Sunteți pe pagina 1din 51

Universit degli Studi di Roma Tre

d i s Dipartimento di Strutture
Camillo Nuti, Ivo Vanzi

Influence of earthquake spatial


variability on the differential
displacements of soil and single
degree of freedom structures
RAPPORTO TECNICO N.1

Luglio 2004

Universit degli
Studi di Roma Tre

Dipartimento di
Strutture

Camillo Nuti, Ivo Vanzi

Influence of earthquake spatial


variability on the differential
displacement of soil and single
degree of freedom structures
RAPPORTO TECNICO N.1

Luglio 2004

Contents

FOREWORD............................................................................................................................................2
PREMESSA .............................................................................................................................................2

PART 1
INFLUENCE OF EARTHQUAKE SPATIAL VARIABILITY ON THE DIFFERENTIAL
DISPLACEMENTS OF SOIL AND SINGLE DEGREE OF FREEDOM STRUCTURES: MODEL
AND CODES PROVISIONS...................................................................................................................3
ABSTRACT .............................................................................................................................................4
1. INTRODUCTION AND SCOPE.........................................................................................................4
2. MODEL AND DEFINITION OF EARTHQUAKES THAT VARY IN SPACE ................................5
2.1 Spatial model of the earthquakes................................................................................................5
2.2 Numeric specification of the earthquake field ............................................................................6
2.3 Soil and structure differential displacements..............................................................................9
3. CODE PROVISIONS FOR DIFFERENTIAL DISPLACEMENTS..................................................13
4. HOW TO OBTAIN POWER SPECTRA COMPATIBLE WITH RESPONSE SPECTRA VIA
MODIFICATION OF THE KANAI-TAJIMI-CLOUGH-PENZIEN POWER SPECTRUM ...............17
5. CONCLUSIONS ................................................................................................................................21
6. REFERENCES...................................................................................................................................22
7. APPENDIX A: UNITS AND LIST OF SYMBOLS..........................................................................23

PART 2
INFLUENCE OF EARTHQUAKE SPATIAL VARIABILITY ON THE DIFFERENTIAL
DISPLACEMENTS OF SOIL AND SINGLE DEGREE OF FREEDOM STRUCTURES:
ANALYSES AND RESULTS ...............................................................................................................26
ABSTRACT ...........................................................................................................................................27
1. INTRODUCTION AND SCOPE.......................................................................................................27
2 ORGANIZATION OF ANALYSES...................................................................................................29
3. RESULTS FOR THE GROUND DIFFERENTIAL DISPLACEMENTS.........................................31
4. RESULTS FOR THE STRUCTURAL DIFFERENTIAL DISPLACEMENTS................................38
5. NUMERICAL EXAMPLE.................................................................................................................43
6. CONCLUSIONS ................................................................................................................................45
7. REFERENCES...................................................................................................................................46
8. APPENDIX A: SUMMARY OF FORMULAS TO COMPUTE THE ASYNCHROUS
EARTHQUAKE RANDOM FIELD AND THE DIFFERENTIAL DISPLACEMENTS .....................46
9. APPENDIX B: UNITS AND LIST OF SYMBOLS ..........................................................................48

Foreword
This report deals with the influence of earthquake spatial variability on the differential displacements
of soil and single degree of freedom structures. It contains the state of the art on the topic, the
description of the physical and mathematical model we have set up, and a detailed parametric analysis
of the model itself. The results of the analyses allow the formulation of simple design rules to account
for earthquake spatial variability.
The report is divided in two parts. The first one contains the state-of-the-art and the model of the
earthquake field; the second one contains the parametric analysis. At the end of the second part, we
have added a summary of the formulas to compute the asynchronous earthquake random field and the
soil and structure differential displacements.
Based on this model, we have developed a computer program to generate asynchronous earthquakes.
The program, named GAS, Generation of ASynchronous earthquakes, takes into account loss of
coherence and possible soil discontinuities. We foresee to publish the program and its users manual in
the next months, always within the DIS reports.

Premessa
Il rapporto ha per oggetto linfluenza della variabilit spaziale del sisma sulla risposta differenziale del
suolo e delle strutture ad un grado di libert. Contiene lo stato dellarte sullargomento, la descrizione
del modello fisico e matematico messo a punto, e una accurata analisi di sensibilit del modello stesso.
I risultati dellanalisi permettono di formulare regole progettuali semplici per tener conto della
variabilit spaziale del sisma.
Il rapporto si compone di due parti. La prima contiene lo stato dellarte e il modello del campo del
sisma; la seconda lanalisi di sensibilit. Alla fine della seconda parte abbiamo aggiunto un riassunto
delle formule che permettono il calcolo del campo di accelerogrammi asincroni e degli spostamenti
differenziali del suolo e della struttura.
A partire da questo modello, abbiamo sviluppato un programma di calcolo per la generazione dei
terremoti asincroni. Il programma, chiamato GAS, Generazione di terremoti Asincroni, tiene conto
della perdita di coerenza e di eventuali discontinuit di suolo. Prevediamo di pubblicare programma e
manuale nei prossimi mesi, sempre allinterno dei rapporti del DIS.

PART 1
INFLUENCE OF EARTHQUAKE SPATIAL
VARIABILITY ON THE DIFFERENTIAL
DISPLACEMENTS OF SOIL AND SINGLE DEGREE OF
FREEDOM STRUCTURES: MODEL AND CODES
PROVISIONS

ABSTRACT
With this study, we aim at developing simple design criteria for inclusion of the effects of spatially
varying ground motions in codes. We deal with the two simplest forms of this problem: differential ground
displacements and differential structural displacements for points and structures separated in space. The structures
considered are linear elastic single degree of freedom (s.d.o.f.) oscillators. Despite these problems may seem
trivial, some of the codes considered appear improvable on this aspect.
We first set up the mathematical model using basic random vibration theory; then we examine the code
provisions and have a first critical look at them.
Incidentally, in the development of the model, we have noticed that the classical modified Kanai - Tajimi
power spectrum representation of the action is incompatible with the shape of the displacement response spectra
given by the codes: for this reason we also propose a modification to its equation so as to make it capable of
representing the decreasing-constant shape of displacement response spectra at high periods.

