E. S. T. P.

Nuclear Civil Engineering

Lecture Notes on Transfer of Response

Spectra and Soil-Structure Interaction

Géodynamique et Structure

January 2018
ii
Contents

Preface ix

1 Transfer of Response Spectra 1

1.1 Motivations and problem statement . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Fundamentals of response spectra . . . . . . . . . . . . . . . . . . . . . . 2
1.2.1 Oscillator under base excitation . . . . . . . . . . . . . . . . . . . 2
1.2.2 The notion of response spectrum . . . . . . . . . . . . . . . . . . . 6
1.2.3 Design response spectra - Properties of response spectra . . . . . . 8
1.2.4 Going back to the initial problem . . . . . . . . . . . . . . . . . . 11
1.3 Solution methodologies for transferred response spectra . . . . . . . . . . . 12
1.3.1 Transferred response spectra using time history analyses . . . . . . 13
1.3.2 Transferred response spectra using results of modal analyses . . . . 14
1.4 Empirical formulas for transferred response spectra . . . . . . . . . . . . . 16

2 Notions of Soil-Structure Interaction 19

2.1 What is Soil-Structure Interaction? . . . . . . . . . . . . . . . . . . . . . . 19
2.2 Illustration of SSI on a simple example . . . . . . . . . . . . . . . . . . . . 23
2.3 Formulation of Soil-Structure Interaction Problem . . . . . . . . . . . . . . 26
2.4 Superposition Theorem for Soil-Structure Interaction . . . . . . . . . . . . 30
2.5 Foundation Impedance . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

3 Exercises 39
3.1 Transfer of Response Spectra . . . . . . . . . . . . . . . . . . . . . . . . . 39
3.2 Effects of soil structure interaction . . . . . . . . . . . . . . . . . . . . . . 44
3.3 Cone model for foundation impedance . . . . . . . . . . . . . . . . . . . . 51

Bibliography 59

iii
iv Contents
List of Figures

1.1 Motivation for calculation of transferred response spectra. . . . . . . . . . . 2

1.2 Problem statement for calculation of transferred spectra. . . . . . . . . . . 2
1.3 Single degree of freedom oscillator under base excitation. . . . . . . . . . . 3
1.4 Accelerogram recorded at Lake Hughes Station (Northridge, 1994). . . . . 4
1.5 Influence of the natural period on the response of the oscillator. . . . . . . . 5
1.6 Influence of the damping ratio on the response of the oscillator. . . . . . . . 6
1.7 Maximum response of oscillator. . . . . . . . . . . . . . . . . . . . . . . . 7
1.8 Relative displacement response spectrum. . . . . . . . . . . . . . . . . . . 7
1.9 Pseudo-acceleration response spectrum . . . . . . . . . . . . . . . . . . . 8
1.10 Design pseudo-acceleration response spectrum. . . . . . . . . . . . . . . . 9
1.11 Design displacement response spectrum. . . . . . . . . . . . . . . . . . . . 10
1.12 Amplification factor for z/H = 0.25. . . . . . . . . . . . . . . . . . . . . 17
1.13 Amplification factor for z/H = 0.50. . . . . . . . . . . . . . . . . . . . . 18
1.14 Amplification factor for z/H = 1.0. . . . . . . . . . . . . . . . . . . . . . 18

2.1 Illustration of soil-structure interaction on the structure response. . . . . . . 21

2.2 Acceleration response spectra at the Humbolt Bay nuclear power station. . . 23
2.3 Idealized model for soil-structure interaction. . . . . . . . . . . . . . . . . 24
2.4 Illustration of the influence of soil-structure interaction. . . . . . . . . . . . 27
2.5 Decomposition of a SSI problem. . . . . . . . . . . . . . . . . . . . . . . . 28
2.6 Examples of inertial and kinematic interaction. . . . . . . . . . . . . . . . 29
2.7 Superposition theorem (K AUSEL ET AL . (1978)). . . . . . . . . . . . . . . 31
2.8 Impedance of a shallow circular footing. . . . . . . . . . . . . . . . . . . . 32
2.9 Impedance of a shallow circular footing. . . . . . . . . . . . . . . . . . . . 35
2.10 Dynamic response of a shallow circular footing subjected to vertical force. . 37
2.11 Foundation impedance of circular footing on elastic layer of limited depth. . 37
2.12 Rheological models with added mass for foundation impedance. . . . . . . 38
2.13 Application of rheological models for foundation impedance. . . . . . . . . 38

3.1 Frame structure with crane. . . . . . . . . . . . . . . . . . . . . . . . . . . 40

3.2 Displacement, pseudo-velocity and pseudo-acceleration response spectra
of input motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.3 Transferred displacement and pseudo-acceleration response spectra with
respect to β. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
3.4 (a) Nuclear reactors at San Onofre (CA) power plant and (b) Simplified
structural model for dynamic analysis. . . . . . . . . . . . . . . . . . . . . 44
3.5 Variation of kF with respect to frequency. . . . . . . . . . . . . . . . . . . 46

v
vi List of Figures

3.6 Design pseudo-acceleration response spectrum. . . . . . . . . . . . . . . . 46

3.7 Simplified soil modeling with (a) Cone model and (b) 1D model. . . . . . . 51
3.8 Dynamic equilibrium of the cone model . . . . . . . . . . . . . . . . . . . 52
3.9 Isolation of rotating machine from mat foundation . . . . . . . . . . . . . . 57
List of Tables

2.1 Impedance coefficients for shallow circular footing . . . . . . . . . . . . . 36

vii
viii List of Tables
Preface

The current notes are intended to serve as supporting text for the lectures on Transfer of
response spectra and on Soil-Structure Interaction presented in the framework of the Nu-
clear Civil Engineering program of the Ecole Spéciale des Travaux Publiques. The task
of preparing these lectures has been assigned by ESTP to Géodynamique et Structure, an
engineering consultancy with over thirty years of experience in the areas of soils, structures
and equipment in relation with cyclic (swell, vibration) and dynamic loading (earthquake,
explosion).
The lectures are organized as follows: for the topic of Transfer of response spectra, a
3-hour lecture is given divided in two parts. The first part (1h30 ) covers the relevant theory
with an emphasis on the notion of response spectrum and its use in structural dynamics.
The second part (1h30 ) is an academic application to be solved by the students with the
guidance of the instructor.
For the topic of soil-structure interaction, two 3-hour lectures have been prepared with
the same structure as for the first topic. The first lecture is focused on a general introduction
to Soil-Structure Interaction and on principal methods to treat this problem in practice. The
second lecture is concerned with the notion of foundation impedance, a fundamental idea
in soil dynamics and the field of foundation vibrations.
The students are responsible for having prepared the corresponding material before
each lecture and for solving the applications with the guidance of the instructor during the
lecture. Presence will be controlled both at the beginning of the theory part and also at the
beginning of the application part of the lecture.
These lecture notes have been partly based on the lecture notes of Professor Alain
Pecker for the course on Structural Dynamics taught at the Ecole Nationale des Ponts et
Chaussées. Dr. Charisis Chatzigogos has been responsible for completing the relevant
theory and for preparing the applications at the end of each lecture.

Montrouge, January 2018

ix
x Preface
Chapter 1

Transfer of Response Spectra

1.1 Motivations and problem statement

One of the common requirements in seismic safety studies of nuclear reactor buildings and
of industrial facilities in general is the calculation of the so-called transferred response
spectra or simply floor spectra to a certain number of points of the examined structure.
This requirement arises from the fact that an industrial building is usually equipped with
elements whose stability needs to be assessed as a safety requirement. Typical examples
are heavy cranes, machines, heavy doors etc. Such elements cannot be explicitly included
in the structural model either for reasons of simplicity or because their exact position and
characteristics are not known at the time of the design.
It is noted that when performing dynamic analyses of industrial buildings, it is a com-
mon practice to model heavy equipment whose position and mass is known, as added
masses in the structural model. Even so, such a simplistic modeling may be sufficient
for describing the effect of these elements on the global dynamic response of the structure,
but it is not sufficient for performing verifications of their stability (for example, verify the
connection of the equipment to the supporting slab). In particular, what the engineer is
interested in, is to have a tool that will allow him to assess the seismic forces exerted on
these elements without being obliged to rerun dynamic analyses on a structural model of
the entire building in which the exact detailing of the equipment has been included.
In order to get a better understanding of the problem let us refer to Figure 1.1. The
figure presents a structure with known characteristics of mass and rigidity expressed by the
mass and stiffness matrices M and K. We know that this structure is equipped with a heavy
equipment of mass m which is attached at point A and we want to verify the connection
of this equipment to its supporting slab when the structure is subjected to a base seismic
excitation. The verification of the connection of the equipment can be done if we know the
motion that is developed at point A due to the seismic excitation at the base. One physical
way of characterizing this motion is by defining the displacement time history uA (t) or
acceleration time history üA (t) at point A. Nevertheless, for engineering applications, it is
more convenient to characterize this motion in terms of the notion of response spectrum.
We will present a detailed definition of this notion in the following paragraphs. The prob-
lem is thus formulated as follows (cf. Figure 1.2): Given the response spectrum of the input
motion, the structural properties and the equipment properties, calculate the transferred
response spectrum of the motion at point A.

1
2 Transfer of Response Spectra

M, K
INPUT
Base excitation
Structural properties
A Oscillator properties
uA
OUTPUT
m Motion at point A

Justification of
equipment connection

ug ug
ug ug

Figure 1.1: Motivation for calculation of transferred response spectra.

Spectral acceleration [g]

Point A

SOIL

Period [s]

Figure 1.2: Problem statement for calculation of transferred spectra.

Let us then open a parenthesis and introduce the notion of response spectrum, one of
the most fundamental ideas in structural dynamics.

1.2 Fundamentals of response spectra

1.2.1 Oscillator under base excitation
The notion of response spectrum is introduced in structural dynamics by studying a single-
degree-of-freedom oscillator subjected to a base excitation as presented in Figure 1.3. The
oscillator consists of a mass m resting on a base that is excited by an acceleration time
history üg . The mass is attached to the base by means of a spring with stiffness k and a
dashpot with damping coefficient c. Accordingly, if we denote by u and u̇, the relative
displacement and velocity between the base and the mass, the force exerted on the mass by
Fundamentals of response spectra 3

the spring and the dashpot at any instant t is:

f (t) = ku(t) + cu̇(t) (1.1)

k
m
c

v
ug u
ug

ug

Figure 1.3: Single degree of freedom oscillator under base excitation.

The equation of dynamic equilibrium of the mass is obtained by identifying all the
forces that are applied on the mass and taking their vectorial sum to be equal to zero.
Besides the reaction force given by equation (1.1), the only other force exerted on the mass
is the inertial force fI given by the product of the mass of the oscillator times the absolute
acceleration that it develops. Denoting by v the absolute displacement of the mass we see
that this is equal to:
v(t) = u(t) + ug (t) (1.2)
Consequently, the inertial force will be:

fI (t) = mv̈(t) = m(ü(t) + üg (t)) (1.3)

and the equation of dynamic equilibrium will be written as follows:

fI (t) + f (t) = 0 =⇒
(1.4)
mü(t) + cu̇(t) + ku(t) = −müg (t)

This equation reveals that the response of the oscillator under a base excitation is analogous
to the response of the oscillator under an equivalent force peff directly applied on the mass
that is equal to:
peff (t) = −müg (t) (1.5)
Equation (1.4)
q can be written in an equivalent form by introducing the natural circular
k √c .
frequency ω = m
and the critical damping ratio ξ = 2 mk
Equation (1.4) thus becomes:

ü(t) + 2ξω u̇(t) + ω 2 u(t) = −üg (t) (1.6)

4 Transfer of Response Spectra

The solution of the ordinary differential equation (1.6) considering zero initial condi-
tions for the displacement and the velocity is given by the following expression, known as
Duhamel integral:
Z t
1
u(t) = − üg (τ )e−ξω(t−τ ) sin [ωD (t − τ )] dt (1.7)
ωD 0

In the above expression, the quantity ωD is given by the equation:

p
ωD = ω 1 − ξ 2 (1.8)

Let us note that in the case of non-zero initial conditions, the solution of differential equa-
tion (1.6) is the sum of a general solution and a particular solution given by the Duhamel
integral. Finally, let us differentiate equation (1.7) to obtain the relative velocity of the
mass. Standard calculus yields:
Z t
u̇(t) = −ξωu(t) − üg (τ )e−ξω(t−τ ) cos [ωD (t − τ )] dt (1.9)
0

Use of the Duhamel integral may be illustrated for the real acceleration time history
recorded at the Lake Hughes Station (horizontal component E-W) during the Northridge
earthquake of 1994. This record is presented in Figure 1.4. Using this record and the
Duhamel integral, we calculate the response of an oscillator characterized by a critical
damping ratio ξ = 2% and a variable circular frequency ω. Equivalently, we may refer to
the natural period of the oscillator T defined by the relationship: T = 2π
ω
. The results for
three different natural periods, namely T = 2s, T = 1s and T = 0.5s are shown in Figure
1.5. The results reveal that the duration, the amplitude and even the form of the response
depend strongly on the natural period of the oscillator. It can generally be noted that the
mass oscillates with a variable amplitude and with a period that is approximately equal to
the natural period of the oscillator.
Acceleration (m/s2)

Time (s)

Figure 1.4: Accelerogram recorded at Lake Hughes Station (Northridge, 1994).

