A Complete Guide to

Engineering Vibrations
Part II: Forced Response, Multi-DOF
Systems,
and Continuum Vibrations

Topics Covered:
ˆ Forced & Harmonic Response ˆ Energetics of Forcing
ˆ Periodic Forcing (Fourier) ˆ Rotating Unbalance
ˆ Harmonic Base Excitation ˆ IC Engine Dynamics
ˆ Impulse, Step & Shock ˆ Arbitrary Excitation
ˆ Multi-DOF Systems ˆ Two-DOF Modal Analysis
ˆ Orthogonality Theorems ˆ Discrete-to-Continuum
ˆ Classical Normal Modes

Continuation of Part I — Chapters 9 through 20

Version 1.0 February 18, 2026

ii
Preface to Part II

Part I established the language of vibrations: ODEs, Lagrangian mechanics, equilibrium,

linearisation, free vibration in all four damping regimes, single and two DOF free dynamics,
Coulomb friction, and viscoelasticity.
Part II puts energy into the system. Real vibrating structures are driven by forces —
engines, earthquakes, wind, rotating imbalances, sudden impacts. We will methodically
build up the complete response to every important class of excitation, from steady-state
harmonic loading through impulses, steps, shocks, and arbitrary discontinuous signals.
We then generalise to N -degree-of-freedom systems, proving the orthogonality of mode
shapes in full rigour, performing complete modal analysis, and finally taking the limit
as the number of DOFs goes to infinity to arrive at the partial differential equations of
continuous structures — strings, rods, and beams.
Throughout, the same formatting conventions as Part I apply: definitions, theorems,
worked examples, key results, and notes and physical interpretations. Chapter and
equation numbering continues directly from Part I.

iii
iv
Contents

Preface to Part II iii

9 Forced Response: General Framework 1

9.1 From Free to Forced Vibration . . . . . . . . . . . . . . . . . . . . . . . . . 1
9.2 Structure of the Complete Solution . . . . . . . . . . . . . . . . . . . . . . 1
9.3 The Frequency Response Function (FRF) . . . . . . . . . . . . . . . . . . . 1
9.4 Fourier Transform and Response in Frequency Domain . . . . . . . . . . . 2
9.5 Time-Domain via Convolution (Recap and Extension) . . . . . . . . . . . . 2

10 Forced Response under Harmonic Excitation and Its Energetics 3

10.1 The Harmonically Forced SDOF Oscillator . . . . . . . . . . . . . . . . . . 3
10.2 Derivation of the Steady-State Response . . . . . . . . . . . . . . . . . . . 3
10.2.1 Complex Exponential Method . . . . . . . . . . . . . . . . . . . . . 3
10.2.2 Non-Dimensional Representation . . . . . . . . . . . . . . . . . . . 4
10.2.3 Complete (Transient + Steady-State) Response . . . . . . . . . . . 4
10.3 Resonance: Physics and Consequences . . . . . . . . . . . . . . . . . . . . 5
10.4 Energetics of Harmonic Forcing . . . . . . . . . . . . . . . . . . . . . . . . 5
10.4.1 Instantaneous and Average Power . . . . . . . . . . . . . . . . . . . 5
10.4.2 Power Balance at Steady State . . . . . . . . . . . . . . . . . . . . 6
10.4.3 Hysteresis Loop (Force–Displacement) . . . . . . . . . . . . . . . . 6
10.4.4 Quality Factor Q . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

11 Periodic Forcing 7
11.1 What is Periodic Forcing? . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
11.2 Fourier Series Decomposition . . . . . . . . . . . . . . . . . . . . . . . . . 7
11.3 Response to Periodic Forcing via Superposition . . . . . . . . . . . . . . . 7
11.4 Example: Rectangular Pulse Train . . . . . . . . . . . . . . . . . . . . . . 8
11.5 Gibbs Phenomenon and Practical Truncation . . . . . . . . . . . . . . . . . 8
11.6 Response in the Frequency Domain . . . . . . . . . . . . . . . . . . . . . . 8

12 Rotating Unbalance 11
12.1 The Problem of Rotor Imbalance . . . . . . . . . . . . . . . . . . . . . . . 11
12.2 Physical Model and EOM . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
12.3 Steady-State Response . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
12.3.1 Key Observations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
12.4 Balancing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

13 Harmonic Base Excitation 15

v
vi Contents

13.1 When the Support Moves . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

13.2 Equation of Motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
13.3 Absolute and Relative Response . . . . . . . . . . . . . . . . . . . . . . . . 15
13.3.1 Absolute Displacement (Motion Transmissibility) . . . . . . . . . . 15
13.3.2 Relative Displacement . . . . . . . . . . . . . . . . . . . . . . . . . 16
13.4 Force Transmissibility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
13.5 Vibration Isolation Design Principles . . . . . . . . . . . . . . . . . . . . . 17

14 Dynamics of the Internal Combustion Engine 19

14.1 Why Engines Vibrate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
14.2 Slider-Crank Mechanism Kinematics . . . . . . . . . . . . . . . . . . . . . 19
14.3 Engine Forcing — Harmonic Orders . . . . . . . . . . . . . . . . . . . . . . 20
14.4 Multi-Cylinder Engine: Firing Order and Balance . . . . . . . . . . . . . . 20
14.5 Torsional Vibration of Crankshaft . . . . . . . . . . . . . . . . . . . . . . . 20
14.6 Engine Mounting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

15 Transient Excitations: Impulse, Step, Shock, and Arbitrary Forcing 23

15.1 Overview of Transient Loading . . . . . . . . . . . . . . . . . . . . . . . . . 23
15.2 Impulsive Excitation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
15.2.1 The Unit Impulse (Dirac Delta) . . . . . . . . . . . . . . . . . . . . 23
15.2.2 Impulse Response Function h(t) . . . . . . . . . . . . . . . . . . . . 23
15.2.3 Response to a Finite-Duration Impulse . . . . . . . . . . . . . . . . 24
15.3 Unit Step Excitation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
15.3.1 The Unit Step Function . . . . . . . . . . . . . . . . . . . . . . . . 24
15.3.2 Step Response Function g(t) . . . . . . . . . . . . . . . . . . . . . . 24
15.4 Shock Excitation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
15.4.1 Shock Response Spectrum . . . . . . . . . . . . . . . . . . . . . . . 25
15.4.2 Half-Sine Shock Pulse . . . . . . . . . . . . . . . . . . . . . . . . . 25
15.4.3 Pulse Duration vs. Natural Period . . . . . . . . . . . . . . . . . . . 25
15.4.4 Applications of SRS . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
15.5 Discontinuous Arbitrary Excitation . . . . . . . . . . . . . . . . . . . . . . 26
15.5.1 The Duhamel Convolution Integral . . . . . . . . . . . . . . . . . . 26
15.5.2 Handling Discontinuities . . . . . . . . . . . . . . . . . . . . . . . . 26
15.5.3 Rectangular Pulse . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
15.5.4 Numerical Convolution . . . . . . . . . . . . . . . . . . . . . . . . . 27
15.5.5 Superposition of Unit Steps (Ramp and Arbitrary) . . . . . . . . . 27

16 Multi-Degree-of-Freedom Oscillator 29
16.1 Extending to N Degrees of Freedom . . . . . . . . . . . . . . . . . . . . . . 29
16.2 Matrix Equation of Motion . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
16.3 Eigenvalue Problem and Natural Frequencies . . . . . . . . . . . . . . . . . 29
16.3.1 Undamped Free Vibration . . . . . . . . . . . . . . . . . . . . . . . 29
16.4 Modal Analysis: The Complete Procedure . . . . . . . . . . . . . . . . . . 30
16.4.1 Step 1: Solve the Eigenvalue Problem . . . . . . . . . . . . . . . . . 30
16.4.2 Step 2: Mass-Normalise the Mode Shapes . . . . . . . . . . . . . . 30
16.4.3 Step 3: Construct the Modal Matrix . . . . . . . . . . . . . . . . . 30
16.4.4 Step 4: Modal Transformation . . . . . . . . . . . . . . . . . . . . . 30
16.4.5 Step 5: Decouple into N SDOF Systems . . . . . . . . . . . . . . . 31
16.4.6 Step 6: Solve and Back-Transform . . . . . . . . . . . . . . . . . . . 31
Contents vii

16.5 Forced Harmonic Response of Multi-DOF System . . . . . . . . . . . . . . 31

17 Two-DOF Oscillator: Complete Forced Analysis 33

17.1 Recap of Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
17.2 Natural Frequencies: Detailed Derivation . . . . . . . . . . . . . . . . . . . 33
17.2.1 Mode Shapes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
17.3 Forced Response via Modal Analysis . . . . . . . . . . . . . . . . . . . . . 34
17.3.1 Antiresonance of Mass 1 . . . . . . . . . . . . . . . . . . . . . . . . 34
17.4 Frequency Response Functions of 2-DOF System . . . . . . . . . . . . . . . 34