1. INTRODUCTION AND SCOPE


Simultaneous recordings of earthquakes [1, 2] in different spatial points indicate that the
earthquake accelerations differ from one point to the other retaining a degree of similarity that can be
quantified in a statistical sense. This experimental observation has led researchers in the last twenty
years to develop accurate models to define the behavior. Departing from the classical work of Luco and
Wong [3] in the mid 80s, different statistical descriptions have been proposed and fitted to the
experimental data [1, 2, 4, 5], with varying degree of complexity and accuracy.
In parallel with the study of the ground behavior, the effects of the earthquake spatial
variability on structures have been the object of systematic investigations. Since random vibration
theory is a powerful tool for dealing with structural behavior under spatially varying earthquakes, it has
been extensively used for assessing the effects on linear structures. A comprehensive method for
response spectrum analyses has been for instance presented by Der Kiureghian and Neuenhofer [6, 7].
For non-linear structures, researchers have normally resorted either to numeric analyses or to
equivalent linearization procedures [8, 9]. The most important outcome of the studies [10, 11, 12], for
what the structural engineer is concerned with, has been unambiguous: apart from a few cases, the fact
that a structure is moved at its supports by non synchronous actions, decreases the stresses induced by
earthquakes as compared to the case with synchronous actions. So, one might wonder, why should a
structural engineer be interested in the effects of spatially varying earthquakes?
First of all, there are cases in which non-synchronism has negatively influenced structural
behavior. Failure of some bridges during the Loma Prieta earthquake is believed to be due to this effect
[10]. Furthermore one should consider that a frequent collapse mode for bridges is deck unseating
which is clearly governed by differential response. And finally, the simple rules provided by the codes
appear improvable. By this we mean that, at least for the problem of ground and statically determinate
structures differential displacements, designers should be able to quantify reliably the effects of nonsynchronism. This does not appear to be the case, for the codes rules usually neglect, implicitly or
explicitly, correlations between motions; further there are clearly cases, e.g. deck unseating, that should
be dealt with a higher safety margin (i.e. lower probability of exceedance), with respect to the other
failure modes, in that it is inherently fragile.
For the above reasons, in this report we concentrate on the two simplest effects of nonsynchronism, of interest to the structural engineer: ground and linear elastic statically determinate
structures differential displacements. In this report, we first develop the model and then examine the
European [14] and draft Italian [15] code provisions. In part 2, given the simplicity of the objects under
investigation, we are able to assess the relative importance of the different parameters governing their
behavior and to propose design rules simple and accurate enough to be adopted in the codes.
In the model development, we have come across an interesting problem: if the earthquake
action is specified in terms of response spectra, which is the case for all the codes, the classical
modified Kanai - Tajimi [16, 17] power spectrum representation of the action, which we need to
quantify the differential displacements, is incompatible with the displacement response spectrum at
high periods. To do away with this, we have corrected the modified Kanai Tajimi power spectrum
and hence, in this report, also propose a modification to its equation so as to make it capable of
representing the descending-constant shape of displacement response spectra at high periods.

2. MODEL AND DEFINITION OF EARTHQUAKES THAT VARY IN SPACE


2.1 Spatial model of the earthquakes
An earthquake acceleration recording at point P in space can be represented via its Fourier
expansion as a sum of sinusoids [5]:
AP ( t ) = BPK cos (K t ) + CPK sin (K t )

(2.1)

In equation (2.1), A is the measured acceleration in point P at time t, k is an index varying


from 1 to the number of circular frequencies k that we are considering, BPk and CPk are the amplitudes
of the k-th cosine and sine functions. Assume the acceleration AP(t) is produced by a wave moving
with velocity V towards a different point in space, say Q, at distance XPQ from P, within a medium
which transmits them without distortion. At point Q, and at time t, we would have:

AQ ( t ) = BPK cos K ( t PQ ) + CPK sin K ( t PQ )

PQ

cos ( )
=
= X PQ

v
V
app

(2.2)

X PQ

In equation (2.2) is the angle between the vector of surface wave propagation and the vector
that goes from P to Q, PQ is the time delay of the signal and vapp is the surface wave velocity. Actual
recordings at different points in space indicate that equation (2.2) is nearly correct. Since the medium
through which the waves travel does distort them, the recording in Q will resemble the one in P, the
more the shorter the distance between them, and will be written as:

AQ ( t ) = BQK cos K ( t PQ ) + CQK sin K ( t PQ )

(2.3)

BPk is correlated with BQk and CPk is correlated with CQk but the Bs and Cs are independent.
The latter is often referred to with the sentence phase angles are random. The statistical properties of
the amplitudes can then be summed up as in equation (2.4) which is saying that the amplitudes BPK and
CQK are statistically independent, for any points P and Q, and any circular frequency K, with the only
exception of BPK and BQK i.e. same circular frequency but different points in space. The same holds for
CPK and CQK.
BPK1 independent on CQK 2 P, Q, K1 , K 2
BPK1 independent on BQK2

P, Q, K1 , K 2 K1

CPK1 independent on CQK2

P, Q, K1 , K 2 K1

BPK correlated with BQK

P, Q, K

CPK correlated with CQK

P, Q, K

(2.4)

The amplitudes, which are themselves random variables, are usually assumed normally
distributed with zero mean and this assumption is experimentally verified [5]. In order to quantify the
acceleration time histories in different points in space, equations (2.1) and (2.3), all is needed is
definition of the correlation between amplitudes and of their dispersion, as measured by the variance
or, equivalently, of the covariance matrix of the amplitudes. Since the amplitudes at different circular
frequencies are independent on each other, and the same holds for the sine and cosine terms, we can
concentrate on a single cosine component from equations (2.1) and (2.3) for further discussion. One
such component, in points P1, P2, , Pl is:

AP1 ( t ) = BP1 cos ( t )

( (

AP2 ( t ) = BP2 cos t P1P2


...........

