Structural DynamicsFEM CourseNotes Chapter16

Structural Dynamics
Introduction
This chapter provides an elementary introduction to time-dependent problems.
We will introduce the basic concepts using the single-degree-of-freedom springmass system. We will include discussion of the stress analysis of the onedimensional bar, beam, truss, and plane frame.
We will provide the basic equations necessary for structural dynamic analysis
and develop both the lumped- and the consistent-mass matrices involved in the
analyses of the bar, beam, truss, and plane frame. We will describe the assembly
of the global mass matrix for truss and plane frame analysis and then present
numerical integration methods for handling the time derivative.
We will provide longhand solutions for the determination of the natural frequencies for bars and beams, and then illustrate the time-step integration process involved with the stress analysis of a bar subjected to a time dependent forcing function.
Dynamics of a Spring-Mass System

In this section we will discuss the motion of a single-degree-of-freedom
spring-mass system as a introduction to the dynamic behavior of bars, truss,
frames. Consider the single-degree-of-freedom spring-mass system subjected to
a time-dependent force F(t) as shown in the figure below. The term k is the stiffness of the spring and m is the mass of the system.
The free-body diagram of the mass is shown below. The spring force T = kx and
the applied force F(t) act on the mass, and the mass-times-acceleration term is
CIVL 7117 Finite Elements Methods in Structural Mechanics
Page 353
shown separately.
Applying Newtons second law of motion, f = ma, to the mass, we obtain the
equation of motion in the x direction:
F (t ) kx = mx
where a dot over a variable indicates differentiation with respect to time;
() = d () / dt . The standard form of the equation is:
mx + kx = F (t )
The above equation is a second-order linear differential equation whose solution
for the displacement consists of a homogeneous solution and a particular solution. The homogeneous solution is the solution obtained when the right-handside is set equal to zero. A number of useful concepts regarding vibrations are
available when considering the free vibration of a mass; that is when F(t) = 0.
Lets define the following term:
k
m
The equation of motion becomes:
2 =
x + 2 x = 0
where is called the natural circular frequency of the free vibration of the
mass (radians per second). Note that the natural frequency depends on the
spring stiffness k and the mass m of the body.
The motion describe by the homogeneous equation of motion is called simple
harmonic motion. A typical displacement/time curve is shown below.
Page 354
where xm denotes the maximum displacement (or amplitude of the vibration).

The time interval required for the mass to complete one full cycle of motion is
called the period of the vibration (in seconds) and is defined as:
The frequency in hertz (Hz = 1/s) is f = 1/ = /(2).

Direct Derivation of the Bar Element
Lets derive the finite element equations for a time-dependent (dynamic)

stress analysis of a one-dimensional bar.
Step 1 - Select Element Type
We will consider the linear bar element shown below.
where the bar is of length L, cross-sectional area A, and mass density (with
typical units of lb-s2/in4), with nodes 1 and 2 subjected to external timedependent loads, fx e (t ) .
Page 355
Step 2 - Select a Displacement Function
A linear displacement function is assumed in the x direction.

u = a1 + a2 x
The number of coefficients in the displacement function, ai, is equal to the total
number of degrees of freedom associated with the element.
We can express the displacement function in terms of the shape functions:
d1x
u = [N1 N2 ] =
d 2 x
where N1 and N2 are the interpolation functions gives as:

N1 = 1
x
L
N2 =
x
L
Step 3 - Define the Strain/Displacement and Stress/Strain Relationships
The stress-displacement relationship is:
{ x } =
{}
du
= [B ] d
dx
where
d
d = 1x
d2 x
The stress-strain relationship is given as:
1
[B ] =
L
1
L
{}
{ x } = [D]{ x } = [D][B]{d}
Step 4 - Derive the Element Stiffness Matrix and Equations
The bar element is typically not in equilibrium under a time-dependent force;

hence, f1x f2x. We must apply Newtons second law of motion, f = ma, to each
node. Write the law of motion as the external force fxe minus the internal force
equal to the nodal mass times acceleration. Therefore:
Page 356
2
2
f e = f + m d1x
f e = f + m d 2 x
1x
1x
1
2x
2x
2
t 2
t 2
where m1 and m2 are obtained by lumping the total mass of the bar equally at the
two nodes such that:
m1 =
AL
2
m2 =
AL
2
In matrix form, the above equations are:
f1x e f1x m1
e = +
f2 x f2 x 0
2d1x
0 t 2
m2 2d2 x
t 2
If we replace {f } with [ k ]{d } we get:
{f (t )} = k {d} + m {d}
e
where the elemental stiffness matrix is:
{}
2 d
1
1
AE

k =
d =

L 1 1
t 2
and the lumped-mass matrix is:
{}
=
m
AL 1 0
2 0 1
Lets derive the consistent-mass matrix for a bar element. The typical
method for deriving the consistent-mass matrix is the principle of virtual work;
however, an even simpler approach is to use DAlemberts principle. The effective body force is:
{ X } = {u}
e
The nodal forces associated with {Xe} are found by using the following:
Page 357
{fb } = [N ]T { X } dV
V
Substituting {Xe} for {X} gives:
{fb } = [N ]T {u} dV
V
The second derivative of the u with respect to time is:
{u} = [N ]{d}
{u} = [N ]{d}
where u and u are the nodal velocities and accelerations, respectively.
{fb } = [N ] [N ]
T
{d} dV = m {d}
where
= [N ] [N ] dV
m
T
The mass matrix is called the consistent mass matrix because it is derived using
the same shape functions use to obtain the stiffness matrix. Substituting the
shape functions in the above mass matrix equations gives:
x
1 L x
=
m
1
x
V
L
L
x
dV
L
or
x
1 L x
= A
m
1
x
0
L
L
L
x
dx
L
or
Page 358
x
1
L
L
= A
m
0
1 x x
L L
x x
L L
dx
2
x
L