18 Orthogonality of Mode Shapes and Orthogonal Vectors 35

18.1 Motivation: Why Orthogonality Matters . . . . . . . . . . . . . . . . . . . 35
18.2 Euclidean Orthogonality: Review . . . . . . . . . . . . . . . . . . . . . . . 35
18.3 M-Orthogonality and K-Orthogonality . . . . . . . . . . . . . . . . . . . . 35
18.4 The Generalized Orthogonality Relations . . . . . . . . . . . . . . . . . . . 36
18.5 Expansion Theorem (Modal Expansion) . . . . . . . . . . . . . . . . . . . 37
18.6 Repeated Eigenvalues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
18.7 Rayleigh Quotient . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37

19 Continuum and Long-Wave Approximation: From Discrete to Contin-

uum 39
19.1 The Limiting Process: N → ∞ . . . . . . . . . . . . . . . . . . . . . . . . 39
19.2 Longitudinal Vibration of a Rod . . . . . . . . . . . . . . . . . . . . . . . . 39
19.3 Transverse Vibration of a Taut String . . . . . . . . . . . . . . . . . . . . . 40
19.4 Transverse Vibration of an Euler–Bernoulli Beam . . . . . . . . . . . . . . 40
19.5 The Long-Wave (Low-Frequency) Approximation . . . . . . . . . . . . . . 40
19.5.1 Dispersion Relation . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
19.5.2 Physical Meaning . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
19.6 Separation of Variables: Normal Modes of the Rod . . . . . . . . . . . . . 41
19.6.1 Boundary Conditions and Eigenfrequencies . . . . . . . . . . . . . . 41
19.7 Orthogonality of Continuum Mode Shapes . . . . . . . . . . . . . . . . . . 41

20 Characteristics of Classical Normal Modes 43

20.1 What is a Normal Mode? . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
20.2 Standing Wave Interpretation . . . . . . . . . . . . . . . . . . . . . . . . . 43
20.3 Properties of Normal Modes . . . . . . . . . . . . . . . . . . . . . . . . . . 44
20.3.1 1. Completeness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
20.3.2 2. Orthogonality . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
20.3.3 3. In-Phase Motion within Each Mode . . . . . . . . . . . . . . . . 44
20.3.4 4. Nodes and Antinodes . . . . . . . . . . . . . . . . . . . . . . . . 45
20.3.5 5. Energy Distribution . . . . . . . . . . . . . . . . . . . . . . . . . 45
20.3.6 6. Frequency Ordering and Modal Density . . . . . . . . . . . . . . 45
20.3.7 7. Rayleigh’s Principle (Stationarity) . . . . . . . . . . . . . . . . . 45
20.3.8 8. Modal Superposition and Independence . . . . . . . . . . . . . . 45
20.4 Damped Normal Modes . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
20.5 Experimental Modal Analysis (EMA) . . . . . . . . . . . . . . . . . . . . . 46
20.6 Summary: Key Properties of Classical Normal Modes . . . . . . . . . . . . 46

A Quick Reference: Part II Key Formulae 47

viii Contents

A.1 Harmonic Forcing (SDOF) . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

A.2 Energetics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
A.3 Rotating Unbalance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
A.4 Base Excitation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
A.5 Impulse and Step Response . . . . . . . . . . . . . . . . . . . . . . . . . . 47
A.6 Duhamel’s Integral . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
A.7 Multi-DOF: Modal Analysis Summary . . . . . . . . . . . . . . . . . . . . 48
A.8 Continuum: Rod and Beam . . . . . . . . . . . . . . . . . . . . . . . . . . 48

B Worked Problems: Part II 49

B.1 Problem 1: Full Transient + Steady-State Response . . . . . . . . . . . . . 49
B.2 Problem 2: 3-DOF Modal Analysis . . . . . . . . . . . . . . . . . . . . . . 49
B.3 Problem 3: Shock Response — Half-Sine Pulse . . . . . . . . . . . . . . . . 50

Further Reading 51
Chapter 9

Forced Response: General Framework

9.1 From Free to Forced Vibration

In Part I we studied the free response — what happens after an initial disturbance with
no ongoing forcing. In reality, most structures and machines are subjected to persistent or
transient external forces. The governing equation is:
mẍ + cẋ + kx = F (t), (9.1)
or, in standard form dividing by m:
F (t)
ẍ + 2ζωn ẋ + ωn2 x = . (9.2)
m

9.2 Structure of the Complete Solution

The complete response is always the sum of two distinct parts:

Key Result: Complete Response Structure

x(t) = xh (t) + xp (t) . (9.3)

| {z } | {z }
transient steady-state

ˆ Transient response xh (t): the homogeneous solution (free vibration at ωn or ωd ).

It decays to zero for ζ > 0.
ˆ Steady-state response xp (t): a particular solution driven by F (t). Its form mirrors
the input.
The transient part satisfies the ICs and “dies out.” After a few cycles, only the steady
state remains. The time to decay depends on ζ.

9.3 The Frequency Response Function (FRF)

For a linear system, the steady-state response to a sinusoidal input is always sinusoidal at
the same frequency. This motivates defining the Frequency Response Function:

1
2 Chapter 9. Forced Response: General Framework

Definition: Frequency Response Function

The FRF (also called receptance or admittance) H(Ω) relates output amplitude to
input amplitude:

X(Ω) 1/k Ω
H(Ω) = = , r= . (9.4)
F0 (1 − r2 ) + 2iζr ωn

Its modulus |H(Ω)| is the dynamic magnification factor (per unit static deflection); its
argument ∠H is the phase lag.

The FRF is a property of the system alone, independent of the particular input. Once H(Ω)
is known for all Ω, the response to any sinusoidal excitation is immediately determined.
Via the Fourier transform, it determines the response to any input.

9.4 Fourier Transform and Response in Frequency

Domain
Definition: Fourier Transform Pair
ˆ ∞ ˆ ∞
1
X(ω) = x(t) e−iωt dt, x(t) = X(ω) eiωt dω. (9.5)
−∞ 2π −∞

Transforming (9.1):
−ω 2 m + iωc + k X(ω) = F (ω),

so the response spectrum is:

X(ω) = H(ω) · F (ω), (9.6)
where H(ω) = 1/(k − mω 2 + icω). The FRF acts as a filter : it amplifies or attenuates
each frequency component of the input differently.

9.5 Time-Domain via Convolution (Recap and Exten-

sion)
The inverse Fourier transform of (9.6) gives the time-domain convolution:
ˆ t
x(t) = h(t − τ ) F (τ ) dτ, (9.7)
−∞

where h(t) is the impulse response function (IRF) — the inverse Fourier transform of
H(ω). This is one of the most powerful results in linear systems theory: the complete
response to any force history is given by a single integral involving h(t).
Chapter 10

Forced Response under Harmonic Ex-

citation and Its Energetics

10.1 The Harmonically Forced SDOF Oscillator

The most important special case of (9.1) is sinusoidal forcing:

mẍ + cẋ + kx = F0 cos Ωt. (10.1)

This models: rotating machinery, acoustic excitation, ocean waves acting on offshore
platforms, ground shaking modelled as a single harmonic, and many other situations.

10.2 Derivation of the Steady-State Response

10.2.1 Complex Exponential Method
Consider the complex equation mx̃¨ + cx̃˙ + kx̃ = F0 eiΩt ; the physical solution is x = Re(x̃).

Assume x̃p = XeiΩt where X is a complex amplitude. Substituting:

(−mΩ2 + icΩ + k)XeiΩt = F0 eiΩt .

Solving for X:

F0 F0 /k
X= = , r = Ω/ωn . (10.2)
k − mΩ2 + icΩ (1 − r2 ) + 2iζr

The real amplitude |X| and phase ψ = −∠X are:

3
4 Chapter 10. Forced Response under Harmonic Excitation and Its Energetics

Key Result: Harmonic Steady-State Response

F0 /k
|X| = p , (10.3)
(1 − r2 )2 + (2ζr)2
 
2ζr
ψ = arctan . (10.4)
1 − r2

The steady-state displacement is xp (t) = |X| cos(Ωt − ψ).

10.2.2 Non-Dimensional Representation

Defining the static deflection Xst = F0 /k and the DMF:

|X| 1
DMF(r, ζ) = =p . (10.5)
Xst (1 − r2 )2 + (2ζr)2

r=1
5
ζ = 0.05
4 ζ = 0.1
DMF = |X|/Xst

ζ = 0.2
ζ = 0.5
3 ζ = 1.0

0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2 2.4
Frequency ratio r = Ω/ωn

Figure 10.1: Dynamic Magnification Factor vs. frequency ratio for various damping ratios.
The dashed line marks resonance r = 1.

10.2.3 Complete (Transient + Steady-State) Response

The complete solution of (10.1) with ICs x(0) = x0 , ẋ(0) = v0 (under-damped case) is:

x(t) = e−ζωn t [A cos ωd t + B sin ωd t] + |X| cos(Ωt − ψ), (10.6)

where A and B are found by applying ICs to the total solution:

A = x0 − |X| cos ψ, (10.7)

v0 + ζωn A − |X|Ω sin ψ
B= . (10.8)
ωd
10.3. Resonance: Physics and Consequences 5

Note:
For most engineering analysis, one waits for the transient to decay (after ∼ 4/(ζωn )
seconds) and analyses the steady-state. But for start-up or shutdown analysis, the
transient is critical—it can produce amplitudes far exceeding the steady-state peak.