( (

APl ( t ) = BPl cos t P1Pl

))

(2.5)

))

where, for simplicity, the subscript k has been dropped. The amplitudes B at l different points are still
collectively normally distributed with mean values equal to zero and covariance matrix :
B N ( 0, )
l

exp bT 1 b =
2

= probability density function of the random variables B


l

pdf B ( b ) = ( 2 ) det ( )

1
2

(2.6)

PQ = covariance between BP and BQ = PQ PP QQ

and can be sampled by remembering that, for collectively normally random variables, conditional
distribution are normal. Amplitudes sampling, which allow to reconstruct the whole earthquake
accelerations random field, can then be done in a slightly different, less elegant but surely simpler,
fashion as reported by Vanmarcke [5]. The sampling scheme for B is:
Sample the value b1 of the random variable B1 at the first considered point P1 from a normal
distribution with mean equal to zero and variance equal to 11
At the second point P2, sample the value b2 of the random variable B2/b1 from a normal
distribution with mean equal to b1 12 22 11 and variance equal to 22 (1 12 2 )

At the i-th point, sample the value bi of the random variable Bi/ b1 b2 bi-1 from a normal

distribution with mean equal to iI II1 bi and variance equal to ( ii iI II1 Ii ) . I is an index
varying from 1 to i-1, iI is the matrix obtained taking the i-th row and the I-ths columns from
and b i is the column vector of the samples up to step i-1, i.e. bi =[b1,b2bi-1]T.
Continue up to point Pl
This process must be repeated for all the values of K, and for the sine and cosine terms, and the
different components summed up as in equation (2.3). Notice that the signal components for different
circular frequencies, once the amplitudes are sampled as described above, can be computed in an
efficient way by doing an inverse Fourier transform, possibly using the fast Fourier algorithm [18].
2.2 Numeric specification of the earthquake field
The amplitudes B and C are defined, from a statistical viewpoint, once the covariance matrix is
specified at each circular frequency . For earthquake fields, the covariance matrix is usually
assembled via independent definition of its diagonal terms, i.e. the variances in each space point and
frequency, and of the correlation coefficients.
The diagonal terms PP are quantified via a power spectrum. A popular choice is the Kanai-Tajimi
power spectrum, as modified by Clough and Penzien [2]:
GPP ( ) = G0

4f + 4 f2 4f 2
2
f

2 ) + 4 f2 4f 2
2

2
g

2 ) + 4 g2 g4 2
2

(2.7)

PP = GPP ( ) d

which we will adopt too, although with a slight modification. The spectrum in equation (2.7) will be
extensively discussed in the final section of part 1. The correlation coefficient between the amplitudes
is expressed via the coherency function, for which the form originally proposed by Uscinski [19] on
theoretical grounds and Luco [3] is retained:


= exp 2 X 2

(2.8)

One can see that the correlation decreases with increasing distance X and circular frequency and
increases with increasing soil mechanical and geometric properties as measured by v/, where is the
incoherence parameter and v is the shear wave velocity. Equation (2.8) has been extensively used [6, 7,
8, 20, 21, 22, 23,] though different functional forms have also been proposed.
We recall that v may be estimated as (Gsoil/soil)0.5 with Gsoil and soil respectively the ground
shear modulus and density and so the value of the incoherence parameter becomes the most difficult
aspect in the coherency function assessment. Zerva [22] and Luco and Mita [29] report for a value in
the range 0.1-0.5. Der Kiureghian and Neuenhofer [7] use for [X./v] a value within 0-2, which is in
line with what expected in the two previous papers. In order to shed some more light on this problem,
we have done some, purportedly rough, numerical comparison of equation (2.8) with some of the
experimental data, trying to keep different soil types within the study. We have actually found many
confirmations of the 0.1-0.5 range, but also some events that showed an even smaller lower bound. For
instance, while all the events recorded at Parkfield, California, and most events recorded in Taiwan on
the SMART array, as reported by [1], fall within the previous range, there are some which do not, e.g.
event 45 reported by Oliveira [2] in Taiwan which shows for a value of about 0.02. The soil in
Parkfield is rocky, while the one in Taiwan is alluvium. For the above reason, we have decided to keep
in the analyses a range for as large as 0.02-0.5.
What happens to the coherency model when there are soil discontinuities? This aspect has
been dealt with in a recent paper [24] but some further remarks seem convenient. With reference to the
following figure, one has to determine equivalent v/ and vapp for use in equations (2.8) and (2.2).

2 X

1 X

P
Soil
column

Soil type 1

Q
Soil type 2

Soil
column
earthquake
motion

earthquake
motion
BEDROCK

Figure 1: geometry for soil discontinuity

For vapp it is straightforward that the equivalent value can be obtained by doing a weighted average on
the distances:
vapp = 1 vapp1 + 2 vapp 2

1 + 2 = 1

(2.9)

Derivation of equivalent v/ is instead slightly trickier. With reference to Figure 1 we can recast (2.8)
in the form:


2
X PQ =
v

PQ = ( X PQ ) = exp K2


= exp K2 (1 X PQ + 2 X PQ )

(2.10)

(1 X PQ +2 X PQ )

and after rearranging terms we have:

PQ = P Q

(
) (

P = (1 X PQ )

1 (1 1 ) X PQ

Q = ( 2 X PQ

2 (1 2 ) X PQ

)
)

(2.11)

Equation (2.11), which is valid in the case of a homogeneous soil, provides a convenient way
to separate the loss of correlation along the distance XPQ.
P can be thought of as the loss of correlation due to the traveling of the waves from P to ,
Q from to Q. From equation (2.11), in the case of two soil types, say 1 and 2, the equivalent
properties of v/ can be obtained from equation (2.11) as:
2

= 1 1 + 2 2
v
v
v
1
2

(2.12)

For instance, with v1/1 =250 m/s, v2/2 =800 m/s, the values v/ of the equivalent soil, predicted by
equation (2.12), are shown in Figure 2, together with the mean value of v/.

Figure 2: values of equivalent v/ when there are discontinuities, with v1/1 =250 m/s, v2/2 =800 m/s

One can see that the soil with the worst mechanical characteristics is the one that commands
in the homogenization. For the case of more than one discontinuity let us refer to the geometry of
Figure 3.