Evaluating the above integral gives:
=
m
AL 2 1
6 1 2
Step 5 - Assemble the Element Equations and Introduce

Boundary Conditions
The global stiffness matrix and the global force vector are assembled using the
nodal force equilibrium equations, and force/deformation and compatibility equations.
{F (t )} = [K ] {d } + [M ]{d}
where
N
[K ] = k (e )
e =1
[M ] = m(e )
e =1
{ }
{F } = f (e )
e =1
Numerical Integration in Time
We now introduce procedures for the discretization of the equations of motion

with respect to time. These procedures will allow the nodal displacements to be
determined at different time increments for a given dynamic system. The general
method used is called direct integration. There are two classifications of direct
integration: explicit and implicit. We will formulate the equations for two direct integration methods. The first, and simplest, is an explicit method known as the
central difference method. The second, more complicated but more versatile
than the central difference method, is an implicit method known as the Newmark-Beta (or Newmarks) method. The versatility of Newmarks method is evidenced by its adaptation in many commercially available computer programs.
Page 359
Central Difference Method
The central difference method is based on finite difference expressions for the
derivatives in the equation of motion. For example, consider the velocity and the
acceleration at time t:
d d i 1
di = i +1
2( t )
d di 1
di = i +1
2( t )
where the subscripts indicate the time step for a given time increment of t. The
acceleration can be expressed in terms of the displacements (using a Taylor series expansion) as:
d 2d i + d i 1
di = i +1
( t )2
We generally want to evaluate the nodal displacements; therefore, we rewrite the

above equation as:
d i +1 = 2d i d i 1 + di ( t )2
The acceleration can be expressed as:
di = M1 (Fi Kd i )
To develop an expression of di+1, first multiply the nodal displacement equation
by M and substitute the above equation for di into this equation.
Page 360
Md i +1 = 2Md i Md i 1 + (Fi Kd i )( t )
Combining terms in the above equations gives:
Md i +1 = ( t ) Fi + 2M ( t ) K d i Md i 1
To start the computation to determine d i +1, di +1, and di +1 we need the displacement
2
at time step i-1. Using the central difference equations for the velocity and acceleration and solving for di-1
( t )2

d i 1 = d i ( t )d i + d i
2
Procedure for solution:

1. Given: d0, d , and Fi(t)
0
2. If the acceleration is not given, solve for d0 at t = 0.

d0 = M1 (F0 Kd0 )
3. Solve for d-1 at t = -t
( t )
d 1 = d0 ( t )d0 + d0
2
4. Solve for d1 at t = t using the value of d-1 from Step 3
2
d1 = M1 ( t ) F0 + 2M ( t ) K d0 Md 1
5. With d0 given and d1 determined in Step 4 solve for d2

2
2
2
d 2 = M1 ( t ) F1 + 2M ( t ) K d1 Md0
6. Solve for d1
d1 = M1 (F1 Kd1 )

7. Solve for d1 using the central difference equation
d d0
d1 = 2
2( t )
Page 361
8. Repeat Steps 5, 6, and 7 to obtain the displacement, acceleration, and

velocity for other time steps.
Page 362
Example Problem
Determine the displacement, acceleration, and velocity at 0.05 second time intervals for up to 0.2 seconds for the one-dimensional spring-mass system shown
in the figure below.
The time-dependent forcing function is given as:
Consider the above spring-mass system as a single degree of freedom problem

represented by the displacement d.

1. At time t = 0:
d0 = 0
d0 = 0
2. The initial acceleration at t = 0:

2,000 100(0)
d0 = M1 (F0 Kd0 ) =
= 62.83 in 2
s
31.83
Page 363

( t )
d 1 = d0 ( t )d0 + d0
2
(0.05)2
d 1 = 0 (0.05)0 +
(62.83) = 0.0785 in
2
4. Solve for d1 at t = t (0.05 seconds) using the value of d-1 from Step 3:
2
2
d1 = M1 ( t ) F0 + 2M ( t ) K d0 Md 1
d1 =
1
2
2
( 0.05 ) 2,000 + 2 ( 31.83 ) ( 0.05 ) (100 ) 0 ( 31.83 )( 0.0785 )
31.82
= 0.0785 in
5. Solve for d2 at t = 0.10 seconds:
2
2
d 2 = M1 ( t ) F1 + 2M ( t ) K d1 Md0
d2 =
1
2
2
( 0.05 ) 1,500 + 2 ( 31.83 ) ( 0.05 ) (100 ) ( 0.0785 ) ( 31.83 )( 0 )
31.82
= 0.274 in
6. Solve for the acceleration d1 at time t = 0.05:
d1 = M1 (F1 Kd1 ) =
1
1,500 100 ( 0.0785 ) = 46.88 in 2
s
31.83

