Appendix B

Some Aspects of Structural

Dynamics

This Appendix deals with some aspects of the dynamic behavior of SDOF and MDOF. It
starts with the formulation of the equation of motion of SDOF systems. Next, the effects
that viscous and Coulomb damping have on the dynamic behavior of SDOF are studied.
The solution of equations of motion by means of the linear acceleration method (Newmark’s
method) is what follows. The energy formulation finishes with the study of SDOF systems.
Then, similar studies are carried out for MDOF systems.
Since Chapters 3 and 4 deal basically with the concepts included in this appendix, other
important topics (e.g., response of viscously damped systems to harmonic input, modal
analysis or dynamics of distributed-parameter systems) are not included in this appendix.

B.1 Single-Degree-of-Freedom (SDOF) Systems

B.1.1 General formulation

Fig. B.1a shows a typical single-story building (SSB). Making the usual static and cinematic
condensations, it can be assumed that the lateral dynamic behavior of this structure can be
described by a single linear degree of freedom (DOF) model. This DOF is the horizontal
displacement x, as illustrated in Fig. B.1b.

B.1.2 Equation of motion with viscous damping

The SDOF mechanical model for the frame of Fig. B.1a is shown in Fig. B.2a.
From the free body diagram of Fig. B.2b, in which the forces acting on the system of
Fig. B.2a are shown, the equation of dynamic equilibrium –equation of motion– can be
expressed as

163
164 B. Some Aspects of Structural Dynamics

x(t ) x(t )
P(t ) EIb

EIc EIc
H
c

L
(a) Single-degree-of-freedom (b) Assumed motion of the frame
(SDOF) frame

Figure B.1 Frame modelled as a single-degre-of-freedom (SDOF) system

FI + FD + FS = P (t) (B.1)

where FI is the inertia force (mẍ), FD is the force due to the viscous damping (cẋ), and FS is
the force due to the deformation of the spring (kx), P (t) represents the external time-varying
force acting on the structure and t is the time. The equation (B.1) can also be written as

mẍ + cẋ + kx = P (t) (B.2)

where m is the mass of the structure, c is the damping coefficient, k is the stiffness coefficient
and x is the displacement, as a function of time t, of the structure. Dots above x represent
the derivatives with respect the time; hence ẋ represents the velocity and ẍ represents the
acceleration.
For the particular case of the structure depicted in Fig. B.1 the stiffness coefficient k can
be calculated with the following expression:

12(6Ib Ic H + Ic2 L)E

k= (B.3)
H 3 (3Ib H + 2Ic L)
where E is the young modulus of the material, Ib and Ic are the moments of inertia of the
beam and the column, respectively, and L and H are the length and the height of the beam
and the column, respectively.
On the other hand, the damping coefficient c plays an important role in the response of
the system shown in Fig. B.2a. The convention widely accepted is that this coefficient can
B.1. Single-Degree-of-Freedom (SDOF) Systems 165

x(t )

k FD (t )
m P(t) P(t)
FFI I(t(t))
c
µg = 0
FS (t )

(a) Basic model of a SDOF system (b) Dynamic equilibrium of forces

Figure B.2 Mechanical model of a SDOF with viscous damping

be considered as a visco-elastic damping, which, when multiplied by the velocity ẋ, yields
the damping force FD .
The effects of viscous and friction damping on the dynamic response of the system of
Fig. B.2a are studied next.

B.1.3 Free vibration with viscous damping

With damping added, as shown in Fig. B.2a, the free vibration (i.e., P (t) = 0) of a SDOF
system can be described for the following differential equation of motion

mẍ + cẋ + kx = 0 (B.4)

and the solution is given by

x = eλt (B.5)

Since Eq. (B.4) is an ordinary second-order differential equation, two valid values of λ are
to be expected. Differentiating Eq. (B.5) and substituting the resultant values of x, ẋ and
ẍ into Eq. (B.4) yields the following characteristic equation:

(mλ2 + cλ + k)eλt = 0 (B.6)

The exponential term in Eq. (B.6) cannot be zero, therefore the expression inside the
parenthesis must be so. Then,

mλ2 + cλ + k = 0 (B.7)
166 B. Some Aspects of Structural Dynamics

or

−c ± (c2 − 4mk)1/2
λ= (B.8)
2m
This last equation can written in the form

µ ¶1/2 µ ¶1/2 ³ ´1/2

c k k c
λ=− ± −1 (B.9)
2(km)1/2 m m 4km
Let be

ω = (k/m)1/2 = undamped natural frequency (B.10)

and

c c
ξ= 1/2
= (B.11)
2(km) 2mω

where ξ is the critic damping rate. The equation of motion (B.1.6) can be rewritten as

ẍ + 2ξω ẋ + ω 2 x = 0 (B.12)

and Eq. (B.9) becomes

λ = −ξω ± ω(ξ 2 − 1)1/2 (B.13)