A second set of results presented in Figure 1.6 yields the response of the oscillator with
natural period T = 0.5s for three different critical damping ratios: ξ = 0%, ξ = 2% and
ξ = 5%. In the case of the undamped system, a steady state response is obtained after a
few seconds of loading in which the amplitude is constant and the largest among all cases.
Fundamentals of response spectra 5

T = 2:0s

Displacement
» = 2%

Time (s)

T = 1:0s T = 0:5s
» = 2% » = 2%
Displacement

Time (s) Time (s)

Figure 1.5: Influence of the natural period on the response of the oscillator.

As the damping ratio increases, the attenuation of the response is more pronounced and the
amplitude of the response smaller.
Once the displacement history u(t) has been determined using the Duhamel integral,
one can easily compute the elastic force fs that is developed on the spring:

fs (t) = ku(t) (1.10)

Observing that k = mω 2 , one gets:

fs (t) = mω 2 u(t) = mA(t) (1.11)

in which, we have introduced the quantity A(t) = ω 2 u(t). This quantity has the dimen-
sions of acceleration and expresses the expected fact that the force is the product of a mass
times an acceleration. This acceleration is called pseudo-acceleration and it must be dis-
tinguished from the absolute acceleration of the mass v̈, which is given by the relationship
(obtained directly from the equation of equilibrium (1.6)):

v̈(t) = ü(t) + üg (t) = −ω 2 u(t) − 2ξω u̇(t) (1.12)

The above equation reveals that the absolute acceleration of the mass v̈ and the pseudo-
acceleration A(t) differ in absolute value by the term 2ξω u̇(t), which is due to damping in
the system. The two quantities are equal if and only if the damping in the system is zero.
In a similar way, one can define the so-called relative pseudo-velocity of the mass V (t)
as follows:
V (t) = ωu(t) (1.13)
This quantity is also to be distinguished from the actual relative velocity of the mass given
by equation (1.9). Actually, one can readily observe that the quantities V (t) and u̇(t) differ
even in the case ξ = 0: in the first case the trigonometric function within the integrand is a
sine and in the second case a cosine.
6 Transfer of Response Spectra

T = 0:5s
» = 0:5%

Displacement
Time (s)

T = 0:5s T = 0:5s
» = 2% » = 5%
Displacement

Time (s) Time (s)

Figure 1.6: Influence of the damping ratio on the response of the oscillator.

1.2.2 The notion of response spectrum

It was shown in the previous section that the response of a single-degree-of-freedom oscil-
lator can be determined in every instant using the Duhamel integral. It was also shown how
this information can be used to determine the elastic forces that are developed in the struc-
ture. However, it is obvious that for the design of a structure, knowledge of the complete
temporal variation of the elastic forces and relative displacements is not necessary and only
the largest absolute value of these quantities need to be known.
This is the motivation of introducing the notion of response spectrum of an excitation
üg (t), undoubtedly one of the most fundamental ideas in structural dynamics.
To illustrate the notion of response spectrum, we return back to the example presented
in Figure 1.4 and we examine the relative displacement of an oscillator characterized by a
period T = 2s and a critical damping ratio ξ =2%. The response of this oscillator under
the excitation of Figure 1.4 is shown in Figure 1.7.
The figure reveals that the maximum relative displacement of the mass m with respect
to its support takes place at the instant t = 8.66s and is equal to 2.5cm. The same type
of calculation can be repeated for a collection of oscillators exhibiting different values of
natural periods T and critical damping ratios ξ. For every examined case of oscillator we
retain only the maximum value of the response, which we shall call spectral displacement
and shall denote as SD . It is obvious that the spectral displacement of an oscillator for a
given excitation is a function of its natural period T and its critical damping ratio ξ. We
write this as follows:
SD (T, ξ) = max{u(T, ξ, t)} (1.14)
t

The obtained values of spectral displacements can be represented as curves that are
functions of the period T for a constant critical damping ratio ξ. Such a set of curves for
the considered excitation is presented in Figure 1.8. This set of curves is characteristic of a
given excitation and is known as the set of displacement response spectra of the excitation
Fundamentals of response spectra 7

umax = 0:25dm

T = 2:0s
» = 2%

Displacement (dm)

Time (s)

Figure 1.7: Maximum response of oscillator.

for different damping ratios. The utility of the displacement response spectrum of an ex-
citation is of paramount importance: if known, the displacement response spectrum allows
determining by direct reference to the appropriate point on the curves and with no further
calculation, the maximum response of any oscillator with characteristics T and ξ. This is
illustrated in Figure 1.8 for the oscillator with T = 2s and ξ = 2% under the excitation of
Figure 1.4.

» = 0%
Relative displacement (m)

» = 2%
» = 5%

SD = 0:25dm

Period (s)

Figure 1.8: Relative displacement response spectrum.

The second piece of information that needs to be known for the design of a structure
is the maximum elastic force developed in the structure, which can be determined using
the pseudo-acceleration A(t) through equation 1.11. This suggests the idea of introducing
the notion of pseudo-acceleration response spectrum, which yields the maximum pseudo-
acceleration of an oscillator under a specific base excitation and is defined as:
SA (T, ξ) = max{ω 2 u(T, ξ, t)} = ω 2 SD (T, ξ) (1.15)
t

The above equation shows that if the displacement response spectrum is known, one
can readily calculate the pseudo-acceleration response spectrum and vice versa. From this
8 Transfer of Response Spectra

point of view, the two spectra are equivalent and passage from one to the other can be
performed without knowledge of the excitation that they characterize. In other words, the
two spectra provide the same kind of information for the considered excitation and none of
them contains more information about the excitation than the other.
The acceleration response spectrum for the excitation of Figure 1.4 is shown in Figure
1.9. Once the spectral pseudo-acceleration SA (T, ξ) for an oscillator has been determined,
the maximum elastic force can be readily calculated by the expression:

max{fs (t)} = mSA (T, ξ) (1.16)

t
Pseudo-acceleration (m/s2)

» = 0%

» = 5%

Period (s)

Figure 1.9: Pseudo-acceleration response spectrum

Let us finally note that a pseudo-velocity spectrum may be defined as follows:

1
SV (T, ξ) = ωSD (T, ξ) = SA (T, ξ) (1.17)
ω
There is hardly any seismic study in which the notion of pseudo-acceleration response
spectrum is not encountered. The use of pseudo-acceleration response spectra is so frequent
that they are commonly referred to simply as response spectra.

1.2.3 Design response spectra - Properties of response spectra

The previous paragraph revealed that if one is interested in the maximum response of a
structure in order to determine the maximum forces that are developed in it, full knowl-
edge of the temporal variation u(t) is not necessary. Instead, only the pseudo-acceleration
response spectrum (or equivalently the displacement response spectrum) need to be known.
Going one step further, the notion of response spectrum can be used as a means to
define an envelope seismic excitation with reference to a particular seismic zone and soil
conditions. It is remarkable to notice that such an envelope excitation cannot be simply de-
fined through a set of acceleration time histories that may characterize the seismic activity
in a region unless the response spectra are introduced. Indeed, both in terms of frequency
content and in terms of amplitudes, the variability of seismic motions is in general very
Fundamentals of response spectra 9

large. Thus, the most severe excitation for one structure, say yi (t) will not be the same as
the most severe excitation, say yj (t) for a different structure. On the contrary, calculation
of the response spectra of these two excitations allows defining an envelope response spec-
trum which will certainly lead to the maximum elastic forces developed on both structures.
It is realized from the above remarks that the envelope response spectrum corresponds
to no physical acceleration time history and obviously it is of no use to search in accelero-
gram databases for an acceleration time history exhibiting an identical response spectrum
as the envelope response spectrum. Where the envelope response spectra are actually very
useful is in design codes for earthquake-resistant design of structures. In all modern earth-
quake design codes, the design seismic motion is characterized by a response spectrum
which is function of the sismicity of the site, the local soil conditions, the damping of the
structure etc., and which is obtained by appropriate statistical treatment of a large number
of accelerograms recorded at the seismic zone of the structure or at sites with similar sis-
micity and soil conditions. Such, so-called design response spectra can be interpreted as
statistical envelope response spectra of the seismic motions that may affect a structure in a
particular site.
A typical design pseudo-acceleration response spectrum for different values of damping
is shown in Figure 1.10. The corresponding displacement response spectrum is shown in
Figure 1.11.
Pseudo-acceleration (m/s2)

Period (s)

Figure 1.10: Design pseudo-acceleration response spectrum.

It is highly instructive to study some properties of the response spectra and in particular
what are the values they take for the two extreme values of the natural period T , namely
T → 0 and T → ∞. We recall that an oscillator with T = 0 is an infinitely rigid oscillator.
Conversely, the case T = ∞ corresponds to an infinitely flexible oscillator.
Let us begin with the case T → 0, by noting that it corresponds to a circular frequency
10 Transfer of Response Spectra

Relative displacement (m)

Period (s)

Figure 1.11: Design displacement response spectrum.

ω  1. The equation of dynamic equilibrium thus becomes:

ü(t) + 2ξω u̇(t) + ω 2 u(t) = −üg (t) =⇒

ω 2 u(t) = −üg (t) =⇒ (1.18)
A(t) = −üg (t)

This is because, the term in ω 2 is much larger than the other two, which can thus be ne-
glected. It turns out that the pseudo-acceleration A(t) is identified with the base excitation.
Thus the spectral acceleration SA (T = 0) is equal to the peak ground acceleration (com-
monly known as PGA), i.e. the maximum (in absolute value) acceleration in the base
excitation acceleration time history. We note that this value is independent of the damping
in the system. We thus conclude, that the response spectra for T → 0 and different damp-
ing rations converge to a unique value which corresponds to the peak acceleration of the
excitation.
Regarding the relative displacement, we can readily observe that through the expres-
sion relating the pseudo-acceleration and the relative displacement response spectrum, we
obtain immediately:
1
lim SD (T ) = lim 2 SA (T ) = 0 (1.19)
T →0 ω→∞ ω

This is an expected result: since the structure is infinitely rigid, it will precisely follow
the movement of the base and its relative displacement with respect to the base will be
zero. Similarly, this result is independent of the damping in the system. We conclude that
the relative displacement response spectra for different damping ratios converge to a zero
spectral displacement when T → 0.
Let us now examine the second limit T → ∞. This situation corresponds to a very
flexible structure exhibiting a circular frequency ω → 0. From the dynamic equilibrium
equation we obtain:
Fundamentals of response spectra 11

ü(t) + 2ξω u̇(t) + ω 2 u(t) = −üg (t) =⇒

ü(t) = −üg (t) =⇒
(1.20)
u(t) = −ug (t) =⇒
SD (T → ∞) = max{ug (t)}
t

In the above equation, the step 3 is obtained from step 2 with double integration. We
conclude that the spectral displacement of a very flexible structure is equal to the peak
ground displacement (PGD), i.e. the maximum (in absolute value) displacement in the
base excitation displacement time history. Again, this result is independent of oscillator
damping. The displacement response spectra for different values of damping tend invari-
ably to a unique limit equal to the peak ground displacement. We can interpret this result
by thinking that the oscillator’s connection to the base is very flexible, thus the oscillator
remains practically immobile and only its base is excited. Thus the maximum relative dis-
placement of the oscillator with respect to the base is simply the absolute displacement of
the base.
Accordingly, the spectral pseudo-acceleration for T → ∞ is obtained from the expres-
sion that links the pseudo-acceleration with the relative displacement:

lim SA (T ) = lim ω 2 SD = 0 (1.21)

T →∞ ω→0

We conclude that the spectral acceleration of a very flexible system tends to zero indepen-
dently of the damping in the system.
Let us finally remark that the above properties of the response spectra are useful be-
cause they provide a means of performing an initial check for a calculated spectrum. For
example, a pseudo-acceleration response spectrum with a zero value at T = 0 or tending
asymptotically to a finite value for large periods means that the calculation has not been
correct.

1.2.4 Going back to the initial problem

The previous paragraphs offered a synopsis of the required theory and definitions related
to the fundamental notion of response spectra. Now, we can revisit the initial problem
presented in Figure 1.1. Given a structure excited at its base with the an excitation char-
acterized by a given design response spectrum SAbase (T, ξ) (notice that the excitation in not
known; only its response spectrum is provided), calculate the transferred response spec-
trum SAA (T, ξ) that will characterize the motion of point A, to which an equipment is at-
tached. Evidently, knowledge of the transferred response spectrum SAA (T ) together with
the natural period and the damping of the equipment allows for an immediate calculation
of the maximum forces developing on the equipment and thus for the verification of the
stability of the equipment.
In practice, the calculation of transferred response spectra needs to be performed for
several values of damping ratios and in the three orthogonal directions, i.e. a separate set
of response spectra needs to be calculated in each direction. This is a subtle point, because
in order to calculate the transferred response spectrum in one direction, say direction x, it
not sufficient to consider a base excitation parallel to the direction x only. Actually, for a
general non-symmetric structure, a base excitation in one direction will generate a response
12 Transfer of Response Spectra

in point A with non-zero components in all three directions. Moreover, given an orthogonal
coordinate system for a structure, it is evident that an earthquake will produce movements
in all three orthogonal directions. We see thus that the following procedure needs to be
followed:
1. Consider three base excitations ug,i (t) parallel to the three directions i = x, y, z. This
forms a triplette of acceleration time histories that represent the complete seismic
excitation of the structure.
2. For each base excitation ug,i (t), calculate the components of the response in point A
uA,ij (t) in the three directions j = x, y, z. For example, the component uA,xx (t) is
the response of point A parallel to x for a base excitation parallel to x, the component
uA,xy (t) is the response parallel to y due to an excitation parallel to x and so on.
3. Note that uA,ij (t) is the absolute displacement time history of point A. Great attention
is needed to this point because very often in structural analysis programs the outcome
of the analysis is the relative displacement in the structure with respect to the base
excitation. If this is the case, the absolute displacement time history needs to be
calculated a posteriori by adding the outcome of the analysis with the base excitation.
4. Sum up the three response components parallel to one direction obtained for the three
excitations to find the overall response in one direction. For example, the overall
component of the response parallel to x is:
uA,x (t) = uA,xx (t) + uA,yx (t) + uA,zx (t) (1.22)
Note that if the calculation is performed through time history analyses, it is possible
to inject simultaneously the three excitation components at the base of the structure.
Then the analysis yields directly the overall component uA,j (t) of the response in
each direction.
5. Calculate the response spectrum SAA,j (T ) of each component uA,j (t), j = x, y, z for
several values of damping ratios.
We see that the number of required calculations in practical applications may increase
significantly, especially if transferred response spectra are to be calculated at a large number
of points and for several damping ratios.