d d0 0.274 0
d1 = 2
=
= 2.74 in
s
2( t )
2 ( 0.05 )
Page 364

velocity for the next time step.
Repeating Step 5:
2
2
d3 = M1 ( t ) F2 + 2M ( t ) K d 2 Md1
d3 =
1
2
2
( 0.05 ) 1,000 + 2 ( 31.83 ) ( 0.05 ) (100 ) ( 0.274 ) ( 31.83 )( 0.0785 )
31.82
= 0.546 in
Repeating Step 6:
d2 = M1 (F2 Kd 2 ) =
1
1,000 100 ( 0.274 ) = 30.56 in 2
s
31.83
Repeating Step 7:
d d1 0.546 0.0785
d2 = 3
=
= 4.68 in
s
2( t )
2 ( 0.05 )
The following table summarizes the results for the remaining time steps as compared with the exact solution.
t (s)
F(t) (lb)
di (in/s2)
di (in/s)
d i (in)
d i (exact)
0.00
2,000
62.83
0.00
0.000
0.0000
0.05
1,500
46.88
2.74
0.0785
0.0718
0.10
1,000
30.56
4.68
0.274
0.2603
0.15
500
13.99
5.79
0.546
0.5252
0.20
-2.68
6.07
0.854
0.8250
0.25
-3.63
5.91
1.154
1.132
Page 365
Newmarks Method
Newmarks equations are given as:
di +1 = di + ( t ) (1 )di + di +1

d i +1 = d i + ( t )di + ( t )2 ( 21 ) di + di +1
where and are parameters. The parameter is typically between 0 and ,
and is often taken to be . For example, if = 0 and = the above equation
reduce to the central difference method.
To find di+1 first multiply the above equation by the mass matrix M and substitute into this the expression for acceleration. Recall the acceleration is:
d0 = M1 (F0 Kd0 )

The expression Mdi+1 is:
Md i +1 = Md i + ( t )Mdi + ( t )2 M ( 21 ) di + ( t )2 [Fi +1 Kd i +1 ]

Combining terms gives:
(M + (t ) K ) d
2
i +1
= ( t )2 Fi +1 + Md i + ( t )Mdi + ( t )2 M ( 21 ) di
Dividing the above equation by (t)2 gives:
K ' d i +1 = F 'i +1
where
K' = K +
1
M
( t )2
F 'i +1 = Fi +1 +
M
d + ( t )di + ( 21 ) ( t )2 di
2 i
( t )
The advantages of using Newmarks method over the central difference method
are that Newmarks method can be made unconditionally stable (if = and =
) and that larger time steps can be used with better results.
Page 366
Procedure for solution of Newmarks Method:
1. Given: d0, d0 , and Fi(t)

2. If the acceleration is not given, solve for d0 at t = 0.
d0 = M1 (F0 Kd0 )

3. Solve the displacement d1 at time t = t
K ' d1 = F '1
Page 367
4. Solve for d1 (original Newmark equation for d i +1 rewritten for di +1 )
d1 =
1
d d0 ( t )d0 ( t )2 ( 21 ) d0
2 1
( t )
5. Solve for d1

d1 = d0 + ( t ) (1 )d0 + d1
velocity for the next time step.
Example Problem
Determine the displacement, acceleration, and velocity at 0.1 second time intervals for up to 0.5 seconds for the one-dimensional spring-mass system shown
in the figure below.
The time-dependent forcing function is given as:
Consider the above spring-mass system as a single degree of freedom problem

represented by the displacement d. Use Newmarks method with = 1/6 and =
.
Page 368

1. At time t = 0: d0 = 0
d0 = 0
2. If the acceleration is not given, solve for d0 at t = 0:

100 70(0)
d0 = M1 (F0 Kd0 ) =
= 56.5 in / s 2
1.77
3. Solve the displacement d1 at time t = 0.1 seconds:
K ' d1 = F '1
K' = K +
1
1
M = 70 + 1
(1.77) = 1,132 lb / in
2
2
( t )
(0.1)
6
F '1 = F1 +
M
d + ( t )d0 + ( 21 ) ( t )2 d0
2 0
( t )
F '1 = 80 +
1.77
0 + (0.1)0 + ( 21 61 ) (0.1)2 ( 56.5 ) = 280 lb
2
1
6 (0.1)
F '1
280
=
= 0.248 in
K ' 1,132
4. Solve for d1 at time t = 0.1 seconds:
d1 =
d1 =
1
d d0 ( t )d0 ( t )2 ( 21 ) d0
2 1
( t )
1
0.248 0 (0.1)0 (0.1)2 ( 21 61 ) 56.5 = 35.4 in 2
2
1
s
6 (0.1)
5. Solve for d1
d1 =
d1 = d0 + ( t ) (1 )d0 + d1