The solution of Eq. (B.12) is

x = C1 eλ1 t + C2 eλ2 t (B.14)

where λ1 and λ2 are the values of λ of Eq. (B.13). The form of the solution depends on the
value of ξ.
If ξ > 1 (overdamped system), then (ξ 2 − 1)1/2 is real and the solution is

n h i h io
x = e−ξωt A exp ω(ξ 2 − 1)1/2 t + B exp −ω(ξ 2 − 1)1/2 t (B.15)

If ξ = 1 (critically damped system), then (ξ 2 − 1)1/2 is zero (i.e., both values of λ are
equal) and Eq. (B.14) is reduced to
B.1. Single-Degree-of-Freedom (SDOF) Systems 167

ξ = 2.0 (Overdam ped)

ξ = 1.0 (Critically dam ped)
2π / ω n
Displacement

ξ = 0.0 (Undam ped)

ξ = 0.5 (Underdam ped)
ξ = 0.1 (Underdam ped)

Tim e

Figure B.3 Free vibration of a SDOF system for different values of the critic damping rate ξ.

x = e−ωt (A + Bt) (B.16)

If ξ < 1 (underdamped system), then (ξ 2 − 1)1/2 is imaginary and is convenient to turn

the exponentials to the equivalent trigonometric functions. This way Eq. (B.14) becomes

n h i h io
x = e−ξωt A cos ω(1 − ξ 2 )1/2 t + B sin −ω(1 − ξ 2 )1/2 t (B.17)

In any case, constants A and B are determined from the initial values of x and ẋ.
Fig. B.3 shows different types of time-history responses for some values of ξ.
If ξ = 1, the system turns back to its equilibrium position without any oscillation. This
is the lowest value of parameter ξ that restricts the oscillation completely, and it is called as
the damping critic case. That is why parameter ξ is known as the critic damping rate.
If ξ > 1, it is the case of over damping, and the system returns to its equilibrium position
without any oscillation, but slower than the case of critic damping.
If 0 < ξ < 1, the system oscillates with a decaying amplitude and a constant frequency
equal to ω(1 − ξ 2 )1/2 , which is little smaller than the oscillation frequency without damping,
ω. This is the case of damping smaller than the critic (under damping), and it is the case
usually found. For this case, it is defined the natural damped frequency, in radians per second,
which is given by
168 B. Some Aspects of Structural Dynamics

ω d = ω(1 − ξ 2 )1/2 (B.18)

and the natural damped period, in seconds, is

2π
Td = (B.19)
ωd

The free response in the case of damping smaller than the critic, in terms of initial
conditions, can be written as

· µ ¶ ¸
−ξωt ẋ0 + ξωx0
x=e x0 cos ω d t + sin ω d t (B.20)
ωd

B.1.4 Free vibration with Coulomb damping

It has been told that damping in actual structures is due to many dissipative mechanisms
that act simultaneously, and a convenient approach is to idealize them by means of an
equivalent viscous damping. Although this approach is sufficient for the analysis of the
majority of structures, can be inappropriate when some special devices of friction have been
incorporated to a building to reduce its vibrations during an earthquake. In this section
the free vibration of systems under the presence of Coulomb friction forces is going to be
analyzed.
Coulomb damping results of the friction generated by the relative sliding of two dry
surfaces. The maximum friction force is equal to F = µN , where µ denotes the static and
dynamic friction coefficients, here considered equal, and N is the normal force between the
sliding surfaces (see Section 2.2). The friction force is assumed to be independent from
velocity once the motion is started. The direction of the friction force is always opposite to
motion, therefore its sign will change as the motion direction changes. For this reason, it is
necessary to establish and to solve two differential equations, one equation valid for motion
in one direction and the other equation valid when motion reverts.
Fig. B.4 shows a mass-spring system with the mass sliding on a dry surface, and the
free-body diagram for the mass which includes the inertia force. The equation that rules the
motion, from right to left, is

FI + FS = F (B.21)

or
B.1. Single-Degree-of-Freedom (SDOF) Systems 169

x(t )

k m
FFI I(t(t))
µg > 0 FS (t )

(a) SDOF system with dry friction (b) Dynamic Equilibrium of forces

Figure B.4 Mechanical model of a SDOF with Coulomb damping

mẍ + kx = F (B.22)

and its solution is

x(t) = A1 cos ωt + B1 sin ωt + xF (B.23)

where xF = F/k. For the mass motion from left to right the ruling equation is

mẍ + kx = −F (B.24)

and its solution is

x(t) = A2 cos ωt + B2 sin ωt − xF (B.25)

Constants A1 , B1 , A2 and B2 depend on the initial conditions of each half-cycle of successive

p
motion; ω = k/m and the constant xF can be interpreted as the static deformation of the
spring due to the friction force F . Each differential equation is linear, but the whole problem
is non-linear because the ruling equation changes each half-cycle of motion.
Considering now the motion of the system of Fig. B.4 it will be started with some initial
conditions and will be continued until the motion stops. At time t = 0, the mass moves a
distance x(0) towards the right direction and it is set free from rest, such that ẋ(0) = 0. For
the first half-cycle of motion, Eq. (B.23) applies with constants A1 and B1 determined from
the initial conditions at t = 0:
170 B. Some Aspects of Structural Dynamics