10.3 Resonance: Physics and Consequences

At r = 1 (excitation frequency equals natural frequency):
ˆ The DMF equals 1/(2ζ), which can be enormous for small ζ.
ˆ The phase lag is exactly 90 regardless of ζ. This is used experimentally to locate
resonance.
ˆ The driving force is entirely in phase with velocity: power delivery is maximised.
The precise frequency at which the amplitude peaks is:
p 1
rpeak = 1 − 2ζ 2 , valid for ζ < √ . (10.9)
2

Physical Interpretation:
At resonance the system absorbs energy most efficiently from the driving force. Even
a tiny force can build up enormous amplitude if damping is low — exemplified by
soldiers breaking step on bridges (Angers, 1850), or the Tacoma Narrows collapse
(1940). In benign applications, resonance is exploited in musical instruments, MRI
machines, and quartz oscillators.

10.4 Energetics of Harmonic Forcing

10.4.1 Instantaneous and Average Power
The instantaneous power delivered by the forcing function is:
P (t) = F (t) ẋ(t).
With F = F0 cos Ωt and ẋ = −|X|Ω sin(Ωt − ψ):
P (t) = −F0 |X|Ω cos Ωt sin(Ωt − ψ).

The time-averaged power over one cycle TΩ = 2π/Ω:

F0 |X|Ω
⟨P ⟩ = sin ψ. (10.10)
2

2ζr
Since sin ψ = p , substituting |X|:
(1 − r2 )2 + (2ζr)2

F02 (2ζr)2
⟨P ⟩ = · . (10.11)
2c (1 − r2 )2 + (2ζr)2
6 Chapter 10. Forced Response under Harmonic Excitation and Its Energetics

10.4.2 Power Balance at Steady State

At steady state, energy input per cycle equals energy dissipated per cycle. The energy
stored in the spring and mass oscillates but its average is constant.

Energy Dissipated per Cycle by Damper

˛ ˆ TΩ
Ediss = cẋ dx = cẋ2 dt = πcΩ|X|2 . (10.12)
0

Energy Input per Cycle by Force

˛ ˆ TΩ
Ein = F dx = F ẋ dt = πF0 |X| sin ψ. (10.13)
0

Setting Ein = Ediss gives F0 sin ψ = cΩ|X|, which is exactly satisfied by the solution
(10.3)–(10.4). This confirms energy balance. ✓

10.4.3 Hysteresis Loop (Force–Displacement)

The plot of F (t) vs. x(t) over one cycle traces a closed ellipse — a hysteresis loop. The
area enclosed by the loop equals the energy dissipated per cycle Ediss = πcΩ|X|2 .

Note: Half-power bandwidth

√
The two frequencies at which the DMF drops to 1/ 2 of its peak value are r1,2 = 1 ∓ ζ
(for small ζ). The half-power bandwidth is:

∆ω
∆r = r2 − r1 ≈ 2ζ, hence ζ ≈ .
2ωn
This is the standard experimental method for measuring damping from a frequency
response test.

10.4.4 Quality Factor Q

Definition: Quality Factor
1 ωn Peak DMF max energy stored
Q= = = = . (10.14)
2ζ ∆ω 1 energy dissipated per radian

High-Q systems oscillate many cycles before losing energy (e.g. quartz crystal: Q ∼ 106 ).
Low-Q systems dissipate energy rapidly (car shock absorber: Q ∼ 1).
Chapter 11

Periodic Forcing

11.1 What is Periodic Forcing?

A force F (t) is periodic with period T if F (t + T ) = F (t) for all t. Real machinery forces
are often periodic (but not purely sinusoidal): a reciprocating engine produces impulses at
regular intervals; a gear mesh produces a periodic pulse train; a cam follower produces
periodic non-sinusoidal loads.

11.2 Fourier Series Decomposition

Theorem: Fourier Series
Any periodic function F (t) with period T and fundamental frequency ω0 = 2π/T can
be written as: ∞
a0 X
F (t) = + [an cos(nω0 t) + bn sin(nω0 t)] , (11.1)
2 n=1

where the Fourier coefficients are:

ˆ
2 T
an = F (t) cos(nω0 t) dt, n = 0, 1, 2, . . . (11.2)
T 0
ˆ
2 T
bn = F (t) sin(nω0 t) dt, n = 1, 2, 3, . . . (11.3)
T 0

In complex (exponential) form:

∞ ˆ T
X 1
F (t) = Cn e inω0 t
, Cn = F (t) e−inω0 t dt. (11.4)
n=−∞
T 0

11.3 Response to Periodic Forcing via Superposition

Since the oscillator is linear, we apply superposition: the total steady-state response is the
sum of the responses to each harmonic:

7
8 Chapter 11. Periodic Forcing

Key Result: Response to Periodic Force

∞
a0 X 1
x(t) = + p [an cos(nω0 t − ψn ) + bn sin(nω0 t − ψn )] ,
2k n=1 (k − mn ω02 )2 + (cnω0 )2
2

  (11.5)
cnω0
where ψn = arctan .
k − mn2 ω02

Physical Interpretation:
Each harmonic of the force is amplified by the system’s own DMF at that harmonic’s
frequency. A harmonic that happens to be near ωn will be dramatically amplified.
This is why gear trains must be designed so that no tooth-mesh harmonic falls near a
structural natural frequency.

11.4 Example: Rectangular Pulse Train

Example: Fourier series of a square wave
A force alternates between +F0 and −F0 with period T :
(
+F0 0 ≤ t < T /2
F (t) =
−F0 T /2 ≤ t < T.

4F0
The Fourier coefficients: a0 = 0, an = 0, bn = for odd n, bn = 0 for even n.
nπ
 
4F0 1 1
F (t) = sin ω0 t + sin 3ω0 t + sin 5ω0 t + · · · .
π 3 5

The response is dominated by the fundamental but receives contributions from all odd
harmonics.

11.5 Gibbs Phenomenon and Practical Truncation

The Fourier series converges to the function at points of continuity, but exhibits overshoot
(≈ 9% of the jump) at discontinuities — the Gibbs phenomenon. In practice, a finite
number N of harmonics is used; typically N ≥ 10 gives adequate accuracy unless the force
has very sharp edges.

11.6 Response in the Frequency Domain

Using the complex Fourier series:
∞
X
xp (t) = Cn · H(nω0 ) einω0 t ,
n=−∞
11.6. Response in the Frequency Domain 9

where H(ω) = 1/(k − mω 2 + icω). This is the discrete-frequency analogue of equation (9.6).
10 Chapter 11. Periodic Forcing
Chapter 12

Rotating Unbalance

12.1 The Problem of Rotor Imbalance

Every rotating machine — motors, fans, turbines, washing machine drums — contains a
rotor. If the centre of mass of the rotor does not lie exactly on the spin axis, the system
is unbalanced. As the rotor spins, the eccentric mass generates a centrifugal force that
excites vibration. Unbalance is the single most common cause of machine vibration in
industry.

12.2 Physical Model and EOM

Consider a machine of total mass M mounted on a spring-damper support. The rotor
contains an eccentric mass m at eccentricity (radius) e from the spin axis. The rotor spins
at angular speed Ω.

m x
e
M −m

k/2c/2 c/2k/2

Figure 12.1: Machine of total mass M with eccentric mass m at eccentricity e, rotating at
Ω.
The position of the eccentric mass is x + e sin Ωt. Newton’s second law applied to the
entire system:

d2
M ẍ + cẋ + kx = m (e sin Ωt − x) · (−1) + (reaction) . . .
dt2

More cleanly: let Xm = displacement of eccentric mass = x + e sin Ωt. Force balance on
the whole system (mass M including m):

M ẍ + cẋ + kx = meΩ2 sin Ωt. (12.1)

11
12 Chapter 12. Rotating Unbalance

The rotating unbalance generates a centrifugal force of amplitude meΩ2 that increases
with Ω2 .

12.3 Steady-State Response

2
Equation (12.1) is a harmonically forced SDOF
p system with F0 = meΩ and msystem = M .
Using the results of Chapter 10 with ωn = k/M :

Key Result: Rotating Unbalance Response

MX r2 Ω
=p , r= . (12.2)
me (1 − r2 )2 + (2ζr)2 ωn
The non-dimensional group M X/(me) is called the unbalance magnification factor.

3
ζ = 0.05
ζ = 0.1
ζ = 0.2
2
M X/(me)

ζ = 0.5

M X/(me) → 1
1

0
0 0.5 1 1.5 2 2.5 3 3.5
r = Ω/ωn

Figure 12.2: Unbalance magnification factor. At high speeds (r ≫ 1) the response tends
to me/M (the mass moves as if the centre of mass is stationary).

12.3.1 Key Observations

ˆ Low speed (r ≪ 1): M X/(me) ≈ r2 ≪ 1. The machine barely moves.
ˆ Resonance (r ≈ 1): Large amplitude. Many machines must rapidly accelerate
through resonance — this is called critical speed passage.
ˆ High speed (r ≫ 1): M X/(me) → 1, i.e. X → me/M . The system centre of mass
tends to remain stationary — the machine self-centres.

