Documente Academic
Documente Profesional
Documente Cultură
Complex Notation
Free Vibrations
W − k( u + δ static ) = mu
&& (1)
which is the equation of motion for the undamped SDOF system. The objective is to solve
the equation of motion to determine the displacement of the mass as a function of time,
u(t), subject to the initial conditions of the system.
One approach to solving this partial differential equation is to assume a solution of the
form:
u( t ) = Ae rt (4)
Take the second derivative with respect to time of Eq. 4 yielding:
u t ) = r 2 Ae rt
&&( (5)
mr 2 Ae rt + kAe rt = 0 (6a)
mr 2 + k = 0 (6b)
k
r = ±i (7)
m
k k
i t −i t
u( t ) = Ae m + Be m (8)
We define the radical term in the exponent to be the circular natural frequency of
vibration of the system:
k
ωn = in rad/sec (9)
m
Note that the circular natural frequency is related to the natural frequency by:
ωn
fn = (10)
2π
1
Tn = (11)
fn
To evaluate the coefficients A and B in Eq. 8, we need two initial conditions. Usually,
these are the initial displacement and velocity of the mass:
u( t = 0) = u 0 (12a)
u& ( t = 0) = u& 0 (12b)
After substituting these expressions into Eq. 8 and its first derivative, we obtain:
u iu& u iu&
u( t ) = 0 + 0 e − iω n t + 0 − 0 e iω n t (13)
2 2ω n 2 2ω n
Figure 1 shows the displacement time history of this SDOF with the initial conditions of
u 0 = 1.0 and u& 0 = 0.0 . Note that the amplitude of the displacement does not diminish with
time because there is no means of attenuating energy within the system.
Figure 1 Displacement Time History for Free Vibrations of an Undamped SDOF System
Damped Free Vibrations
or
c k
&&u + u& + u = 0 (16b)
m m
u( t ) = Ae rt (17)
Take the first and second derivative with respect to time of Eq. 17 yielding:
u t ) = r 2 Ae rt
&&( (18b)
c k
r 2 Ae rt + rAe rt + Ae rt = 0 (19a)
m m
−c
2
c k
r= ± − (20)
2m 2m m
u( t ) = Ae r1t + Be r2 t (21)
We define a critical value of c such that the term inside the radical equals 0:
c crit = 2 km (22)
c c
β= = (23)
c crit 2 mω n
c
= ω nβ (24)
2m
r = − ω n β ± iω n 1 − β 2 (25)
[
u( t ) = e −ω nβt Ae −iω n 1− β 2 t
+ Be iω n 1− β 2 t
] (26)
To evaluate the coefficients A and B in Eq. 26, we need two initial conditions. Usually,
these are the initial displacement and velocity of the mass:
u( t = 0) = u 0 (27a)
After substituting these expressions into Eq. 26 and its first derivative, we obtain:
u i( u& 0 − ω n βu 0 ) −iω n u i(u& 0 − ω n βu 0 ) iω n
u( t ) = e −ω n βt + e 1− β 2 t
+ − e
1− β 2 t
0 0
2 2
2 2ω n 1 − β 2 2ω n 1 − β
(28)
1. β = 1 (critically damped)
r = −ω n (29)
and the partial differential equation has repeated roots. As a result, the solution takes the
form:
[ ]
u( t ) = u 0 + ( u& 0 + ω n u 0 )t e − ω n t (30)
Figure 2 shows the response of a critically damped SDOF for three initial conditions. The
initial displacement is equal to 1.0, but the initial velocity of the mass is 1.0, 0.0, and –1.0
for the three different cases. Notice that the motion quickly diminishes to zero because of
the large damping in the system.
The overdamped case is similar to the critically damped case. Eq. 28 can be used directly
since the roots are not repeated. Figure 3 below shows the response of an overdamped
SDOF for the same three initial conditions in Fig. 2. As in the case of the critically damped
SDOF, the displacement quickly diminishes to zero because of the large damping in the
system.