1.3 Solution methodologies for transferred response spec-

tra
In this section, we discuss the most important solution methodologies for calculating trans-
ferred response spectra. These are:
1. Using time history analyses.
2. Using results of dynamic modal analyses.
3. Using empirical relationships suggested in design norms.
The following paragraphs discuss these three methodologies.
Solution methodologies for transferred response spectra 13

1.3.1 Transferred response spectra using time history analyses

The first solution methodology is inspired directly from the definition of transferred re-
sponse spectra. It consists in performing time history analyses of the structure subjected to
a given base excitation in order to calculate the time history of the response at the points
where a transferred spectrum is sought. Once these time histories have been calculated, it
is straightforward to calculate the corresponding response spectrum.
We emphasize once more one of the subtle points of this method: the quantity that needs
to be calculated is the absolute displacement or acceleration time history of the examined
points and not the relative one. This can be understood by examining the transferred spec-
trum at a point of a very rigid structure. The transferred spectrum in this case must be
identical to the input motion response spectrum and this is guarranteed if the considered
transferred motion is identical to the input motion. In order for this to be true, it is obvious
that the considered transferred motion must be the absolute motion of the point. In this
case, the relative displacement of the point is evidently zero.
The great advantage of this method is that it is applicable to any structural system
either linear or non-linear. Moreover, it is intuitive in the sense that the time history of
the response is actually calculated and thus one can have a more direct view of how the
structure will respond to a given base excitation. Finally, it is the most general and rigorous
method and it is more accurate when compared to simpler methods such as method 2 and
method 3.
However, this method is seldom used in practical applications for the following reasons:

• First of all, time history analyses are computationally expensive and they require high
level of modeling skills (selection of records, time integration algorithm, modeling
of damping etc).

• Since the base excitation is usually given in the form of a design response spectrum,
it is necessary to find acceleration time histories that match this spectrum. As it was
explained in the previous paragraph, there are two main difficulties related to this
requirement. First, natural acceleration time histories recorded after real earthquakes
cannot be used, since the design spectrum corresponds to a smoothened envelope
spectrum. To overcome this difficulty, it is usually proposed to select natural accel-
eration time histories, compatible with the characteristics of the examined seismic
event, and then artificially modify them so that they match, as much as possible,
the design spectrum. There exist sophisticated numerical techniques that allow for
this matching to be performed, but the process gets more and more cumbersome as
the number of damping ratios and frequency values (for which spectral matching
is requested) becomes larger. The second difficulty is that the analyses have to be
repeated with more than one record, in order to get a grasp of the variability of seis-
mic motion. Different codes prescribe different recommendations, but a minimum of
three to five records should always be used.

• As it was also explained in the previous paragraph, an excitation parallel to one direc-
tion, say x, will generally yield a structural response parallel to all three directions.
Moreover, a realistic earthquake excitation has non-zero components in all three or-
thogonal directions (two horizontal, one vertical). Consequently, the input motion
14 Transfer of Response Spectra

that has to be injected at the base of the structure must actually be a triplette of accel-
eration or displacement time histories representing the vertical and the two horizontal
components of the ground motion.

• Even if the modeler can afford performing the necessary calculations for generation
of spectrum-compatible records and for transient dynamic analyses, in certain cases
the structural model itself may not be available. This is the case, for example, of old
buildings to which a new equipment is to be added.

• Time history analyses are actually mandatory only in the case of pronounced non-
linear behavior in the examined structure. More efficient methods can be used for the
calculation of transferred response spectra in the case of linear structural response or
in case the nonlinear response remains limited.

1.3.2 Transferred response spectra using results of modal analyses

This methodology is a very efficient alternative for linear systems. It allows deriving trans-
ferred response spectra using simply the input ground motion response spectrum together
with certain modal characteristics of the primary structure and optionally of the equipment.
Of course it is obvious that the great advantage of this methodology is that is does not
require neither artificial accelerograms matching the design spectrum nor ever an entire
structural model of the structure; only certain modal characteristics are required. These are
the following:
ωi
1. The natural frequencies of vibration of the structure fi = 2π , i = 1, 2, ..., n, where n
the number of retained natural frequencies from modal analysis.

2. The modal damping ratios of the structure defined as:

ϕTi Cϕi
ξi = (1.23)
2mi ωi
In the above equation, the vectorial quantities ϕi represent the modal shapes of the
structure, the scalar quantities mi are the generalized masses defined as:

mi = ϕTi Mϕi (1.24)

and the square matrices M and C are the mass and damping matrices respectively.
The number of modal damping ratios is also equal to n. The modal damping ratios
are dimensionless quantities.

3. The modal masses, which depend on the modal shapes and the excitation direction
and are given by the following expression:

(ϕi Mrj )2
m?ij = , j = x, y, z (1.25)
mi
In the above equation the quantity rj is called the excitation direction vector. Its
length is equal to N , where N is the number of degrees of freedom of the structural
system. Its entries take the values 0 or 1: they are equal to 1 if the considered
Solution methodologies for transferred response spectra 15

degree of freedom is parallel to the considered excitation direction j; otherwise they

are equal to zero (for example, the terms corresponding to the rotational degrees of
freedom are always 0). The number of modal masses is equal to 3n. The modal
masses are positive quantities that have the units of mass.

4. The quantities known as participation factors, which are defined by the relationships:
ϕi Mrj
P Fij = , j = x, y, z (1.26)
mi
The participation factors are dimensionless quantities that could be either positive,
zero or negative. Thus, special care has to be taken so that the participation factors
are given with their sign. The number of required participation factors is equal to 3n.

5. The components of the modes shapes in the three directions only for the points in
which a transferred spectrum is requested. In other words, out of the N components
of the mode shapes, one has to retain 3 components for a transferred spectrum at one
point, 6 components for transferred spectra at two points and so on. Of course, these
components must be given for the n retained modes shapes. Consequently, the total
number of mode shapes components to be retained is 3nk, where k is the number of
points for which a transferred spectrum is required. The mode shapes components are
usually dimensionless, following a particular normalization scheme for the modes.
There is no restriction in the selection of this normalization scheme. However, the
definition of all the other necessary quantities listed above should comply with the
same normalization scheme.

6. Regarding the input motion, one has only to define the pseudo-acceleration response
spectrum in a certain frequency range and for a number of damping ratios. It is
usually recommended that the frequency range, in which the input motion spectrum
is given, is large enough to contain all n retained natural frequencies fi from modal
analysis. Similarly, the values of damping ratios should cover a range large enough
to contain all n modal damping ratios ξi . This is so in order to avoid extrapolations
in the evaluation of input motion spectral values, which entail an important degree of
uncertainty and should generally be avoided.

7. Finally, it may be necessary to provide the mass of the equipment if it is comparable

with some characteristic mass in the structure (eg. the mass of the slab to which it is
attached). Usually, the mass of the equipment is neglected.

Certain comments deserve to be made regarding this solution method for transferred
spectra. First of all, how all these input data are combined together to produce the trans-
ferred response spectra? Actually, the theory behind this method makes use of advanced
results from such scientific fields as modal synthesis techniques, theory of random pro-
cesses and perturbation theory, that are beyond the scope of these notes. However, use of
this method for real applications has been rendered relatively straightforward due to avail-
able specialized software such as FSG, which has been developed by one of the pioneers of
this method, Armen der Kiureghian with his coworkers, at the University of California at
Berkeley. A standard reference for the theory implemented in FSG is the article of I GUSA
AND D ER K IUREGHIAN (1985).
16 Transfer of Response Spectra

Second, how accurate is this method, compared to the most rigorous method of time
history analyses? The answer is that its accuracy depends greatly on the number of natural
frequencies n to be retained from modal analysis. It is known from structural dynamics that
a structure with N degrees of freedom will possess N natural frequencies. However, not
all of these frequencies need to be retained in order to perform a modal - spectral analysis
and only few of the modes are sufficient. A rule of thumb is to retain n  N natural
frequencies, such that the sum of their modal masses is larger than 90% of the total mass
of the structure Mj (in a certain direction):
n
X
m?ij > 0.9Mj (1.27)
i=1

If this condition is not satisfied, then the retained set of modes and natural frequencies
is considered incomplete and the accuracy of the obtained results may be questionable.
In this case, it is a standard practice to artificially complete the modal set using the no-
tion of pseudo-mode. More details about the definition of the pseudo-mode can be found
in the lecture notes of the structural dynamics course of ENPC available at the address
http : //www.enpc.fr/fr/formations/ecole_virt/cours/Pecker2009_2010/.
Last but not least, it should be reminded that, although only certain modal character-
istics are necessary in this solution method, a modal analysis of the structure is always a
prerequisite.

1.4 Empirical formulas for transferred response spectra

In this paragraph we examine simplified empirical methods, as proposed in certain design
norms, that can be used to have an initial estimate of the expected result. These simplified
procedures are based on the hypothesis that the structure and the equipment respond on
the excitation solely with the natural mode of excitation of the structure. In these methods,
maximum acceleration developed by the equipment ae is given by the product of the peak
ground acceleration PGA times an amplification coefficient KT :

ae = KT PGA (1.28)

The amplification coefficient KT is then expressed as the product of two functions rep-
resenting the geometrical effect and the effect of frequency matching between the natural
frequency of the structure and the natural frequency of the oscillator:
z  
Te
KT = f ×g (1.29)
H Ts

In the above expression, regarding the arguments of f , z represents the height of the
equipment position and H the height of the building. Regarding the arguments of g, Te
represents the natural frequency of the equipment and Ts , the natural frequency of the
building. Qualitatively, in what concerns f , one sees that it should be close to 1 if z is
small with respect to H: the equipment is close to the soil surface and thus it develops
an acceleration similar to the PGA. As the ratio Hz tends to one, function f is increasing
because the structure acceleration transferred to the equipment is maximum at the top of
Empirical formulas for transferred response spectra 17

the building for a vibration according to the first natural period of vibration. Similarly, in
what concerns g, one should expect to find a maximum when the ratio of the two periods
is around one: this can be interpreted as a resonance phenomenon because the vibration
of the structure according to the first natural period being transmitted to the equipment
can be seen as a harmonic excitation on the equipment with period Ts . The function g is
also a function of the damping of the oscillator: the larger the damping, the smaller the
amplification around the resonant periods.
The following figures present the values of the amplification factor KT proposed in the
following design norms:

• The French norm PS90 considering a value of damping equal to 2%.

• The French norm PS92 considering a value of damping equal to 5%.

• The European standard EC8 considering a value of damping equal to 5%.

• The American code FEMA 368 considering a value of damping equal to 5%.

• A draft proposal of a special committee of AFPS (Association Française du Génie

Parasismique)

Three figures are given for three values of the ratio Hz , namely 0.25, 0.5 and 1.0. We
can note that the maximum amplification factor for 2% damping is around 13 for Hz = 0.25
and it reaches up to 26 for Hz = 1.0. Similarly, when the ratio TTes is either too small or too
Chaire3 GCN
large, the amplification factor is around 1.5 for Hz = 0.25 and around to 4 for Hz = 1.0.
Génie Civil Nucléaire
KT
AMPLIFICATION FACTOR

Te =Ts

Figure 1.12: Amplification factor for z/H = 0.25.

18 Transfer of Response Spectra

Chaire GCN
Génie Civil Nucléaire
KT
AMPLIFICATION FACTOR

Te =Ts

Figure 1.13: Amplification factor for z/H = 0.50.

Chaire GCN
Génie Civil Nucléaire
KT
AMPLIFICATION FACTOR

Te =Ts

Figure 1.14: Amplification factor for z/H = 1.0.