d1 = 0 + (0.1) (1 21 )56.5 +
1
2
( 35.4 ) = 4.59 in s

velocity for the next time step (t = 0.2 s).
Page 369
Repeating Step 3:
F '2 = F2 +
F '2 = 60 +
M
d + ( t )d1 + ( 21 ) ( t )2 d1
2 1
( t )
1.77
0.248 + (0.1)4.59 + ( 21 61 ) (0.1)2 ( 35.4 ) = 934 lb
2
1
6 (0.1)
F '1 934
=
= 0.825 in
K ' 1,132
Repeating Step 4:
d1 =
d2 =
1
d d1 ( t )d1 ( t )2 ( 21 ) d1
2 2
( t )
d2 =
1
0.825 0.248 (0.1)4.59 (0.1)2 ( 21 61 ) 35.4 = 1.27 in 2
2
1
s
6 (0.1)
Repeating Step 5:
d2 = d1 + ( t ) (1 )d1 + d2
d2 = 4.59 + (0.1) (1 21 )35.4 +
1
2
(1.27 ) = 6.42 in s
The following table summarizes the results for the time steps through t = 0.5 seconds.
t (s)
F(t) lb
di (in)
d (in/s2)
d (in/s)
i
100
56.6
0.1
80
35.4
4.59
0.248
0.2
60
1.27
6.42
0.825
0.3
48.6
-26.2
5.17
1.36
0.4
45.7
-42.2
1.75
1.72
0.5
42.9
-42.2
-2.45
1.68
Page 370
Natural Frequencies of a One-Dimensional Bar
Before solving the structural stress dynamic analysis problem, lets consider
how to determine the natural frequencies of continuous elements. Natural frequencies are necessary in vibration analysis and important when choosing a
proper time step for a structural dynamics analysis.
Natural frequencies are obtained by solving the following equation:
Md + Kd = 0
The standard solution for d is given as d (t ) = d ' e it where d ' is the part of the
nodal displacement matrix called natural modes that is assumed to independent
of time, i is the standard imaginary number, and is a natural frequency.
Differentiating the above equation twice with respect to time gives:
d = d ' 2 e it
Substituting the above expressions for d and d into the equation of motion gives:
M 2d ' e it + Kd ' e it = 0
Combining terms gives:
e it K 2M d ' = 0
Since eit is not zero, then:
(K M ) d ' = 0
2
The above equations are a set of linear homogeneous equations in terms of displacement mode d ' . There exist a non-trivial solution if and only if the determinant of the coefficient matrix of d ' is zero.
K 2M = 0
Page 371
Example Problem
Determine the first two natural frequencies for the bar shown in the figure below. Assume the bar has a length 2L, modulus of elasticity E, mass density ,
and cross-sectional area A.
Lets discretize the bar into two elements each of length L as shown below. We
need to develop the stiffness matrix and the mass matrix (either the lumpedmass of the consistent-mass matrix). In general, the consistent-mass matrix has
resulted in solutions that compare more closely to available analytical and experimental results than those found using the lumped-mass matrix. However,
when performing a long hand solution, the consistent-mass matrix is more difficult and tedious to compute; therefore, we will use the lumped-mass matrix.
The elemental stiffness matrices are:

1
1 1
k(1) = AE

L 1 1
k(2) = AE

L
1 1
1 1
The global stiffness matrix is:

1 1 0
AE
[K ] = L 1 2 1
0 1 1
Page 372
The lumped-mass matrices are:

1
(1) =
m
AL 1 0
2 0 1
The global lumped-mass matrix is:
[M ] =
(2) =
m
AL 1 0
2 0 1
1 0 0
0 2 0
2
0 0 1
AL
Substituting the above stiffness and lumped-mass matrices into the natural frequency equation
(K M ) d ' = 0
2
and applying the boundary condition d1x = 0 (or d '1 = 0 ) gives:

AE 2 1
0 d '2 0
2 AL 2
1 1
0 1 d ' = 0
L
2
Set the determinant of the coefficient matrix equal to zero as:

AE
L
2 1
AL 2 0
1 1
0 1 = 0
2
where = 2. Dividing the above equation by AL and letting = E

2
L2
gives:
=0
2
Evaluating the determinant of the above equations gives:
= 2 2
or
1 = 0.60
2 = 3.41
Page 373
For comparison, the exact solution gives = 0.616, whereas the consistentmass approach yields = 0.648 . Therefore, for bar elements, the lumped-mass
approach can yield results as good as, or even better than, the results from the
consistent-mass approach. However, the consistent-mass approach can be
mathematically proven to yield an upper bound on the frequencies, whereas the
lumped-mass approach has no mathematical proof of boundedness.
The first and second natural frequencies are given as:
1 = 1 = 0.77
2 = 2 = 1.85
The term may be computed as:
E
30 106
= 2 =
= 4.12 106 s 2
2
L (0.00073)(100)
Therefore, first and second natural frequencies are:
1 = 1.56 103 rad / s
2 = 3.76 103 rad / s
In general, an n-degree-of-freedom discrete system has n natural modes and

frequencies. A continuous system actually has an infinite number of natural
modes and frequencies. The lowest modes and frequencies are approximated
most often; the higher frequencies are damped out more rapidly and are usually
less important.
Substituting 1 into the following equation
AE
L
gives:
0 d '2 0
2 1
2 AL 2
=
1 1
2 0 1 d '3 0
1.4 d '2(1) d '3(1) = 0
d '2(1) + 0.7 d '3(1) = 0
where the superscripts indicate the natural frequency. It is customary to specify

the value of one of the natural modes d ' for a given i or i and solve for the
(1)
remaining values. For example, if d '(1)
3 = 1 than the solution for d '2 = 0.7 . Similarly, if we substitute 2 and let d '(2)
3 = 1 the solution of the above equations gives
Page 374
d '(2)
2 = 0.7 . The modal response for the first and second natural frequencies are
shown in the figure below.
The first mode means that the bar is completely in tension or compression, depending on the excitation direction. The second mode means that bar is in compression and tension or in tension and compression.
Time-Dependent One-Dimensional Bar Example

Consider the one-dimensional bar system shown in the figure below.
Assume the boundary condition d1x = 0 and the initial conditions d0 = 0 and d0 =
0. Let = 0.00073 lb-s2/in.4, A = a in.2, E = 30 x 106 psi, and L = 100 in. The bar
will be discretized into two elements as shown below.
Page 375
The elemental stiffness matrices are:

1
1 1
k(1) = AE

L 1 1
1 1
k(2) = AE

L 1 1
The global stiffness matrix is:

1 1 0
AE
[K ] = L 1 2 1
0 1 1
The lumped-mass matrices are:
1
(1) =
m
AL 1 0
2 0 1
(2) =
m
AL 1 0
2 0 1

1 0 0
[M ] = 2 0 2 0
0 0 1
Substitute the global stiffness and mass matrices into the global dynamic equations gives:
AL
1 1 0 d1x
1 0 0 d1x R1
AE
d + AL 0 2 0 d = 0
1
2
1
2x
2x
L
2