1 X

2 X

1 X

X
Q

P
Soil
column

Soil
type
1

Soil
type
N

Soil
type
N-1

Soil
type
2

Soil
column

earthquake
motion

earthquake
motion
BEDROCK

Figure 3: geometry for several soil discontinuities

By the same token, one can prove that:


N

PQ = h

(2.13)

h =1

h = ( h X PQ )

h (1 h ) X PQ

and:

v
h =1 h

(2.14)

while for vapp it is again straightforward that:


N

vapp = h vapp _ h
h =1

h =1

(2.15)

=1

2.3 Soil and structure differential displacements


In this section we compute the differential response to a spatially varying earthquake defined
as in equation (2.1) and (2.3). The problem geometry is depicted in Figure 4.

Soil
column

M
Structure 1

Structure 2

Linear elastic
s.d.o.f. system

Linear elastic
s.d.o.f. system

Q
Soil
column
earthquake
motion

earthquake
motion
BEDROCK

Figure 4: geometry of the problem

The figure is made with reference to a typical application problem, a simply supported bridge
deck for which one has to compute the seating length. Structures 1 and 2 are different but both linear
elastic. The goal is to compute the differential response between P and Q, points on the soil, and
between L and M, located on top of single degree of freedom oscillators.
In order to do this we will use basic random vibration theory [25], using as inputs the
accelerations considered frequency component by frequency component (we will consider together sine
and cosine terms though). If we drop, for simplicity, the subscript k, one component of the acceleration
time history in Q can be written as:

AQ ( t ) = BQ cos ( t PQ ) + CQ sin ( t PQ )

(2.16)

We rewrite now the above equation in terms of a sum of complex numbers because this is
useful in the mathematical passages that follow. Since:
cos ( ) = 0.5 exp ( j ) + exp ( j )
sin ( ) = 0.5 j exp ( j ) exp ( j )

(2.17)

equation (2.16) can be rewritten as:


AQ ( t ) = 0.5 BQ exp ( j ) + exp ( j ) j 0.5 CQ exp ( j ) exp ( j )

= ( t PQ )

(2.18)

The total displacement in M, ZM(t) is the sum of the ground displacement ZQ(t) and of the s.d.o.f.
system displacement with respect to the ground ZMQ(t):

10

Z M (t ) = ZQ (t ) + Z MQ (t )
ZQ (t ) =

(2.19)

AQ ( t )

We also know that if structure 2, which has damping equal to and circular frequency equal
to , is subjected to the acceleration exp[j(t-PQ)], with j the imaginary unit, its displacement ZMQ(t)
relative to the ground is:
Z MQ (t ) = H ( ) exp( j t ) exp ( j PQ )
H ( ) =

1
2
2
( ) + j ( 2 2

(2.20)

Substituting equation (2.20) into (2.19), we can recast (2.19) in the form:

1
1
Z M (t ) = 0.5 BQ E H 2 + E * H * 2 +

1
1
j 0.5 CQ E H 2 E * H * 2 =

1
1
= AQ real E H 2 + BQ imag E H 2

E = exp ( j )

(2.21)

= ( t PQ )

where real(.) and imag(.) are respectively the real and imaginary parts of a complex number. Notice
that if we multiply the terms inside equation (2.21) by two coefficients n1 and n2 we can use a single
formula to obtain both the structural absolute and relative displacements and the ground displacement:

n
n
Z (Q , M , MQ ) ( t , n1 , n2 ) = AQ real E n1 H 22 + BQ imag E n1 H 22

(2.22)
Z M (t ) = Z (Q , M , MQ ) ( t , n1 = 1, n2 = 1)
Z Q (t ) = Z (Q , M , MQ ) ( t , n1 = 0, n2 = 1)
Z MQ (t ) = Z (Q , M , MQ ) ( t , n1 = 1, n2 = 0 )

The absolute and relative displacements in points M and Q, equation (2.22), can be finally written as:

Z (Q , M , MQ ) ( t , n1 , n2 ) = N Q BQ cos ( t PQ ) + Q + CQ sin ( t PQ ) + Q

after noting that:

11

(2.23)


n
E n1 H 22 = NQ exp j ( t PQ ) + Q

NQ =

n2 F2 2 + F32
F12 + F22

(2.24)

n2 F2
F2
atan
F1
F3

Q = atan

F1 = 2 2
F2 = 2
F3 = n1 2 + n2 F1

At this point we can formulate the differential response between points L and M or P and Q as:
Z * ( t , m1 , m2 , n1 , n2 ) = Z (Q ,M , MQ ) ( t , n1 , n2 ) Z ( P , L , PL ) ( t , m1 , m2 ) = W R
R = BP , CP , BQ , CQ

(2.25)

N P cos ( t + P )

N P sin ( t + P )
W =
N Q cos ( t PQ ) + Q

N Q sin ( t PQ ) + Q

(
(

)
)

Since Z* is a linear combination of zero-mean normal random variables, the amplitudes of the sine and
cosine terms within the vector R, it is still a zero-mean gaussian random variable. Its variance is equal
to:
var Z * ( t , m1 , m2 , n1 , n2 ) = W W T

GPP

0
=
PQ GPP GQQ

PQ GPP GQQ

GPP

GQQ

PQ GPP GQQ

PQ GPP GQQ

GQQ

(2.26)

Performing the matrix multiplication in (2.26) it can be shown that the variance of Z* is independent of
time and so, finally, we will have the following distribution:
Z * ( t , m1 , m2 , n1 , n2 ) = Z * ( m1 , m2 , n1 , n2 ) N ( = 0, Z2* )

Z2 = GZ K
*

2
2
GZ * K = PK
+ QK
PQK

(2.27)

2
2
PK
= N PK
GPPK
2
2
QK
= N QK
GQQK

PQK = 2 PK QK PQK cos ( PK QK K PQ )

Equation (2.27), together with the definitions in (2.24), allows to compute the variance of all the
differential displacements between P, Q, L and M, namely total displacements (m1=1, m2=1, n1=1,
n2=1), ground displacements (m1=0, m2=1, n1=0, n2=1), structural displacement with respect to ground

12

(m1=1, m2=0, n1=1, n2=0) or any mixed quantity, e.g. total displacement in P with respect to ground
displacement in Q (m1=1, m2=1, n1=1, n2=0).
Equation (2.27) gives the power spectrum of differential displacements. It is worth noticing
that the expression of GZ*K is saying that the correlation between the response in L and M is equal to the
correlation between amplitudes PQK times the cosine of the phase shift between signals. In fact, for any
two random variables, say x and y, the variance of their difference, say z=x-y, is equal to 2z=2x+2y2xyxy, with xy the correlation between x and y. By comparing this expression with equation (2.27)
one can see what has just been stated.
After the power spectrum of differential displacement is computed, one can find the
distribution of maxima using the peak factor formulation [26], by setting :
Z s*, p = Z * rs , p