Envlope of the displacement

response with Coulomb
damping
Displacement

µ N/k

-µ N/k

Free vibration with

Coulomb damping

ξ = 0.0 (Undamped)
T = 2π /ω

Time

Figure B.5 Free vibration of a SDOF system with Coulomb dry friction

A1 = x(0) − xF B1 = 0

Substituting these values into Eq. (B.23) yields

x(t) = [x(0) − xF ] cos ωt + xF 0 ≤ t ≤ π /ω (B.26)

This is plotted in Fig. B.5; it is a cosine function with an amplitude = x(0) − xF and
shifted in the positive x direction by xF . Eq. (B.26) is valid until the velocity becomes zero
again at t = π/ω = T /2 (see Fig. B.5); at this instant x = −x(0) + 2xF .
Starting now in this position on the left end, the mass now moves to the right with its
motion described by Eq. (B.25). Constants A2 and B2 are determined from the established
conditions at the beginning of this half-cycle:

A2 = x(0) − 3xF B2 = 0

Substituting these into Eq. (B.25) yields

x(t) = [x(0) − 3xF ] cos ωt − xF π /ω ≤ t ≤ 2π /ω (B.27)

B.1. Single-Degree-of-Freedom (SDOF) Systems 171

This is plotted in Fig. (B.5); it is a cosine function with a reduced amplitude = x(0)−3xF
and shifted in the negative x direction by xF . Eq. (B.27) is valid until the velocity becomes
zero again at t = 2π/ω = T (see Fig. B.5); at this time x = x(0) − 4xF .
At t = 2π/ω the motion reverses and it is described by Eq. (B.23), which, after evaluating
constants A1 and B1 , becomes

x(t) = [x(0) − 5xF ] cos ωt + xF 2π /ω ≤ t ≤ 3π /ω (B.28)

This is a cosine function with a reduced amplitude = x(0) − 5xF and shifted, as before,
in the positive direction x by xF .
The time taken for each half-cycle is π/ω and the duration of a full cycle, the natural
period of vibration , is

2π
T = (B.29)
ω
It can be observed that the natural period of a system with Coulomb damping is the same
as for the system without any damping. On the contrary, the viscous damping has the effect
of lengthening the natural period [Eqs. (B.18) and (B.19)].
At each cycle of motion, the amplitude is reduced by 4xF ; i.e., the displacements xk and
xk+1 at two successive maxima are related by

xk+1 = xk − 4xF (B.30)

Therefore the envelopes of the displacement-time curves are straight lines, as shown in Fig.
B.5, instead of the exponential functions for systems with viscous damping.
If at end of any half-cycle the motion has an amplitude lesser than xF , this means that
in this point the spring force acting on the mass is less than the friction force, kx < F , and
motion ceases. In Fig. B.5 this occurs at the end of the second and a half cycle. The final
rest position of the mass is shifted from its original equilibrium position and represents a
permanent deformation in which the friction force and the spring force are locked in. If the
system is shaken, this will usually jar it sufficiently to restore equilibrium .
Damping in real structures is due partially to Coulomb friction [71]. This is because
only this mechanism can stop completely the free vibration motion. If damping were purely
viscous, theoretically the motion will keep on forever, although it had infinitesimally small
amplitudes. This is merely an academic point of view, but it is basic for a full understanding
of damping mechanisms.
The diverse damping mechanisms that prevail in real structures are rarely modeled in-
dividually. In particular, the Coulomb frictional forces that must exist are not considered
172 B. Some Aspects of Structural Dynamics

explicitly unless frictional devices have been incorporated in the structure itself. Even when
these devices are used, it is possible to use an equivalent viscous damping to obtain approxi-
mately the dynamic response [61, 72]. This equivalent coefficient is determined by equalizing
the areas of the hysteresis loops (dissipated energy).

B.1.5 Steady-state response to a harmonic input with Coulomb damping

In this subsection the steady-state response of a single-degree-of-freedom model with Coulomb
damping and subjected to a harmonic driving force P (t) = P0 cos ω̄ t is obtained. A closed-
form expression for the amplitude of the harmonic response is provided. This case and the
one considered in the previous subsection (free vibration) are the only ones (with Coulomb
damping) where analytical (closed-form) solutions can be derived. The analysis presented
here relies on the assumption that there are no stops, hence, it can be mostly applied to
vibrations occurring near resonance.
During the steady-state motion there are no stops and, hence, the absolute value of the
friction force is equal to µ N . The equation of motion is given by

m ẍ + (−1)sgn ẋ µ N + k x = P0 cos ω̄ t (B.31)