12.4 Balancing
P
Static balancing: ensures the centre of mass is on the spin axis ( mi ei = 0).
Dynamic balancing: additionally ensures the inertia tensor is diagonal w.r.t. the spin
axis, eliminating moments.
12.4. Balancing 13

In practice, balancing machines measure residual vibration and compute the required
correction masses and their angular positions.
14 Chapter 12. Rotating Unbalance
Chapter 13

Harmonic Base Excitation

13.1 When the Support Moves

In many applications the forcing does not act directly on the mass — instead, the support
(ground, vehicle chassis, machine frame) vibrates. Examples:
ˆ Vehicle suspension: road roughness excites the axle, which excites the car body.
ˆ Earthquake engineering: ground motion excites building foundations.
ˆ Sensitive instrument isolation: the instrument is to be protected from floor vibrations.
ˆ Piping systems: support movements excite attached pipes.

13.2 Equation of Motion

Let the base undergo displacement y(t) = Y sin Ωt. The spring and damper connect the
mass m to the base, so their deflections are proportional to relative displacement z = x − y.
Applying Newton to the mass:

mẍ = −c(ẋ − ẏ) − k(x − y), (13.1)

which rearranges to:

mẍ + cẋ + kx = cẏ + ky = cY Ω cos Ωt + kY sin Ωt. (13.2)

p
The right-hand
√ side can be written as Y k 2 + (cΩ)2 sin(Ωt + ϕ0 ) — a single sinusoid of
amplitude Y k 2 + c2 Ω2 .

13.3 Absolute and Relative Response

13.3.1 Absolute Displacement (Motion Transmissibility)
The ratio of mass amplitude X to base amplitude Y is the motion transmissibility Td :

15
16 Chapter 13. Harmonic Base Excitation

Key Result: Motion Transmissibility

s
X 1 + (2ζr)2
Td = = . (13.3)
Y (1 − r2 )2 + (2ζr)2

13.3.2 Relative Displacement

Let z = x − y. The relative displacement EOM is:

mz̈ + cż + kz = −mÿ = mY Ω2 sin Ωt. (13.4)

This is mathematically identical to the rotating unbalance EOM (12.1) with munbal e ←
mY ! Hence:
Z r2
=p , (13.5)
Y (1 − r2 )2 + (2ζr)2
where Z is the relative amplitude. Relative displacement is the quantity sensed by
seismometers and strain gauges attached to the structure.

13.4 Force Transmissibility

The force transmissibility is the ratio of the force transmitted to the base to the applied
force on the mass (relevant for vibration isolation mounts):
s
Ftransmitted 1 + (2ζr)2
TF = = = Td . (13.6)
Fapplied (1 − r2 )2 + (2ζr)2

Remarkably, the motion transmissibility and force transmissibility are identical.

√
r= 2
4
ζ = 0.05
Transmissibility Td

ζ = 0.2
3
ζ = 0.5
ζ = 1.0
2

0
0 0.5 1 1.5 2 2.5 3
r = Ω/ωn
√
Figure
√ 13.1: Transmissibility vs. frequency ratio. All curves cross Td = 1 at r = 2. For
r > 2, Td < 1 — the system isolates vibration; higher damping worsens isolation in this
regime.
13.5. Vibration Isolation Design Principles 17

13.5 Vibration Isolation Design Principles

√ √
ˆ Isolation condition: r > 2, i.e. ωn < Ω/ 2. The mount’s natural frequency must
be well below the excitation frequency.
ˆ Soft mounts (low k, low ωn ) give better high-frequency isolation but larger static
deflection.
ˆ Damping trade-off: More damping reduces
√ the resonance peak (dangerous during
start-up) but worsens isolation at r > 2.
ˆ A common design rule: ωn ≤ 0.3 Ω for Td ≈ 0.1 (90% isolation).
18 Chapter 13. Harmonic Base Excitation
Chapter 14

Dynamics of the Internal Combustion

Engine

14.1 Why Engines Vibrate

The internal combustion (IC) engine is one of the most vibration-rich machines in engi-
neering. Forces arise from:
1. Gas pressure on the piston during combustion (impulsive per cycle).
2. Reciprocating inertia of pistons, connecting rods, and valves.
3. Rotating unbalance of the crankshaft.
4. Torsional excitation of the crankshaft from cyclic torque pulses.

14.2 Slider-Crank Mechanism Kinematics

r L
θ
Piston
Crank centre

xp

Figure 14.1: Slider-crank mechanism: crank radius r, connecting rod length L, crank angle
θ = Ωt.
The piston displacement from TDC (top dead centre) is:
 
λ
xp (θ) = r (1 − cos θ) + (1 − cos 2θ) + · · · , (14.1)
4

where
p λ = r/L is the crank ratio (typically 0.25–0.35). The approximation uses
2
1 − λ2 sin2 θ ≈ 1 − λ2 sin2 θ.
The piston acceleration (forcing function):

ẍp = −rΩ2 [cos θ + λ cos 2θ + · · · ] . (14.2)

19
20 Chapter 14. Dynamics of the Internal Combustion Engine

14.3 Engine Forcing — Harmonic Orders

The inertia force from one piston is:

Finertia = −mr ẍp = mr rΩ2 [cos Ωt + λ cos 2Ωt + · · · ] , (14.3)

where mr is the reciprocating mass. The harmonics of the excitation are integer multiples
of the shaft speed Ω:
ˆ 1st order (primary): cos Ωt — amplitude mr rΩ2 .
ˆ 2nd order (secondary): cos 2Ωt — amplitude λmr rΩ2 .
ˆ Higher orders exist but are small.

14.4 Multi-Cylinder Engine: Firing Order and Bal-

ance
For an n-cylinder engine with cranks equally spaced at 2π/n (4-stroke: firing every 4π/n),
the total forces sum:
n−1  
2
X 2πj
Ftotal = mr rΩ cos Ωt − .
j=0
n

Theorem: Balance Condition

The k-th harmonic of a n-cylinder engine is balanced (zero net force) if:
n−1
X
e2πijk/n = 0 ⇔ k is not a multiple of n.
j=0

Engine Balanced harmonics Lowest unbalanced order

4-cyl in-line 1st, 2nd, 3rd 4th (negligible)
6-cyl in-line 1st through 5th 6th (very small)
V8 (90°) 1st, 2nd 3rd

14.5 Torsional Vibration of Crankshaft

The crankshaft is a distributed elastic system subjected to periodic torque pulses. For
a simple two-inertia model (engine flywheel J1 , gearbox inertia J2 , connected by shaft
stiffness kT ):

J1 θ̈1 + kT (θ1 − θ2 ) = T (t), J2 θ̈2 − kT (θ1 − θ2 ) = 0. (14.4)

The natural frequency of the torsional mode is:

s  
tors 1 1
ωn = kT + . (14.5)
J1 J2
14.6. Engine Mounting 21

Torsional resonance (when an engine harmonic nΩ = ωntors ) causes severe crankshaft

stresses and is a critical design constraint. It is controlled by torsional dampers fitted to
the crankshaft nose.

14.6 Engine Mounting

The engine is connected to the vehicle chassis via engine mounts (rubber isolators).
Their design uses the transmissibility analysis of Chapter 13:
Ωidle
ωnmount ≤ √ .
2
For a 4-stroke 4-cylinder engine at 700 rpm idle: Ωidle = 700 × 2π/60 = 73.3 rad/s, 2nd
order firing frequency = 146.6 rad/s. The mount must satisfy ωn ≤ 104 rad/s to isolate
the firing frequency.
22 Chapter 14. Dynamics of the Internal Combustion Engine
Chapter 15

Transient Excitations: Impulse, Step,

Shock, and Arbitrary Forcing

15.1 Overview of Transient Loading

Harmonic and periodic excitations are steady-state phenomena. Many real events are
transient: a hammer blow, a door slam, an earthquake, a landing gear touchdown, a
gun recoil. For these we need the full time-history response, not just the steady-state
amplitude.
All transient methods rest on three foundations:
1. The unit impulse response h(t),
2. The unit step response g(t),
3. The convolution integral.

15.2 Impulsive Excitation

15.2.1 The Unit Impulse (Dirac Delta)
Definition: Dirac Delta Function
The unit impulse δ(t − t0 ) is a generalised function satisfying:
ˆ ∞
δ(t − t0 ) = 0 for t ̸= t0 , δ(t − t0 ) dt = 1.
−∞
´
It represents an infinitely short, infinitely intense force with unit impulse ( F dt = 1).

15.2.2 Impulse Response Function h(t)

Applying F (t) = δ(t) to the system at rest (x(0) = 0, ẋ(0− ) = 0):
The impulse delivers velocity ∆v = 1/m to the mass at t = 0+ . This is an initial velocity
problem with x(0+ ) = 0, ẋ(0+ ) = 1/m.