(2.28)

where Z*s,p is the displacement value which is not exceeded with probability p during an earthquake of
duration s. Typical values of the peak factor rs,p lie within 1.20-3.5; rs,p is computed as set out in [26],
in which proper account is taken for the non-stationarity of the response via the use of the equivalent
damping:
s ( s ) = 1 exp ( 2 s ) = equivalent damping
GZ * ( s ) = power spectrum of Z * computed with s ( s )

{, } = {inertia radius, center} of the power spectrum GZ ( s )


*

rt =

G ( s ) d G ( 0.5 s ) d
Z*

(2.29)

Z*

rg = 1

s0 = s exp 2 ( rt 1)
rp =

s0
1

2 log ( p )

)}

rs , p = 2 log 2 rp 1 exp rg log ( 2 rp )

3. CODE PROVISIONS FOR DIFFERENTIAL DISPLACEMENTS


For the problem of the structural differential displacement considering non-synchronism,
codes allow, together with a refined analysis, a simplified one. In the simplified analysis, the codes deal
with non-synchronism as two separate problems, one concerning the ground differential displacement
only, e.g. between points P and Q in Figure 4, and the other the structural differential displacements
considering synchronous soil motion, e.g. between points L and M with nihil differential displacement
between P and Q. The outcomes from the two analyses are summed up disregarding their correlation,
yielding results that appear on the safe side (we are summing up the modules of correlated
displacements). This procedure is justified in view of the fact that the structural analyses are anyway
done for stress calculation and that codes have to service designers and hence be simple.
For a comparison with the codes, we have considered:
The European seismic code Ec8 [14], whose use has been, in some parts, including the effects of
spatially varying ground motions, recently recommended in Italy by the Civil Protection [27]. We
will refer to this code in the following with the word EC8/ICPC, meaning EuroCode 8 / Italian
Civil Protection Code.
The draft Italian seismic code for bridges [15]. We will refer to this code in the following with the
word DICB, meaning Draft Italian Code for Bridges. We anticipate that the code, for this part, has
been drafted following mainly the results from the present study
The codes, for the ground differential displacements, state they be computed as:

13

I
I _ MAX
= X PQ pga C
uPQ
uPQ

vapp 2

EC8/ICPC
2
I _ MAX
2
0.025 pga ( P TPC TPD ) + ( Q TQC TQD )
uPQ

(3.1)

q
II
II _ MAX

= pga q1 + q2 log ( X ) 3 uPQ


uPQ

DICB
2
2
II _ MAX
uPQ
1.25 0.025 pga ( P TPC TPD ) + ( Q TQC TQD )

with:
X PQ = distance between points P and Q

P = soil coefficient in point P


TPC , TPD

(3.2)

pga = peak ground acceleration


= periods defining the response spectra in point P
vapp =surface wave velocity

q1 , q2 , q3 = coefficients depending on the soils in P and Q

The numeric values of the quantities in equation (3.2) will be defined below. The differential
displacements of the two points of the structure, uLM, are computed with a response spectrum analysis,
with synchronous motion. This means that for two statically determinate oscillators, the complete
quadratic combination [28], CQC, rule will be used.

uLM = S d (TL ) + Sd (TM ) + S d (TL ) Sd (TM ) (LM


2

CQC )
(LM
=

CQC )

+ (ML

CQC )

8 2 (1 + rf LM ) rf LM 3 2

(1 rf )

2 2

LM

(3.3)

+ 4 2 (1 + rf LM ) rf LM
2

2 2
rf LM =

TL TM

Notice that for well spaced modes CQC gives the same results as the more popular square
root of sum of squares, SRSS, rule. The total differential displacements prescribed by the codes are
then:

{U

I
LM

{U

I
= uPQ
+ uLM } EC8/ICPC

II
LM

(3.4)

II
= uPQ
+ uLM } DICB

It is worth noticing that the DICB excplicitly permits to compute uLM also with the SRSS rule and that,
in case of fragile failures, e.g. deck unseating, advises to multiply the final results, be it the soil or
structural differential displacements, by 1.25.
The parameters in equation (3.1) and (3.4) are shown in Table 1 for the EC8/ICPC code and Table 2 for
the DICB. For the Italian draft code, we have added the soil description at the end of Table 1. For both
codes we are obviously referring to the ultimate limit state.
Table 1: value of the response spectra parameters for EC8/ICPC and properties of the soil types in the Draft Italian
Code for bridges (N.A.: not applicable)

parameter / ground type

description

rock
1
0.4
2
>800
3000

TC (sec)
TD (sec)
v (m/s)
vapp (m/s)

B
EC8/ICPC
gravel, sand, clay
1.25
0.5
2
360-800
2000

14

like B
1.25
0.5
2
180-360
2000

like B
1.35
0.8
2
<180
1500

B,C or D on A
1.25
0.5
2
100-360
1500

description
v (m/s)

DICB
gravel, sand, clay
180-360

rock
360-800

N.A.

B on A
100-180

Table 2: value of the parameters qi for the DICB (the dot . symbol indicates that the value must be computed by
symmetry)

Soil P/Q
Parameter
A
B
C

A
0
.
.

B
100.q1
0.7
0
.

3.3
2.0
0

17
.
.

B
105.q2
1.4
14.3
.

17
29
266

2.5
.
.

B
q3
3.9
2.9
.