In this equation P0 and ω̄ are the amplitude and the frequency of the harmonic input.
Provided that only the steady-state response is of interest, to analyze half a cycle is
enough. During such interval, if ẋ < 0, the above equation can be written as

m ẍ + k x = P0 cos ω̄ t + µ N (B.32)

This is a second-order, linear, ordinary differential equation. Its general solution can be
written as the sum of the general solution of the homogeneous equation plus any particular
solution of the complete equation:

P0 1 µN
x (t) = C1 sin ω̄ t + C2 cos ω̄ t + 2 cos (ω̄ t − φ) + (B.33)
k 1−β k

Where β is the ratio between the input frequency and the natural one (β = ω̄ω , where
q
k
ω= m is the natural frequency of the undamped system) and φ is the phase angle. It is
remarkable that both parts of the response do not damp out, i.e., no transient component
exists.
There are four unknowns that can be obtained by imposing the conditions at the begin-
ning and at the end of the half cycle. Such unknowns are the constants C1 and C2 , the phase
angle φ and the amplitude x0 . The initial and final conditions are:
B.1. Single-Degree-of-Freedom (SDOF) Systems 173

x0/(P0/k)

µN/P0

ω /ω

Figure B.6 Amplification of motion of the sliding system in the region of resonance

³ π´ ³ π´
x (t0 ) = x0 , ẋ (t0 ) = 0, x t0 + = −x0 , ẋ t0 + =0 (B.34)
ω̄ ω̄
Since these conditions are symmetric, it implies that the friction force has to reverse its
sign at the end of the half cycle.
By substituting the above conditions in the general solution of the equation of motion
during the half cycle, the following closed-form solution for the sought amplitude x0 is
obtained:

v
u µ ¶2 Ã 1 π
!2
P0 u
t 1 µ N β sin β
x0 = ¡ ¢2 − (B.35)
k 1 − β2 P0 1 + cos πβ
s
³ ´2 µ 1
sin π
¶2
P0 1 µN β β
As k can be considered as the static response, D = 2 − P0 1+cos π is
(1−β 2 ) β

the dynamic magnification factor: D = x0 /(P0 /k). The ratio µPN

indicates the amount of
0
Coulomb damping compared to the amplitude of the excitation. The bigger the sliding force
µ N, the smaller the amplitude of D. Fig. B.6 displays the plots of D vs. β for different
values of the ratio µPN
0
.
If µPN
0
≤ π4 , the amplitude of peak at resonance (β = 1, φ = π2 ) becomes infinitely large.
This conclusion can be confirmed (and explained too) by energy balance relations: if the
174 B. Some Aspects of Structural Dynamics

energy EI fuelled into the structure by the input is bigger than the energy EF dissipated by
friction (and the difference EI − EF keeps constant with time or, at least, does not decreases
too fast), the amplitude grows indefinitely.
The input energy is equal to the work done by the driving force P (t):

Z 2π Z 2π
ω̄ ω̄
EI = P (t) ẋ (t) dt = P0 cos (ω̄ t) x0 ω̄ cos (ω̄ t) dt = P0 x0 π (B.36)
0 0

As the hysteresis loop is rectangular (see Fig. 2.2) the energy dissipated by friction is
EF = 4 x0 µ N . Comparison between the values of EI and of EF shows that EI > EF if
µN π
P0 < 4 . This is due to the fact that the energy dissipated by friction is proportional to x0
instead of to x20 as for viscous damping [73, 74].
In practical cases µPN 0
< π4 , hence, friction dissipators or isolators are not useful by
themselves to cut the resonance peak. Some viscous damped should be also considered for
this purpose. This is particularly relevant for narrow-band inputs.

B.1.6 Solution of the equation of motion

Eq. (B.1.6) can be solved by means of many procedures, but generally a numerical solution
is preferred; this way the external force P (t) can be of any type.

B.1.6.1 Direct analysis in time domain for non-linear systems

The analysis in time domain consists in the solution for Eq. (B.1.6) by means of diverse
algorithms. Among the most utilized are the step-by-step numerical integration methods
[1, 61, 75]. One of these procedures is described next.

B.1.6.2 Incremental equation of dynamic equilibrium

The structure to be considered in this discussion is the SDOF system shown in Fig. B.2a.
The forces acting on the mass of the system are indicated in Fig. B.2b, and the general
nonlinear characteristics of the spring and damping forces are shown in Fig. B.7a and B.7b,
respectively, while an arbitrary applied loading is sketched in Fig. B.7c.
At any time instant t the dynamic equilibrium of the forces acting on the mass m requires

FI (t) + FD (t) + FS (t) = P (t) (B.37)

while a short time ∆t later the equation would be

FI (t + ∆t) + FD (t + ∆t) + FS (t + ∆t) = P (t + ∆t) (B.38)

B.1. Single-Degree-of-Freedom (SDOF) Systems 175

FD (x& ) FS (x)
∆FD (t ) FS (t + ∆t )
FD (t + ∆t ) ∆FS (t )
Slope = c(t) FS (t )
FD (t ) Slope = k(t)
∆x& (t ) ∆x(t )

x& (t ) x& (t + ∆t ) x& x (t ) x(t + ∆t ) x

(a) Non-linear damping (b) Non-linear stiffness

P(t)

