MECHANICAL VIBRATIONS

A Lecturebook

Charles M. Krousgrill
Je↵rey F. Rhoads

c 2018 Charles M. Krousgrill and Je↵rey F. Rhoads

About the Authors:

Charles M. Krousgrill: Charles M. Krousgrill is a Professor in the School of Mechanical Engineer-

ing at Purdue University. He received his M.S. in mechanical engineering from Purdue University
in 1975, and his M.S. and Ph.D. in applied mechanics from the California Institute of Technology
in 1976 and 1980, respectively. During his more than 30 years at Purdue University, few, if any,
individuals have had a greater impact on the university’s undergraduate students and the institu-
tion’s commitment to engineering education. Dr. Krousgrill’s e↵orts in this regard have garnered
numerous awards and international accolades. To date, these include the Purdue University School
of Mechanical Engineering’s Harry L. Solberg Best Teacher Award (eight times), the Purdue Uni-
versity College of Engineering’s Potter Best Teacher Award (four times), the Purdue University
Murphy Best Teacher Award, the 2010 Purdue University Helping Students Learn Award, the 2010
Purdue Alumni Association Special Boilermaker Award and the American Society of Engineering
Education’s 2011 Archie Higdon Distinguished Educator Award – the de facto lifetime achieve-
ment award for educational accomplishments in the field of mechanics. In January 2018, it was
announced that Dr. Krousgrill had been awarded the named 150th Anniversary Professorship by
Purdue University. Outside of the classroom, Dr. Krousgrill conducts research in the general areas
of dynamics and mechanical vibration.

Je↵rey F. Rhoads: Je↵rey F. Rhoads is a Professor in the School of Mechanical Engineering at

Purdue University and is affiliated with both the Birck Nanotechnology Center and Ray W. Herrick
Laboratories at the same institution. He also serves as the Director of Practice for MEERCat
Purdue: The Mechanical Engineering Education Research Center at Purdue University and the
Associate Director of PERC: The Purdue Energetics Research Center. Dr. Rhoads received his B.S.,
M.S., and Ph.D. degrees, each in mechanical engineering, from Michigan State University in 2002,
2004, and 2007, respectively. Dr. Rhoads’ current research interests include the predictive design,
analysis, and implementation of resonant micro/nanoelectromechanical systems (MEMS/NEMS)
for use in chemical and biological sensing, electromechanical signal processing, and computing;
the dynamics of parametrically-excited systems and coupled oscillators; the thermomechanics of
energetic materials (including explosives, pyrotechnics, and propellants); additive manufacturing;
and mechanics education. Dr. Rhoads is a Member of the American Society for Engineering
Education (ASEE) and a Fellow of the American Society of Mechanical Engineers (ASME), where
he serves on the Design Engineering Division’s Technical Committee on Vibration and Sound. Dr.
Rhoads is a recipient of numerous research and teaching awards, including the National Science
Foundation’s Faculty Early Career Development (CAREER) Award; the Purdue University School
of Mechanical Engineering’s Harry L. Solberg Best Teacher Award (twice), the Robert W. Fox
Outstanding Instructor Award, the and B.F.S. Schaefer Outstanding Young Faculty Scholar Award;
the ASEE Mechanics Division’s Ferdinand P. Beer and E. Russell Johnston, Jr. Outstanding New
Mechanics Educator Award; and the ASME C. D. Mote Jr., Early Career Award. In 2014, Dr.
Rhoads was included in ASEE Prism Magazine’s 20 Under 40.
 Chapter I
Chapter I

Equations of motion for discrete

systems
Introduction
In this chapter we will discuss several di↵erent approaches for obtaining the equations of motion
(EOM’s) of systems for which the mass, damping and sti↵ness components appear in discrete
components. These EOM’s will be ordinary di↵erential equations which describe the motion of the

Chapter I
system. For the most part, we will be interested in describing small-amplitude motion. Because
of this, we will be able to deal with a linearized form of these di↵erential equations, where the
linearization is performed about equilibrium states.
We will be considering a number of approaches in deriving the EOM’s:

• Newton-Euler formulation – This is a vector-based approach beginning with forces and ac-
celerations written in vector form. In doing so, we will need to deal with all of the forces
acting on the system. In particular, we will include forces of reaction and forces of constraint
along with the applied forces. It is often the case that we will not care about quantifying
the reaction and constraint forces, and consequently, we will eliminate these forces from the
EOM’s before attempting to solve the EOM’s. This elimination can be a tedious task.

• Power equation – The power equation, as we will see, is based on a work/energy (scalar)
description of the motion. An important consequence of this is that many of the forces of
reaction and constraint will naturally not appear in the energy equation. However, since we
will start with a single work/energy equation for the system of interest, we will obtain only
a single EOM regardless of the number of coordinates needed to completely describe the mo-
tion. Hence, the power equation will be useful to us only for systems having a single degree
of freedom (DOF), that is, systems for which only one coordinate is needed to describe the
motion.

• Lagrange’s equations – The Lagrangian formulation is a means by which we will be able to

separate out from the power equation the correct number of EOM’s needed to describe the
motion of the system. That is, for a system having N DOFs, the Lagrangian formulation will
produce N EOM’s.

• Linearized equations of motion – For small amplitude oscillations about an equilibrium state,
we will be able to replace nonlinear terms appearing in the EOM’s obtained from the La-
grangian approach by their linearized approximations. This will produce a set of linear
ordinary di↵erential equations that we will use later on for determining the response of the
system. We will be able to obtain these linear EOM’s directly from the kinetic energy, poten-
tial energy and Rayleigh dissipation functions without directly using Lagrange’s equations.
This formulation will allow us to observe symmetry properties of the mass and sti↵ness ma-
trices. These symmetry properties will prove useful to use later on in the course in when we
use orthogonality properties of the modal vectors of the problem.