C
2.9
2.9
1.1

The codes basically group the grounds into two homogeneous types, rock and loose and
cohesive soils (gravel, sand, clay) and name them respectively A and B, C, D in EC8 and A and B in the
draft code. They add another soil type for the case of soft soil on top of rock, and name it E in EC8 and
C in the draft code. When different soil types are present in P and Q, the EC8/ICPC states the most
unfavorable values of the parameters be used, whereas the DICB explicitly deals with this case, via the
values of the parameters q.
The acceleration and displacement spectra prescribed by the codes are respectively shown in
Figure 5, and Figure 6. The soil differential displacements, in natural and logarithmic scale, are shown
in Figure 7 and Figure 8

Figure 5: acceleration response spectra given by the codes

15

Figure 6: displacement response spectra given by the codes

Figure 7: soil differential displacement prescribed by the codes. Thicker lines are for EC8/ICPC

16

Figure 8: soil differential displacement prescribed by the codes. Thicker lines are for EC8/ICPC

In Figure 6 we have also highlighted the values of TE and TF, the periods that define the start
and the end of the decrease in the spectral displacement. We will refer to them in the next section.
From Figure 5 and Figure 6 one can see that soils A and D of EC8/ICPC coincide, as far as
response spectra are concerned, with soils A and C of the DICB while soils B, C and E of EC8/ICPC
are almost coincident with soil B of the DICB. In the analyses we will then consider only three ground
types: A, B and D as defined in EC8/ICPC, which are the most different situations encountered.
If we examine the differential displacements, the similarities between the codes get much
more blurred. While for the maxima differential displacements the difference is only the 1.25 factor in
the last of equations (3.1), the trend is completely different. EC8/ DICB increases linearly up to the
maximum, the DICB grows in a parabolic fashion. Besides, in the range of distances where most civil
engineering structures are, between 5 and 100 m, from buildings columns to long bridges piers, the
differences are very large: for instance, at 20 m distance, EC8/ICPC formulas give 2 mm or less while
DICB forecasts differential displacements from 2 mm to about 40 mm, depending on the soils
coupling. Such large differences, which are obviously crucial in engineering applications, must be
carefully investigated.
The systematic investigation of the dependence of differential displacements on the governing
variables is the object of part 2: we anticipate that the outcome of the study is that EC8/ICPC is always
unconservative, with very strong underestimation exactly in the 5-100 m range.

4. HOW TO OBTAIN POWER SPECTRA COMPATIBLE WITH RESPONSE SPECTRA VIA


MODIFICATION OF THE KANAI-TAJIMI-CLOUGH-PENZIEN POWER SPECTRUM
If we want to use the formulation in equation (2.28) to compute the differential displacements,
we need the action to be specified in terms of power spectrum. For several reasons, namely for
comparison of our results with those obtained by other researchers and to do away with unforeseen
numerical difficulties, it is convenient to use a power spectrum defined in closed-form and which has
been already adopted in similar studies: the natural choice would then certainly be the modified KanaiTajimi [16, 17].
For a direct code comparison, however, we have to start from the earthquake specification in
terms of response spectra: we then need to go from one representation to the other. We have done this
in two steps: first we have computed a numerically defined (i.e. no closed-form expression) power
spectrum G(num)() corresponding to the displacement and acceleration response spectra, and then we

17

have fitted the parameters of the modified Kanai-Tajimi power spectrum, equation (2.7), taking
G(num)() as target.
For the first step, we have used the procedure outlined in [7] and summed up in
equation (4.1):
G ( ) ( ) = G (
i

i 1)

S d ( )

( i 1)
S d ( )

( )

= circular frequency
G

(i )

( ) = ground acceleration power spectrum at i-th step and at


S d ( ) = target displacement response spectrum at

(4.1)

= s.d.o.f. system natural circular frequency


S d(i ) ( ) = displacement response spectrum at computed from G (i ) ( ) =
+

()
G ( ) H ( )

= rs , p

H() is the frequency response function in equation (2.20) while rs,p is the Vanmarcke peak
factor in equation (2.29). The trial and error procedure tries to converge to G(num)(), value of G(i)() at
the last step, starting with a guess value for the power spectrum G(0)(), then computing the maximum
response to the power spectrum of the s.d.o.f.s systems described by H(), and finally correcting the
trial power spectrum, frequency by frequency, via comparison of the target and obtained displacement
response spectra. Usually, a small number of iterations, about 3-5, are sufficient to converge at a stable
result. Der Kiureghian [7] also gives suggestion for the selection of the guess value for the power
spectrum at the first iteration G(0)() but the procedure is little dependent on this - one may also start
from a constant power spectrum. It is worth noticing that because of numerical instability, the target
acceleration/displacement response spectra in equation (4.1) must be conveniently smoothed before
use. The procedure converges very well and the error between the target displacement/acceleration
response spectra and the displacement/acceleration response spectra computed from the numeric power
spectrum G(num)() is disregardable.
For step two, let us first recall the birth of the Kanai-Tajimi power spectrum. The acceleration
spectrum was originally introduced in 1957-1960 by Kanai [16] and Tajimi [17] in the following form:
( )
= G0
GKT
A

4f + 4 f2 4f 2
2
f

2 2

+ 4
2
f

4
f

(4.2)
2

Equation (4.2) is a gaussian white noise with constant value equal to G0 passed through a linear filter,
representing the ground, with circular frequency f and damping f. The problem with this
representation is that the displacement power spectrum, obtained dividing the acceleration power
spectrum by 4, tend to + with 0 and that the integral of the displacement power spectrum, i.e.
the variance of displacements, is infinite too. This is not physically acceptable. Clough and Penzien
[18] introduced a second filter to do away with this and proposed the following widely adopted form
for the accelerations power spectrum:
(A)
(A)
(A)
GKT-CP
= GKT
CP
(A)
CP
=

2
g

(4.3)

2 ) + 4 g2 g4 2
2

Equation (4.3) is the so-called modified Kanai-Tajimi spectrum and its parameters are the scale factor
G0, the central frequencies of the filters, f and g, and their damping, f and g. We will refer to this
spectrum as the Kanai-Tajimi-Clough-Penzien (KT-CP) in what follows.
We tried to have equation (4.3) converge to the numeric power spectrum G(num)() via a
standard minimum error procedure. After several iterations, we checked the match between the
obtained KT-CP and the numeric spectrum and found that everything was all right in terms of power

18

spectra of accelerations, but the same could not be said for the power spectra of displacements. In fact,
while the KT-CP power spectrum tends to G0/g4 with 0, the target numeric displacement power
spectrum G(num)() tends to 0 in that range, see Figure 9.