Chapter 2

Notions of Soil-Structure Interaction

2.1 What is Soil-Structure Interaction?

Let one of the emblematic contributors of the scientific field coined under the term Soil-
Structure Interaction provide us with a definition. We read in the abstract of Eduardo
Kausel’s paper “Early history of soil structure interaction” (cf. K AUSEL (2010)):
“Soil-structure interaction is an interdisciplinary field of endeavor which
lies at the intersection of soil and structural mechanics, soil and structural
dynamics, earthquake engineering, geophysics and geomechanics, material
science, computational and numerical methods, and diverse other technical
disciplines. Its origins trace back to the late 19th century, evolved and ma-
tured gradually in the ensuing decades and during the first half of the 20th
century, and progressed rapidly in the second half stimulated mainly by the
needs of the nuclear power and off-shore industries, by the debut of powerful
computers and simulation tools such as finite elements, and by the needs for
improvements in seismic safety.”
The above definition together with the juxtaposition in this very term of the words
soil and structure imply that Soil-Structure Interaction is referring to static and dynamic
phenomena arising in a mixed system consisting of a compliant soil medium and a stiffer
super-structure. Nonetheless, the discipline has grown to the extent of encompassing a
large number of diverse aspects and it is indeed not straightforward to provide a brief yet
consistent definition. Instead one can form a list of the principal problems encompassed by
the theory of SSI:
1. Calculate the response of a soil domain to external static or dynamic sources acting
near or on the soil surface. The sources may be concentrated or distributed loads. In
the case of dynamic loading they can be harmonic in time or suddenly applied with
an arbitrary variation in time. Solutions of such generic problems form the corpus
of what is known as fundamental solutions or Green’s functions. Such solutions
are called fundamental because they can be used as tools to construct solutions to
more complex problems. The most well-known fundamental solution, which is of
paramount importance not only in the theory of SSI but also in geophysics, acoustics
etc., is the solution to the problem of the response (displacement and stress) of an
infinite elastic medium when subjected to a time-varying point load. This solution is

19
20 Notions of Soil-Structure Interaction

due to S TOKES (1849). The theory of how to find these solutions is beyond the scope
of these notes: if needed, we will be exploiting available repositories of fundamental
solutions without further proofs.
2. Calculate the response of the soil to ground-borne vibrations elicited by earthquakes
or some other sources, such as fast moving trains, even before any structures lie in
their path. The problems in this category constitute the so-called free-field problem.
The characterization “free-field” means that only the soil is treated without the pres-
ence of any structure.
3. Calculate the response of rigid, ideally massless structures to ground-borne waves
passing underneath. Such problems are known as wave passage problems and de-
scribe the phenomenon that is known as kinematic interaction. The term kinematic
refers to the fact that the structure is here considered massless and it is only because
of the contrast between the rigidities of the structure and the soil that a different
response will be obtained with respect to the free-field problem.
4. Calculate the response of ideally massless foundation elements such as footings, piles
or caissons embedded in compliant soils and directly elicited by static, harmonic or
transient loads. Problems in this category refer to what is known as determination of
the static and dynamic stiffnesses of a foundation system or more simply, determina-
tion of the dynamic impedance of the foundation.
5. Evaluate the additional deformation of the soil in the neighborhood of an actual struc-
ture caused by feedback of the structure’s own inertia. Here we are interested in the
effect of the structure’s inertia (i.e. mass) so the described phenomenon is known as
inertial interaction.
6. Develop original analytical and numerical methods in order to analyze any of the
above problems. Numerical methods commonly used in SSI include finite element
models with transmitting boundaries, boundary element methods etc.
Let us take a further look into the notions of soil-structure interaction revealed above
by examining the problem illustrated in Figure 3.2 (extract from G AZETAS AND M YLON -
AKIS (1998)). The problem concerns the general situation of a structure founded on an
embedded foundation supported on piles, which is subjected to a seismic excitation. The
source of this excitation is the earthquake hypocenter which may be located in a very large
distance and depth away from the structure. At the instant of the seismic event, numerous
seismic waves will start propagating from the seismic source towards the soil surface and
the structure. These incident waves form a complicated wave field composed of different
types of waves, such as shear waves (S waves), dilatational waves (P waves) and surface
waves (Rayleigh waves, Love waves).
The nature of the incoming waves will be partly dictated by the seismological character-
istics of the seismic event and partly by the geometry, stiffness and damping characteristics
of the soil deposit. The variation of soil properties from the source towards the ground
surface will tend to modify the wave field due to wave scattering, diffractions, reflections
etc. The modified motion of the ground surface aff 1 with respect to the motion near the
1
The motion is usually described by acceleration time histories in the three spatial directions. This is why
the symbol a is used to represent the motion at a certain point.
What is Soil-Structure Interaction? 21

Figure 2.1: Illustration of soil-structure interaction on the structure response.

seismic source ar is known as free-field motion at the site of the foundation. Determination
of the free field motion aff is in itself a challenging task because, as pointed out by LYSMER
(1978), the design motion is usually specified at only one location, (for example an out-
crop of the bedrock underlying the soil deposit), and the complete composition of the wave
field cannot be back-calculated from this incomplete information; that is, the problem is
mathematically ill-posed. Assumptions have to be made regarding the exact wave compo-
sition of the free-field motion and it can be stated that no globally adopted methodology
has yet been proposed to deal with this problem.
Let us now consider the motion around the structure and its foundation: the seismi-
cally deforming soil will force the piles, the embedded foundation and subsequently the
supported structure to vibrate. However, the motion that will be transmitted in the founda-
tion (and this is true even without the presence of the superstructure) will be different from
the free-field motion because of the differences mainly in rigidity and secondly in mass
between the soil on one hand, and the foundation elements (piles, embedded foundation)
on the other; the incident waves are reflected and scattered by the foundation and piles. In
parallel, under the influence of this modified wave field, these elements are stressed, devel-
oping curvatures and bending moments. The modification of the free-field motion due to
the presence of foundation elements with different rigidity and mass than the soil, is known
as kinematic interaction.
Once this modified motion due to kinematic interaction is induced at the foundation
level, it is transmitted in the superstructure and generates oscillations in it. The vibration of
the superstructure is accompanied by inertia forces and overturning moments developing at
22 Notions of Soil-Structure Interaction

its base. These time-varying forces and overturning moments are thus retransmitted back
to the foundation elements and result in the generation of new waves that originate from the
base of the structure and emanate within the surrounding soil inducing a further modifica-
tion of the wave field with respect to the free-field motion. This further modification of the
motion due to the inertial response of the superstructure is known as inertial interaction.
We can emphasize on two interesting aspects related to inertial interaction: first, the
vibration of the superstructure is influenced by the presence of a surrounding medium such
as the soil which is commonly more flexible than the superstructure. In other words, if a
structural model were to be developed for the superstructure only, it should describe cor-
rectly the support of the structure onto the soil by some sort of springs that reproduce the
real flexibility of the soil. Second, the transmission of waves from the structure to the soil
can be interpreted as transmission of energy: indeed, a certain amount of the energy given
in the superstructure from the seismic event is returned back to the soil medium and prop-
agates towards its infinite extremities. This leak of energy towards the unbounded regions
of the soil domain is an essential feature of soil-structure interaction and it is known as ra-
diation damping. Returning back to the structural model of the superstructure, one realizes
that replacing the surrounding soil by equivalent springs is not sufficient. These springs
have to be coupled with energy absorbing elements (dampers) reproducing the effects of
radiation damping. Only in the ideal case of a soil infinitely stiffer than the superstructure,
one can have a structural model for the superstructure in which the base is considered fixed:
no displacements are transmitted from the structure’s vibration to the surrounding soil, no
energy leaks towards infinity within the soil medium and there is no inertial interaction at
all.
It is instructive to see how one should proceed for the verification of the foundation
elements of the examined structure with and without consideration of soil-structure inter-
action. In the standard design methodology which neglects soil-structure interaction one
would perform the following steps:
• Given the characteristics of the soil deposit and a design motion defined in a specified
location within the soil deposit, calculate the seismic motion at the surface of the
ground aff . (problem confronted by soil dynamics).
• Calculate the response of the superstructure under the free-field motion aff consider-
ing that it is fixed at its base (problem confronted by structural dynamics). Note that
in this step two assumptions are made: first, the fact that the structure is solicited by
the free-field motion means that the kinematic interaction is neglected. Second, con-
sidering fixed-base conditions means that the inertial interaction is neglected because
no motion and energy is retransmitted from the superstructure back to the soil.
• Infer the reaction forces and reaction bending moments that are developed at the base
of the superstructure.
• Verify the structural safety of the foundation elements for the maximum value of
these reaction forces and reaction bending moments (problem confronted by geotech-
nical engineering).
It is seen that neglecting soil-structure interaction simplifies the original problem into sim-
pler problems that can be routinely confronted with standard methods of soil dynamics,
structural dynamics and geotechnical engineering.
Illustration of SSI on a simple example 23

On the contrary, for a design methodology that takes into account soil-structure inter-
action, it is obvious that one should be able to evaluate the combined effect of kinematic
and inertial interaction on the foundation elements. One cannot a priori decompose the
problem into the above steps because actually interaction means that each step is coupled
with the others. It will be seen in the following paragraphs, that under certain hypotheses
(pertaining to the constitutive behavior of the soil and the structure), a decomposition of
the the SSI problem is feasible but this decomposition is obviously not similar to the one
described above. Moreover, in the general case the problem has to be treated in its entirety
implying a model comprising both the structure and the surrounding soil. In such a model,
one has to compute the propagation of the seismic wave field and the induced vibrations in
the superstructure and the surrounding soil simultaneously.
Spectral acceleration [g]

Free field

Structure lvl. 0

Frequency [Hz]

Figure 2.2: Acceleration response spectra at the Humbolt Bay nuclear power station.

Let us close this introductory section with a recorded manifestation of SSI in a real
structure. This is the very well known case study of the Humbolt Bay nuclear power station
in California. This structure was significantly affected by a strong earthquake in 1977,
during which several accelerograms were recorded both in points within the structure and
in the ground surface away from the structure, the latter being representative of the free-
field motion. Figure 2.2 represents the acceleration response spectra for a point in the
free-field and a point in the structure which is on the same elevation. The figure reveals
the strong soil-structure interaction effect leading to a significant attenuation of the spectral
accelerations for all frequencies larger than 1Hz.
The following paragraph is intended to provide some initial quantitative aspects of SSI
and in particular, inertial interaction.

2.2 Illustration of SSI on a simple example

The effect of soil structure interaction and in particular, inertial interaction, can be il-
lustrated with the simple idealized model of the structure depicted in Figure 3.3 (W OLF
(1985)).
The example concerns a typical concrete bridge pylon founded on a shallow footing and
subjected to a horizontal harmonic displacement at its base. Considering that the vertical
24 Notions of Soil-Structure Interaction

u g = U g exp ( iωt )

Figure 2.3: Idealized model for soil-structure interaction.

motion of this structure is negligible, the pylon can be modeled by a concentrated mass
m, a spring with coefficient k that reproduces the flexural stiffness of the pylon and a
dashpot with damping coefficient c to model the material damping of concrete. The mass
is placed at an elevation h above the foundation. The connection between the structure
and the foundation is ensured by a rigid beam. The foundation rests on the soil deposit
and its interaction with the soil is modeled by springs and dashpots in the horizontal and
the rotational degrees of freedom of the footing. The springs represent the translational
and rotational stiffness of the footing within the the supporting medium and the dashpots
reflect the radiation of the seismic waves away from the foundation. The system depicted
in Figure 3.3 possesses 3 degrees of freedom:
• The horizontal displacement of the mass um .
• The horizontal displacement of the foundation uf .
• The rotation of the foundation θ.
The structure is subjected to a horizontal harmonic support displacement with circular
frequency ω and amplitude Ug . The dynamic equilibrium equations of the system can be
easily established using the method of Lagrange equations with the following generalized
coordinates:
• q1 = um : the relative displacement of the mass with respect to point A (cf. Figure
3.3).
• q2 = uf : the horizontal displacement of the foundation.
• q3 = θ : the rotation of the foundation.
The absolute displacement ut of the mass is related to the previous parameters by :

ut = ug + uf + um + hθ (2.1)

The kinetic energy of the system is given by

1  2
T = m u̇g + u̇f + u̇m + hθ̇ (2.2)
2
Illustration of SSI on a simple example 25

and its potential energy by:

1
ku2m + kh u2f + kθ θ2


V = (2.3)
2
Finally, the work of the non conservative forces that are developing in the dashpots is:
 
δW = − cu̇m δum + ch u̇f δuf + cθ θ̇δθ (2.4)

The dynamic equilibrium equations are derived from the Lagrange equations:
 
d ∂T ∂T ∂V
− + = Qi (2.5)
dt ∂ q̇i ∂qi ∂qi

in which, the quantities qi and q̇i are taken as independent variables and the quantities
P Qi
are inferred by writing the work of non-conservative forces in the form δW = i Qi δqi .
Developing equations (2.5), one obtains the dynamic equilibrium equations in the time
domain as follows:

m(üf + üm + hθ̈) + cu̇m + kum = −müg

m(üf + üm + hθ̈) + ch u̇f + kh uf = −müg (2.6)

mh(üf + üm + hθ̈) + +cθ θ̇ + kθ θ = −mhüg


Equations (2.6) form a system of ordinary differential equations with unknown quantities
um (t), uf (t) and θ(t). This system is more easily handled in the frequency domain be-
cause the ordinary differential equations are transformed to ordinary algebraic equations.
To switch from the time domain to the frequency domain, we initially introduce the dimen-
sionless damping ratios:
ωC ωCh ωCθ
ξ= , ξh = , ξθ = , (2.7)
2k 2kh 2kθ
We then search for solutions written in the form um (t) = Um exp(iωt), uf (t) = Uf exp(iωt)
and θ(t) = Θ exp(iωt). With these manipulations, equations (2.6) become:

2 2
−mω (Uf + Um + hΘ) + k(1 + 2iξ)Um = mω Ug

−mω 2 (Uf + Um + hΘ) + kh (1 + 2iξh )Uf = mω 2 Ug (2.8)
 2 2
−mhω (Uf + Um + hΘ) + kθ (1 + 2iξθ )Θ = mhω Ug


For the design of the examined structure, we are mainly interested in the amplitude of the
response of the bridge deck Um . We thus proceed in an elimination of the variables Uf and
Θ between the three equations (2.8). Introducing the notations mωs2 = k, mωs2 = kh and
mh2 ωθ2 = kθ and observing that ξ, ξh , ξθ  1, we arrive at the following expression for the
relative displacement of the bridge deck Um :
ω2 ω2 ω2 ω2
 
1 + 2iξ − 2 − 2 (1 + 2iξ − 2iξh ) − 2 (1 + 2iξ − 2iξθ ) Um = 2 Ug (2.9)
ωs ωh ωθ ωs
In order to be able to assess the influence of SSI on the response of the bridge deck,
we will search the dynamic characteristics of a single-degree-of-freedom oscillator that
26 Notions of Soil-Structure Interaction

would yield the same response as our interacting model. In other words, we search the
equivalent characteristics of the pylon that we should use in a fixed-base analysis so that
the results are the same as in an analysis with consideration of SSI. If we denote by ω̃ the
natural frequency of this equivalent oscillator, by ξ˜ its dimensionless damping ratio and by
ũg the amplitude of the harmonic motion to which it should be subjected, we obtain for its
response the following expression:

ω2 ω2
 
˜
1 + 2iξ − 2 Um = 2 ũg (2.10)
ω̃ ω̃

By direct comparison of equations (2.9) and (2.10) we see that the following relations must
hold for the bridge deck amplitude Um to be the same in the two cases:

1 1 1 1 ˜ ω̃ 2 ω̃ 2 ω̃ 2 ω̃ 2
= + + ξ = ξ + ξh + ξ ũ
θ g = ug (2.11)
ω̃ 2 ωs2 ωh2 ωθ2 ωs2 ωh2 ωθ2 ωs2

Examination of equations (2.11) shows that the effect of soil structure interaction is:

• To decrease the frequency of vibration of the fixed base structure (ω̃ < ωs );

• To increase the damping ratio of the system (ξ˜ < ξ) with respect to the fixed base
structure;

• To decrease the amplitude of the effective input motion at the base of the structure
(ũg < ug ).