• Flexibility matrix and influence coefficients – Later on in the course we will need to use the in-
verse of the sti↵ness matrix (called the flexibility matrix) for some of the numerical eigenvalue
extraction methods. Here we will discuss a direct approach for finding the flexibility matrix
using the so-called ’influence coefficients. These influence coefficients are generally easier to
find than the inverse of the sti↵ness and will provide some physical insight into the problem.
 I.1 Equations
EOM’s: Newton-Euler of Motion Using the Newton-Euler Equations
equations
Background
Background: For a single rigid body executing planar motion, we have the
following set of Newton-Euler equations:

!F = ma G

!M A = IA "

G
where

G = center of mass of body

For a single Arigid body executing planar motion, we have the following set of Newton-Euler equa-
= EITHER center of mass OR fixed point of the rigid body
tions:
X ! F = resultant force vector acting on rigid body
F~ =!m~Ma A = resultant moment vector acting about point A
G
X I = mass moment of inertia about point A
~ AA= IA ↵
M ~
a G = acceleration vector of the center of mass G
where
! = angular acceleration vector of the rigid body
G = center of mass of the body
Remarks:A = EITHER the center of mass OR a fixed point on the rigid body
X draw appropriate free body diagrams (FBD’s) before you attempt
a) ALWAYS
F~ =the
to develop the resultant
EOM’s for the force
system.acting on the body
X consider the coordinates used (such as positive direction, whether
b) Carefully
measuredM~ Afrom
= the
fixedresultant moment
reference or moving about point
reference, A acting
whether on the body
measured
from equilibrium or unstretched positions) before drawing FBD’s.
c) You IAneed
= the mass ALL
to include moment
forcesof inertia
acting of body
on the the body about down
when writing pointthe
A
EOM’s (including forces of reaction). Later on when we use the power
~aG = and
equation the Lagrange’s
acceleration vectorwefor
equations the
will be center of mass
able to ignore G that do
forces
no virtual work when deriving EOM’s.
↵
~ = the angular acceleration vector for the body
d) You should have exactly N EOM’s for a system having N degrees of
freedom (we will discuss the concept of number of degrees of freedom in a
later lecture). Typically you will have more than N EOM’s; however, you
will be able to eliminate forces of constraint (reaction forces) and enforce
kinematic constraints to reduce this number of EOM’s to N.
Remarks:
e) Never attempt to write down the EOM’s by inspection.
• ALWAYS draw appropriate free body diagrams (FBD’s) before you attempt to develop the
EOM’s for the system.
I.2
• Carefully consider the coordinates used (such as sign conventions, whether measured from
fixed reference or moving reference, whether measured from equilibrium or unstretched posi-
tions) before drawing FBD’s.

• You need to include ALL of the forces acting on the body when writing down the EOM’s
(including forces of reaction). Later on when we use the power equation and Lagrange’s
equations we will be able to ignore forces that do no virtual work when deriving EOM’s.
• In the end, you should have exactly N EOM’s for a system having N degrees of freedom (we
will discuss the concept of number of degrees of freedom in a later lecture). Typically you
will have more than N Newton/Euler equations at the start; however, you will be able to
eliminate forces of constraint (reaction forces) and enforce kinematic constraints to reduce

Chapter I
this number of EOM’s to N .

• Never attempt to write down the EOM’s by inspection.

Example I.1.1

A homogeneous cylinder having a mass of M and radius r rolls without slipping on an inclined
surface. A spring having a sti↵ness of k is attached between the center G of the cylinder and ground.

• Determine the equation of motion (EOM) corresponding to the coordinate x, where x de-
scribes the position on the incline for point G with x being measured from the position
of G when the spring is unstretched.

• Determine the EOM corresponding to the coordinate z, where z describes the position on
the incline for point G with z being measured from the equilibrium position of G.

unstretched position
x equilibrium position

k z k

m G
m G
r
no slip r
no slip
θ θ

figure&a)& figure&b)&
Chapter I
Example I.1.2

Two particles, 1 and 2, (each of mass m) are interconnected by a set of three springs (each of sti↵-
ness k) and three dashpots (each having a damping constant of c) and slide on a smooth horizontal
surface. A force f acts on particle 2.

Find the equations of motion (EOM’s) for this problem corresponding to three di↵erent sets of
coordinates:

• Using x1 and x2 with both x1 and x2 being measured positive to the right from fixed reference
points with the springs being unstretched when x1 = x2 = 0.

• Using y1 and y2 with y1 being measured positive to the right from a fixed reference point, y2
being measured positive to the left from a fixed reference point and with the springs being
unstretched when y1 = y2 = 0.

• Using z1 and z2 with z1 being measured positive to the right from a fixed reference point,
z2 representing the stretch/compression of the middle spring (z2 > 0 when the spring is
stretched) and with the springs being unstretched when z1 = z2 = 0.

x1 x2
f
1 2
figure&a)&

y1 y2
f
1 2
figure&b)&

z1 z2
f
1 2
figure&c)&

figure&a)& figure&b)&
Chapter I
Example I.1.3

Particle A (having a mass of m) slides on a smooth horizontal surface with a thin homogeneous bar
(having a length of L and mass M ) being attached to A with a smooth pin, as shown. A spring and
dashpot connect A to ground. The coordinate x describes the motion of A (x is measured positive
to the right and x = 0 when the spring is unstretched). The coordinate ✓ describes the orientation
of the bar, with ✓ being positive for a counterclockwise rotation from the vertical.

Find the EOM’s of the system using the coordinates of x and ✓.

k
A
c

G
θ
Chapter I
I.2 EOM’s: Power equation
Objectives
It is desired to be able to obtain the di↵erential equation of motion for a single degree-of-freedom
system without having to account for “workless” forces acting on the system.