Figure 9: comparison between displacement power spectra


More precisely, if:

e =

2
e

(4.4)

2
f =
f

with TE and TF defined in Figure 6, then:


G num ( ) = 0
G

num

( ) = GKT CP ( )

0, f

(4.5)

and for values between f and e the numeric spectrum has an intermediate behavior. For this
reason we used and propose a further correction to the Kanai-Tajimi-Clough-Penzien power spectrum
as follows:
( )
( )
( )
Gmodified-KT-CP
= GKT
CP
0(
A

(A)

A)

2 E F
= 0.5 + atan

E F

(4.6)

The reason why the inverse tangent function has been chosen is not theoretical but simply that
the function has two asymptotes, is smooth and is anti-symmetric .The correction factor 0(A) is plotted
in Figure 10 for F=1.5, E=2.5, =10. is a factor that controls the speed with which the inverse
tangent function goes to the asymptotic values 0 and 1 and should be equal or higher than about 10.

19

Figure 10: correction factor 0(A)

The correction in equation (4.6) has appeared to work well on all soils tested. An example
final result in terms of target and obtained displacement power spectra is shown in Figure 9.
The target response spectra, together with those obtained from the KT-CP and modified KTCP power spectra, are shown in Figure 11 (acceleration) and Figure 12 (displacement):

Figure 11: comparison among acceleration response spectrum

20

Figure 12: comparison among displacement response spectrum


From Figure 11 (acceleration) and Figure 12 (displacement) one can see that both the KT-CP
and the modified KT-CP match the target acceleration response spectrum but only the modified KT-CP
matches the target displacement response spectrum, at least for high periods.
For completeness, in Table 3 we show the values usually assumed for the Kanai-TajimiClough-Penzien in the different soil conditions [26] and the values that we have computed for the
modified Kanai-Tajimi-Clough-Penzien in the soil conditions corresponding to soils A, B and D of the
EC8 [14]. The constants g1 and g2 are needed to compute the scale factor G0 as a function of the
earthquake duration s as:
G0 =

pga 2
g1 log ( s ) + g 2

(4.7)

Table 3: parameters of the Kanai-Tajimi-Clough-Penzien and modified Kanai-Tajimi-Clough-Penzien


power spectra for different soils (N.A.: not applicable)
Soil / Parameter
Firm
Medium
Soft
A of EC8
B of EC8
D of EC8

g1
g (%)
f
f (%)
Kanai-Tajimi-Clough-Penzien, equation (4.3)
1.5
60
15
60
184
1
60
10
40
125.5
0.5
60
5
20
90.2
Modified Kanai-Tajimi-Clough-Penzien, equation (4.6)
2.8
97
23
43
142.8
2.8
94
18.5
53
65.5
2.8
87
11.1
61
30

g2

557.2
286.3
95.8

N.A.
N.A.
N.A.

225.8
98.2
39.3

10.0
10.0
10.0

5. CONCLUSIONS
Based on well known expressions for spatial variability of seismic motion, we have developed
a theoretical model, founded on basic random vibration theory, to compute the differential
displacements of points on the grounds and on top of s.d.o.f. linear elastic system. The model has been
specialized for the seismic input specified by Eurocode 8 [14], Italian code set up by Civil Protection

21

[27] and new draft Code Provision for bridges [15]. We have explicitly included the case of contiguous
different soils, finding that the expression of the coherency function governing the problem is the same
as for the case of homogeneous soil. We give the expression of the equivalent parameter in equations
(2.13), (2.14), (2.15) (see also Figure 1 and Figure 3). The expression for the coherence obtained is
clear, simple to use and solve many of the uncertainties in the definition of the random fields for
differential displacements of soil.
With the exception of the Italian code for Bridges [15], which has been drafted following
mainly the results from this study, it is shown that codes seem improvable on this aspect, both from the
qualitative and quantitative viewpoint. This is surprising since the theoretical model for relative
displacements is rather straightforward to set-up, though it requires some mathematics in the
developments. It is however a matter of fact that long structures suffer differential displacements that
are the main cause of spectacular failures as the serial unseating of simply supported bridges deck.
We have given the correct closed form equations of relative displacement between statically
determined structures, obtaining a unified formulation for the statistic of relative displacement between
the top of two structures, top of one structure and base of the other, base of two structures.
The solution is given in terms of the well known formula: the response is the standard
deviation of the process (of the quantity in question: i.e. relative top to top displacements, base to base
displacements etc.) times a peak response factor which is a function of the probability of exceedance
(see equations (2.28) and (2.29)). In general the median value is considered (50 % probability of
exceedance). In part 2 a simpler and very general formulation will be proposed to pass from the median
to higher fractiles.
A further result, as we have shown in section 4, is that the widely adopted Kanai-TajimiClough Penzien power spectrum must be modified in order to reproduce the displacements at large
periods required by code displacement response spectra.
By now we have prepared all the steps needed to do a sensitivity analysis of the differential
displacements with the hope of being able to control the extremely large number of variables
influencing it. A further difficulty is introduced by the fact that the parameters of the coherency models
seem very disperse and the data collected up to today seem inadequate to draw a precise judgment on
their value.
In part 2 we will do these steps and will see that, due to some fortunate regularities in the
differential response, we will be able both to draw final conclusions and to mathematically describe the
differential response in closed form as a function of almost all the governing variables.