P (t + ∆t )
∆P (t )
P(t)
∆t

t t + ∆t t

(c) Applied external load

Figure B.7 Stiffness, damping and driving forces for a non-linear system

Subtracting Eq. (B.37) from Eq. (B.38) then yields the incremental form of the equation of
motion for the time interval ∆t:

∆FI (t) + ∆FD (t) + ∆FS (t) = ∆P (t) (B.39)

The incremental forces in this equation may be expressed as follows:

∆FI (t) = FI (t + ∆t) − FI (t) = m∆ẍ(t)

∆FD (t) = FD (t + ∆t) − FD (t) = c(t)∆ẋ(t)
∆FS (t) = FS (t + ∆t) − FS (t) = k(t)∆x(t) (B.40)
∆P (t) = P (t + ∆t) − P (t)

in which it is tacitly assumed that the mass remains constant, and where the terms c(t)
and k(t) represent the damping and stiffness properties corresponding to the velocity and
displacement existing during the time interval, as indicated in Figs. B.7a and B.7b, respec-
tively. In practice, the secant slopes indicated could be evaluated only by iteration because
the velocity and displacement at the end of the time increment depend on these properties;
176 B. Some Aspects of Structural Dynamics

for this reason the tangent slopes defined at the beginning of the time intervals are frequently
used instead:

µ ¶ µ ¶
dFD dFS
c(t) ≈ k(t) ≈ (B.41)
dẋ t dx t

Substituting the force expressions of Eqs. (B.40) into Eq. (B.39) leads to the final form
of the incremental equilibrium equations for time t:

m∆ẍ(t) + c(t)∆ẋ(t) + k(t)∆x(t) = ∆P (t) (B.42)

The material properties considered in this type of analysis may include any form of
nonlinearity. Thus, there is no need for the spring force FS to be dependent only on the
displacement, as in a nonlinear elastic material. A nonlinear hysteretic material may equally
be well specified, in which the force depends on the past history of deformation as well
as the current value of displacement. The only requirement is that the stiffness properties
be completely defined by the past as well as the current state of deformation. Moreover,
it is evident that the implicit assumption of a constant mass is arbitrary; it also could be
represented as a time-varying quantity.

B.1.6.3 Step-by-step integration

As said before, a number of procedures is available for the numerical integration of Eq.
(B.42). The technique employed here is simple in concept but has been found to yield
excellent results with relatively little computational effort. The basic assumption of the
process is that the acceleration varies linearly during each time increment while the properties
of the system remain constant during this interval. The motion of the mass during the time
interval is indicated in graphical form in Fig. B.8, together with equations for the assumed
linear variation of the acceleration an the corresponding quadratic and cubic variations of
the velocity and displacement, respectively. Evaluating these latter expressions at the end
of the interval (τ = ∆t) leads to the following equations for the increments of velocity and
displacement:

∆t
∆ẋ(t) = ẍ(t)∆t + ∆ẍ(t) (B.43a)
2
∆t2 ∆t2
∆x(t) = ẋ(t)∆t + ẍ(t) + ∆ẍ(t) (B.43b)
2 6
Now it will be convenient to use the incremental displacement as the basic variable of the
analysis; hence Eq. (B.43a) is solved for the incremental acceleration, and this expression is
substituted into Eq. (B.43b) to obtain
B.1. Single-Degree-of-Freedom (SDOF) Systems 177

6 6
∆ẍ(t) = ∆x(t) − ẋ(t) − 3ẍ(t) (B.44a)
∆t2 ∆t
3 ∆t
∆ẋ(t) = ∆x(t) − 3ẋ(t) − ẍ(t) (B.44b)
∆t 2
Substituting Eqs. (B.44) into Eq. (B.42) leads to the following form of the equation of
motion:

· ¸ · ¸
6 6 3 ∆t
m ∆x(t) − ẋ(t) − 3ẍ(t) + c(t) ∆x(t) − 3ẋ(t) − ẍ(t) +
∆t2 ∆t ∆t 2
k(t)∆x(t) = ∆P (t)