Background
The work-energy equation for a system of n rigid bodies is given by:

T + U = T0 + U0 + W (nc)

where
1 1
Tj = kinetic energy of the jth body = mvAj 2 + IAj ! 2
2 2
n
X
T = kinetic energy of the system = Tj
i=j

U = potential energy due to conservative forces

XZ 2
W (nc)
= work done by nonconservative forces = F~ (j) · d~
⇢j
j 1

Aj = EITHER a fixed point OR the center of mass of the jth body

vAj = speed of point Aj
! = angular speed of the body
T0 , U0 = initial values for the kinetic and potential energies, respectively

pa th of point j

j

d ρ j (ta ngent to pa th)
Aj

ρj

O (fixed point)
Results
The “power equation” results from taking a time derivative of the work-energy equation:

dW (nc) dT dU

Chapter I
P ower = = +
dt dt dt

Remarks
• As we will see in the following examples, the power equation will produce a second-order dif-
ferential EOM for a system for which only a single coordinate is needed to describe its motion
(i.e., systems having a single degree of freedom). The power equation has the advantage over
the Newton-Euler approach in that one can ignore forces that do no nonconservative work on
the system.

• A particle is a rigid body for which either the dimensions are small (IG ⇡ 0) or is under pure
translation (! = 0). Therefore, the kinetic energy for a particle is given by T = 12 mvG 2 .

• Note that the initial values for the kinetic and potential energies (T0 and U0 , respectively)
drop out when di↵erentiating the work-energy with respect to time.
Example I.2.1

A homogeneous thin bar of mass M and length L is pinned to ground at point O. Using the power
equation, find the equation of motion (EOM) for the bar corresponding to the coordinate of ✓ where
✓ is measured counterclockwise from the vertical.

O
g
G

θ
Chapter I
Example I.2.2

Use the power equation to find the EOM for the system shown corresponding to the coordinate x,
where the coordinate x is defined such that x = 0 when the spring is unstretched.

k
m f (t)
c
Chapter I
Example I.2.3

Use the power equation to find the EOM of the system shown corresponding to the coordinate .
The mass of the particle is M , and the mass of the homogeneous cylinder (outer radius r) is m.
The bar connecting the particle and the center of the cylinder has a mass that is negligible. The
spring is unstretched when = 0.

k m O negligible ma ss f (t)
M
r
no slip smooth

C
Chapter I
I.3 EOM’s: Generalized coordinates, generalized forces and gen-
eralized mass coefficients
Objectives
Here we will rewrite the power equation from the last section in a way that the set of chosen
coordinates and their time derivatives are explicitly apparent. In this form, sets of “generalized
forces” and ”generalized mass coefficients” are introduced. The results of this section will lead us
directly in the Lagrangian formulation of the EOM’s found in the next section.

Background

θk

Ak


rk

O (fixed point)

• For a system of n planar rigid bodies, let Ak be either the center of mass or a fixed point on
the kth body. Assume that the position vectors, ~rk , and angular rotations, ✓k , are written
in terms of a set of N generalized coordinates qi (i = 1, 2, ..., N ). It is assumed that these
generalized coordinates completely describe the configuration of the system for all time. It is
also assumed that ~rk and ✓k are not explicit functions of time. From this, we can write for
k = 1, 2, ..., n:

~rk = ~rk (q1 , q2 , ..., qN )

✓k = ✓k (q1 , q2 , ..., qN )
• For a function b = b (q1 , q2 , ..., qN ), we have the chain rule of di↵erentials:

X @b N
@b @b @b
db = dq1 + dq2 + ... + dqN = dqi
@q1 @q2 @qN @qi
i=1

• For a function b = b (q1 , q2 , ..., qN ), we have the chain rule of di↵erentiation:

N
X @b
db @b @b @b
= ḃ = q̇1 + q̇2 + ... + q̇N = q̇i
dt @q1 @q2 @qN @qi
i=1
Results
N N
1 XX
T = mik q̇i q̇k
2
i=1 k=1

Chapter I
N 
X ✓ ◆
d @T @T
dT = dqi
dt @ q̇i @qi
i=1
N
X @U
dU = dqi
@qi
i=1
N
X
dW (nc) = Qi dqi
i=1

where Qi = “generalized force” corresponding the generalized coordinate qi and mik = “generalized
mass coefficient”, each corresponding to the generalized coordinates qi and qk and given by:
M
X @~
⇢j
Qi = F~j ·
@qi
j=1

n 
X @~rj @~rj @✓j @✓j
mik = mj · + Ij ·
@qi @qk @qi @qk
j=1

where M is the number of applied forces on the system.

Derivation
Recall the following set of Newton-Euler equations for planar motion of a rigid body:

F~j = mj~aGj = mj ~r¨j

Mj = Ij ↵j = Ij ✓¨j
where

F~j = the resultant force acting on the jth body

Mj = the resultant moment about point Aj acting on the jth body

Gj = center of mass of the jth body
Aj = EITHER the center of mass OR a fixed point on the jth rigid body
Say we take the dot product of the first equation with ~rj , multiply the second equation by d✓j , sum
those results over all n bodies and add the results, producing:
n h
X i Xn h i
mj ~r¨j · d~rj + Ij ✓¨j d✓j = F~j · d~rj + Mj d✓j
j=1 j=1

We recognize the right hand side of the above equation as being the total di↵erential work done
on the system. Alternately, we can write this right hand side of the equation as the di↵erence
between the di↵erential work done by nonconservative forces and the di↵erential of the potential
of conservative forces:
n h
X i
F~j · d~rj + Mj d✓j = dW (nc) dU
j=1

Therefore, the left hand side of this equation, according to the work-energy equation,

dT + dU = dW (nc)

must be the di↵erential kinetic energy for the system:

n h
X i
dT = mj ~r¨j · d~rj + Ij ✓¨j d✓j
j=1

In this section of the notes, we will work to simplify the expressions for the three terms of dT , dU
and dW (nc) . Consider the following steps:

• From the chain rule of di↵erentials:

N
X @~rj
d~rj = dqi
@qi
i=1

N
X @✓j
d✓j = dqi
@qi
i=1
Substituting these expressions into the expression for dT gives:

n
" N
! N
!#
X X @~rj X @✓j
dT = mj ~r¨j · dqi + Ij ✓¨j dqi
@qi @qi

Chapter I
j=1 i=1 i=1
2 3
N
X n ✓
X ◆
@~rj @✓j 5
= 4 mj ~r¨j · + Ij ✓¨j dqi
@qi @qi
i=1 j=1

Note that through the product rule of di↵erentiation:

✓ ◆ ✓ ◆
¨ @~rj d ˙ @~rj ˙~rj · d @~rj
~rj · = ~rj ·
@qi dt @qi dt @qi
✓ ◆
d ˙ @~rj @~r˙j
= ~rj · ~r˙j ·
dt @qi @qi

It can be shown that (see Appendix I):

@~rj @~r˙j
=
@qi @ q̇i

Substituting into the above gives:

!
@~rj d @ ˙j
~
r @~r˙j
~r¨j · = ~r˙j · ~r˙j ·
@qi dt @ q̇i @qi

1 d @ ⇣˙ ˙ ⌘ 1 @ ⇣˙ ˙ ⌘
= ~rj · ~rj ~rj · ~rj
2 dt @ q̇i 2 @qi

In a similar way, it can be shown that:

" #
@✓j 1 d @ ✓˙j2 1 @ ✓˙j2
✓¨j =
@qi 2 dt @ q̇i 2 @qi

Substituting these two results into the above equation for dT gives:
2 3 2 3
N
X n ✓
X ◆ N
X n ✓
X ◆
d 4 @ 1 1 4 @ 1 1
dT = mj ~r˙j · ~r˙j + Ij ✓˙j2 5 dqi mj ~r˙j · ~r˙j + Ij ✓˙j2 5 dqi
dt @ q̇i 2 2 @qi 2 2
i=1 j=1 i=1 j=1
N 
X ✓ ◆
d @T @T
= dqi
dt @ q̇i @qi
i=1

where:
n ✓
X ◆
1 1 2
T = mj ~r˙j · ~r˙j + Ij ✓˙j
2 2
j=1
• Using the chain rule of di↵erentiation:

n n
1X 1X ˙2
T = mj ~r˙j · ~r˙j + I j ✓j
2 2
j=1 j=1
n N
! N
! n N
! N !
1X X @~rj X @~rj 1X X @✓j X @✓j
= mj q̇i · q̇k + Ij q̇i q̇k
2 @qi @qk 2 @qi @qk
j=1 i=1 k=1 j=1 i=1 k=1
0 1 0 1
XN XN X n N X
X N n
X
1 @ @~rj @~rj A 1 @ @✓j @✓j A
= mj · q̇i q̇k + Ij q̇i q̇k
2 @qi @qk 2 @qi @qk
i=1 k=1 j=1 i=1 k=1 j=1
N X
X N
1
= mik q̇i q̇k
2
i=1 k=1

where

n 
X @~rj @~rj @✓j @✓j
mik = mj · + Ij
@qi @qk @qi @qk
j=1

are the “generalized mass coefficients”. Two things to note from the above. First, the kinetic
energy expression T is a function of both the N generalized coordinates qj (through the mik
coefficients) and their time derivatives q̇j . Secondly, the coefficients mik are “symmetric”;
that is, mik = mki .

• The potential energy is a function of only the spatial configuration of the system. Therefore,
U can be expressed completely in terms of the generalized coordinates qi and will not involve
their time derivatives q̇i : U = U (q1 , q2 , ..., qN ). Using the chain rule of di↵erentials, we can
therefore write:

N
X @U
dU = dqi
@qi
i=1

• Suppose that we have M nonconservative forces F~j (j = 1, 2, ..., M ) acting at the following
M locations within the system: ⇢
~j = ⇢
~j (q1 , q2 , ..., qN ). Using the chain rule of di↵erentials
gives us:

N
X @~
⇢j
d~
⇢j = dqi
@qi
i=1
 Using this in the di↵erential work term produces:
M
X
dW (nc)
= F~j · d~
⇢j

Chapter I
j=1
M N
!
X X @~
⇢j
= F~j · dqi
@qi
j=1 i=1
0 1
N
X XM
@~
⇢ jA
= @ F~j · dqi
@qi
i=1 j=1
N
X
= Qi dqi
i=1

where
M
X @~
⇢j
Qi = F~j ·
@qi
j=1

= “generalized force” corresponding to the generalized coordinate qi

• In summary, the di↵erential form of the power equation:

dT + dU = dW (nc)

can be written in the following explicit form:

N 
X ✓ ◆
d @T @T @U
+ Qi dqi = 0
dt @ q̇i @qi @qi
i=1
Example I.3.1

Forces f1 , f2 and f3 act on the three-particle system shown. Generalized coordinates x1 , x2 and
x3 are used to describe the motion of these particles, where x2 and x3 are relative coordinates.
• Find the generalized forces corresponding to the generalized coordinates x1 , x2 and x3 for the
forces f1 , f2 and f3 .
• Find the generalized mass coefficients corresponding to the generalized coordinates x1 , x2
and x3 .

x1 x2 x3
y f1 f2 f3
1 2 3

m m m
x
Chapter I
I.4 EOM’s: Lagrange’s equations
Objectives
Our goal here is to develop a systematic method for deriving the equations of motion for a system
having N degrees of freedom. We will start with the explicit form of the power equation derived
in the last section. From the power equation we will obtain a set of N EOM’s for the system.

Background
In the last section of the notes, we saw that the power equation for a system described by N
generalized coordinates qi (i = 1, 2, ..., N ) can be written as:

N 
X ✓ ◆
d @T @T @U
+ Qi dqi = 0
dt @ q̇i @qi @qi
i=1

where T and U are the kinetic and potential energies for the system, and Qi is the generalized force
corresponding to the coordinate qi :
M
X @~
⇢j
Qi = F~j ·
@qi
j=1
Derivation and results
It is assumed that the generalized coordinates qi (i = 1, 2, ..., N ) used above completely describe the
configuration of the system at all instances in time. However, we have said nothing at this point
as to whether the coordinates chosen are independent. That is, some of the coordinates could be

Chapter I
related by constraints that have not yet been enforced, and as a result, the coordinates are not
!
independent.