The previous conclusions are illustrated in Figure 2.4, which presents a numerical appli-
cation of the above developments. The equivalent springs and dashpots for the foundation
have been calculated for a circular foundation of radius r resting on the surface of a homo-
geneous elastic half-space with shear wave velocity Vs and mass density ρ. The variations
of the quantities ωω̃s , ξ˜ and ũugg are given as a functions of the dimensionless parameters:

h ωs h m
h̄ = , s= , m̄ = 3 (2.12)
r Vs ρr

Figure 2.4 clearly shows that soil structure interaction is more pronounced for soft soil
conditions (increasing s) and for heavy structures (increasing m).

2.3 Formulation of Soil-Structure Interaction Problem

Before examining the different ways to take into account SSI in a dynamic analysis, it is
useful, and illustrative, to formulate the problem in a general sense. That formulation is
presented within the framework of the finite element method. In fact, the complexity of the
problem to solve is beyond the capability of closed-form solutions and numerical solutions
are required. As a matter of fact, other numerical techniques can be used, such as boundary
element techniques; nevertheless, the concepts that are presented below are general and not
restricted to finite element solutions; The dynamic equilibrium equations are obtained with
reference to Figure 2.5, which is a schematic representation of a SSI problem.
Formulation of Soil-Structure Interaction Problem 27

Figure 2.4: Illustration of the influence of soil-structure interaction.

28 Notions of Soil-Structure Interaction

Figure 2.5: Decomposition of a SSI problem.

In the following, we refer to a mixed system composed of a superstructure and a part

of the surrounding soil as in Figure 2.5. The soil domain is limited by a fictitious boundary
onto which we apply a loading vector qf (t) together with appropriate boundary conditions
(transmitting boundaries that prohibit spurious wave reflections). Denoting by M, C and
K the mass, damping and stiffness matrices of the entire system and by u(t) the unknown
time-varying displacement vector, the dynamic equilibrium equations are:

Mü(t) + Cu̇(t) + Ku(t) = qf (t) (2.13)

In the following we drop the explicit representation of time-varying quantities for con-
venience. The source of the incident motion (earthquake focus) is usually not included
within the finite element model; therefore, the loading vector qf has non zero values on
the boundary of the model. This boundary is assumed to be sufficiently remote from the
structure in order that the motion at the boundary is not influenced by the presence of the
structure. In the following, we will assume that the constitutive behavior of the soil and the
structure is linear, which guarantees that the principle of superposition holds.
Without the structure, the equations of motion are identical to equations (2.13) with
indices ’f ’ representing the quantities related to the free field soil:

Mf üf + Cf u̇f + Kf uf = qf (2.14)

It is worth noting that equation (2.14) can only be solved provided some hypotheses
are made with respect to the nature and direction of the incident motion. Introducing the
interaction displacement ui defined by:

u = ui + uf (2.15)

and subtracting (2.14) from (2.13) it is easily seen that the equation satisfied by ui is:

Müi + Cu̇i + Kui = qi (2.16)

with:
qi = (M − Mf ) üf + (C − Cf ) u̇f + (K − Kf ) üf (2.17)
The load vector qi is determined from the free field displacements. For linear systems,
equations (2.14) to (2.17) show that the SSI problem is decomposed in the sum of a site
Formulation of Soil-Structure Interaction Problem 29

free field response analysis (equation (2.14)) and of a source problem (equation (2.16)),
in which the load vector qi has only non zero components at the nodes common to the
structure and to the soil (basement) as illustrated in Figure 2.5. Once these two easier
problems are solved, the total displacement u for the SSI problem may be recuperated by
equation (2.15).
Equation (2.17) clearly points out that there is interaction as soon as there is a difference
in stiffness and/or mass between the soil and the structure. To simplify the demonstration
let us leave out the damping term in equation (2.17) to obtain:

qi = (M − Mf ) üf + (K − Kf ) üf (2.18)

We will restrict our attention to two extreme cases represented in Figure 2.6.

Inertial interaction only Kinematic interaction only

ρf ρs
Ef Es

Vertically propagating S wave ρ f = ρs

E f  Es

Figure 2.6: Examples of inertial and kinematic interaction.

The first case is that of a structure at the surface of a horizontally layered soil profile
subjected to the vertical propagation of shear waves. Under these conditions, in the free
field, all the points at the soil surface move in phase; the presence of the structure, even if
it is stiffer from the soil does alter the rigidity of the deposit to respond in the vertically
propagating shear waves as long as the embedment of the structure is zero. Consequently
the second term in the right-hand side of equation (2.18) vanishes; the load vector reduces
to:
qi = (M − Mf ) üf (2.19)
The forces qi at the base of the structure give rise to a support motion, equivalent to an
inertial force field in the superstructure. Consequently, interaction is only generated by
inertial forces in the structure; this phenomenon was defined in 2.1 as inertial interaction.
Let us consider now an embedded structure, whose mass is zero above the ground,
whose mass density ρf is equal to the mass density of the soil ρs and whose modulus of
elasticity Ef is much larger than the modulus of the soil Es . It is obvious that in this case,
the mass contrast is zero and the rigidity contrast significant. The load vector becomes:

qi = (K − Kf ) uf (2.20)

These forces arise only from the difference in stiffness for the embedded part between
the structure and the soil. Even with the same mass, there is interaction; this kind of
interaction was named kinematic interaction in section 2.1. It arises from the stiffness of
30 Notions of Soil-Structure Interaction

the foundation that prevents it from following the displacements imposed by the soil. This
kind of interaction may be equal to zero as shown previously for surface foundations, or
negligible under certain circumstances, like very flexible piled foundations. However, for
stiff embedded structures it may be very significant.
In the most general situation, soil structure interaction arises from both phenomena:
inertial and kinematic interaction. Figure 2.5 and the previous developments illustrate the
two broad approaches for evaluating SSI; Figure 2.5(a) corresponds to the direct methods,
the solution of which is obtained by a direct solution of equation (2.13). This approach does
not involve any superposition and is therefore well suited for non linear systems. On the
other hand it requires very advanced modeling capabilities and significant computational
cost to the extent that it is only occasionally used in practice for structures of particular
importance or when the effects of non-linearity have to be accounted for. Alternatively, the
so-called substructure methods take advantage of the decomposition of Figure 2.5(b) and
2.5(c), or of similar decompositions, to solve the global problem in successive steps. These
methods are obviously only applicable to linear problems. They are much lighter compu-
tationally than the direct methods and they constitute the standard choice for confronting
SSI problems in engineering practice.

2.4 Superposition Theorem for Soil-Structure Interaction

Decomposition of SSI in inertial interaction and kinematic interaction, as exposed in the
previous paragraph is not only convenient to illustrate the fundamentals of SSI but, as it
was stated, gives rise to solution techniques called substructure methods. The validity
of these methods relies on a superposition theorem established in the works of ROESSET
ET AL . (1973) and K AUSEL ET AL . (1978) and which has been known since as Kausel’s
superposition theorem. This theorem states that the response of the system depicted in
Figure 2.7(a), subjected to a base acceleration üg can be obtained by:
• Either with a direct solution of equation
Mü + Ku = −M1üg (2.21)
with:
– u the vector of relative displacement with respect to the base of the model
– 1 a unit vector giving the direction of load application
– M the mass matrix
– K the stiffness matrix
• Or in two steps by defining the vector of relative displacement u as the sum of kine-
matic interaction displacement uk and of inertial interaction displacements uin :
u = uk + uin (2.22)
and by solving the simultaneous system of differential equations:
(
Mf ük + Kuk = −Mf 1üg
(2.23)
Müin + Kuin = −Ms (ük + 1üg )
Superposition Theorem for Soil-Structure Interaction 31

Figure 2.7: Superposition theorem (K AUSEL ET AL . (1978)).

in which Mf and Ms represent the mass matrices of the soil substructure and of the
structure with M = Mf + Ms .

Equivalence between equation (2.21) and equations (2.23) is established by simple ad-
dition taking into account equation (2.22) and the previous definition of the mass matrices.
Equation (2.23)(a) gives the response of a massless structure to the incident motion. Its
solution provides the kinematic interaction motions that are used as input motions for the
solution of equation (2.23)(b). In the solution of equation (2.23)(b) the soil can be modeled
with finite elements or equivalently by a stiffness matrix representing the condensation
of all the foundation-soil degrees of freedom at the interface; this condensation is only
possible in the frequency domain. In that framework, the stiffness matrix is formed with
the complex valued moduli taking into account material damping. The stiffness matrix is
composed of a real part (the stiffness) and of an imaginary part representing the energy
dissipation arising either from material damping and/or from radiation damping. The terms
in the matrix are frequency dependent.
For a rigid foundation, it is legitimate to replace the (N × N ) stiffness matrix (N
being the number of degrees of freedom at the interface) by a (6 × 6) matrix providing
the rigid body motions of the foundation. This matrix is called the impedance matrix and
can be conceptually viewed as an assemblage of springs and dashpots. It follows that the
kinematic interaction motions are the rigid body motions of the massless structure. The
notion of impedance matrix will be further discussed in section 2.5.
The structure of equation (2.23)(b) reveals that the solution can be interpreted as the
vector of relative displacements with respect to a fictitious support subjected to the rigid
body motions of the foundation. Therefore, under the assumption of a rigid foundation, it
is pertinent to split the global problem into there sub-problems:

1. Determination of the motion of the massless rigid foundation subjected to the seismic
design motion; this step represents the solution of equation (2.23)(a).
32 Notions of Soil-Structure Interaction

Figure 2.8: Impedance of a shallow circular footing.

2. Determination of the foundation impedance matrix; this matrix is composed of a real

and an imaginary component, both being frequency dependent;

3. Calculation of the dynamic response of the structure connected to the foundation

impedances and subjected, at its support, to the kinematic interaction motions calcu-
lated in step 1.

As long as the foundation is perfectly rigid, this three-steps approach is strictly equiva-
lent to the resolution of the global problem (equation (2.21)). The advantage of this decom-
position is obvious if one of the successive steps can be simplified or ignored: the first step
always exists except for a surface foundation resting at the surface of a horizontal layered
soil profile subjected to the upward propagation of shear waves; in this situation, solution
to step 1 is identical to solving the free field site response since there is no kinematic inter-
action as it was explained in section 2.3. Solution to the second step can be simplified for
common geometries by using published results in the literature. The third step is always
required; however it is simpler and more common to structural engineers since it resorts to
classical dynamic analyses.