0 1 1 d3 x
0 0 1 d3 x F3 (t )
where R1 denotes the unknown reaction at node 1.
For this example, we will used the central difference method, because it is
easier to apply, for the numerical time integration. It has been mathematically
shown that the time step t must be less than or equal to two divided by the
highest natural frequency.
Page 376
max
For practical results, we should use a time step defined by:

t
3 2
4 max
An alternative guide (used only for a bar) for choosing the approximate time step
is:
t =
L
cx
where L is the element length, and c x =
Ex
is called the longitudinal wave
velocity. Evaluating the time step estimates gives:
t =
3 2
1.5
= 0.40 103 s
=
3
4 max 3.76 10
L
100
=
= 0.48 103 s
cx
30 106
0.00073
Guided by these estimates for time step, we will select t = 0.25 x 10-3 s.
t =

1. Given: d1x = 0 (fixed end), all nodal displacements, velocities are zero
at time t = 0, d0 = 0 and d0 = 0, also d1x = 0 at all times.
2. Solve for d at t = 0.
0
d0 = M1 (F0 Kd0 )
d
2 12 0 0 AE 2 1 0
d0 = 2 x =
0 1 1,000 L 1 1 0

AL
d
3 x t =0

Applying the boundary conditions d1x = 0 and d1x = 0 and simplifying gives:
Page 377
2000 0 0 in
d
d0 = 2 x =
=
2

AL
d
3 x 0
1 27,400 s
( t )
d 1 = d0 ( t )d0 + d0
2
Applying the initial conditions for d0 and d0 and d1x from Step 2 gives:
2
0
d 2 x
(0.25 10 3 )2 0
3
in
0
(0.25
10
)(0)
=
=
3
d
2
27,400
0.856
10
3 x 1
4. Solve for d1 at t = t using the value of d-1 from Step 3
2
2
d1 = M1 ( t ) F0 + 2M ( t ) K d0 Md 1
2
0 2(0.073) 2 0
d 2 x
2 12 0
3
0.25
10
2
1,000
0 1
d3 x 1 0.073 0 1
2
0
2 1 0 0.073 2 0
0.25 10 3 30 10 4
3
2 0 1 0.856 10
1 1 0
)(
Simplifying the above equation gives:

0
0
d 2 x

2 12 0
=
3
3
d3 x 1 0.073 0 1 0.0625 10 0.0312 10
The nodal displacements at t = 0.25 x 10-3 are:
0
d 2 x
in
=
3
d
0.858
10
3 x 1
5. With d0 given and d1 determined in Step 4 solve for d2
2
2
d 2 = M1 ( t ) F1 + 2M ( t ) K d1 Md0
}
Page 378
2 0
2(0.073) 2
d 2 x
2 12 0
3
0.25
10
=
2
1000
0
d3 x 2 0.073 0 1
2
0
2 1
0.073 2
0.25 10 3 30 10 4
3
2 0
1 1 0.858 10
)(
0
1
0 0

1 0
0
d 2 x
0.0161 103
2 12 0
=
0 1 0.0625 103
3
d
0.073

0.0466 10
3 x 2
0.221 10 3
d 2 x
in
=
3
d3 x 2 2.99 10
6. Solve for d1

d1 = M1 (F1 Kd1 )
d2 x
0
1
2 12 0 0
4 2
(30
10
)
=
3
d3 x 1 0.073 0 1 1000
1 1 0.858 10
d2 x 3,526 in
=
2
d3 x 1 20,345 s
d d0
d1 = 2
2( t )
Page 379
0.221 10 3 0

2.99
10
d1 =
2 0.25 10 3
d2 x 0.442 in
=
s
d3 x 1 5.98

velocity for other time steps.
Repeating Step 5:
2
2
d3 = M1 ( t ) F2 + 2M ( t ) K d 2 Md1
d 2 x
2 12 0
3
=

0.25 10
d3 x 3 0.073 0 1
0.25 10
)(
2
0 2(0.073) 2 0
2
1000

0 1
0
2 1 0.221 10 3 0.073 2 0
30 10
3
2 0 1 0.858 103
1 1 2.99 10
4

0
d 2 x
0.080 10 3
2 12 0
=

0 1 0.0625 10 3 +
3
d
0.073

0.135 10
3 x 3
1.096 10 3
d 2 x
in
=
3
d
5.397
10
3 x 3
Repeating Step 6:
d2 = M1 (F2 Kd 2 )
d2 x
1 0.221 10 3
2 12 0 0
4 2
(30
10
)
=

3
1 1 2.99 10
d3 x 2 0.073 0 1 1000
Page 380

d2 x 10,500
=
in s 2
d3 x 2 4,600
Repeating Step 7:
d d1
d2 = 3
2( t )
1.096 103
0
3
3
0.858
10
5.397
10
d2 =
2 0.25 10 3
d2 x 2.192 in
=
s
d3 x 2 9.078
Beam Element Mass Matrices and Natural Frequencies
We will develop the lumped- and consistent-mass matrices for time-dependent

beam analysis. Consider the beam element shown in the figure below.
The basic equations of motion are:
{F (t )} = [K ] {d } + [M ]{d}
where the stiffness matrix is:
d1y
12
6L
EI
k = 3
L 12
6L
d2y
6L
4L2
12
6L
6L
2L2
12
6L
6L
2L2
6L
4L2
and the lumped-mass matrix is:

Page 381
d1y
1 d 2 y 2
AL 0
=
m
2 0
0 0 0
0 0 0
0 1 0
0 0 0
The mass in lumped equally into each transitional degree of freedom; however,
the inertial effects associated with any possible rotational degrees of freedom is
assumed to be zero. A value for these rotational degrees of freedom could be
assigned by calculating the mass moment of inertia about each end node using
basic dynamics as:
I=
AL3
24
The consistent-mass matrix can be obtained by applying
= [N ] [N ] dV
m
T
N1
N
L
= 2 [N1 N2 N3
m
0 A
N3
N4
where
N1 =
1
2 x 3 3 x 2L + L3
3
L
N3 =
1
2 x 3 + 3 x 2L
3
L
N4 ] dA dx
N2 =
1 3
x L 2 x 2L2 + xL3
3
L
N4 =
1 3
x L x 2L2
3
L
Substituting the shape functions into the above mass expression and integrating
gives:
Page 382
156 22L
2
AL 22L 4L
]=
[m
420 54
13L
2
13L 3L
54
13L
13L 3L2
156 22L
22L 4L2
Example Problem
Determine the first natural frequency for the beam shown in the figure below.
Assume the bar has a length 2L, modulus of elasticity E, mass density , and
cross-sectional area A.
Lets discretize the beam into two elements each of length L. We will use the
lumped-mass matrix. We can obtained the natural frequencies by using the following equation.
K 2M = 0
The boundary conditions are d1x = d3x = 0 and 1 = 3 = 0. Therefore the global
stiffness matrix is:
d 2 y 2
EI 24 0
L3 0 8L2
K=
M=
AL 2 0
2 0 0
Substituting the global stiffness and mass matrices into the global dynamic equations gives:
EI
L3
24 0
1 0
2
0 8L2 AL 0 0 = 0
Page 383
Dividing by AL and simplify
2 =
24EI
AL4
or
4.90
L2
EI
A
The exact solution for the first natural frequency is:
5.59
L2
EI
A
Example Problem
Determine the first natural frequency for the beam shown in the figure below.
Assume the bar has a length L = 30 in, modulus of elasticity E = 3 x 107 psi,
mass density = 0.00073 lb-s2/in, and cross-sectional area A 1 in2, moment of
inertia I = 0.0833 in4, and Poissons ratio = 0.3.
Lets discretize the beam into two elements each of length L = 15 in. We will use
the lumped-mass matrix. We can obtained the natural frequencies by using the
following equation.
K 2M = 0
The problem is similar to the previous problem. The solution for the first natural
frequency is:
3.148
L2
EI
A
The exact solution for the first natural frequency is:

Page 384
3.516
L2
EI
A
According to vibration theory for a clamped-free beam, the higher natural frequencies to the first natural frequency is given as:
2
= 6.2669
1
3
= 17.5475
1
The figure below shows the first, second, and third mode shapes corresponding
to the first three natural frequencies for the cantilever beam.
Page 385
The table below shows various finite element solutions compared to the exact solution.
1, (rad/s)
2, (rad/s)
228
1,434
Using 2 elements
205
1,286
Using 6 elements
226
1,372
Using 10 elements
227.5
1,410
Using 30 elements
228.5
1,430
Using 60 elements
228.5
1,432
Exact Solution
Finite Element Solution
Truss and Plane Frame Analysis

The dynamics of trusses and plane frames are preformed by extending the
concepts of bar and beam element. The truss element requires the same transformation of the mass matrix from local to global coordinates as that used for the
stiffness matrix given as:
T
m = TT m
Truss Elements
Since the motion of the element is now in two- or three-dimension, the bar element mass matrix must be reformulated to account for the axial and transverse
inertial properties in the x and y directions.
Page 386
Considering two-dimensional motion, the axial and the transverse displacement

are given as:
d1x

x 0 d1y
0
u 1 L x
=

L x 0 x d2 x
v L 0

d 2 y
The shape functions for the matrix are:
0
1 L x
L x
0
[N ] = L
0
0 x
The consistent-mass matrix can be obtained by applying:

] = [N ]T [N ]dV
[m
V
AL 0
m =
6 1
0 1 0
2 0 1
0 2 0
1 0 2
The lumped-mass matrix for two-dimensional motion is obtained by simply lumping mass at each node and remembering that mass is the same in both the x and
y directions, The lumped-mass matrix is:
1
AL 0
=
m
2 0
0 0 0
1 0 0
0 1 0
0 0 1
Frame Elements
The plane frame element requires combining the bar and beam elements to
obtain the local mass matrix. There are six degrees of freedom associated with a
plane frame element.
Page 387
The plane frame analysis requires first expanding and then combining the bar
and beam mass matrices to obtain the local mass matrix. The bar and beam
mass matrices are expanded to a 6 x 6 and superimposed. Combining the local
axis consistent-mass matrices for the bar and beam elements gives:
2 6
0
0
] = AL
[m
1
6
0
0
156
22L
420
22L
420
420
0
54
420
13 L
4L
420
0
13 L
420
420
3 L
420
0
0
2
0
0
0
54
420
13 L
420
0
156
420
22 L
420
13 L
420
3 L2
420
0
22L
420
4 L2
420
The resulting lumped-mass matrix for a plane frame element is give as:
d1x d1y d 2 x d 2 y 2
1
1
0
AL 0
]=
[m
2 0
0
0 0 0 0 0
1 0 0 0 0
0 0 0 0 0
0 0 1 0 0
0 0 0 1 0
0 0 0 0 0
The global mass matrix for the plane frame element arbitrarily oriented in x-y coordinates is transformed by:
Page 388
T
m = TT m
where the transformation matrix is given as:
C S 0 0 0 0
S C 0 0 0 0
0 0 1 0 0 0
T =
0 0 0 C S 0
0 0 0 S C 0
0 0 0 0 0 1
Long-hand solution to the truss and frame problem are quite tedious and lengthy;
therefore, we will use a computer problem to generate approximation for the motion of truss and frame structures.
Plane Stress/Strain Elements

