functions 2. admissible functions 3. coordinate functions
Success or failure in applying the assumed series solution for the assumed field
(say deflection w), largely depends on the choice of proper coordinate functions.
In the majority of cases a few terms of the series give a sufficiently accurate result
if the coordinate functions are the eigen-functions of the problem. In other cases,
satisfactory result can be obtained when the coordinate functions are chosen from
a set of orthogonal functions.
The set of n-vectors v1, v2, …, vn are linearly independent if their
linear combination,
c1v1 + c2v2 + … + cnvn ( = 0 ) is zero only when all the arbitrary
scalars (ci –s ) are zero. In this case, the vectors will form an
orthogonal set, such that
1, for i = j
( i j)
v v dξ
∫ =
0, for i ≠ j
It is known in vector algebra that to form a basis of a n-
dimensional vector space, the basis vectors must come from a set
of linearly independent vectors, which again forms an orthogonal
set. Similarly the solution space for the assumed unknown field
(say deflection w) is complete, when they are formed through a set
of orthogonal functions.
Kashi Nath Saha, Mech. Engg. Dept., Jadavpur Univ., Kol-32 12/05/10 04:04
Expression of strain energy due to bending of a beam / column:
Assumption – The beam / column is slender, i.e., the effect of shear deformation
can be neglected.
Kashi Nath Saha, Mech. Engg. Dept., Jadavpur Univ., Kol-32 12/05/10 04:04
Expression of potential energy of a column:
Kashi Nath Saha, Mech. Engg. Dept., Jadavpur Univ., Kol-32 12/05/10 04:04
Kashi Nath Saha, Mech. Engg. Dept., Jadavpur Univ., Kol-32 12/05/10 04:04
Energy approach for stability problem
Kashi Nath Saha, Mech. Engg. Dept., Jadavpur Univ., Kol-32 12/05/10 04:04
Kashi Nath Saha, Mech. Engg. Dept., Jadavpur Univ., Kol-32 12/05/10 04:04
Energy approach for stability problem
Kashi Nath Saha, Mech. Engg. Dept., Jadavpur Univ., Kol-32 12/05/10 04:04
Buckling mode shapes:
Kashi Nath Saha, Mech. Engg. Dept., Jadavpur Univ., Kol-32 12/05/10 04:04
Energy approach for dynamic problem
1 ∂u ∂u
Uo = strainenergydensity= 12 σ ε = 12 (E ε x )ε x = E
2 ∂x ∂x
V = potentialenergy= strainenergy= U
T = kineticenergy= 12 mu 2
Hamilton’s principle:
t2
0 = ∫ (δT − δV ) dt
t1
t2 L ∂u ∂
0 = ∫ ∫ ρ Adx u δu − A E ( δu ) dx dt
t1
0
∂x ∂x
L ∂ ∂ ∂u ∂u L
( ) δu dt + ∫ ( ρ A u ) δu tt12 dx
t2 L
0 = ∫ ∫ − ρ Adx u δu + A E δu dx − A E
∂t ∂x ∂x ∂x
t1 0 0 0
t2 L ∂ ∂ ∂u ∂u L
0 = ∫ ∫ − ( )
ρ A u + A E δu dx − A E
δu dt
t1
0 ∂t ∂x ∂ x ∂x 0
Kashi Nath Saha, Mech. Engg. Dept., Jadavpur Univ., Kol-32 12/05/10 04:04
Fixed-free bar – Free vibration
E
α=
∂2u 2 ∂ u
2
ρ
For free vibration: =α
∂t 2
∂x 2 = wave speed
General solution: u ( x, t ) = A( x) cos(ω n t )
iπ E
Hence ωn = are the frequencies (eigenvalues)
2L ρ
(i = 1, 3, 5, )
iπ x
sin are the eigenfunctions
2 L
Kashi Nath Saha, Mech. Engg. Dept., Jadavpur Univ., Kol-32 12/05/10 04:04
Ritz method – Free vibration
Start with Hamilton’s principle after I.B.P. in time:
L ∂ ∂u ∂
∫0 − ( ρ A u ) δu − A E ( δu ) dx dt
t2
0=∫
∂t ∂x ∂x
t1
Seek an approximate solution to u(x, t):
In time: harmonic function ⇒ cos(ω t) (ω = ω n)
In space: X(x) = a1φ 1(x)
where: a1 = constant to be determined
φ 1 (x) = known function of position
φ 1 (x) must satisfy the following:
Satisfy the homogeneous form of the EBC.
u(0) = 0 in this case.
Be sufficiently differentiable as required by HP.