Consider an example of the simple pendulum shown below. The kinetic and potential energies for

x
L

y m P

this system in terms of the (x, y) coordinates for the particle P are given by:
1 1
T = m ẋ2 + ẏ 2 = m q̇12 + q̇22
2 2
U= mgy = mgq2
where q1 = x and q2 = y. Note that the chosen generalized coordinates are related by the constraint
q12 + q22 = L2 , which by di↵erentiating with respect to time produces: q̇12 = q22 q̇22 / L2 q22 . Al-
though q1 and q2 completely describe the configuration of the system, they are NOT independent.
If we enforce the above constraint, T and U can be written in terms of q2 alone as:

1 q2
T = m 2 2 2 + 1 q̇22
2 L q2
U= mgq2
This single degree-of-freedom system is now described in terms of a single coordinate.

An alternate (and better) choice of coordinates would be to describe the system in terms of the
angle ✓. From the figure we see that we have the following constraints: x = Lsin✓ and y = Lcos✓.
Substituting these constraints into the original T and U gives:
1 h i 1 1
T = m L2 ✓˙2 cos2 ✓ + L2 ✓˙2 cos2 ✓ = mL2 ✓˙2 = mL2 q̇32
2 2 2
U= mgLcos✓ = mgLcosq3
where q3 = ✓. Again, we have used a single coordinate in describing the motion of a single degree-
of-freedom system.

At this point, let us assume that all of the generalized coordinates are independent: any given
coordinate cannot be expressed in terms of the remaining N -1 coordinates. With this being the
case, N now represents the total number of degrees of freedom in the system. Since the coordinates
are independent, the general form of our power equation:
N 
X ✓ ◆
d @T @T @U
+ Qi dqi = 0
dt @ q̇i @qi @qi
i=1

becomes N independent EOM’s of the form:

✓ ◆
d @T @T @U
+ = Qi
dt @ q̇i @qi @qi

for i = 1, 2, ..., N . The above are known as the set of Lagrange’s equations for an N degree-of-
freedom system.

Recall that the potential energy term U includes the work done by conservative forces (such as
gravitational and spring forces). The generalized force terms Qi includes the work done by all other
forces. The contribution of damping terms naturally appears within the generalized force terms.
It is possible to make the contribution of damping forces more explicit in Lagrange’s equations
through a so-called Rayleigh dissipation function R. The Rayleigh dissipation function for a single
linear dashpot can be written as:
1
Rj = cj ˙ 2j
2

where cj is the damping coefficient and ˙ j is the relative speed across the two ends of the dashpot.
If the system has r dashpots, then the total Rayleigh dissipation function is:
r
X r
1X ˙2
R= Rj = cj j
2
j=1 j=1

Including damping produces the following modified form of Lagrange’s equations:

✓ ◆
d @T @T @R @U
+ + = Qi
dt @ q̇i @qi @ q̇i @qi

for i = 1, 2, ..., N . Here, the generalized forces Qj include all nonconservative forces except those
from the damping terms included in R.
Steps in using Lagrange’s equations
1. Number of degrees of freedom (DOFs) Determine the number of degrees of freedom
(DOF’s) in the problem. To do so, carefully consider the least number of generalized coor-
dinates N that are needed to completely describe the configuration of the system at any time.

Chapter I
2. Generalized coordinates. Choose your set of generalized coordinates qj (t) for j = 1, 2, ..., N .
Be sure that you have chosen an independent set of coordinates. (Ask the question: Can you
change each coordinate individually while holding the other coordinates fixed and not violate
any motion constraints on the system?) If your coordinates are not independent, there are
constraints that must exist among your coordinates. Enforce the constraints at this point
before proceeding. Also, reconsider your decision on the number of DOF’s in 1. above before
continuing: What is the correct number of DOFs?

3. Free body diagrams. Draw free body diagrams (FBD’s) for all bodies in your system.
These FBD’s will be necessary later on when you derive the generalized forces acting on the
system.

4. Kinetic energy expression, T.

• Write down the velocity vector corresponding to point Aj (either center of mass or fixed
point) for each body (j = 1, 2, ..., n):

d~rj
~vj =
dt

If Aj is a fixed point, then, of course, ~vj = 0.

• Write down the angular velocity corresponding to each body (j = 1, 2, ..., n):

d✓j
!j =
dt

• Form the kinetic energy expression for each body:

1 1
Tj = mj ~vj · ~vj + Ij !j2
2 2

(Note that the first term in the above expression is a dot product of the velocity vector
with itself. Take care to write down the vector expression for velocity before writing
down the kinetic energy expression. Also, Ij is the mass moment of inertia of the jth
body for point the choice of point Aj . If the body is to be treated as a particle, then
Ij = 0.)

• Form the total kinetic energy for the system by summing up the kinetic energy expres-
sions for the n rigid bodies:
n
X
T = Tj
j=1
 • Expand the expression for T to explicitly show the appearance of the q̇j terms:
N N
1 XX
T = mik q̇i q̇k
2
i=1 k=1

From this, identify the mass matrix elements mik . This step will help to simplify the
di↵erentiation process later on.