3. β < 1 (underdamped)
The case of most interest to us in soil dynamics problems is that in which the fraction of
critical damping is less than 1.0. Equation 28 may be used again to calculate the
displacement as a function of time. An example of a typical displacement time history for
an underdamped SDOF is shown in Fig. 4.
Notice that the displacement time history continues for many cycles of motion (i.e.
oscillatory). The motion can be considered to be a harmonic function modulated by a
decreasing exponential function. The origin of these two functions can be seen in Eq. 28.
Figure 4 Displacement Time History of an Underdamped SDOF System
ωd = ωn 1− β (31)
along the corresponding damped natural frequency and damped natural period, fd and Td,
respectively. For small values of β, ωd ≈ ωn.
Another widely used measure of the damping in a viscous system is the logarithmic
decrement:
1 u( t ) 2πβ
δ= ln = ≅ 2 πβ for small β (32)
n u( t + nTd ) 1 − β2
Parallel and Series Springs
It will sometimes be necessary to consider springs acting in parallel or series. This is easily
handled by determining the effective stiffness of a single, equivalent spring as shown
below.
m
u(t)
k1
1 1 1
= +
k eff k 1 k 2
k2
m
u(t)
k eff = k 1 + k 2
k1 k2
Forced Vibrations
[
u( t ) = e −ω nβt Ae −iω n 1− β 2 t
+ Be iω n 1− β 2 t
]+ u particular ( t) (34b)
Assume that the particular solution has the same form as the external force:
where C is a constant. As before we need to take the first and second derivatives with
respect to time to substitute into the equation of motion.
P
u particular ( t ) = e iΩt (39)
k − Ω m + iΩ c
2
1 2t 1 2t P
u ( t ) = e −ωn βt Ae − ωn −β + Be ωn −β +
1
i i
e iΩt (41)
k Ω 2 Ω
1− + 2iβ
2
ωn ωn
The transient component will include the transient response of the system due to the initial
displacement and velocity of the mass and the transient response of the system due to the
sudden application of the external force. The coefficients A and B may be determined by
setting Eq. 41 and its first time derivative equal to u 0 and u& 0 , respectively. The resulting
expressions for A and B are:
iω β + ω 1 − β 2 u − iω β − ω 1 − β 2 + Ω
n n 0 iu& 0 n n
A= + + P 1
2
2ω n 1 − β 2 2ω n 1 − β 2 2ω n 1 − β 2 k Ω Ω
1− + 2iβ
ωn2 ωn
(42a)
− iω β + ω 1 − β 2 u iω β − ω 1 − β 2 − Ω
n n 0 − iu& 0 n n
B= + + P 1
2
2ω n 1 − β 2 2ω n 1 − β 2 2ω n 1 − β 2 k Ω Ω
1− + 2iβ
ωn2 ωn
(42b)
Figure 5 shows the response of a single degree of freedom system for which the external
force has a circular frequency, Ω, equal to 1 rad/sec and the initial displacement and initial
velocity of the mass are both equal to zero.
Note that during the first 20 to 25 seconds the response of the system includes both a
transient component (the general solution) and a steady-state component (the particular
solution). After about 25 seconds the transient response has diminished to near zero and
the overall response is dominated by the steady-state response. Thus for many forced
vibration problems of interest, we can neglect the transient component of the solution
since it is not significant after the first few cycles of motion.
Figure 5 Response of SDOF System to Forced Vibrations
One way to characterize the response of the SDOF system is to relate the amplitude and
phase of the displacement to the external force. Once steady-state conditions are reached,
the ratio of the amplitudes is about 1.5 (displacement) to 1.0 (force). The displacement is
about 180 degrees out of phase with respect to the displacement. Later, we will see that
examining the amplitude and phase as a function of frequency is a convenient way to
characterize a single-degree-of-freedom system.