Kashi Nath Saha, Mech. Engg. Dept., Jadavpur Univ., Kol-32 12/05/10 04:04
Fixed-free bar – General solution
Free vibration: E = wave speed
α=
ρ
EBC: u (0) =0
∂u ∂u
NBC: AE x =L =0 ∴ x =L =0
∂x ∂x
General solution:
∞
u ( x, t ) = ∑[ A cos( p t ) + B sin( p t )] [C
i =1
i i i i i cos ( pi x α ) + Di sin ( pi x α ) ]
∞
EBC ⇒ u (0, t ) = ∑ C [ A cos( p t ) + B sin( p t )] = 0
i =1
i i i i i ∴ Ci = 0
∂u ∞
Di pi
NBC ⇒ x =L = ∑ cos ( pi L α ) [ Ai cos( pi t ) + Bi sin( pi t )] = 0
∂x i =1 α
p L pi L π 3π 5π
Either Di = 0 (trivialsolution) or cos i = 0 = or or
α α 2 2 2
iπα
For any time dependent problem: pi = (i = 1, 3, 5, )
2L
∞
iπ x iπαt iπαt
u ( x, t ) = ∑ sin Ai cos + Bi sin
i =1, 3, 5 , 2L 2L 2 L
Kashi Nath Saha, Mech. Engg. Dept., Jadavpur Univ., Kol-32 12/05/10 04:04
One-term Ritz approximation
Pick: φ1 ( x) = x ⇒ u ( x, t ) = a1 φ1 ( x) cos(ω t ) = a1 x cos(ω t )
Alsoapproximat e: δu = φ1 ( x) cos(ω t ) = x cos(ω t )
Substituting:
t2 L
[ ]
0 = ∫ ∫ a1 ρω 2 A ( x)( x) − A E (1)(1) dx cos 2 (ω t ) dt
t1
0
ω 2 ∫ ρ A x 2 dx a1 = ∫ A E dx a1
0
L
0
L
(
in matrixform: ω 2 [ M ] { a} = [ K ] { a} )
2 AL
3
3 E 3
ρω = A E L ⇒ ω 2 = 2 = 2 α 2
3 L ρ L φ1RITZ = x
Hence 3 α
ωRITZ = α = 1.732 π x
φ1EXACT = sin
L L 2L
π α
ωEXACT = α = 1.571
2L L
Ritz estimate is higher than the exact
Only get one frequency
If we pick a different basis/trial/approximation function φ 1, we would
get a different result.
Kashi Nath Saha, Mech. Engg. Dept., Jadavpur Univ., Kol-32 12/05/10 04:04
One-term Ritz approximation (cont.)
π x dφ1 π π x
What if we pick: φ1 ( x ) = sin ⇒ = cos
2L dx 2L 2L
u ( x, t ) = a1 φ1 ( x) cos(ω t ) = a1 sin ( π x 2 L ) cos(ω t )
Alsoapproximat
e: δu = φ1 ( x) cos(ω t ) = sin ( π x 2 L ) cos(ω t )
t2 L ∂ ∂u ∂
Substituting: 0 = ∫ ∫ − ( ρ A u ) δu − A E ( δu ) dx dt
∂t ∂x ∂x
t1 0
L
2
π x π 2 π x
t2
0=∫ ∫0 a1 ρ ω A sin
2 2
− A E cos dx cos 2 (ω t ) dt
t1
2L 2L 2 L
π E π
Hence ωRITZ = = α = ωEXACT
2L ρ 2L
Both mode shape and natural frequency are exact.
But all other functions we pick will never give us a frequency lower
than the exact.
Kashi Nath Saha, Mech. Engg. Dept., Jadavpur Univ., Kol-32 12/05/10 04:04
Two-term Ritz approximation
dX
Let: X ( x) = a1 x + a2 x 2 ⇒ = a1 + 2a2 x
dx
∫ [ ρ Aω ]
t2 L
e δu = φ1 = x : 0 = ∫
If approximat 2
(a1 x + a2 x 2 ) x − AE (a1 + 2a2 x)(1) dx dt
t1
0
∫ [ ρ Aω ]
t2 L
e δu = x 2 :
If approximat 0=∫ 2
(a1 x + a2 x 2 ) x 2 − A E (a1 + 2a2 x)(2 x ) dx dt
t1
0
3 11 ∫0 (1)(1)dx = L
K =
L4
L L
M 12 = M 21 = ∫0 ( x )( x)dx = = = ∫0 =
2 2
K
12 K 21 ( 2 x )(1) dx L
4 3
L5 K = L (2 x)( 2 x)dx = 4 L
∫0
L
M 22 = ∫0 ( x )( x )dx =
2 2
5
22
3
Kashi Nath Saha, Mech. Engg. Dept., Jadavpur Univ., Kol-32 12/05/10 04:04
Two-term Ritz approximation (cont.)
E
Substitution of: λ = ω 2 and α 2 =
ρ
(α 2 L − λ L3 3) (α 2 L2 − λ L4 4) a1 0
leads to 2 2 =
(α L − λ L 4) (α 4 L 3 − λ L 5) a2 0
4 2 3 5
Solving characteristic polynomial (for det[ ]=0) yields 2 frequencies:
(ω1 ) RITZ = 1.5767 α L and (ω2 ) RITZ = 5.67 α L
(ω1 ) EXACT = 1.5708 α L and (ω2 ) EXACT = 4.7123 α L
Let a1 = 1:
Kashi Nath Saha, Mech. Engg. Dept., Jadavpur Univ., Kol-32 12/05/10 04:04
MODE SHAPES OF BEAM VIBRATION
Kashi Nath Saha, Mech. Engg. Dept., Jadavpur Univ., Kol-32 12/05/10 04:04
Mode shape plots for SFFF and CCSS boundary
conditions
Kashi Nath Saha, Mech. Engg. Dept., Jadavpur Univ., Kol-32 12/05/10 04:04
Mode shape plots for SFFF and CCSS boundary
conditions
Kashi Nath Saha, Mech. Engg. Dept., Jadavpur Univ., Kol-32 12/05/10 04:04