5. Potential energy expression, U.

• Write down the potential energy for each spring in the system:
1 2
(Usp )j = kj j
2
• Write down the gravitational potential energy for each body in the system:

(Ugr )j = mj ghj

where hj is the elevation of the center of mass of the body above the datum line that
you have chosen. Please note that if hj > 0, the center of mass is above the datum line,
whereas hj < 0 corresponds to the center of mass below the datum line.
• Form the total potential energy for the system:

X n
X
U= (Usp )j + (Ugr )j
j=1

6. Rayleigh dissipation expression, R.

• Write down the Rayleigh dissipation for each dashpot in the system:
1 2
Rj = c j ˙ j
2
Form the total Rayleigh dissipation function for the system:
r
X
R= Rj
j=1

7. Generalized forces, Qi .
• Reconsider your FBD’s of the system. Determine which forces that do virtual work on
the system. A reminder of some forces that you will NOT include:
– Forces due to springs, gravitation attraction and dashpots (since you have already
included these in U and R).
– Contact forces at smooth interfaces. These will do no work on the system.
– Frictional forces required for rolling without slipping. These do no work on the
system.
 – Constraint forces due to rigid connections between bodies. These do work on the
individual bodies but NOT on the entire system.
• Write down position vectors for the point of application of forces that do virtual work
on the system:

Chapter I
⇢
~j = ⇢
~j (q1 , q2 , ..., qN )

• Find the di↵erential change in the position vectors for the point of application of the
forces:
N
X @~
⇢j
d~
⇢j = dqi
@qi
i=1

• Find the di↵erential work for each force:

N
X @~
⇢j
dWj = F~j · d~
⇢j = F~j · dqi
@qi
i=1

• Find the di↵erential work for entire system, expand and identify the generalized forces
Qi (i = 1, 2, , N ) as the coefficients of the individual dqi terms in the following expression
for dW :
M
X N
X
dW = Wj = Qi dqi
j=1 i=1

8. Equations of motion.
• Form the partial derivatives of T with respect to qi and q̇i (i = 1, 2, , N ) . Recall that in
forming the partial derivatives with respect to variables qi and q̇i , only the EXPLICIT
appearance of the variables are a↵ected by the partial di↵erentiation.
• Form the total time derivative for the terms:
✓ ◆
d @T
dt @ q̇i

Here, when finding the total time derivative, both EXPLICIT and IMPLICIT (through
the qi and q̇i terms) appearance of time t is a↵ected by the di↵erentiation.
• Form the partial derivatives of U and R with respect to qi and q̇i , respectively
• Combine results to form the N Lagrange’s equations for the system (i = 1, 2, ..., N ):
✓ ◆
d @T @T @R @U
+ + = Qi
dt @ q̇i @qi @ q̇i @qi

where N is the number of DOF’s in the system.

Concluding remarks on Lagrange’s equations: some advanced ideas
Recall that we made two important assumptions in deriving the preceding form of Lagrange’s equa-
tions: i) the position vectors and rotation angles are not explicit functions of time (“scleronomic”
systems), and ii) we have chosen an independent set of coordinates. The following discusses the
consequences of these assumptions and what needs to be done when the assumptions are not valid
for a particular system of interest.

Rheonomic systems. “Rheonomic” systems are those for which the position vectors and rotation
angles are explicit functions of time: ~rj = ~rj (q1 , q2 , ..., qN , t) and ✓j = ✓j (q1 , q2 , ..., qN , t). With the
explicit time dependence, the chain rule of di↵erentials gives us:
N
X @~rj @~rj
d~rj = dqi + dt
@qi @t
i=1

With this, we see that the di↵erential work for an applied force includes an additional term due to
this explicit time dependence:
"N #
X @~
⇢ j @~
⇢ j
dWj = F~j · d~
⇢j = F~j · dqi + dt
@qi @t
i=1

For rheonomic systems, Lagrange’s equations are typically developed in terms of “virtual” displace-
ments ~rj , where:
N
X @~rj
~rj = qi
@qi
i=1

Note that virtual displacements are di↵erential displacements where the explicit time dependence
is frozen while applying the chain rule. In fact, for scleronomic systems, virtual and di↵erential
displacements are exactly the same: ~rj = d~rj .

Using virtual displacements gives us “virtual work”, Wj :

"N # N  N
X @~⇢j X @~
⇢j X
~
W j = Fj · ⇢ ~
~ j = Fj · qi = F~j · qi = Q j qi
@qi @qi
i=1 i=1 i=1

and, as a result, the generalized forces Qi are the same as before. Therefore, the Lagrangian formu-
lation developed in this section for scleronomic systems are still applicable for rheonomic systems
provided that we use virtual work (rather than di↵erential work) in finding our generalized forces.

Example of a rheonomic system

One of the most common instances of rheonomic systems are those on which “base excitation” is
applied. Consider the two-DOF system shown below where the base “B” of the system is given a
prescribed motion of y(t) = y0 sin!t. The motion of particles 1 and 2 is described by the relative
coordinates x1 and x2 . With that, the position vectors for these two particles are given by:

~r1 = [y0 sin!t + x1 ] î = ~r1 (x1 , t)

 ~r2 = [y0 sin!t + x1 + x2 ] î = ~r2 (x1 , x2 , t)
where we see here that due to the prescribed motion of base B, the position vectors for the two
particles are explicit functions of time.

Chapter I
y(t) x1 x2
ĵ
f1 f2
1 2

iˆ
B

Using the chain rule of di↵erentials:

d~r1 = [y0 !cos!tdt + dx1 ] î

d~r2 = [y0 !cos!tdt + dx1 + dx2 ] î

and the definition of a virtual displacement:

~r1 = [ x1 ] î

~r2 = [ x1 + x2 ] î
we can write the work and virtual work done by the applied forces:

dU = f~1 · d~r1 + f~2 · d~r2 = (y0 !cos!t) (f1 + f2 ) dt + (f1 + f2 ) dx1 + f2 dx2