Rotating Mass Excitation
For machinery vibrations problems involved rotating machinery, the dynamic forces are
often the result of out-of-balance or eccentric masses as shown in the figure below.
me
Ω
The dynamic force resulting from the eccentric mass is given by:
P( t ) = m e eΩ 2 exp(iΩt ) (43)
Notice that the magnitude of the force is frequency dependent unlike the constant force
excitation defined in Eq. 33.
Transfer (Frequency Response) Functions
Output
H (Ω ) = (44)
Input
By normalizing the output of the system with respect to the input, we emphasize the
characteristics and response of the system over the characteristics of the output or input.
Let's define a transfer function between the steady-state displacement output and force
input of a SDOF system undergoing forced vibrations:
Displacement u ss ( t )
H (Ω ) = = (45a)
Force Pe iΩt
Pe iΩt
H (Ω ) = k − Ω miΩt+ iΩc
2
(45b)
Pe
1
H (Ω ) = (45c)
k − Ω m + iΩ c
2
1 1
H ( Ω) = (45d)
k Ω 2
Ω
1 − 2 + 2i β
ωn ωn
Notice that the transfer function is a complex-valued quantity meaning the response of the
SDOF system can be characterized by a magnitude and phase. Figure 6 shows the
magnitude and phase plots for a SDOF system expressed as a function of the normalized
frequency, Ω ωn .
Figure 6 Magnitude and Phase of Transfer Function
Equations 45c and 45d and Figure 6 can be used to provide insight into the parameters
that control the response of a SDOF in different frequency ranges. Note in Equations 45c
and d that when Ω→0, the transfer function reduces to:
1
H (Ω = 0) = (46)
k
Thus, the stiffness of the system controls the response at low frequency. As Ω→ωn, the
transfer function reduces to:
1 1
H (Ω = ω n ) = = (47)
iΩc 2ikβ
and the response of the system is controlled to a large extent by the damping in the
system. Finally, as Ω becomes large, the transfer function becomes:
1
H (Ω → ∞ ) ∝ (48)
− Ω2 m
and the response of the system is largely controlled by the mass (the inertia) of the system.
The transfer function defined above was expressed in terms of the displacement. Other
response quantities such as the velocity and acceleration of the mass can also be used to
define a transfer function for various applications. The names associated with each of
these transfer or frequency response functions are given in Table 1.
The right hand side is set equal to zero because there are no external forces applied to the
mass. Rearranging yields:
&& + cu& + ku = cz& + kz
mu (50)
z( t ) = Ae iΩt (51)
u( t ) = Be iΩt (52)
After differentiating, substituting, and solving, we obtain the solution for the steady-state
displacement of the mass:
k + iΩ c
u( t ) = Ae iΩt (53)
k − Ω 2 m + iΩ c
We can also define a transfer function between the displacement of the mass and the input
ground displacement:
u( t )
H (Ω ) = (54a)
z( t )
k + iΩ c
Ae iΩt
H (Ω) = k − Ω m +iΩiΩ
2
c (54b)
Ae t
k + iΩ c
H (Ω ) = (54c)
k − Ω 2 m + iΩ c
Ω
1 + 2 iβ
ωn
H (Ω ) = (54d)
Ω2 Ω
1 − 2 + 2 iβ
ωn ωn
As before we can express the complex-valued transfer function in terms of the magnitude
and phase:
Figure 7 Magnitude and Phase of Transfer Function for Ground Displacement
In many earthquake engineering problems,
the ground motion input at the base of the
system is specified in terms of a ground
m u(t)
acceleration.