2.5 Foundation Impedance

It was shown in the previous paragraph, that the general SSI problem for linear systems
can be decomposed into three sub-problems. The first, related to kinematic interaction, is
usually neglected in practice, especially if the foundation is composed of surface footings.
The third subproblem was shown to result in a classical structural dynamic analysis. Our
attention, is then oriented towards the second step in the sub-structuring procedure which
concerns the calculation of the dynamic impedance matrix of the foundation.
In order to illustrate the notion of foundation impedance, we consider the simple case
of a rigid circular footing resting on the surface of a homogeneous and isotropic elastic
half-space representing the soil as in Figure 2.8.
The footing is characterized by its radius r0 and the elastic half-space by its mass den-
sity ρ and its elastic characteristics expressed with a couple of elastic moduli as, for in-
Foundation Impedance 33

stance, its shear modulus G and its Poisson’s ratio ν. Another useful elastic q parameter is
the shear wave velocity VS of the half-space defined by the relation: VS = Gρ .
By definition, the impedance of the foundation is equal to the reaction force developed
on the foundation considered massless, when it is subjected to a unitary harmonic displace-
ment (or rotation) following one of its six degrees of freedom. Inversely and since the
footing is considered massless, the impedance represents the ratio of the force applied on
the footing (equal to the soil reaction) divided by the induced displacement. It is clear from
these considerations that since the footing possesses 6 degrees of freedom as a rigid body
(3 translations and 3 rotations), the impedance matrix S̃(ω) will be of dimension 6 × 6.
It is noted that in the case of a footing of arbitrary shape that moreover can be embed-
ded, the six degrees of freedom are coupled and the impedance matrix S̃(ω) will be full.
In the case of a footing possessing symmetries, some of the degrees of freedom will be
decoupled and the corresponding off-diagonal terms of S̃(ω) will be zero. In the case of
the examined circular footing, four degrees of freedom suffice for the description of the
system: horizontal and vertical translation, rotation about a vertical axis and rotation about
a horizontal axis. It is also noted that the vertical translation and the rotation about a ver-
tical axis are totally decoupled from the other two degrees of freedom. On the contrary,
the horizontal translation and the rotation about a horizontal axis are coupled. However, it
has been shown that the off-diagonal coupling term for a surface footing is relatively small
compared to the diagonal terms and can thus be neglected. Under these assumptions the
impedance matrix is a 4 × 4 diagonal matrix. Each term of the main diagonal represents the
ratio of the applied force on the footing divided by the displacement in the same degree of
freedom. In the following, one of the diagonal terms of the impedance matrix, say the one
corresponding to vertical translation, will be considered separately. Conclusions will hold
for the other degrees of freedom as well. The term is generically denoted by K and will be
referred to simply as impedance.
Before studying the foundation impedance, it is useful to draw the analogies between
the notion of impedance and the notion of single-degree-of-freedom oscillator. We thus
consider such an oscillator subjected to a harmonic force P eiωt . The resulting displacement
of the oscillator is given by the expression:

P eiωt
z(t) = (2.24)
(k − mω 2 ) + iωc
By definition, the impedance of the single-degree-of-freedom oscillator is thus:

P eiωt
K=
z(t)
= (k − mω 2 ) + iωc (2.25)
"  2 #
ω ω
=k 1− + 2iξ
ωn ωn

In the above expression we have adopted the usual notations for a simple oscillator: ωn and
ξ represent the circular frequency and the critical damping coefficient.
Examination of (2.25) shows that the impedance is the product of:
• A term corresponding to the static stiffness k.
34 Notions of Soil-Structure Interaction

• A term representing the dynamic effect. This term consists of a real part and an
imaginary part which is due the fact that the force and the displacement are out of
phase. This difference in phase is due to dissipation in the system (damping term non
zero).
We note in parallel that the real part of the impedance as is equation (2.25) can become
negative if the frequency is large enough.
In analogy to (2.25) it can be shown that the foundation impedance can be written in
the following general form:

K = KS [k1 (ω) + ia0 c1 (ω)] (2.26)

In the previous equation we define the parameter a0 as follows:

ωr0
a0 = (2.27)
VS
It is seen from equation (2.26) that the foundation impedance consists of the multiplying
term KS , which corresponds to the static stiffness (i.e. stiffness for excitation applied very
slowly on the footing) and a term representing the dynamic effect of the excitation. This
dynamic term is complex: it consists of a real and an imaginary part. The coefficients a0 ,
k1 and c1 are dimensionless and depend on the cyclic frequency ω.
It is instructive to give a physical interpretation of the notion of impedance. Equation
(2.26) gives by definition the soil reaction applied on the footing:

R = KS [k1 (ω) + ia0 c1 (ω)] z (2.28)

Given that the excitation is harmonic, it holds:

ż = iωz (2.29)

Equation (2.25) can thus be written as follows:

KS r0 c1 (ω)
R = KS k1 (ω)z + ż (2.30)
VS
It is seen that the soil reaction is composed of two components, proportional to the
footing displacement and the footing velocity. We can attribute to these two components
the physical notion of a spring and a dashpot characterized by the following properties,
functions of the cyclic frequency ω:
KS r0 c1 (ω)
K̄ = KS k1 (ω)z, C̄ = ż (2.31)
VS
The presence of a dashpot with non-zero damping coefficient implies that the energy
dissipation within the system is non-zero even though the soil has been considered as an
elastic half-space, i.e. as a non-dissipative medium. The dissipation within the system
is due to the energy transport obtained from the waves that emanate from the foundation
during the harmonic excitation and propagate towards the infinite extremities of the elastic
half-space. For this reason it is called radiation damping2 and it corresponds to the notion
discussed in 2.1.
2
Radiation damping may also be referred to as geometric damping since it is of purely geometric origin
Foundation Impedance 35

Figure 2.9: Impedance of a shallow circular footing.

The calculation of impedance functions such as in equation (2.31) for realistic founda-
tion configurations require either advanced analytical methods for the most simple cases or
elaborate numerical methods for configurations with arbitrary foundation geometries and
soil properties. These analytical or numerical methods are beyond the scope of these notes.
The engineering practice is greatly facilitated by well-established solutions available in the
literature that cover the most common foundation geometries and soil profiles.
An an example of such published solutions, Figure 2.9 presents the impedance functions
of a rigid circular footing that rests on the surface of a homogeneous, isotropic elastic half-
space. The figure presents the dynamic parts of the impedance only, for the four degrees
of freedom of the footing. It is verified that the terms k1 and c1 vary with the circular
frequency ω. We also note that some of the stiffness terms become indeed negative for
high ω.
It is noted that if the impedance functions were independent of the circular frequency
ω, their introduction within the model of the superstructure (which is the last step in the
implementation of the superposition theorem) would be immediate: it would suffice to
introduce a spring and a dashpot for each degree of freedom, with the appropriate stiffness
and damping coefficients. Unfortunately, this is not possible in the general case because of
the dependence of the impedance functions on the circular frequency ω. This dependence
prohibits of solving the interaction problem in the time domain and obliges to make a
switch to the frequency domain, where the dependence of the impedance functions on the
circular frequency ω can be readily incorporated.
There exist however certain configurations for which the approximation of a spring and
a dashpot with coefficients independent of frequency is possible. In such cases, the analysis
36 Notions of Soil-Structure Interaction

Equivalent Impedance Coefficients

Vibration mode Real part k1 Imaginary part c1

Vertical 1.0 0.854

Horizontal 1.0 0.576

0.30
Rocking 1.0 3(1−ν)Ix
1+
8ρr 5
VS
q 0
Iz
r0 Kθ
Torsion 1.0 1+ 2Iz5
ρr0

Table 2.1: Impedance coefficients for shallow circular footing

can also be carried out in the time domain. One case for which this is possible is actually
the circular footing studied in this section. Comparison of the “exact” response of the
footing with the response of a footing on a spring and dashpot with constant coefficients
reveals that it is possible to approximate in a very satisfactory way the exact response as
long as appropriately chosen values are adopted for the coefficients. For example, Table
2.1 presents a possible choice of parameters and Figure 2.10 compares the exact response
of the footing with the one obtained from the simplified model for the case of excitation
parallel to the vertical degree of freedom.
In the expressions of Table 2.1, the quantities I represent the moments of inertia of the
footing around the vertical and the horizontal axis.
The case of the shallow circular footing on a homogeneous, isotropic elastic half-space
is a configuration simple enough (both in terms of footing geometry and in terms of soil
profile) that the approximation of frequency-independent spring and dashpot coefficients
for the foundation impedance is satisfactory from a practical point of view. For more com-
plicated configurations however, this approximation ceases to be possible. As an illustrative
example, Figure 2.11 presents the variation of the real and imaginary part of the impedance
for the vertical degree of freedom of a circular footing resting on the surface of an elas-
tic soil layer of limited depth. The figure depicts the erratic variation with frequency of
both the real and the imaginary part of the impedance. Moreover, it reveals that for small
values of frequency and finite soil depth, the imaginary part tends to zero, in other words,
there is no dissipation in the system. This phenomenon can be explained by the fact that
the waves emanating from the foundation are totally reflected at the infinite boundary; the
energy cannot get transported to the infinity but it is sent back towards the footing. Energy
transport can only take place when the frequency of excitation is larger than a characteristic
frequency value, denoted as f0 . It can be shown that f0 corresponds to the fundamental fre-
VP VS
quency of the soil layer which is equal to 4H for the vertical mode and 4H for the horizontal
mode, where H is the depth of the layer. Beyond this limiting value, the foundation vibra-
tion is accompanied by generation of surface waves that indeed propagate in the horizontal
direction parallel to the soil surface and thus give raise to a non zero geometric damping.
It is obvious that the abrupt variations of the impedance functions as in Figure 2.11
Foundation Impedance 37

Figure 2.10: Dynamic response of a shallow circular footing subjected to vertical force.

Figure 2.11: Foundation impedance of circular footing on elastic layer of limited depth.
38 Notions of Soil-Structure Interaction

Figure 2.12: Rheological models with added mass for foundation impedance.

Figure 2.13: Application of rheological models for foundation impedance.

cannot be reproduced by frequency-independent springs and dashpots. Nevertheless, it is

still possible to approximate the footing response with a rheological model, more com-
plicated than a simple spring and dashpot, but still affordable from the practical point of
view. The idea behind the development of such approximative rheological models is given
by observing (as in equation (2.25)) that the frequency dependence for the real part of the
impedance is due to the mass of the oscillator. This remark suggests the idea to add a
fictitious mass to the simple spring and dashpot model whose role is to reproduce this vari-
ation. Some authors have suggested that this fictitious additional mass is a part of the soil
around the footing that is mobilized during the excitation and vibrates in phase with the
footing. Notwithstanding the physical intuition of this idea, it should be born in mind that
this mass is actually a pure mathematical artifact that is only introduced to achieve a better
approximation of the real footing response.
Figure 2.12 presents two examples of rheological models with additional masses that
lead to frequency-dependent impedance functions. The model parameters are calibrated
in order to reproduce the exact impedance functions. Such a calibration is shown in Fig-
ure 2.13 which presents the impedance of a real foundation (solid line) together with the
response obtained from a simple rheological model with added mass. In particular, it is
interesting to note that the simple rheological model can reproduce negative stiffness terms
beyond the value of 1.5Hz.
Chapter 3

Exercises

3.1 Transfer of Response Spectra

The purpose of this exercise is to demonstrate the process of response spectra transfer for
the design of secondary equipment in buildings. For this reason, we examine the single-
storey frame shown in 3.1. The dynamic response of this structure is modelled by a single-
degree-of-freedom oscillator characterized by its mass m0 and its stiffness k0 . The damping
of the structure is neglected. The structure is subjected to an impulsive base excitation as
follows:

üg (t) = Aδ(t)

In the above expression A is the magnitude of the acceleration impulse and δ(t) is the
Dirac delta function.

I. Calculate the horizontal response at the roof of the structure under this excitation.

II. Find the maximum value of the response as a function of the natural period of the
structure.

III. Draw the acceleration, velocity and displacement response spectra of the oscillator
for the considered base excitation.

In the following, we consider that a crane with mass m1 , negligible with respect to the
mass of the structure, is attached at the roof of the structure. The rigidity of the connection
of the equipment is k1 (we can additionally neglect damping, or consider a critical damping
ratio ξ1 ). It is supposed that the presence of the crane does not influence the dynamic
response of the structure. It is thus assumed that an acceleration time history identical to
the vibration of the roof as calculated in question I, is applied at the connection between
the roof and the crane.

IV. For this vibration, calculate the response of the crane.

39
40 Exercises

V. Determine the maximum value of the response as a function of the natural period of
the crane.

VI. Draw the acceleration, velocity and displacement response spectra of the crane that
correspond to the calculated response.

VII. Compare the response spectrum of the impulsive excitation at the base of the struc-
ture with the transferred response spectrum at the position of the crane.

NOTE. Recall that the Dirac delta function is defined as follows:

(
+∞ t = 0
δ(t) =
0 t 6= 0

For the Dirac delta function and any function f (t), the following identity holds:
Z t
f (t − τ )δ(τ ) dτ = f (t)
0

Figure 3.1: Frame structure with crane.

Solution
I. The single-storey frame is modelled as a single-degree-of-freedom oscillator, which is
excited by the base excitation:
üg (t) = Aδ(t)
in which, the quantity A has units of velocity and δ(t) is the Dirac delta function.
The equation of dynamic equilibrium of this structure is:

m0 üA + c0 u̇A + k0 uA = −m0 üg (t)

Transfer of Response Spectra 41

Neglecting the damping in the structure and introducing the circular frequency ω0 , the
above equation becomes:
üA + ω 2 uA = −üg (t)
q
k0
in which ω0 = m 0
. The solution of the above differential equation is given by the Duhamel
integral as follows:
Z t
1
uA = − ÿ(τ )eξω0 (t−τ ) sin [ωD (t − τ )] dτ
ωD 0
p
with ÿ(t) = üg (t) = Aδ(t), ξ = 0 and ωD = ω0 1 − ξ 2 = ω0 . It is thus seen that the
Duhamel integral can be written as follows:
Z t
1
uA = − δ(τ ) sin [ω0 (t − τ )] dτ =
ω0 0
A t
Z
=− δ(τ ) sin [ω0 (t − τ )] dτ =
ω0 0
A
= − sin ω0 t
ω0

It turns out that point A executes a harmonic vibration with amplitude ωA0 and period T0 =
q
2π m0
ω0 = 2π k0 . We verify that the amplitude of motion is expressed in units of displacement.

Note. In the above reasoning pertaining to use of the Duhamel integral, it is implied that
zero initial conditions apply for both the displacement and the velocity of point A.