Finally, transferring all terms associated with the known initial conditions to the right-hand
side gives

k̃(t)∆x(t) = ∆P̃ (t) (B.45)

in which

6 3
k̃(t) = k(t) + 2
m+ c(t) (B.46a)
∆t ∆t · ¸
6
∆P̃ (t) = ∆P (t) − m∆ẍg (t) + m ẋ(t) + 3ẍ(t) +
∆t
· ¸
∆t
c(t) 3ẋ(t) + ẍ(t) (B.46b)
2
It will be noted that Eq. (B.45) is equivalent to a static incremental-equilibrium rela-
tionship, and may be solved for the incremental displacement by dividing the incremental
load by the stiffness. The dynamic behavior is accounted for by the inclusion of inertial and
damping effects in the effective-load and stiffness terms. After Eq. (B.45) is solved for the
displacement increment, this value is substituted into Eq. (B.44) to obtain the incremental
velocity. Then, initial conditions for the next time step result from the addition of these
incremental values to the velocity and displacement at the beginning of the time step.
This numerical procedure includes two significant approximations: (1) the acceleration
varies linearly and (2) the damping and stiffness properties remain constant during the time
step. In general, neither of these assumptions is entirely correct, even though the errors
are small if the time step is short. Therefore, errors generally will arise in the incremental-
equilibrium relationship which might tend to accumulate from step to step, and this ac-
cumulation should be avoided by imposing the total-equilibrium condition at each step of
the analysis. This may be accomplished conveniently by expressing the accelerations at the
beginning of the time step in terms of the total external load minus the total damping and
spring forces (Eq. (B.47)).
178 B. Some Aspects of Structural Dynamics

A cceleratio n ∆ &x&
(linear) &x&(τ ) = &x&( t ) + τ
∆t
∆ &x&(t )
&x&(t )

V elocity ∆ &x& τ 2
(quadratic) x& (τ ) = x& ( t ) + &x&( t )τ +
∆t 2
∆ x& (t )
x& (t )

D isplacem ent τ 2 ∆&x& τ 3

(cubic) x (τ ) = x (t ) + x& (t )τ + &x&(t ) +
2 ∆t 6

∆ x (t )

x (t )

τ
t t + ∆t

Figure B.8 Schematic diagram of the method of linear acceleration

B.1. Single-Degree-of-Freedom (SDOF) Systems 179

B.1.6.4 Summary of the procedure

For any given time increment, the analysis procedure consists of the following operations:
(1) Initial velocity and displacement values ẋ(t) and x(t) are known, either from values
at the end of the preceding increment or as initial conditions of the problem.
(2) With these values and the specified nonlinear properties of the structure, the damping
c(t) and the stiffness k(t) for the interval, as well as current values of the damping FD (t)
and strain FS (t) forces are found, e.g., from Figs. B.7a and b.
(3) The initial acceleration is given by

1
ẍ(t) = [P (t) − FD (t) − FS (t)] (B.47)
m
This is merely a rearrangement of the equation of equilibrium (B.37) for time t.
(4) The effective load increment ∆P̃ (t) and effective stiffness k̃(t) are computed from
Eqs. (B.46).
(5) The increment of displacement is given by Eq. (B.45), and with it the increment of
velocity is found from Eq. (B.44b).
(6) Finally the velocity and displacement at the end of the increment are obtained from

ẋ(t + ∆t) = ẋ(t) + ∆ẋ(t) (B.48a)

x(t + ∆t) = x(t) + ∆x(t) (B.48b)

When step 6 has been completed, the analysis for this time increment is finished, and
the entire process may be repeated for the next time interval. Obviously the process can
be carried out consecutively for any desired number of time increments; thus the complete
response history can be evaluated for any SDOF having any prescribed nonlinear properties.
Linear systems can also be treated by the same process, of course; in this case the damping
and stiffness properties remain constant so the analysis procedure is somewhat simpler.
It is interesting to notice that no iterative calculations are required during the process.

B.1.7 Energy formulation

Multiplying Eq. (B.1.6) by ẋdt and integrating through time leads to the following energy
balance expression

EK + ED + ES = EI (B.49)

where
180 B. Some Aspects of Structural Dynamics

Z t
1
EK = mẍẋdt = mẋ2
0 2
Z t Z t
ED = cẋẋdt = cẋ2 dt
0 0
Z t
1 2
ES = kxẋdt = kx (B.50)
0 2
Z t
EI = P (t)ẋdt
0

The terms included in the left member of Eq. (B.49) represent the kinetic energy (EK ),
the energy dissipated by means of the viscous damping (ED ) and the strain energy (ES ).
The addition of these terms must be equal to the input energy introduced to the structure
(EI ) due to any lateral force.

B.2 Multi-Degree-of-Freedom (MDOF) Systems

B.2.1 General formulation for buildings

Fig. B.9a represents a typical frame belonging to a symmetric multi-story building (MSB).
For simplicity, the contributions of partitions, installations and nonstructural elements to
the lateral stiffness of the building are neglected and, moreover, this frame will be treated as
if the floor and the ceiling were infinitely rigid (for bending) compared to the columns. This
is the usual approximation of the so-called ’shear buildings’. Also, the axial deformation
of members will be neglected, as well as the second order effect of the axial force over the
columns stiffness. Hence, the lateral stiffness of the building is due entirely to the bending
stiffness of columns. Fig. B.9b shows the forces acting on a bent column. The degrees of
freedom of these types of models are equal to the number of floors, N , thus the motion
of each floor is defined by its horizontal displacement. However, if the building were not
symmetric (e.g., if there were eccentricities between the centers of mass and rigidity), its
dynamic behavior could not be described by a 2D model (one DOF per floor); but rather by
a 3D model with 3 DOF per floor (two horizontal displacements and one twist angle).
The mass is assumed to be distributed through the building, but here it will be treated
as if it were concentrated on each floor –’lumped mass’ assumption. The sum of all masses
compresses the total mass of the building from the half height of the upper floor until the
half height of the lower floor, including nonstructural components and a percentage of the
variable loads. This way the building will be treated as if it were a weightless frame with
lumped masses attached to it on every floor, as shown in Fig. B.10. Masses are constrained
to move horizontally only, and each one is rigidly joined to the frame so the horizontal inertia
B.2. Multi-Degree-of-Freedom (MDOF) Systems 181