z t ) = Ce iΩt
&&( (55)
k c
Furthermore, the response parameter of
interest is the relative displacement between
the base and the mass. The latter is important
because it is the relative displacements that
are proportional to the forces induced in the
structure. &&z( t )
y( t ) = u( t ) − z( t ) (56)
m(&&y + &&)
z + cy& + ky = 0 (55a)
or
&& + cy& + ky = − mz
my && (57b)
−m
y( t ) = Ce iΩt (58)
k − Ω 2 m + iΩ c
Finally, we can define the transfer function between the ground acceleration and the
relative displacement of the mass and base:
y( t )
H (Ω ) = (59a)
&&(
z t)
−m
H ( Ω) = (59b)
k − Ω m + iΩ c
1
−
ω 2n
H (Ω ) = (59c)
Ω2 Ω
1 − 2 + 2iβ
ωn ωn
Thus far we have assumed that the inputs and outputs of single-degree-of-freedom
(SDOF) systems have been harmonic functions of the form:
u( t ) = Ae iΩt (60a)
p( t ) = Be iΩt (60b)
This form is convenient for deriving transfer functions for SDOF systems, but is of
limited practical value by itself because many actual inputs and outputs are more
complex. We must combine the solutions we have developed for harmonic excitation
with another tool, the Fourier Transform, to allow us to determine the response of SDOF
systems to an arbitrary excitation. It is important to note that the linear (or equivalent
linear) nature of most soil dynamics problems enables us to use this approach.
Fourier Transforms
The Fourier Transform is a way to decompose arbitrary displacement and/or force time
histories into their harmonic components. Once decomposed, transfer functions can be
applied to the individual harmonic components. Finally, an inverse Fourier Transform can
be used to “reassemble” the individual harmonic responses to obtain the response time
history.
Let's assume we have a displacement or force time history which we denote as x(t). We
can express x(t) as a Fourier series:
∞
x( t ) = a 0 + ∑ (a n cos nΩ 0 t + b n sin nΩ 0 t ) (61a)
n =1
where
T
1
a 0 = ∫ x( t ) dt (61b)
T0
x( t ) cos(nΩ 0 t ) dt
2
T ∫0
an = (61c)
T
b n = ∫ x( t ) sin(nΩ 0 t ) dt
2
(61d)
T0
∞
x( t ) = c n ∑e
n =−∞
inΩ 0 t
(62)
∞
x ( t ) = ∫ X(Ω) e iΩt dΩ (63a)
−∞
X( Ω ) = ∫ x( t ) e − iΩt
dt (63b)
−∞
Equation 63b is the continuous Fourier transform used to convert a time history x(t) to the
frequency domain in which the signal is portrayed as a function of frequency. Note that
X(Ω) is a complex-valued function. Similarly, Eq. 63a is the inverse Fourier transform
enabling us to go from the frequency domain to the time domain.
N −1
X( k∆f ) = ∑ x( n∆t ) e − i 2 πk∆f∆t (64a)
n=0
1 N −1
x( n∆t ) = ∑ X( k∆f ) e i 2 πk∆f∆t (64b)
N n=0
where N is the total number of points in the time history and k and n are and frequency
and time indices, respectively.
The fast Fourier transform (FFT or IFFT) is simply a numerically efficient version of the
DFT. The algorithm used to calculate the FFT is most efficient when the number of
points being transformed is equal to an integer power of 2 (i.e., 2n where n is an integer).
Thus it is common to increase the increase the number of points to the next larger integer
power of 2 by "padding" the end of the record with a sufficient number of zeros.
There are several features of the Fourier Transform that are helpful to know when using it
as a tool to solve problems. Suppose that a force or displacement time history contains N
points (where N = 2n) with sampling interval ∆t and that the total length of the time
history is T. When the FFT algorithm is applied to that time history to transform it to the
frequency domain, the result will be N values in the frequency domain separated by:
1
∆f = (65)
T
The Nyquist frequency is the maximum frequency that can be accurately resolved by a
sampling interval ∆t and is equal to:
1
f Nyquist = (66)
2∆t
Finally, for real-valued time histories (the most common case) only the first N/2 + 1
points in the frequency domain are unique. The remaining points are complex conjugates
of values in the first half of the record. One must be extremely careful in selecting the
correct frequencies and mathematical operations for values in the second half of the
frequency domain record to obtain correct results from the inverse FFT algorithm.