The plane stress/strain constant-strain triangle consistent-mass matrix is obtained using the shape functions given below as:
[N ] = 0i
0
Ni
Nj
0
0
Nj
Nm
0
0
Nm
The consistent-mass matrix can be obtained by applying:

] = [N ]T [N ]dV
[m
V
where dV = tdA
The CST global consistent-mass Matrix is:
Page 389
2
0
tA 1
[m ] =
12 0
1
0 1 0 1 0
2 0 1 0 1
0 2 0 1 0
1 0 2 0 1
0 1 0 2 0
1 0 1 0 2
Example Problem
Determine the motion of the frame structure shown below.
Assume the modulus of elasticity E = 3 x 107 psi. The mass densities are obtained by dividing the total mass of each floor by the cross-sectional area and
length the element. For example, consider the element 6:
M6 =
2
W6 (104 psf )( 30 ft )(15 ft )
=
= 121 lb s
in
g
386.4 in s 2
121 lbs in
lb s 2 4
0.0136
6 =
=
in
(24.7 in 2 )(360 in )
2
Page 390
Use Newmarks method with = and = .

The following is the input file for WinFElt.
problem description
title=dynamic frame analysis nodes=8 elements=9 analysis=transient
analysis parameters
beta=0.25 gamma=0.5 alpha=0.0 duration=0.8 dt=0.05
nodes=[8,6,3] dofs=[Tx] mass-mode=lumped
nodes
1 x=0
y=0
constraint=fixed
2 x=360 y=0
3 x=0
y=180 constraint=free force=f1
4 x=360
5 x=0
y=300
force=f2
6 x=360
7 x=0
y=420
force=f3
8 x=360
beam elements
1 nodes=[1,3] material=wall_bottom
2 nodes=[3,5] material=wall_top
3 nodes=[5,7]
4 nodes=[7,8] material=floor_top
load=top_wt
5 nodes=[5,6] material=floor_bottom load=bottom_wt
6 nodes=[3,4]
load=bottom_wt
7 nodes=[8,6] material=wall_top
8 nodes=[6,4]
9 nodes=[4,2] material=wall_bottom
material properties
wall_bottom A=13.2 Ix=249 E=30e6 rho=0.0049
wall_top
A=6.2 Ix=107 E=30e6 rho=0.0104
floor_top
A=12.3 Ix=133 E=30e6 rho=0.01315
floor_bottom A=24.7 Ix=237 E=30e6 rho=0.0136
distributed loads
top_wt
direction=perpendicular values=(1,-62.5) (2,-62.5)
bottom_wt direction=perpendicular values=(1,-130) (2,-130)
forces
f1 Fx=1000*(t < 0.2 ? 25*t : 5)
f2 Fx=800*(t < 0.2 ? 25*t : 5)
f3 Fx=500*(t < 0.2 ? 25*t : 5)
constraints
fixed Tx=c Ty=c Rz=c
free Tx=u Ty=u Rz=u
end
Page 391
The following is the WinFElt output

-----------------------------------------------------------------time
Tx(8)
Tx(6)
Tx(4)
-----------------------------------------------------------------0
0
0
0
0.05
0.0054775
0.0046834
0.0047332
0.1
0.032795
0.028894
0.026946
0.15
0.10231
0.092341
0.078059
0.2
0.23314
0.21232
0.16267
0.25
0.43808
0.39636
0.27818
0.3
0.71526
0.63623
0.41441
0.35
1.0528
0.91484
0.56253
0.4
1.4335
1.2132
0.71742
0.45
1.8341
1.5124
0.87255
0.5
2.2269
1.7954
1.0185
0.55
2.5809
2.0464
1.1461
0.6
2.8674
2.2528
1.2504
0.65
3.0641
2.4041
1.3297
0.7
3.1589
2.4924
1.3824
0.75
3.1498
2.511
1.4031
0.8
3.0433
2.4556
1.3844
Page 392
Example Problem
Determine the motion of the frame structure shown below. This problem is the
same as the previous example, except for the loading function F(t) and the time
duration.
Page 393
The following is the WinFElt output

-----------------------------------------------------------------time
Tx(8)
Tx(6)
Tx(4)
-----------------------------------------------------------------0
0
0
0
0.05
0.0054775
0.0046834
0.0047332
0.1
0.032795
0.028894
0.026946
0.15
0.10231
0.092341
0.078059
0.2
0.21123
0.19359
0.14374
0.25
0.32881
0.29952
0.18933
0.3
0.43722
0.38244
0.20996
0.35
0.52949
0.43491
0.22408
0.4
0.59182
0.45828
0.23644
0.45
0.61612
0.45612
0.23879
0.5
0.59851
0.42714
0.22158
0.55
0.53509
0.37319
0.1881
0.6
0.42314
0.30076
0.1489
0.65
0.27506
0.21612
0.11208
0.7
0.11427
0.11995
0.07504
0.75
-0.041956
0.012366
0.028051
0.8
-0.18758
-0.10174
-0.034383
0.85
-0.31656
-0.21548
-0.10737
0.9
-0.41638
-0.31995
-0.17587
0.95
-0.48337
-0.40479
-0.22728
1
-0.52583
-0.45861
-0.25632
1.05
-0.55228
-0.47809
-0.26688
1.1
-0.55737
-0.46626
-0.26286
1.15
-0.5321
-0.42613
-0.24354
1.2
-0.4741
-0.3581
-0.20324
1.25
-0.38481
-0.26715
-0.13927
1.3
-0.26441
-0.16487
-0.062339
1.35
-0.11615
-0.058684
0.0067415
1.4
0.046209
0.05035
0.051067
1.45
0.20774
0.16045
0.076603
1.5
0.35642
0.26056
0.10478
1.55
0.48124
0.34101
0.15095
1.6
0.56682
0.40072
0.20559
1.65
0.60441
0.44251
0.24564
1.7
0.59778
0.46276
0.25798
1.75
0.55525
0.45409
0.24836
1.8
0.48211
0.41509
0.22947
1.85
0.38454
0.3486
0.20213
1.9
0.27609
0.25766
0.15744
1.95
0.16472
0.14464
0.088264
2
0.046915
0.019038
0.0033795
Page 394
Problems
22. Do problems 16.5 and 16.11 on pages 611-613 in your textbook A First
Course in the Finite Element Method by D. Logan.
23. Do problems 16.14 and 16.16 on pages 613 - 614 in your textbook A First
Course in the Finite Element Method by D. Logan using WinFElt.
Page 395