23
24 Chapter 15. Transient Excitations: Impulse, Step, Shock, and Arbitrary Forcing

Key Result: Impulse Response Function

For an under-damped system (0 < ζ < 1):

e−ζωn t
h(t) = sin(ωd t) · u(t), (15.1)
mωd

where u(t) is the unit step function (u(t) = 1 for t ≥ 0, 0 otherwise).

sin(ωn t)
For the undamped case: h(t) = .
mωn

15.2.3 Response to a Finite-Duration Impulse

A rectangular impulse of magnitude F0 and duration ϵ imparts momentum F0 ϵ. For
F0 ϵ −ζωn t
ϵ ≪ Tn , the response is approximately x(t) ≈ e sin ωd t.
mωd
Example: Hammer blow on a structure
A 2 kg mass (ωn = 50 rad/s, ζ = 0.05) receives a hammer blow of 10 N·s impulse. The
peak response:
I 10 10
Xmax ≈ = √ = = 0.100 m.
mωd 2 × 50 1 − 0.0025 99.94

The time of first peak: t∗ ≈ π/ωd ≈ 62.8 ms.

15.3 Unit Step Excitation

15.3.1 The Unit Step Function
F (t) = F0 u(t) is a force that is zero for t < 0 and F0 for t ≥ 0 — representing a suddenly
applied constant load (e.g. suddenly hanging a weight on a spring, or a blast over-pressure
that persists).

15.3.2 Step Response Function g(t)

The step response is the integral of the impulse response:
ˆ t
g(t) = h(τ ) dτ.
0

Key Result: Step Response

For an under-damped system, with force F0 suddenly applied:
" #
F0 e−ζωn t ζ
x(t) = 1− p cos(ωd t − ϕ) , ϕ = arctan p . (15.2)
k 1 − ζ2 1 − ζ2
15.4. Shock Excitation 25

The static deflection is F0 /k. The response oscillates about it, and the maximum
overshoot (first peak) is:
F0  √ 
−πζ/ 1−ζ 2
xmax = 1+e . (15.3)
k

For ζ = 0: xmax = 2F0 /k — the mass overshoots twice the static deflection. This is a
classic result: a suddenly applied load causes twice the displacement of the same load
applied slowly.

Physical Interpretation:
The factor of 2 overshoot for the undamped step response is one of the most important
results in structural dynamics. It means that a structure must be designed for twice
the static load if that load is applied suddenly (e.g. emergency brake application,
sudden gust).

15.4 Shock Excitation

A shock is a transient force of finite duration and arbitrary shape. The shock response
spectrum (SRS) quantifies the severity of a shock in terms of the maximum response it
induces in a SDOF system as a function of ωn .

15.4.1 Shock Response Spectrum

Definition: Shock Response Spectrum
For a given acceleration pulse ÿ(t) of duration td , the SRS is:

SRS(ωn ) = max|x(t)|ωn , (15.4)

t≥0

where x(t) is the response of a SDOF system with natural frequency ωn (typically
undamped or with standard ζ = 0.05) to the shock.

The SRS is evaluated over the entire duration including the residual response after the
shock ends (t > td ).

15.4.2 Half-Sine Shock Pulse

A half-sine pulse: F (t) = F0 sin(πt/td ) for 0 ≤ t ≤ td , zero otherwise. The primary SRS
has a peak near td /Tn ≈ 0.4 (where Tn = 2π/ωn ) and asymptotes to 2F0 /k for Tn ≫ td
(effectively a step) and to F0 /k for Tn ≪ td (quasi-static).

15.4.3 Pulse Duration vs. Natural Period

The ratio td /Tn is the key parameter:
ˆ td /Tn ≫ 1 (slow pulse): quasi-static, xmax ≈ F0 /k.
ˆ td /Tn ≈ 0.5: resonance-like amplification, maximum severity.
26 Chapter 15. Transient Excitations: Impulse, Step, Shock, and Arbitrary Forcing

ˆ td /Tn ≪ 1 (impulsive): response governed by impulse momentum, xmax ≈ I/(mωn ).

15.4.4 Applications of SRS

ˆ Military equipment specifications (MIL-STD-810): components must survive specified
SRS levels.
ˆ Earthquake engineering: ground motion is characterised by its acceleration SRS.

ˆ Packaging design: drop-test SRS determines required shock isolation.

15.5 Discontinuous Arbitrary Excitation

15.5.1 The Duhamel Convolution Integral
For any forcing F (t), the response of a system at rest is given by the convolution integral
(Duhamel’s formula):

ˆ t ˆ t
x(t) = F (τ ) h(t − τ ) dτ = h(τ ) F (t − τ ) dτ, (15.5)
0 0

where h(t) is the impulse response function (15.1). The integral treats F (t) as a superposi-
tion of infinitesimal impulses applied at time τ , each generating a response F (τ ) dτ ·h(t−τ ).

15.5.2 Handling Discontinuities

When F (t) has jump discontinuities (e.g. a rectangular pulse, a triangular pulse, a suddenly
removed load), the convolution integral is split at each discontinuity:
ˆ t1 ˆ t2
x(t) = F1 (τ )h(t − τ ) dτ + F2 (τ )h(t − τ ) dτ + · · ·
0 t1

15.5.3 Rectangular Pulse

Example: Rectangular pulse: F = F0 for 0 ≤ t ≤ t1 , then F = 0
During the pulse (0 ≤ t ≤ t1 ): Treated as a step response:
" !#
F0 ζ
x(t) = 1 − e−ζωn t cos ωd t + p sin ωd t .
k 1 − ζ2

After the pulse (t > t1 ): The system responds freely from its state (x1 , ẋ1 ) at t = t1 :
 
−ζωn (t−t1 ) ẋ1 + ζωn x1
x(t) = e x1 cos ωd (t − t1 ) + sin ωd (t − t1 ) .
ωd
15.5. Discontinuous Arbitrary Excitation 27

15.5.4 Numerical Convolution

For measured force histories F (ti ) at discrete time steps ∆t, the convolution integral is
evaluated numerically:
Xn
x(tn ) ≈ ∆t F (tj ) h(tn − tj ). (15.6)
j=0

This is computationally efficient using the Fast Fourier Transform (FFT): compute
F (ω) and H(ω) via FFT, multiply, then inverse-FFT to get x(t). Complexity reduces
from O(N 2 ) to O(N log N ).

15.5.5 Superposition of Unit Steps (Ramp and Arbitrary)

Any forcing can also be decomposed into a series of step changes of magnitude ∆Fj at
times tj . Each step generates a step response; the total is the sum:
X
x(t) = ∆Fj g(t − tj ) u(t − tj ),
j

where g(t) is the unit step response. This is the Heaviside superposition method.
28 Chapter 15. Transient Excitations: Impulse, Step, Shock, and Arbitrary Forcing
Chapter 16

Multi-Degree-of-Freedom Oscillator

16.1 Extending to N Degrees of Freedom

Real structures are continuous bodies with infinitely many DOFs. In practice, we discretise
them into N DOFs using:
ˆ Lumped parameter models: point masses connected by springs and dampers.
ˆ Finite element models: the structure is divided into elements, each contributing
stiffness and mass to global matrices.

16.2 Matrix Equation of Motion

The most general N -DOF damped system under forcing {F (t)} is:

[M]{ẍ} + [C]{ẋ} + [K]{x} = {F(t)}, (16.1)

where:
ˆ [M] (N × N , symmetric positive definite): mass matrix.
ˆ [C] (N × N , symmetric positive semi-definite): damping matrix.
ˆ [K] (N × N , symmetric positive semi-definite): stiffness matrix.
ˆ {x(t)} (N × 1): generalised displacement vector.
ˆ {F(t)} (N × 1): generalised force vector.
All three matrices are symmetric for systems with non-gyroscopic, non-circulatory forces.

16.3 Eigenvalue Problem and Natural Frequencies

16.3.1 Undamped Free Vibration
Setting [C] = 0 and {F} = 0, assume {x} = {ϕ}eiωt :

[K] − ω 2 [M] {ϕ} = {0}.

(16.2)

29
30 Chapter 16. Multi-Degree-of-Freedom Oscillator

Non-trivial solutions exist when:

det [K] − ω 2 [M] = 0.

(16.3)

This is a polynomial of degree N in ω 2 , yielding N eigenvalues ω12 ≤ ω22 ≤ · · · ≤ ωN

2
(all
real, non-negative for positive semi-definite [K]).

Note: Zero natural frequencies

If [K] is singular (e.g. an unconstrained system that can translate rigidly), one or
more ωr2 = 0. These correspond to rigid body modes — the structure translates or
rotates without deformation. A free–free beam has 6 rigid body modes (3 translations,
3 rotations).

16.4 Modal Analysis: The Complete Procedure

16.4.1 Step 1: Solve the Eigenvalue Problem
Compute all N pairs (ωr2 , {ϕ}(r) ), r = 1, . . . , N .

16.4.2 Step 2: Mass-Normalise the Mode Shapes

Scale each mode shape so that:

{ϕ}(r)T [M]{ϕ}(r) = 1 =⇒ m∗r = 1. (16.4)

Then automatically {ϕ}(r)T [K]{ϕ}(r) = ωr2 .

16.4.3 Step 3: Construct the Modal Matrix

[Φ] = {ϕ}(1) | {ϕ}(2) | · · · | {ϕ}(N ) ,
 
(N × N ).

16.4.4 Step 4: Modal Transformation

Transform physical coordinates to modal (principal) coordinates {q}:

{x} = [Φ]{q}.