II. We showed that:

A
uA = − sin ω0 t =⇒
ω0
A A
max{uA } = = T0
ω0 2π
The quantity max{uA } expresses the spectral displacement of the motion at point A.
It is seen that the displacement response spectrum of the motion of point A is linear with
respect to the period T0 of the structure. Accordingly, the following remarks can be made:

• For a rigid structure: T0 → 0: max{uA } → 0: the relative displacement is zero as

expected.

• For a very flexible structure: T0 → +∞: max{uA } → +∞: the structure departs with
a constant velocity A and the displacement is theoretically unbounded.

III. We showed in the previous question that the displacement response spectrum is
linear with respect to the period:
A A
Sd = max{uA } = = T0
ω0 2π
Correspondingly, the velocity response spectrum is by definition equal to:

Sv = ω0 Sd = A
42 Exercises

SD SV SA

A
A
2π
2πA

1. T T 1. T

Figure 3.2: Displacement, pseudo-velocity and pseudo-acceleration response spectra of

input motion

It is thus seen, that the velocity response spectrum is constant with respect to the period.
The pseudo-acceleration response spectrum is also calculated easily from the dis-
placement response spectrum as follows:
2πA
Sv = ω02 Sd =
T0
It is seen that the pseudo-acceleration response spectrum is inversely proportional to
the period of the structure.

IV. Similarly to the previous question, the crane is modelled as a single-degree-of-freedom

which is subjected to a base excitation given by the vibration of point A. Thus the equation
of dynamic equilibrium of the crane is written:

m1 ücr + c1 u̇cr + k1 ucr = −m1 üA (t)

with uA (t) = − ωA0 sin ω0 t. Double differention of this expression yields the acceleration of
point A: üA (t) = −Aω0 sin ω0 t.
In the following, we will neglect the damping of the crane for simplicity. The circular
frequency of the structure ω0 is considered fixed and the circular frequency of the crane
ω1 is considered variable. Under these assumptions, the equation of dynamic equilibrium
becomes:
ücr + ω12 ucr = −üA (t)
Considering zero initial conditions, the solution of the above differential equation is
given by the Duhamel integral:
Z t
1
ucr (t) = − ÿ(τ )eξω1 (t−τ ) sin [ωD (t − τ )] dτ
ωD 0
√
with ÿ(τ ) = Aω0 sin ω0 τ , ξ1 = 0 and ωD = ω1 1 − ξ1 = ω1 . So, we obtain:

Z t
1
ucr (t) = − Aω0 sin ω0 τ sin [ω1 (t − τ )] dτ
ω1 0
Aω0 t
Z
=− sin ω0 τ sin [ω1 (t − τ )] dτ
ω1 0
Transfer of Response Spectra 43

SD SA

ū Aω0

1. β 1. β

Figure 3.3: Transferred displacement and pseudo-acceleration response spectra with re-
spect to β.

Rt
The evaluation of the integral I = 0 sin ω0 τ sin [ω1 (t − τ )] dτ depends on the values of
ω0 and ω1 :
• CASE A. If ω1 6= ω0 :
1
I= (ω0 sin ω1 t − ω1 sin ω0 t)
ω02 − ω12
Then, the displacement time history of the crane is:
Aω0
ucr (t) = − I
ω1
ω0 A
Introducing the auxiliary quantities β = ω1 and ū = ω0 , one obtains:
β2
ucr (t) = ū (sin ω0 t − β sin ω1 t)
1 − β2
• CASE B. If ω1 = ω0 :
1
I= (sin ω1 t − ω1 t cos ω1 t)
2ω1
and, introducing the quantities β and ū as before, the displacement time history of
the crane is written:
ū
ucr (t) = (sin ω1 t − ω1 t cos ω1 t)
2

V. We search the maximum of the response calculated in the previous paragraph. We

can remark the following:
• If ω1  ω0 =⇒ β → 0: ucr (t) → 0.
• If ω1 = ω0 =⇒ ucr (t) = ū2 (sin ω1 t − ω1 t cos ω1 t) =⇒ max{ucr (t)} increases
infinitely. This can be thought of as a resonance state since the crane and the sup-
porting structure have the same natural period.
• If ω1 → 0: we have to go back to the initial equation and notice that:
ücr + ω12 ucr = −üA =⇒
ücr = −üA =⇒
max{ucr } = max{uA } = ū
44 Exercises

3.2 Effects of soil structure interaction

The purpose of this exercise is to demonstrate the effects of soil-structure interaction on the
dynamic response of buildings and to show how these are taken into account in the design
process. The exercise is inspired from the two reactors in the San Onofre nuclear power
plant in California shown in Figure 3.4a, for which it is known that their fixed-base natural
period is of the order of 0.15s whereas their natural period with account of soil flexibility
is 0.5s.
The dynamic response of the reactors in the horizontal direction will be studied with the
simplified model shown in figure 3.4b. The superstructure is modeled as a single-degree-
of-freedom-oscillator with mass mS and stiffness kS .
The structure is founded on a very rigid circular slab of radius r and mass mF which lies
on the surface of a homogeneous and isotropic soil layer with shear modulus G, Poisson’s
ratio ν and mass density ρ overlain by a rigid bedrock. The height of the soil layer is
H = 2r.
For the dependence of the real part of the horizontal impedance kF on the frequency of
excitation, we will consider that it varies linearly from the static value at 0Hz to zero at the
value α = 1.75, where α the dimensionless parameter defined by the expression:
ωr
α=
Vs
In the above expression ω is the circular frequency of excitation and Vs the shear wave
velocity of the soil layer. The variation of kF with respect to the frequency of excitation is
shown in Figure 3.5. The static horizontal frequency will be given by the expression:
8Gr  r
kF,st = 1 + 0.5
2−ν H
The structure is subjected to a seismic excitation characterized by the spectrum pre-
sented in Figure 3.6, defined at 5% damping for the following control periods and numeri-
cal parameters: TB =0.25s, TC =0.45s, TD =3.0s, RA =1.0, RM =2.5, P GA =2.5.

(a) (b)

Figure 3.4: (a) Nuclear reactors at San Onofre (CA) power plant and (b) Simplified struc-
tural model for dynamic analysis.

The spectral pseudo-acceleration is given as SA = P GA × Re(T ) where :

Effects of soil structure interaction 45

• Branch AB: Re(T ) = RA + (RM − RA)(T /TB )

• Branch BC: Re(T ) = RM
• Branch CD: Re(T ) = RM (TC /T )
• Branch DE: Re(T ) = RM (TC /T )(TD /T )

I. Calculate the maximum shear force exerted at the base of the structure considering
fixed base conditions. The structural damping will be taken equal to 5%.

II. Calculate the maximum shear force exerted at the base of the structure considering the
soil flexibility and the contribution of only the first mode of vibration. In order to account
for the uncertainty in soil properties three values of soil shear modulus will be considered
namely G together with Gmin = 23 G and Gmax = 32 G. Modal damping will be taken equal
to 5%. The foundation impedance should be calibrated to fit the natural period of vibration
of the foundation-superstructure system.

III. Perform the same calculation as in question II considering the contributions of both
modes. Perform the calculation using the following numerical values:
• mS = 100kt
• kS = 175000MN/m
• r = 12.0m
• mF = 0.2mS
• G = 500MPa
• ν = 0.3
• ρ = 0.002kt/m3

Solution
I. Considering fixed-base conditions implies that the natural period of vibration of the
structure is given by the expression:
r
100 × 106
r
mS
T0 = 2π = 2π = 0.1502s
kS 175 × 103 × 106
From the response spectrumm for T0 = 0.15s, we see that we are on the branch AB.
The pseudo-acceleration corresponding to T0 is thus:

SA (T0 ) = P GA × Re(T0 ) = (3.1)

 
T0
= 2.5 RA + (RM − RA) = (3.2)
TB
= 4.753m/s2 (3.3)
46 Exercises

Figure 3.5: Variation of kF with respect to frequency.

Figure 3.6: Design pseudo-acceleration response spectrum.

Effects of soil structure interaction 47

The maximum relative displacement of the mass is given by:

SA
SD (T0 ) =
ω02
2π
in which ω0 = T0 . We thus obtain:

SD (T0 ) = 0.0027m

The elastic force corresponding to this displacement is:

Fmax = ks SD (T0 ) =
= mS ω02 SA (T0 ) = 475.3M N/m

II. We next introduce the soil-structure interaction effect, which is taken into account with
a frequency-dependent spring and dashpot. The system thus exhibits two degrees-of-
freedom:

• uS : horizontal displacement of the structure

• uF : horizontal displacement of the foundation

 
 uS 
These degrees-of-freedom are assembled into the displacement vector: u = 
 .

uB
The equation of dynamic equilibrium is written in matrix form as follows:

Mü + Ku = p

In order to determine the stiffness and the mass matrices M and K, we work with the
Euler-Lagrange equations:
 
d ∂T ∂T ∂V δW
− + =
dt ∂ q̇i ∂qi ∂qi δqi

in which the following quantities are defined:

• qi : degrees of freedom in the system

• T : kinetic energy in the system

• V : potential energy in the system

• W : work of the non-conservative forces in the system (eg. external forces not ob-
tained from a potential, damping etc.)

We write the expressions of the kinetic energy:

1 1
T = mS (u̇S + u̇g )2 + mF (u̇F + u̇g )2
2 2
Taking the derivatives with respect to uS , uF , u̇S and u̇F (notice that in the derivation of
Euler-Lagrange equations, qi and q̇i are taken as independent variables), one obtains:
∂T
• ∂uS =0
48 Exercises

∂T
• ∂uF =0
 
∂T d ∂T
• ∂ u̇S = mS (u̇S + u̇g ) =⇒ dt ∂ u̇S = mS (üS + üg )
 
∂T d ∂T
• ∂ u̇F = mS (u̇F + u̇g ) =⇒ dt ∂ u̇F = mF (üF + üg )

The potential energy is written as follows:

1 1
V = kS (uS − uF )2 + kF u2F
2 2
The derivatives with respect to uS and uF , one obtains:
∂V
• ∂uS = kS (uS − uF )
∂V
• ∂uF = kF uF − kS (uS − uF )
The Euler-Lagrange equations thus become:
(
mS üS + kS (us − uF ) = −mS üg
mF üF − kS us + (kF + kS )uF = −mF üg

The above equations can be written in matrix form as follows:

Mü + Ku = −Mrüg

with the following definitions:

 
mS 0 
• Mass matrix: M = 



0 mF
 
 kS −kS 
• Stiffness matrix: K = 



−kS kS + kF
 
1
• Direction vector: r = 
 

1

Having determined the equations of dynamic equilibrium of the system, we proceed in

calculating the natural frequencies and mode shapes. This is done by examining the ho-
mogeneous equation:
Mü + Ku = 0
and looking for solutions of the form:

u = ϕi sin(ωi t + θ)

where ωi is the i-th natural frequency and ϕi is the i-th mode shape. Replacing this solution
into the system of differential equations, one obtains:

K − ωi2 M ϕi = 0


Effects of soil structure interaction 49

This is a linear system with non-trivial solutions, if:

det K − ωi2 M = 0

Developing the determinant, one obtains:

(kS − ωi2 mS )(kS + kF − ωi2 mF ) − kS2 = 0

which is a polynomial equation with respect to ωi with two real roots ω1 and ω2 with ω1 < ω2 .
Note that this equation is non-linear with respect to ωi not only because of the polynomial
of degree 4 that the determinant gives rise to but also because the term kF is dependent
on frequency. In order to solve this equation, we thus perform an iteration assuring that kF
is compatible with the first natural circular frequency ω1 .
Having determined, the natural frequencies ω1 and ω2 , the mode shapes are calculated
by the equation:
K − ωi2 M ϕi = 0

Expanding this equation, we get:

    
2
kS − ωi mS −kS   ϕi,S  0
  = 
    
−kS kS + kF − ωi2 mF ϕi,F 0

We note the mode shapes can be determined up to a multiplicative constant. We can thus
take the first component equal to 1, by writing:
 
1
ϕi,S = 1 =⇒ ϕi = 
 

λi

It is also worth noting that each natural circular frequency satisfies:

ϕT
i Kϕi
ωi2 =
ϕT
i Mϕi

The mode shapes are used to solve the system of differential equations expressing the
dynamic equilibrium in the system. Initially, we note that the modes form a complete or-
thogonal basis for the vector space in which u is defined. We can thus write:
X
ui (t) = ϕi qi (t)

The system of differential equations expressing equilibrium thus becomes:

X X
M ϕi q̈i (t) + K ϕi qi (t) = −Mrüg

Multiplying by left with every mode shape ϕT

i and using the orthogonality property of mode
shapes allows obtaining n = 2 decoupled differential equations of motion:

ϕT T T
i Mϕi q̈i (t) + ϕi Kϕi qi (t) = −ϕi Mrüg
50 Exercises

We proceed by dividing the above expressions with ϕT

i Mϕi and introducing the natural
ϕTi Mr
circular frequencies and the participation factors defined as: αi = ϕT
. The equations
i Mϕi
of equilibrium thus become:

q̈i (t) + ωi2 qi (t) = −αi üg (t)

The solution of this equation can be solved for the entire time histories qi (t) if the time
history of the excitation ug (t) is known using the Duhamel integral. Alternatively, if the
excitation is defined by means of a response spectrum (as is usually done in practice), we
can only know the maxima of the quantites qi (t). These maxima are given by reading the
spectral values:
1
max{qi } = αi SD (ξi , ωi ) = 2 αi SA (ξi , ωi )
t ωi
where SD (ξi , ωi ) is the spectral displacement and SA (ξi , ωi ) the spectral pseudo-acceleration.
Notice that we have not introduced damping so far in our equations for simplicity. We can
however rewrite the equations introducing the damping matrix C. In the general case, this
matrix does not satisfy the orthogonality condition with respect to the mode shapes as is
the case for the matrices M and K. In practical applications we admit that the damping
matrix is also orthogonal with respect to the mode shapes. Under this assumption, we can
define the modal damping ratios:
ϕT Cϕi
ξi = i
2mi ωi
Usually, the damping ratios are defined a priori for every mode shape. Here, a damping
ratio equal to 5% will be considered for every mode shape. The spectral values should
thus be read at the response spectrum defined at 5% damping.
The maximum displacements of the system in each mode are written:
1
umax,i = ϕi qmax,i = ϕi αi SA (ξi , ωi )
ωi2

The maximum elastic forces in each mode:

fmax,i = Kumax,i

The maximum shear force at the base of the structure (above the foundation) in each mode
is:
Vmax,i = fS,max,i − fF,max,i
Finally, the composition of mode shapes can be made with the SRSS rule. For example,
the maximum shear force is:
q
2
Vmax = Vmax,1 2
+ Vmax,2

The numerical application can be performed with standard mathematical modeling soft-
ware, such as MATLAB, Octave, Mathematica, NumericalPython etc.
Cone model for foundation impedance 51

3.3 Cone model for foundation impedance

We wish to estimate the vibration of a mat foundation that supports a rotating machine that
exhibits a load eccentricity as shown in Figure 3.7. The machine is fixed on the mat founda-
tion: this necessitates the determination of the dynamic impedance of the mat foundation.
In order to achieve this, we introduce a simplified model, known as cone model in which
the soil is represented by a cone of angle α and homogeneous mechanical properties E, ν
and ρ. We suppose that the constitutive law of the soil is linear elastic, i.e. the shear stress
τ is linked to the shear strain γ with the relationship:

τ = Gγ

where G is the soil shear modulus.