Rigid beams

∆i
HN
Mi
12EI i
Vi = 3
∆i
H i +1 Hi
Hi

Hi
Vi
6 EI i
Mi = ∆i
H1 Hi
2

(a) Idealized shear-building (b) Forces on a deformed column

Figure B.9 Multi-story building (MSB) idealized as a shear building

force enters into the equations of dynamic equilibrium, but there is no vertical motion of the
mass.
On the other hand, the damping is difficult to evaluate, here and in almost all dynamic
systems. In the case of SDOF systems the damping was considered as proportional to
the velocity, i.e., viscous damping, and it is mathematically convenient because leads to
simple, consistent solutions in vibration problems of real systems affected by small amplitude
vibrations. The damping force will be considered as the force exerted by a valve attached
to the mass, producing a force proportional to the mass and opposite to the direction of
motion. The multi-degree-of-freedom system complicates the concept, since the damping
force proportional to the velocity exerted on a mass could be taken as proportional to the
velocity of this mass, relative to a fixed reference frame, ’absolute damping’, or relative to the
other masses of the system, ’relative damping’. This work deals with the relative damping,
which will be analyzed by the effects that valves exert on each floor, each one generating a
damping force proportional to the velocity of the horizontal motion between the floors.

B.2.2 Equations of motion

With the assumptions made in the previous subsection, it is possible to derive the equations
of motion considering the free-body diagrams of the masses and writing the equations of
182 B. Some Aspects of Structural Dynamics

Figure B.10 Multi-story building (MSB) modelled as a shear building with lumped masses

dynamic equilibrium for each of them. From Fig. B.11, let m1 , m2 and m3 be the masses
designated from bottom to top; let c1 , c2 and c3 be the three damping coefficients; and let
k1 , k2 and k3 be the stiffness of each floor, each one representing the sum of the stiffnesses
of all columns in each floor. Besides, let P1 (t), P2 (t) and P3 (t) be the lateral dynamic forces
acting on the masses. Fig. B.12 shows the mechanical model of the frame of Fig. B.11 and
Fig. B.13 shows the free-body diagrams of the three masses in dynamic equilibrium.
Writing the equation of motion for each mass, leads to a system of three linear differential
equations of motion:

P3 (t) − m3 ẍ3 − c3 (ẋ3 − ẋ2 )

−k3 (x3 − x2 ) = 0

P2 (t) − m2 ẍ2 + c3 (ẋ3 − ẋ2 ) − c2 (ẋ2 − ẋ1 ) (B.51)

+k3 (x3 − x2 ) − k2 (x2 − x1 ) = 0

P1 (t) − m1 ẍ1 + c2 (ẋ2 − ẋ1 )

−c1 ẋ1 + k2 (x2 − x1 ) − k1 x1 = 0
B.2. Multi-Degree-of-Freedom (MDOF) Systems 183

Rigid beams
P3 (t ) m3

H3
P2 (t ) m2

H2
P1 (t ) m1

H1

Figure B.11 3-story building with lumped masses

Figure B.12 Mechanical model for a 3-story shear building

m3 &x&3
P3 (t )
k 3 ( x3 − x2 ) c3 ( x&3 − x& 2 )

m2 &x&2
P2 (t )
k 2 ( x2 − x1 ) c2 ( x& 2 − x&1 )

m1 &x&1
P1 (t )
k1 x1 c1 x&1

Figure B.13 Dynamic forces acting on the three masses

184 B. Some Aspects of Structural Dynamics

These equations can be rewritten as

m1 ẍ1 + (c1 + c2 )ẋ1 − c2 ẋ2 + (k1 + k2 )x1

−k2 x2 = P1 (t)

m2 ẍ2 − c2 ẋ1 + (c2 + c3 )ẋ2 − c3 ẋ3 − k2 x1 (B.52)

+(k2 + k3 )x2 − k3 x3 = P2 (t)

m3 ẍ3 − c3 ẋ2 + c3 ẋ3 − k3 x2 + k3 x3 = P3 (t)