Pre-multiply (16.1) by [Φ]T . By orthogonality (Chapter 18):

[Φ]T [M][Φ] = [I], [Φ]T [K][Φ] = [Λ] = diag(ω12 , . . . , ωN

2
). (16.5)

For Rayleigh damping [C] = α[M] + β[K]:

α βωr
[Φ]T [C][Φ] = diag(2ζr ωr ), ζr = + . (16.6)
2ωr 2
16.5. Forced Harmonic Response of Multi-DOF System 31

16.4.5 Step 5: Decouple into N SDOF Systems

The modal equations are:

q̈r + 2ζr ωr q̇r + ωr2 qr = fr (t), r = 1, . . . , N, (16.7)

where {f } = [Φ]T {F} are the modal forces. Each equation is a standard SDOF oscillator!

16.4.6 Step 6: Solve and Back-Transform

Solve each (16.7) using any SDOF method (exact solution, Duhamel’s integral, etc.). Then
recover physical response:
XN
{x(t)} = {ϕ}(r) qr (t). (16.8)
r=1

Physical Interpretation:
Modal superposition is the workhorse of structural dynamics. A 10,000-DOF finite
element model has 10,000 modes, but in practice only the lowest few modes (say, 20–50)
contribute significantly to the response. Modal truncation keeps only the M ≪ N
significant modes, reducing computation enormously while maintaining accuracy.

16.5 Forced Harmonic Response of Multi-DOF Sys-

tem
For {F(t)} = {F0 }eiΩt , the steady-state response is {x} = {X}eiΩt :

[K] − Ω2 [M] + iΩ[C] {X} = {F0 }.

(16.9)

In modal coordinates, using (16.7):

N
X {ϕ}(r) {ϕ}(r)T {F0 }
{X} = . (16.10)
r=1
ωr2 − Ω2 + 2iζr ωr Ω

The frequency response matrix (or dynamic flexibility matrix) is:

N
X {ϕ}(r) {ϕ}(r)T
[H(Ω)] = . (16.11)
r=1
ωr2 − Ω2 + 2iζr ωr Ω
32 Chapter 16. Multi-Degree-of-Freedom Oscillator
Chapter 17

Two-DOF Oscillator: Complete Forced

Analysis

17.1 Recap of Model

The two-DOF system from Part I has EOM:
       
m1 0 ẍ1 k1 + k2 −k2 x1 F1 (t)
+ = .
0 m2 ẍ2 −k2 k2 + k3 x2 F2 (t)

17.2 Natural Frequencies: Detailed Derivation

For the symmetric case m1 = m2 = m, k1 = k3 = k, k2 = κ:
 
2 k + κ − mω 2 −κ
[K] − ω [M] = .
−κ k + κ − mω 2

Characteristic equation:

(k + κ − mω 2 )2 − κ2 = 0 =⇒ mω 2 = k, mω 2 = k + 2κ.

Key Result: Natural Frequencies of Symmetric 2-DOF

k k + 2κ
ω12 = , ω22 = . (17.1)
m m

17.2.1 Mode Shapes

Mode 1 (ω12 = k/m): (k + κ − k)ϕ11 = κϕ21 =⇒ ϕ11 = ϕ21 .
In-phase mode: both masses move identically. The coupling spring κ is not stretched.
Mode 2 (ω22 = (k + 2κ)/m): −κϕ12 = (k + κ − k − 2κ)ϕ12 =⇒ ϕ12 = −ϕ22 .
Out-of-phase mode: masses move in opposite directions. The coupling spring is
maximally stretched.

33
34 Chapter 17. Two-DOF Oscillator: Complete Forced Analysis

Mode
+1 1 (In-phase)
+1 Mode
+12 (Out-of-phase)

m m m m
−1

Figure 17.1: Mode shapes for the symmetric 2-DOF system.

17.3 Forced Response via Modal Analysis

With F1 (t) = F0 cos Ωt applied to mass 1 only (F2 = 0):
Modal forces: fr = {ϕ}(r)T {F0 , 0}T = ϕ1r F0 .
The modal equations are:
q̈r + ωr2 qr = ϕ1r F0 cos Ωt.
Steady-state modal responses:
ϕ1r F0
qr(ss) = cos Ωt (undamped).
ωr2 − Ω2

Physical responses:
2
X ϕ21r F0
x1 (t) = cos Ωt , (17.2)
ω 2 − Ω2
r=1 r
2
X ϕ1r ϕ2r F0
x2 (t) = cos Ωt . (17.3)
r=1
ωr2 − Ω2

17.3.1 Antiresonance of Mass 1

Recall the tuned mass damper: when Ω = ω1 , the amplitude of mass 1 is zero if ϕ11 /(ω12 −
Ω2 ) + ϕ12 /(ω22 − Ω2 ) = 0. This is achieved when Ω2 = κ/m (the natural frequency of
the absorber mass alone). This is the vibration absorber principle — mathematically an
antiresonance in the FRF of mass 1.

17.4 Frequency Response Functions of 2-DOF System

The FRF matrix from (16.11):
2
X ϕjr ϕkr
Hjk (Ω) = .
ω2
r=1 r
− Ω2 + 2iζr ωr Ω

Key features of H11 (Ω):

ˆ Two resonance peaks at ω1 and ω2 .
ˆ One antiresonance between the peaks (for lightly damped systems).
ˆ Phase shifts by 180 at each resonance.
Chapter 18

Orthogonality of Mode Shapes and

Orthogonal Vectors

18.1 Motivation: Why Orthogonality Matters

Modal analysis works because the transformation [Φ] diagonalises [M] and [K] simulta-
neously. This is only possible because the eigenvectors (mode shapes) are orthogonal
with respect to [M] and [K]. Without orthogonality, the N coupled ODEs would remain
coupled after the transformation.

18.2 Euclidean Orthogonality: Review

Definition: Euclidean Orthogonality
Two vectors u, v ∈ RN are orthogonal if their dot product is zero:
N
X
T
u·v =u v = ui vi = 0.
i=1

Geometrically, orthogonal vectors are perpendicular. Orthogonal bases allow any vector
to be decomposed into independent components—coordinates with respect to one basis
vector do not depend on other basis vectors.

18.3 M-Orthogonality and K-Orthogonality

Mode shapes are not Euclidean-orthogonal in general. They are orthogonal in a weighted
inner product:

Definition: M-Orthogonality
Two vectors u and v are [M]-orthogonal if:

⟨u, v⟩M = uT [M]v = 0.

35
36 Chapter 18. Orthogonality of Mode Shapes and Orthogonal Vectors

Theorem: Orthogonality of Mode Shapes

Let {ϕ}(r) and {ϕ}(s) be eigenvectors of [K]{ϕ} = ω 2 [M]{ϕ} corresponding to distinct
eigenvalues ωr2 ̸= ωs2 . Then:

{ϕ}(r)T [M]{ϕ}(s) = 0, (18.1)

{ϕ}(r)T [K]{ϕ}(s) = 0. (18.2)

Derivation: Proof of M-orthogonality

From the eigenvalue problem:

[K]{ϕ}(r) = ωr2 [M]{ϕ}(r) ...(i)

[K]{ϕ}(s) = ωs2 [M]{ϕ}(s) ...(ii)

Pre-multiply (i) by {ϕ}(s)T :

{ϕ}(s)T [K]{ϕ}(r) = ωr2 {ϕ}(s)T [M]{ϕ}(r) .

Pre-multiply (ii) by {ϕ}(r)T :

{ϕ}(r)T [K]{ϕ}(s) = ωs2 {ϕ}(r)T [M]{ϕ}(s) .

Since [M] and [K] are symmetric, both left-hand sides are equal: {ϕ}(s)T [K]{ϕ}(r) =
{ϕ}(r)T [K]{ϕ}(s) .
Subtracting the two equations:

0 = (ωr2 − ωs2 ) {ϕ}(r)T [M]{ϕ}(s) .

Since ωr2 ̸= ωs2 , we must have {ϕ}(r)T [M]{ϕ}(s) = 0. □

[K]-orthogonality follows immediately from substituting back.

18.4 The Generalized Orthogonality Relations

Collecting all modes into [Φ], the orthogonality relations in matrix form are:

Key Result: Modal Orthogonality in Matrix Form

[Φ]T [M][Φ] = [M∗ ] = diag(m∗1 , m∗2 , . . . , m∗N ), (18.3)

∗
[Φ]T [K][Φ] = [K∗ ] = diag(k1∗ , k2∗ , . . . , kN ), (18.4)
where m∗r is the modal mass and kr∗ = ωr2 m∗r is the modal stiffness.
18.5. Expansion Theorem (Modal Expansion) 37

18.5 Expansion Theorem (Modal Expansion)

Theorem: Modal Expansion Theorem
Any vector {f } ∈ RN can be uniquely expanded in the basis of mass-normalised mode
shapes:
N
X
{f } = αr {ϕ}(r) , αr = {ϕ}(r)T [M]{f }. (18.5)
r=1

Physical Interpretation:
This is the vector-space analogue of a Fourier series. Just as any periodic function can
be written as a sum of sines and cosines, any force vector can be written as a sum of
mode shapes — each amplified by its modal force. The mass matrix plays the role of
the integration measure (weight function), and the mode shapes play the role of the
orthogonal basis functions.