Using Fourier Transforms we can decompose a complex time history into its harmonic
components. This enables us to apply the transfer functions we developed for harmonic
motion. We will use a 3-step approach:
Use the FFT to calculate the amplitude and phase spectra corresponding to a given input
time history. The input time history may be a force (active loading) or a ground motion
(passive loading).
p(t)→FFT→P(Ω)
Calculate the response of the SDOF system in the frequency domain using the transfer
(frequency response) function:
U(Ω)→IFFT→u(t)
e − ωn βt t ωnβτ
u (t ) = ∫e p(τ)sin (ω d (t − τ )) dτ (68)
mω d 0
where τ is a dummy variable. Interested readers are referred to Weaver et al. (1990) or
other texts on engineering vibrations for the derivation of Duhamel’s integral. Equation
68 may also be viewed as a convolution integral in which the forcing function p(t) is
convolved with the impulse response of a SDOF system.
Note that the forces exerted by the dashpot and the spring depend on the relative velocity
and displacement between the two masses, respectively. For the second mass, the
equation of motion is:
We can combine the two equations of motion into a system of equations and express
them in matrix form as follows:
m1 0 !u!1 c1 − c1 u! 1 k 1 − k 1 u 1 p1 (t )
0 + + =
m 2 !u! 2 − c1 c1 + c 2 u! 2 − k 1 k 1 + k 2 u 2 p 2 (t )
(70a)
or
!! + Cu! + Ku = p
Mu (70b)
One approach to determining the motion of the two masses is to formulate the solution
directly. Assume that the displacements of the masses and the external forces are of the
form:
u = Ue iΩt (71a)
and
After substituting Eqs. 71a and 71b into Eq. 70b and rearranging, we obtain:
[− Ω M + iΩC + K ]U = P
2
(72)
u
[
U = 1 = − Ω 2 M + iΩC + K ]
−1
P (73)
u 2
U = BP (74a)
where
1 K 22 − Ω 2 M 22 + iΩC 22 − K 12 − iΩC12
B(Ω ) = (74b)
D − K 21 − iΩC 21 K 11 − Ω M 11 + iΩC11
2
and
( )( )
D = K 11 − Ω 2 M11 + iΩC11 K 22 − Ω 2 M 22 + iΩC 22 − (K 12 + iΩC12 )2 (74c)
Equations 74b and 74c are the complex-valued frequency response functions for the
2DOF system.
Another solution to Eq. 70b is given by the normal mode solution or modal analysis. To
obtain the normal mode solution, it is convenient to assume proportional damping of the
form:
C = β K + γM (75)
The normal mode solution is calculated by solving the generalized eigenvalue problem:
(K − ω M ) Φ = 0
2
(76)
where ω 2 is a diagonal matrix containing the generalized eigenvalues and Φ contains the
eigenvectors. Each rth eigenvalue is the square of the circular natural frequency of the rth
mode of vibration of the N degree of freedom system. The corresponding eigenvector
gives the mode shape of the rth mode of vibration.
The generalized stiffness, mass, and damping matrices are given by:
m = Φ T MΦ (77a)
k = Φ T KΦ (77b)
c = Φ T CΦ (77c)
βω r γ
ζr = + r = 1" N (78)
2 2ω r
and the damped circular natural frequency of each mode is given by:
ωdr = ω r 1 − ζ 2r (79)
Finally, the frequency response functions for the multiple degree of freedom system are
defined as:
(Φ jr )(Φ kr )
B jk (Ω ) = ∑
N
(80)
r =1 k r − Ω 2 m r + iΩc r
Complex Notation
Consider a function of the form:
x (t ) = Xe iΩt (1)
X = a + bi (2a)
X = X e iθ (2b)
x (t ) = X e i (Ωt +θ ) (3a)
Thus, using x (t ) = Xe iΩt and taking either the real or imaginary component is equivalent
to using either x (t ) = X cos(Ωt + θ) or x (t ) = X sin (Ωt + θ) , respectively. The primary
advantage of using the exponential form is that derivatives with respect to time are
conveniently manipulated.