And these can be rewritten in matrix form as

Mẍ + Cẋ + Kx(t) = P(t) (B.53)

where

 
m1 0 0
 
M =  0 m2 0  = inertia (mass) matrix
0 0 m3
 
c1 + c2 −c2 0
 
C =  −c2 c2 + c3 −c3  = damping matrix
0 −c3 c3

 
k1 + k2 −k2 0
 
K =  −k2 k2 + k3 −k3  = stiffness matrix
0 −k3 k3
 
x1 (t)
 
x(t) =  x2 (t)  = displacement vector
x3 (t)
 
P1 (t)
 
P(t) =  P2 (t)  = external force vector
P3 (t)
B.2. Multi-Degree-of-Freedom (MDOF) Systems 185

B.2.3 Properties of system matrices

Some properties of system matrices M, C and K can be observed, and it is useful to take it
into account to avoid potential errors and to take advantages in the calculation process.
First, M, C and K are symmetric matrices. In this case, M is a diagonal matrix, but
this it is not always true. The symmetry of matrix K comes from Maxwell’s reciprocity
theorem, which is based on the superposition principle and on stress-energy considerations.
Comparable arguments using dissipated energy in viscous damping instead of strain energy
can be utilized to establish the symmetry of C, and considerations based on the kinetic
energy can lead to the symmetry of M.
Second, diagonal elements of M and K are always positive, due to the defined positive
character of strain and kinetic energies. In a similar way, diagonal elements of C are always
positive because any velocity must be accompanied by energy dissipation in the viscous
damping, or, in a better case, for no change in the energy –the damping cannot increase the
energy of the system. The possibility that a part of the system does not have any damping
makes possible that some diagonal elements are zero.
Third, the biggest element of M is always in the diagonal, as well as the biggest element
of K. Some non-diagonal elements can be bigger than some diagonal elements, but the
biggest element of the whole matrix will always be in the diagonal of these matrices.
Fourth, for building structures that are not modelled as shear buildings, the stiffness
matrix K can have non-zero elements even in the farthest corners. However, the solution
presented next applies still.

B.2.4 Solution of the equations of motion using the linear acceleration

method
A number of numerical methods can be employed to solve the differential equations of motion
of MDOF systems. The procedure is relatively simple, but since the equations are generally
coupled, as in this case, they must be solved simultaneously.
For example, considering a multi-story frame with N stories, the equation of motion
(B.53) at any time tk can be written as

ẍk = M−1 [P(tk ) − Cẋk − Kxk ] (B.54)

Since the inertia matrix is diagonal, Eq. (B.54) can also be written as

 
X X
ẍi = Pi − cij ẋj − kij xj  /mi i = 1, 2, . . . , N (B.55)
j j
186 B. Some Aspects of Structural Dynamics

for the i-th floor at time tk .

If M were not diagonal, the following more general form could be used:

à !
X X X
ẍi = m−1
ij Pi − cjl ẋl − kjl xl i = 1, 2, . . . , N (B.56)
j l l

In this last equation the term m−1ij refers to the element of file i and column j of matrix
−1 −1
M , i.e., mij is not equal, generally, to 1/mij .
Knowing the displacement xk and the velocity ẋk vectors at a time tk (normally, the
initial conditions would be the starting point), ẍk is calculated from Eq. (B.54). Now, a
time interval ∆t is chosen and then:

1. Assume a set of values for ẍ∗k+1 , the acceleration vector at the end of the interval. The
initial acceleration ẍk could be the first estimate of ẍ∗k+1 .

2. Calculate
∆t2
xk+1 = xk + ∆tẋk + (2ẍk + ẍ∗k+1 ) (B.57)
6

3. Calculate
∆t
ẋk+1 = ẋk + (ẍk + ẍ∗k+1 ) (B.58)
2

4. Calculate

ẍk+1 = M−1 [P(tk+1 ) − Cẋk+1 − Kxk+1 ] (B.59)

5. (a) If ẍk+1 6= ẍ∗k+1 , the calculated values of ẍk+1 must be taken as an improved estimate
of ẍ∗k+1 . Go back to step 2.
(b) If ẍk+1 = ẍ∗k+1 (with a prescribed tolerance), the iteration has converged. Go to
the next time interval, starting from step 1.

Note that the procedure is applicable for linear and nonlinear systems.

B.2.5 Energy formulation

Pre-multiplying and post-multiplying Eq. (B.53) by ẋT and dt, respectively, and integrating
through time, the following energy balance equation is obtained:

EK + ED + ES = EI (B.60)
B.2. Multi-Degree-of-Freedom (MDOF) Systems 187

where

Z t
1
EK = ẋT Mẍ dt = ẋT Mẋ
0 2
Z t
ED = ẋT Cẋ dt
0
Z t
1
ES = ẋT Kx dt = xT Kx (B.61)
0 2
Z t
EI = ẋT P(t) dt
0

The terms included in the left members of Eqs. (B.61) represent the kinetic energy
(EK ), the dissipated energy due to the viscous damping (ED ) and the strain energy (elastic
or inelastic) (ES ). The addition of these terms must be equal to the input energy (EI )
introduced to the structure due to any lateral load.