18.6 Repeated Eigenvalues

When ωr2 = ωs2 (repeated natural frequency, occurring in symmetric structures), the
eigenvectors are not unique. However, we can always choose an [M]-orthogonal basis
within the degenerate eigenspace using the Gram–Schmidt procedure.

18.7 Rayleigh Quotient

The Rayleigh quotient provides an estimate of the natural frequency for any trial vector
{v}:

{v}T [K]{v}
R({v}) = ≥ ω12 . (18.6)
{v}T [M]{v}

The minimum of R over all trial vectors gives the lowest natural frequency ω12 (Rayleigh–Ritz
method). This is stationary when {v} = {ϕ}(1) . For any trial vector, R provides an upper
bound on ω12 .
38 Chapter 18. Orthogonality of Mode Shapes and Orthogonal Vectors
Chapter 19

Continuum and Long-Wave Approxi-

mation: From Discrete to Continuum

19.1 The Limiting Process: N → ∞

So far, structures have been modelled as N lumped masses and springs. Real structures
are continuous: every point has mass and stiffness. By taking N → ∞ while shrinking
element size ∆x → 0, the discrete EOM transitions to a partial differential equation
(PDE).

19.2 Longitudinal Vibration of a Rod

Consider a uniform rod of length L, cross-sectional area A, elastic modulus E, and mass
density ρ. Discretise it into N elements, each of length ∆x = L/N , mass ∆m = ρA∆x,
and stiffness ∆k = EA/∆x.

The EOM for the n-th element displacement un :

EA EA
ρA∆x ün = (un+1 − un ) − (un − un−1 ).
∆x ∆x

Dividing by ∆x and taking ∆x → 0 (un → u(x, t)):

∂ 2u ∂ 2u
ρA = EA , (19.1)
∂t2 ∂x2

or
s
2 2
∂ u ∂ u E
2
= c2L 2 , cL = . (19.2)
∂t ∂x ρ

This is the one-dimensional wave equation. cL is the longitudinal wave speed.

39
40
Chapter 19. Continuum and Long-Wave Approximation: From Discrete to Continuum

19.3 Transverse Vibration of a Taut String

A string of tension T and linear mass density µ vibrating transversely v(x, t):
s
∂ 2v ∂ 2v T
µ 2 =T , c= . (19.3)
∂t ∂x2 µ

Mathematically
p identical to the rod equation — a wave equation with wave speed c =
T /µ.

19.4 Transverse Vibration of an Euler–Bernoulli Beam

For a beam of bending stiffness EI, mass per unit length m̄ = ρA, and transverse
displacement w(x, t):
∂ 2w ∂ 4w
m̄ + EI = p(x, t), (19.4)
∂t2 ∂x4

where p(x, t) is the distributed load. Note the fourth spatial derivative — unlike the wave
equation (second order). This leads to dispersive wave propagation: different frequencies
travel at different speeds.

19.5 The Long-Wave (Low-Frequency) Approxima-

tion
19.5.1 Dispersion Relation
Assume harmonic wave solution u(x, t) = U ei(kx−ωt) (wavenumber k, frequency ω). Substi-
tuting into the equations:

System Dispersion relation Wave speed c = ω/k

Rod (wave eq.) ω 2 = c2L k 2 c=pcL (constant; non-dispersive)
Beam (E-B) ω 2 = (EI/m̄)k 4 c = EI/m̄ k (dispersive)

The long-wave approximation retains only the leading term of the dispersion relation.
For a periodic discrete lattice of spacing d, the exact dispersion is:
 
2 4ks 2 kd
ω = sin .
m 2

For kd ≪ 1 (wavelength ≫ d): sin(kd/2) ≈ kd/2, giving ω 2 ≈ (ks d2 /m)k 2 — identical to

the wave equation with c2L = ks d2 /m. The continuum model is valid when the wavelength
greatly exceeds the lattice spacing.
19.6. Separation of Variables: Normal Modes of the Rod 41

19.5.2 Physical Meaning

Physical Interpretation:
The continuum limit is the long-wave approximation: it is accurate when the wavelength
of vibration is much larger than the smallest structural scale (grain size, element size,
lattice spacing). This is typically satisfied for low-frequency engineering vibration.
High-frequency phonons, impact waves, or lattice dynamics require the full discrete
model.

19.6 Separation of Variables: Normal Modes of the

Rod
Assume u(x, t) = U (x) T (t). Substituting into (19.2):

T̈ U ′′
= c2L = −ω 2 = const.
T U

This gives two ODEs:

T̈ + ω 2 T = 0 =⇒ T (t) = A cos ωt + B sin ωt, (19.5)

ω2 ωx ωx
U ′′ + 2 U = 0 =⇒ U (x) = C cos + D sin . (19.6)
cL cL cL

19.6.1 Boundary Conditions and Eigenfrequencies

Fixed–free rod (U (0) = 0, U ′ (L) = 0): D = 0 from U (0) = 0... wait, U (0) = 0 ⇒ C = 0,
so U = D sin(ωx/cL ); then U ′ (L) = 0 ⇒ cos(ωL/cL ) = 0:

ωn L (2n − 1)π (2n − 1)πcL

= , n = 1, 2, 3, . . . =⇒ ωn = . (19.7)
cL 2 2L

Fixed–fixed rod (U (0) = U (L) = 0): C = 0, sin(ωL/cL ) = 0:

nπcL nπx
ωn = , Un (x) = sin , n = 1, 2, 3, . . . (19.8)
L L

The mode shapes Un (x) = sin(nπx/L) are trigonometric functions — the continuum
analogue of the discrete eigenvectors.

19.7 Orthogonality of Continuum Mode Shapes

42
Chapter 19. Continuum and Long-Wave Approximation: From Discrete to Continuum

Theorem: Continuum Orthogonality

For the rod with appropriate boundary conditions, the mode shapes satisfy:
ˆ L (
0 n ̸= m
ρA Un (x) Um (x) dx = , (19.9)
0 ρAL/2 n = m

i.e. the integral of the product of mode shapes (weighted by mass density) is zero for
distinct modes.

This is the continuum counterpart of the discrete [M]-orthogonality. The mass matrix
[M] is replaced by the mass density ρA and the sum is replaced by an integral.
Chapter 20

Characteristics of Classical Normal

Modes

20.1 What is a Normal Mode?

Definition: Normal Mode

A normal mode (or natural mode or principal mode) of a vibrating system is a
motion in which:
1. Every point in the system moves harmonically at the same frequency ωr .
2. Every point passes through its equilibrium position simultaneously (all in phase,
or all out of phase — no phase variation across the structure within a single
mode).
3. The shape of the deformation is constant in time (only the amplitude scales with
cos ωr t).

20.2 Standing Wave Interpretation

Normal modes are standing waves. They arise from the superposition of two travelling
waves of equal amplitude moving in opposite directions. The nodes (points of zero
displacement) are fixed in space for all time.

For the fixed-fixed rod, mode n has nodes at x = jL/n for j = 1, . . . , n − 1.

43
44 Chapter 20. Characteristics of Classical Normal Modes

1 Mode 1: sin(πx/L)
Mode 2: sin(2πx/L)

Mode shape Un (x)

0.5 Mode 3: sin(3πx/L)
Mode 4: sin(4πx/L)

−0.5

−1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
x/L

Figure 20.1: First four normal modes of a fixed-fixed rod. Mode n has n − 1 internal
nodes.

20.3 Properties of Normal Modes

20.3.1 1. Completeness

Theorem: Completeness of Normal Modes

The set of all normal mode shapes {Un (x)}∞
n=1 forms a complete basis for the space
of admissible displacement functions. Any physically admissible deformation can be
represented as:
∞
X
u(x, t) = Un (x) qn (t), (20.1)
n=1

where qn (t) are the time-dependent modal coordinates.

20.3.2 2. Orthogonality
Already proven (Chapter 18 for discrete, Chapter 19 for continuum). Orthogonality
guarantees the decoupling of modal equations.

20.3.3 3. In-Phase Motion within Each Mode

In a classical (undamped, conservative) normal mode, all points of the structure move
either in-phase (+1 relative to the reference) or out-of-phase (−1 relative to the reference).
There are no intermediate phases. The entire deformed shape can be described by a single
sign pattern.

This is the mathematical content of: the mode shape eigenvector has all real entries.
(Complex mode shapes arise when damping is non-proportional.)
20.4. Damped Normal Modes 45

20.3.4 4. Nodes and Antinodes

Definition: Node and Antinode
A node is a point of zero displacement in a given mode. An antinode is a point of
maximum displacement amplitude. Mode n of a fixed-fixed structure has n − 1 nodes
between the boundaries.

20.3.5 5. Energy Distribution

In mode r with amplitude qr :
ˆ Kinetic energy: Tr = 21 m∗r q̇r2
ˆ Potential energy: Vr = 12 kr∗ qr2 = 21 m∗r ωr2 qr2
ˆ Total energy: Er = 12 m∗r ωr2 Q2r (constant, where Qr is amplitude)
Energy in mode r stays in mode r — there is no inter-modal energy transfer in a linear
undamped system.