(a) (b)

Figure 3.7: Simplified soil modeling with (a) Cone model and (b) 1D model.

The geometry of the mat foundation is circular with radius R. We will consider that
the cone deforms on tension-compression mode when the vibration of the foundation is
vertical and on shear mode when the vibration is horizontal. The rotating machine applies
a force on the mat foundation that can be expressed as follows:
(
fx = mdΩ2 cos Ωt
fz = mdΩ2 sin Ωt

The machine rotates with a frequency of Ω =50Hz. The quantities m and d represent
respectively the rotating mass and its eccentricity. The masses of the machine and the mat
foundation are M1 =100t and M2 =1000t respectively.

QUESTIONS

I. Derive the differential equations that govern the vibration of an infinitely thin soil layer
situated at a depth z in the vertical and horizontal direction. Compare them.

In the following, only the horizontal vibration will be examined.

52 Exercises

z
x
α ∂N dz
z0 N (z) +
ρ A ( z ) uz ∂z 2

dz
2R

∂N dz
N (z) −
∂z 2

Figure 3.8: Dynamic equilibrium of the cone model

II. Write down the boundary conditions and calculate the dynamic impedance of the mat
foundation. The results will be normalized with respect to the static stiffness of a circular
footing of radius R in the horizontal direction, given by the equation Kst = 8GR2−ν
. Which
do you think is the most appropriate value for the angle α ?

III. Compare the cone model (α > 0) with the 1D model (α = 0).

IV. Derive the equation of motion of the mat foundation supporting the machine which
is rotating with circular frequency Ω. Discuss the amplitude and phase of this motion with
respect to Ω.

V. Write the equation of motion of the mat foundation and the machine if the machine is
isolated from the mat foundation with an elastic spring of stiffness K 0 and damping C 0 .

Solution
I. The equation of equilibrium in the vertical direction is written as follows:

∂N dz ∂N dz
N (z) − + ρA(z)üz (z)dz = N (z) + =⇒
∂z 2 ∂z 2
∂N
− dz + ρA(z)üz (z)dz = 0 =⇒
∂z
∂N
− + ρA(z)üz (z) = 0
∂z

Introducing the constitutive law of the elastic material of the cone, we obtain:

∂uz
N (z) = EA(z)
∂z
Cone model for foundation impedance 53

The equation of equilibrium thus becomes:

 
∂ ∂uz
− EA(z) + ρA(z)üz (z) = 0 =⇒
∂z ∂z
−EA(z)u00z − EA0 (z)u0z + ρA(z)üz = 0 =⇒
EA(z)u00z + EA0 (z)u0z − ρA(z)üz = 0
The surface area of a horizontal section of the cone is given by the expression:
A(z) = π(z tan α)2
from which it is seen that:
A0 (z) = 2zπ tan2 α
and:
A0 (z) 2
=
A(z) z
It turns out that the equation of dynamic equilibrium in the vertical direction is written as:
2 1
u00z + u0z − 2 üz = 0
z cp
with the quantity cp representing the velocity of propagation of compression-extention
waves, which is equal to: s
E
cp =
ρ
The equation of equilibrium in the horizontal direction is similarly written as follows:
∂T dz ∂T dz
T (z) − + ρA(z)üz (z)dz = T (z) + =⇒
∂z 2 ∂z 2
∂T
− dz + ρA(z)üz (z)dz = 0 =⇒
∂z
∂T
− + ρA(z)üz (z) = 0
∂z
Introducing the constitutive law of the elastic material of the cone, we obtain:
∂uz
T (z) = GA(z)
∂z
The equation of equilibrium thus becomes:
 
∂ ∂uz
− GA(z) + ρA(z)üz (z) = 0 =⇒
∂z ∂z
−GA(z)u00z − GA0 (z)u0z + ρA(z)üz = 0 =⇒
GA(z)u00z + GA0 (z)u0z − ρA(z)üz = 0
It turns out that the equation of dynamic equilibrium in the vertical direction is written
as:
2 1
u00z + u0z − 2 üz = 0 (3.4)
z ct
with the quantity ct representing the velocity of propagation of compression-extention
waves, which is equal to: s
G
ct =
ρ
54 Exercises

II. The boundary condition applicable to the partial differential equation of dynamic equi-
librium in the horizontal direction will be provided by the force equilibrium at the surface of
the soil, in other words at z = z0 . The force equilibrium will be written as:

T (z0 ) = F (z0 ) =⇒
A(z0 )Gu0x = F (z0 )

In order to solve the partial differential equation of dynamic equilibrium in the horizontal
direction, we introduce the following change of variables:

Φ = zux

With this transformation equation eq:EQ1 becomes:

1
Φ00 − Φ̈ = 0
c2t

We reckognize the classical wave equation. Considering that we have a harmonic excita-
tion in the foundation, we will look for solutions of the wave equation written in the following
form:
iω
Φ(z, t) = Φ0 exp [(z − z0 ) + ct t]
ct
There will only be waves propagating in the −z direction. The displacement will thus be
written:  
Φ(z, t) Φ0 iω
ux (z, t) = = exp [(z − z0 ) + ct t]
z z ct
from which, it is obtained that:
   
Φ0 iω Φ0 iω
u0x (z, t) = − 2 exp [(z − z0 ) + ct t]
z ct z ct

The boundary condition at the soil surface is thus written:

F (z0 ) = T (z0 ) =⇒
F (z0 ) = GA(z0 )u0x (z0 , t) =⇒
 
2 Φ0 iω Φ0
F (z0 ) = GπR − 2 exp (iωt)
z ct z
R Φ0
At the soil surface we have z0 = − tan α . We can pose u0 = z0 . With these substitu-
tions, we obtain:
 
u0 u0 ωi
F (z0 ) = GπR2 − + exp (iωt) =⇒
z ct
 
ωRi
F (z0 ) = GπR tan α 1 + u0 exp (iωt) =⇒
ct tan α
F (z0 ) = K(ω)u0 exp (iωt)

in which, we have introduced the impedance of the foundation:

 
ωRi
K(ω) = GπR tan α 1 +
ct tan α
Cone model for foundation impedance 55

For ω → 0, the above equation must provide the static impedance of the foundation. This
is a condition permitting to determine the unknown angle α. We have:
GπR tan α = Kst =⇒
8GR
GπR tan α = =⇒
2−ν
8
tan α =
(2 − ν)π
With this value of α, the impedance will be written as follows:
 
ωRi (2 − ν)π
K(ω) = Kst 1 +
ct 8
It turns out that the soil can be replaced by a spring with stiffness coefficient equal to Kst
and a dashpot with damping coefficient equal to:
Kst R (2 − ν)π
C=
ct 8
We see that both the equivalent spring and the equivalent damping coefficient turn out to
be frequency-independent. Obviously, this is not a general result but holds only for the
particular examined case.

III. In the particular case α = 0, the infinite cone representing the soil domain, becomes
an infinite cylinder. The equation of dynamic equilibrium in the system becomes:
∂T
dz = ρAüx (z)dz =⇒
∂z
∂
(GAγ) − ρAüx (z) = 0 =⇒
∂z
Gu00x − ρüx = 0 =⇒
1
u00x − 2 üx = 0
ct
This is the classical wave equation, the solution of which for a harmonic excitation is:
 
iω
ux = u0 exp [(z − z0 ) + ct t]
ct
We can thus calculate the derivative:
 
0 iω iω
ux = u0 exp [(z − z0 ) + ct t]
ct ct
The boundary condition is written as in the previous case:
F (z0 ) = GAu0x (z0 ) =⇒
iω
F (z0 ) = GπR2 u0 exp (iωt)
c
From the above expression it is seen that the impedance is equal to:
iω
K(ω) = GπR2
ct
In other words, the soil is replaced by a dashpot only, whose damping coefficient is equal
to:
GπR2
C=
ct
56 Exercises

IV. The equation of dynamic equilibrium is written as:

M üx + K(Ω)ux = fx
with:
fx = mdΩ2 cos Ωt
M = M1 + M2
and  
ΩRi (2 − ν)π
K(Ω) = Kst 1+
ct 8
We will only retain the stationary response of the mat foundation. We can thus search
solutions of the above differential equation of the following form:
ux (t) = λ sin Ωt + µ cos Ωt
With substitution of the above equation into the ODE and term-by-term identification, we
find:
mdΩ2
λ = 0µ =
K(Ω) − M Ω2
The displacement is thus written as:
mdΩ2 fx
ux (t) = 2
cos Ωt =
K(Ω) − M Ω K(Ω) − M Ω2
Let us examine the response of the system for certain characteristic cases of Ω:
• For Ω → 0, we see that the quantity K(Ω) − M Ω2 → Kst . The displacement is thus:
fx
ux (t) =
Kst
and this is obviously, the quasi-static response of the system.
• For Ω → +∞ we see that the quantity K(Ω) − M Ω2 → −M Ω2 . The displacement is
thus:
fx md
ux (t) = − 2
=− cos Ωt
MΩ M
We obtain an displacement amplitude that does not depend on the frequency Ω,
but on the mass of the oscillator M . This is known as quasi-inertial response, be-
cause the elastic force is negligible with respect to the inertia force. Note, that in
the quasi-inertial response there is opposition of phase between the force and the
displacement.
Kst
• For Ω = M . This situation corresponds at a resonance state. The quantity:
π(2 − ν)ΩR
K(Ω) − M Ω2 = iKst
8ct

and the displacement is written as:

8ct
ux (t) = −ifx
Kst ΩRπ(2 − ν)
We see that there is a difference in phase equal to π2 . The displacement is retarded
π
2 with respect to the applied force.
Cone model for foundation impedance 57

M1 um

K′
C′
M2 uf

K f (ω )

C f (ω )

Figure 3.9: Isolation of rotating machine from mat foundation

V. If we isolate the rotating machine from the mat foundation, we obtain a system with two
degrees of freedom: horizontal displacement of the machine um and horizontal displace-
ment of the mat foundation uf . Denoting the real part of the foundation impedance as
Kf (ω) and the imaginary part of the foundation impedance as ωCf (ω), it is straightforward
to find that the equation of dynamic equilibrium of the system is:

Mü + Cu̇ + Ku = f
     
um  M1 0    C0 −C 0
with the following: u = 
 
, M = 

, C = 
 
 and K =

uf 0 M2 −C 0 C 0 + Cf
 
0 −K  0
K
 .
 
−K 0 K 0 + Kf
58 Exercises
Bibliography

G Gazetas and G Mylonakis. Seismic soil-structure interaction: new evidence and emerg-
ing issues. In Geotechnical Earthquake Engineering and Soil Dynamics, pages 1119–
1174. ASCE, 1998.

T Igusa and A Der Kiureghian. Generation of floor response spectra including oscillator-
structure interaction. Journal of Earthquake Engineering and Structural Dynamics, 13:
661–675, 1985.

E Kausel. Early history of soil-structure interaction. Soil Dynamics and Earthquake Engi-
neering, 30(9):822–832, 2010.

E Kausel, A Whitman, J Murray, and F Elsabee. The spring method for embedded founda-
tions. Nuclear Engineering and Design, 48, 1978.

J Lysmer. Analytical procedures in soil dynamics. In State of the art, ASCE Conference on
Soil Dynamics and Earthquake Engineering, Pasadena, CA (USA), 1978.

JM Roesset, RV Whitman, and R Dobry. Modal analysis for structures with foundation
interaction. Journal of Structural Engineering Division, ASCE, 99(ST3):389–416, 1973.

GG Stokes. On the dynamical theory of diffraction. Transactions of the Cambridge Philo-

sophical Society, 9:1–62, 1849.

JP Wolf. Dynamic Soil-structure interaction. Prentice Hall Inc., Englewood Cliffs, New
Jersey, 1985.

59