U = f~1 · ~r1 + f~2 · ~r2 = (f1 + f2 ) x1 + f2 x2

The di↵erence between the work and virtual work expressions above is that virtual work does
not include the influence of the prescribed displacement of the base within the work expression,
whereas it must be included in the work expression. This is how using virtual displacements in
the Lagrangian formulation simplifies the EOM’s and is the recommended formulation for deriving
EOM’s for rheonomic systems.
Nonholonomic systems. Consider the following system made up of particles A and B. Particle


vB

cart, C
A B  θ
vC A
m m
smooth knife − edge
Iˆ iˆ
ĵ

side view top view

Ĵ

A moves within a slot cut into cart C (of mass M ), as shown, where C is moving with a prescribed
speed (vC ) and direction (X). The slot is perpendicular to the prescribed direction of motion for
C. Particle B is attached to A with a massless, rigid rod of length L. B rides along on the same
horizontal surface as C on a knife-edge support. This support allows for smooth sliding of B along
the direction of the rod but does not allow slip perpendicular to the rod. Rotation about the
contact point of the knife-edge with the ground is possible. Enforcing these constraints on a rigid
body kinematics equation relating the velocities of A and B gives:
~vB = ~vA + ! ~ ⇥ ~rB/A
⇣ ⌘ ⇣ ⌘
vB î = ẊA Iˆ + ẎA Iˆ + ✓˙k̂ ⇥ Lî
= ẊA Iˆ + ẎA Iˆ L✓˙ĵ
From the figure we see that:
Iˆ = cos✓î sin✓ĵ
Jˆ = sin✓î + cos✓ĵ
Therefore:
⇣ ⌘ ⇣ ⌘
vB î = ẊA cos✓ + ẎA sin✓ î + ẊA sin✓ + ẎA cos✓ L✓˙ ĵ

Balancing coefficients gives the following two constraints:

vB = ẊA cos✓ + ẎA sin✓ = vC cos✓ + ẎA sin✓
0= ẊA sin✓ + ẎA cos✓ L✓˙ = vC sin✓ + ẎA cos✓ L✓˙
Using the first constraint equation above in the kinetic energy expression for the system gives:
1 2 1 2 1 2
T = mvA + mvB + M vC
2 2 2
1 ⇣ 2 ⌘ 1 ⇣ ⌘ 1
= m vC + ẎA2 + m vC 2
cos2 ✓ + ẎA2 sin2 ✓ + M vC
2
2 2 2
From this we see that the kinetic energy is a function of both coordinates YA and ✓. The second
constraint above shows that these coordinates are NOT independent; however, due to the nature
of this constraint, we are not able to enforce it. (Do you see the problem? The constraint is
non-integrable; that is, we cannot relate ✓ to YA without actually solving the problem first.) The

Chapter I
Lagrangian formulation developed here relies on independent coordinates, and therefore we cannot
use this form of Lagrange’s equations to develop the EOM for this single DOF system.

Systems having non-integrable constraints (such as the example above) are known as “nonholo-
nomic”. Since we cannot enforce these constraints apriori, Lagrange’s equations cannot be applied
to such systems. There are a number of ways to modify Lagrange’s equations to handle nonholo-
nomic systems. We will not pursue that here. You are encouraged to read advanced dynamics
textbooks if you are interested in learning more. Should we encounter such systems in this course,
we will avoid this complication and develop the EOM’s through the Newton-Euler formulation.
Example I.4.1

Find the EOM for the simple pendulum using ✓ as the generalized coordinate.

y
O
g
L
θ
x
P
m
negligible ma ss
Chapter I
Example I.4.2

Particles 1 and 2 (having masses of m and M , respectively) move in a HORIZONTAL plane.

Generalized coordinates x1 and x2 describe the position of the system where x2 is measured relative
to particle 1. Forces f and F act in the x-direction on particles 1 and 2, respectively. The springs
are unstretched when x1 = x2 = 0. Find the EOM’s of the system using generalized coordinates
x1 and x2 . Ignore gravity. Consider all of the surfaces to be smooth.

x1
y
K M c 1

x2 m x

f F
k 2
C θ
Chapter I
Example I.4.3

Find the EOM’s of system shown below using x, and ✓ as generalized coordinates. The spring
is unstretched when x = = 0. All three bodies are to be considered to be homogeneous in their
mass distribution, with each body have a mass of m.

y
φ x
x
1 2
r k A

O g
no slip smooth

L
θ 3
Chapter I
Example I.4.4

If particle 0 is given a prescribed motion of y(t) = y0 sin⌦t, find the EOM’s for the system. Note
that x1 and x2 are relative coordinates. The springs are unstretched when y = x1 = x2 = 0.

y(t) x1 x2

f1 f2
1 2

k m k m k
Chapter I
I.5 EOM’s: Linearization
Objectives
The EOM’s for a system are typically nonlinear, with the nonlinearities arising from geometric and
material e↵ects. In this course, we will study small oscillations of systems. For this, we will use
a linearized set of free response EOM’s arising from Lagrange’s equations. In this section of the
notes, we will develop a systematic approach for this linearization process for free motion of the
system.

Background
• The kinetic energy for discrete models is of the form:
N N
1 XX
T = mik q̇i q̇k
2
i=1 k=1

where the symmetric mass coefficients mik are given by:

n 
X @~rj @~rj @✓j @✓j
mik = mj · + Ij
@qi @qk @qi @qk
j=1

• For free response, the system is at an equilibrium state ~q0 when (i = 1, 2, ..., N ):
✓ ◆
@U
=0
@qi q~0

• The Taylor series expansion of a function g(~q ) about the point ~q0 is given by:
N 
X N N 
@g 1 X X @2g
g(~q ) = g(~q0 ) + (qi q0i ) + (qi q0i ) (qk q0k ) + ...
@qi q~0 2 @qi qk q~0
i=1 i=1 k=1

Results
Small motion about an equilibrium state of a free system is governed by the following linearized
EOM’s:

[M ] ~z¨ + [C] ~z˙ + [K] ~z = ~0

where

~z (t) = ~q (t) ~q0

Mik = (mik )q~0 = Mki

✓ 2 ◆
@ R
Cik = = Cki
@qi @qk q~0
✓ 2 ◆
@ U
Kik = = Kki
@qi @qk q~0
Derivation and remarks