Structural DynamicsFEM CourseNotes Chapter16

Încărcat de

Informații document

Drepturi de autor

Formate disponibile

Partajați acest document

Partajați sau inserați document

Opțiuni de partajare

Vi se pare util acest document?

Este necorespunzător acest conținut?

Drepturi de autor:

Formate disponibile

Structural DynamicsFEM CourseNotes Chapter16

Încărcat de

Drepturi de autor:

Formate disponibile

Structural Dynamics

Dynamics of a Spring-Mass System

CIVL 7117 Finite Elements Methods in Structural Mechanics

where xm denotes the maximum displacement (or amplitude of the vibration).

The frequency in hertz (Hz = 1/s) is f = 1/ = /(2).

Lets derive the finite element equations for a time-dependent (dynamic)

We will consider the linear bar element shown below.

CIVL 7117 Finite Elements Methods in Structural Mechanics

Step 2 - Select a Displacement Function

A linear displacement function is assumed in the x direction.

where N1 and N2 are the interpolation functions gives as:

Step 3 - Define the Strain/Displacement and Stress/Strain Relationships

The stress-displacement relationship is:

The bar element is typically not in equilibrium under a time-dependent force;

CIVL 7117 Finite Elements Methods in Structural Mechanics

In matrix form, the above equations are:

If we replace {f } with [ k ]{d } we get:

where the elemental stiffness matrix is:

CIVL 7117 Finite Elements Methods in Structural Mechanics

Substituting {Xe} for {X} gives:

The second derivative of the u with respect to time is:

CIVL 7117 Finite Elements Methods in Structural Mechanics

Step 5 - Assemble the Element Equations and Introduce

Numerical Integration in Time

We now introduce procedures for the discretization of the equations of motion

CIVL 7117 Finite Elements Methods in Structural Mechanics

Central Difference Method

We generally want to evaluate the nodal displacements; therefore, we rewrite the

Combining terms in the above equations gives:

Procedure for solution:

2. If the acceleration is not given, solve for d0 at t = 0.

5. With d0 given and d1 determined in Step 4 solve for d2

6. Solve for d1

d1 = M1 (F1 Kd1 )

8. Repeat Steps 5, 6, and 7 to obtain the displacement, acceleration, and

CIVL 7117 Finite Elements Methods in Structural Mechanics

The time-dependent forcing function is given as:

Consider the above spring-mass system as a single degree of freedom problem

Procedure for solution:

2. The initial acceleration at t = 0:

3. Solve for d-1 at t = -t

7. Solve for d1 using the central difference equation

8. Repeat Steps 5, 6, and 7 to obtain the displacement, acceleration, and

CIVL 7117 Finite Elements Methods in Structural Mechanics

di +1 = di + ( t ) (1 )di + di +1

d0 = M1 (F0 Kd0 )

Md i +1 = Md i + ( t )Mdi + ( t )2 M ( 21 ) di + ( t )2 [Fi +1 Kd i +1 ]

Dividing the above equation by (t)2 gives:

Procedure for solution of Newmarks Method:

1. Given: d0, d0 , and Fi(t)

d0 = M1 (F0 Kd0 )

CIVL 7117 Finite Elements Methods in Structural Mechanics

5. Solve for d1

The time-dependent forcing function is given as:

Consider the above spring-mass system as a single degree of freedom problem

Procedure for solution:

2. If the acceleration is not given, solve for d0 at t = 0:

d1 = d0 + ( t ) (1 )d0 + d1

6. Repeat Steps 3, 4, and 5 to obtain the displacement, acceleration, and

CIVL 7117 Finite Elements Methods in Structural Mechanics

Natural Frequencies of a One-Dimensional Bar

CIVL 7117 Finite Elements Methods in Structural Mechanics

The elemental stiffness matrices are:

The global stiffness matrix is:

2. If the acceleration is not given, solve for d0 at t = 0.

6. Solve for d1

d1 = M1 (F1 Kd1 )

7. Solve for d1 using the central difference equation

di +1 = di + ( t ) (1 )di + di +1

d0 = M1 (F0 Kd0 )

Md i +1 = Md i + ( t )Mdi + ( t )2 M ( 21 ) di + ( t )2 [Fi +1 Kd i +1 ]

1. Given: d0, d0 , and Fi(t)

d0 = M1 (F0 Kd0 )

5. Solve for d1

2. If the acceleration is not given, solve for d0 at t = 0:

d1 = d0 + ( t ) (1 )d0 + d1

d0 = M1 (F0 Kd0 )

6. Solve for d1