20.3.6 6. Frequency Ordering and Modal Density

Natural frequencies are ordered: ω1 ≤ ω2 ≤ · · · ≤ ωN . The lowest mode has the simplest
shape (fewest nodes, longest wavelength, lowest energy stored per unit amplitude). Higher
modes have increasingly complex shapes and more nodes.
For a rod of length L: ωn = nπcL /L, so frequencies are equally spaced. For a beam:
ωn ∝ n2 (frequencies grow as the square of mode number, because of the k 4 dispersion).

20.3.7 7. Rayleigh’s Principle (Stationarity)

Theorem: Rayleigh’s Principle
The natural frequencies are stationary values of the Rayleigh quotient. In particular:
ˆ ω12 = minv R(v) — the fundamental frequency is the global minimum.
ˆ ωr2 = minv⊥M {ϕ}(1) ,...,{ϕ}(r−1) R(v) — each higher frequency is a constrained mini-
mum.

20.3.8 8. Modal Superposition and Independence

Because of completeness and orthogonality, any free or forced vibration is a superposition
of normal modes. Each mode contributes independently. Exciting one mode does not
excite others (in the linear, undamped case). This independence is why modal testing
works: one can identify individual natural frequencies, mode shapes, and damping ratios
separately by exciting and measuring near each resonance.

20.4 Damped Normal Modes

For proportionally damped systems, the mode shapes are unchanged from the undamped
case, but each mode has its own damping ratio ζr . The modes are still real (in-phase
46 Chapter 20. Characteristics of Classical Normal Modes

within each mode).

For non-proportionally damped systems, the eigenvalue problem becomes:

λ2 [M]{ψ} + λ[C]{ψ} + [K]{ψ} = {0}.

The eigenvalues λr and eigenvectors {ψ}(r) are complex. Complex modes indicate that
different parts of the structure are not in phase with each other — physically corresponding
to travelling wave components in the vibration.

20.5 Experimental Modal Analysis (EMA)

Normal modes are not just theoretical constructs — they are directly measurable:
1. Apply a measured force F (t) at point j (using an instrumented hammer or shaker).
2. Measure the response xk (t) at point k (using accelerometers).
3. Compute the FRF Hkj (Ω) = Xk /Fj .
4. Fit the FRF to the modal expansion (16.11) to extract ωr , ζr , and ϕkr ϕjr (modal
residues).
This process — experimental modal analysis — is used to validate finite element
models, detect structural damage, and characterise the dynamic behaviour of any structure
from a bridge to a spacecraft.

20.6 Summary: Key Properties of Classical Normal

Modes
Property Statement
Frequency Distinct, real, non-negative
Mode shape Real-valued; constant shape, harmonic in time
Phase All points in-phase or antiphase
Nodes Mode n has n − 1 internal nodes
Orthogonality [M]- and [K]-orthogonal
Completeness Any admissible motion = sum of modes
Energy Stored in each mode independently
Decoupling Modal equations are uncoupled SDOF oscillators
Chapter A

Quick Reference: Part II Key Formu-

lae

A.1 Harmonic Forcing (SDOF)

F0 /k 2ζr Ω
|X| = p , ψ = arctan , r= .
(1 − r2 )2 + (2ζr)2 1 − r2 ωn
p 1
Peak DMF at r∗ = 1 − 2ζ 2 , DMFmax = p .
2ζ 1 − ζ 2

A.2 Energetics
F02 (2ζr)2 1
⟨P ⟩ = · , Ediss = πcΩ|X|2 , Q= .
2c (1 − r2 )2 + (2ζr)2 2ζ

A.3 Rotating Unbalance

MX r2
=p .
me (1 − r2 )2 + (2ζr)2

A.4 Base Excitation

s
X 1 + (2ζr)2
Td = = TF = .
Y (1 − r2 )2 + (2ζr)2

A.5 Impulse and Step Response

e−ζωn t F0  √ 
2
h(t) = sin ωd t, xstep
max = 1 + e−πζ/ 1−ζ .
mωd k

47
48 Appendix A. Quick Reference: Part II Key Formulae

A.6 Duhamel’s Integral

ˆ t
x(t) = F (τ ) h(t − τ ) dτ.
0

A.7 Multi-DOF: Modal Analysis Summary

[Φ]T [M][Φ] = [I], [Φ]T [K][Φ] = diag(ωr2 ).
X
q̈r + 2ζr ωr q̇r + ωr2 qr = {ϕ}(r)T {F}, {x} = {ϕ}(r) qr .
r

A.8 Continuum: Rod and Beam

∂ 2u 2
2 ∂ u
p
= c L , cL = E/ρ.
∂t2 ∂x2
∂ 2w ∂ 4w
m̄ 2 + EI 4 = 0.
∂t ∂x
Rod fixed-fixed: ωn = nπcL /L, mode shapes sin(nπx/L).
Chapter B

Worked Problems: Part II

B.1 Problem 1: Full Transient + Steady-State Re-

sponse
A SDOF system (m = 2 kg, k = 800 N/m, ζ = 0.1) is excited by F = 100 cos(15t) N from
rest. Find the complete response.
Solution:
p √
ωn = 800/2 = 20 rad/s, ωd = 20 1 − 0.01 = 19.9 rad/s, r = 15/20 = 0.75, Xst =
100/800 = 0.125 m.
Steady-state amplitude:
0.125 0.125 0.125
|X| = p =√ = = 0.2703 m.
(1 − 0.5625)2 + (2 × 0.1 × 0.75)2 0.1914 + 0.0225 0.4625

Phase: ψ = arctan(0.15/0.4375) = 18.96.

Steady-state: xp = 0.2703 cos(15t − 0.331).
Apply ICs: x(0) = 0 ⇒ A = −0.2703 cos(0.331) = −0.2558, ẋ(0) = 0 ⇒ B = [Ω|X| sin ψ +
ζωn A]/ωd = 0.1045.
Complete response: x(t) = e−2t (−0.2558 cos 19.9t+0.1045 sin 19.9t)+0.2703 cos(15t − 0.331).

B.2 Problem 2: 3-DOF Modal Analysis

Three equal masses m on springs k (chain: fixed wall – k – m1 – k – m2 – k – m3 – k –
fixed wall):  
2 −1 0
[M] = m[I], [K] = k −1 2 −1 .
0 −1 2
Natural frequencies ωr2 = λr k/m where λr are eigenvalues of [K]/k:
√ √
λ1 = 2 − 2, λ2 = 2, λ3 = 2 + 2.

49
50 Appendix B. Worked Problems: Part II
q √ p p p
ω1 = (2 − 2)k/m = 0.765 k/m, ω2 = 2k/m, ω3 = 1.848 k/m.
√ √
Mode shapes (normalised): {ϕ}(1) ∝ {1, 2, 1}T , {ϕ}(2) ∝ {1, 0, −1}T , {ϕ}(3) ∝ {1, − 2, 1}T .
These exhibit the characteristic even/odd symmetry patterns of a symmetric chain.

B.3 Problem 3: Shock Response — Half-Sine Pulse

A 5 kg component (ωn = 100 rad/s) is subjected to a half-sine acceleration pulse: a(t) =
50 sin(200πt) m/s2 for 0 ≤ t ≤ 5 ms, then zero. Find the peak response during and after
the pulse (undamped).
During the pulse (F0 = 5 × 50 = 250 N, Ω = 200π rad/s):
250 1 250 1
r = 200π/100 = 2π ≈ 6.28, |X|SS = = = 0.131 mm.
k |1 − r2 | 5 × 104 |1 − 39.5|

Residual (after td ): The phase at td determines the residual amplitude. For the half-sine, the
residual SRS peaks at td /Tn ≈ 0.5. Here td = 5 ms, Tn = 2π/100 = 62.8´ td ms, so td /Tn = 0.08
(impulsive regime). Peak response ≈ impulse approximation: I = 0 250 sin(200πτ )dτ =
2
250 × 200π = 0.796 N·s; xmax = I/(mωn ) = 0.796/(5 × 100) = 1.59 mm.
Further Reading — Part II

[1] Rao, S. S., Mechanical Vibrations, 6th ed., Pearson, 2018. — Chapters 3–6 (forced
response), 6–7 (multi-DOF).
[2] Inman, D. J., Engineering Vibration, 4th ed., Pearson, 2014. — Chapters 2 (harmonic),
3 (general forcing), 4 (multi-DOF).
[3] Ewins, D. J., Modal Testing: Theory, Practice and Application, 2nd ed., Research
Studies Press, 2000. — The definitive reference on experimental modal analysis.
[4] Craig, R. R. & Kurdila, A. J., Fundamentals of Structural Dynamics, 2nd ed., Wiley,
2006. — Excellent on multi-DOF and continuum.
[5] Blevins, R. D., Formulas for Natural Frequency and Mode Shape, Van Nostrand
Reinhold, 1979. — Comprehensive tables for beams, plates, shells.
[6] Norton, R. L., Machine Design, 5th ed., Pearson, 2014. — IC engine dynamics,
rotating balance.
[7] Harris, C. M. & Piersol, A. G., Harris’ Shock and Vibration Handbook, 5th ed.,
McGraw-Hill, 2002. — Shock spectra, transient response.

51