Chapter I
Example I.5.1

In the system shown, the springs are unstretched when x11 = x22 = 0. Find the mass, damping and
sti↵ness matrices for small motion about the equilibrium state corresponding to the generalized
coordinates shown. Note that x2 is a relative coordinate.

x1 2m x2

m g
k k 2 y

c c 1
x

L negligible mass

θ
3 f

~r1 = x1 î
~r1 = x1 î
~r2 = (x1 + x2 )î
~r2 = (x1 + x2 )î
~r3 = (x1 + x2 + Lsin✓)î + ( Lcos✓)ĵ
~r3 = (x1 + x2 + Lsin✓)î + ( Lcos✓)ĵ
~v1 = ẋ1 î
~v1 = ẋ1 î
~v2 = (ẋ1 + ẋ2 )î
~v2 = (ẋ1 + ẋ2 )î ˙ ˙
~v3 = (ẋ1 + ẋ2 + L✓cos✓)î + (L✓sin✓) ĵ
˙ ˙
~v3 = (ẋ1 + ẋ2 + L✓cos✓)î + (L✓sin✓)ĵ
Chapter I
I.6 EOM’s: Influence coefficients and the flexibility matrix
Objectives
Later on in our investigation of approximate methods for finding natural frequencies and modal
vectors, we will find it necessary to obtain the inverse of the sti↵ness matrix [K]. This inverse,
[K] 1 , is known as the “flexibility matrix”. In this section, we will explore a method for obtaining
the flexibility matrix directly without the need for performing the inversion of an N ⇥ N matrix.

Background
• Consider a generalized force Qj = 1 (unit magnitude) corresponding to generalized coordinate
qj . Let qi be the displacement of the ith generalized coordinate as a result of the applied force
Qj . The “influence coefficient” aij is defined as displacement of the generalized coordinate qi

qi qj
Qj

due to the application of a unit magnitude generalized force Qj .

• We have seen that the sti↵ness matrix for a conservative system obtained from the Lagrangian
formulation is symmetric; that is, [K]T = [K]. Since the order of the transpose and inverse
operations on a non-singular matrix is reversible, it can be seen that [K] 1 is also symmetric:
⇣ ⌘T ⇣ ⌘ 1
1
[K] = [K]T = [K] 1

Results
⇣ ⌘
1
• The influence coefficients aij form the elements of the flexibility matrix; that is, [K] =
ij
aij .

• For a linearly elastic system, the influence coefficients aij defined above are symmetric: aji .
In words, this says that a unit force applied at j will produce the same displacement at i
as the displacement at j due to a unit force at i. This is known as “Maxwell’s reciprocity
theorem”.
 • Method for determing the influence coefficients:
– Apply a unit generalized force Qj with Qk = 0 for k 6= j.

– Calculate/measure the change in the displacements of the generalized coordinates qi (i =

Chapter I
1, 2, ...N ). These displacements are the influence coefficients aij (i = 1, 2, ...N ).

– Use the symmetry of aij to simplify your calculations; that is, although a symmetric
N ⇥ N matrix has N 2 terms, it has only N (N + 1)/2 unique terms. Furthermore, aij
may be easier to compute than aji in some instances.

Derivation and remarks

Example I.6.1
Influence Coefficients and the Flexibilit
Consider a homogeneous, clamped beam having a length of L and flexural rigidity (Sections
of EI. Find the
6.1-6.2)
flexibility matrix for the beam using the transverse deflections at the three particles as generalized
coordinates. Note that from mechanics of materials, we know that the deflection at x, u(x), due to
a force F applied at location a x is given by:
Example 1 - flexibility matrix for cantilevered beam
F ha homogeneous,
Consider
i
clamped beam having a length of 3L and flexural rigidity
u(x) = 2a3 3a2 (a x) + (a x)3
of EI.6EI
Find the flexibility matrix for the beam using the transverse deflections at the
three particles as generalized coordinates.

F
a
L/3 L/3 L/3

x u(x)

Solution
From mechanics of materials, we know that the deflection at x, u(x), due to a force F
applied at location a is given by:

F 3 2 3
u(x) = 6EI [2a - 3a (a - x) + (a - x) ] ; x≤a

I.26
Chapter I
Example I.6.2

Find the flexibility matrix for the three-DOF system using the absolute coordinates shown. The
springs are unstretched when x1 = x2 = x3 = 0.

x1 x2 x3

k 2k k
1 2 3
smooth
Chapter I
I.7 EOM’s: Summary
Newton-Euler formulation
• Vector approach dealing with each body individually.
• Must deal with ALL forces acting on each body.
• Often times the reduction of the equations to a number equal to the number of DOF’s can
be tedious.
• Dealing with the signs and direction of spring forces can be tedious.

Power equations
• Scalar approach dealing with the system as a whole.
• Need not deal with forces which do no work on the system as a whole.
• Dealing with the signs and direction of spring forces is more straight-forward.
• Useful for single-DOF systems only.

Lagrange’s equations
• Scalar approach dealing with the system as a whole.
• Need not deal with forces which do no virtual work on the system as a whole.
• Dealing with the signs and direction of spring forces is more straight-forward.
• Valid for systems with any number of DOF’s provided that one uses independent generalized
coordinates.
• Development of generalized forces is a straight-forward procedure when using expressions for
virtual work:
– write down a vector expression for each force, F~j .
– write down a vector expression for the di↵erential (or virtual) displacement for the point
of application of each force, d~
⇢j
– then from
M
X M
X
dW = F~j · d~
⇢j = Qj dqj
j=1 j=1

we can identify the corresponding generalized forces Qj .

Linearized EOM’s
• General form of the linearized EOM’s for free response: [M ] ~z¨ + [C] ~z˙ + [K] ~z = ~0
• The mass and sti↵ness matrices are always symmetric (unlike those found from Newton-
Euler).
• These EOM’s describe small motion about an equilibrium state: ~z (t) = ~q (t) ~q0 .