6. REFERENCES
1. Abrahamson, N. A., Schneider, J. F., Stepp, J. C., Empirical spatial coherency functions for
application to soil-structure interaction analyses, Earthquake Spectra, 7, 1, Feb. 1991, pages 1-27
2. Oliveira, C. S., Hao, H., Penzien, J., Ground motion modeling for multiple-input structural
analysis, Structural Safety, 10, 1-3, May 1991, pages 79-93
3. Luco, J. E., Wong, H. L., Response of a rigid foundation to a spatially random ground motion,
Earthquake Engineering & Structural Dynamics, 14, 6, Nov.-Dec. 1986, pages 891-908
4. Santa-Cruz, S., Heredia-Zavoni, E., Harichandran, R. S., Low-frequency behavior of coherency for
strong ground motions in Mexico City and Japan, 12th World Conference on Earthquake
Engineering, New Zealand Society for Earthquake Engineering, Upper Hutt, New Zealand, 2000,
Paper No. 0076
5. Vanmarcke, E. H., Fenton, G. A., Conditioned simulation of local fields of earthquake ground
motion, Structural Safety, 10, 1-3, May 1991, pages 247-264
6. Der Kiureghian, A., Neuenhofer, A., Response spectrum method for multi-support seismic
excitations, Earthquake Engineering & Structural Dynamics, 21, 8, Aug. 1992, pages 713-740
7. Der Kiureghian, A., Neuenhofer, A., A response spectrum method for multiple-support seismic
excitations, UCB/EERC-91/08, Berkeley: Earthquake Engineering Research Center, University of
California, Aug. 1991, 66 pages
8. Monti, G., Nuti, C., Pinto, P.E., Vanzi, I., Effects of non Synchronous Seismic Input on the
Inelastic Response of Bridges, II international workshop on seismic design of bridges,
Queenstown, New Zeland, 1994
9. Monti, G., Nuti, C., Pinto, P. E., Nonlinear response of bridges under multisupport excitation,
Journal of Structural Engineering, 122, 10, Oct. 1996, pages 1147-1159
10. Hao, H., A parametric study of the required seating length for bridge decks during earthquake,
Earthquake Engineering & Structural Dynamics, 27, 1, Jan. 1998, pages 91-103

22

11. Harichandran, R. S., Hawwari, A., Sweidan, B. N., Response of long-span bridges to spatially
varying ground motion, Journal of Structural Engineering, 122, 5, May 1996, pages 476-484
12. Sextos, A.G., Pitilakis K.D., Kappos A. J., Inelastic dynamic analysis of RC bridges accounting
for spatial variability of ground motion, site effects and soil-structure interaction phenomena. Part
1: Methodology and analytical tools. Part 2: Parametric study, Earthquake Engineering &
Structural Dynamics, 32, 4, Apr. 2003, pages 607-652
13. Housner, G. et al., Competing against time, (Ed. C.C. Thiel jr.), Report to Governor G.
Deukmejian from the Governors Board of Inquiry on the 1989 Loma Prieta earthquake, Dep. Of
General Services, State of California, North Highlands, Ca, 1990
14. Comit Europen de Normalisation, CEN, Eurocode 8: design of structures for earthquake
resistance, Draft n. 2, doc cen/tc250/sc8/n320, May 2002
15. Consiglio Superiore dei Lavori Pubblici, Aggiornamento delle Norme Tecniche per le Costruzioni
in zone sismiche, draft version of October 2003, 2003 (in Italian, available at
http://host.uniroma3.it/master/mica/mica.asp
>
normativa
and
at
http://www.
infrastrutturetrasporti.it/main/facciamo/consup/consup.html > normativa)
16. Kanai, K., Semi-empirical formula for the seismic characteristics of the ground, University of
Tokio Bulletin of Earthquake Research Institute, vol. 35, pp. 309-325, 1957
17. Tajimi, H., A statistical method of determining the maximum response of a building structure
during an earthquake, Proceedings of the Second World Conference on Earthquake Engineering,
Tokyo, Vol. II, 1960, pages 781-797
18. Clough, R. W., Penzien, J., Dynamics of structures, McGraw-Hill, Inc., New York, 1975, 634
pages
19. Uscinski, B. J., The elements of wave propagation in random media, Mc Graw-Hill, New York,
1977
20. Zerva, A., Effect of spatial variability and propagation of seismic ground motions on the response
of multiply supported structures, Probabilistic Engineering Mechanics, 6, 3-4, Sept.-Dec. 1991,
pages 212-221
21. Zerva, A., Seismic ground motion simulations from a class of spatial variability models,
Earthquake Engineering & Structural Dynamics, 21, 4, Apr. 1992, pages 351-361
22. Zerva, A., Response of multi-span beams to spatially incoherent seismic ground motions,
Earthquake Engineering & Structural Dynamics, 19, 6, Aug. 1990, pages 819-832
23. Zerva, A., Development of differential response spectra from spacial variability models,
Proceedings of the Tenth World Conference on Earthquake Engineering, A. A. Balkema,
Rotterdam, Vol. 9, 1992, pages 5469-5474
24. Lupoi, A., Franchin, P., Monti, G., Pinto., P.E., Relevance of spatial variability of ground motion
on seismic design of bridges, Concrete Structures in Seismic Regions: FIB 2003 Symposium,
Athens, Greece, 2003, 12 pages
25. Crandall, S.H., Mark, W.D., Random vibration in mechanical systems, Academic Press, New
York, 1973, 1963, 166 pages
26. Vanmarcke, E.H., Fenton, G.A., Heredia-Zavoni, E., SIMQKE-II, conditioned earthquake ground
motion simulator : user's manual, version 2.1, Princeton University, [Princeton, N.J.], 1999, 25
pages
27. Presidenza del Consiglio dei Ministri, Primi elementi in materia di criteri generali per la
classificazione sismica del territorio nazionale e di normative tecniche per le costruzioni in zona
sismica, Ordinanza, March 2003, (in Italian).
28. Wilson, E.L.; Der Kiureghian, A.; Bayo, E., A replacement for the SRSS method in seismic
analysis, Earthquake Engineering & Structural Dynamics, 9, pp. 187-192, 1981
29. Luco, J. E., Mita, A., Response of circular foundation to spatially random ground motion, Journal
of Engineering Mechanics, 113, 1, Jan. 1987, pages 1-15

7. APPENDIX A: UNITS AND LIST OF SYMBOLS


Unless stated otherwise, all units belong to the MKS (meter, mass kilogram, second) system.
Log is the natural (base e) logarithm.
Symbol
bPK
cPK
f

Meaning
A sampled value of BPK
A sampled value of CPK
Factor to compute the fractiles of the response as a function of the exceedance probability
1-p

23

g
h
i
j
k
l
m1, m2
n1, n2
p
r
s
t
u
v
vapp
AP(t)
BPK
CPK
D
E
G
Gsoil
H
I
L
M
N
O
P
Q
R
S
T
U
V
W
XPQ
ZP, ZMQ
Z*

,f,g,

PQ
soil

PQ

k
i

Constants in Table 3
Index of the soil type
Index of one of the l points on the ground surface
Imaginary unit
Index of the ground acceleration and displacements circular frequencies
Number of points on the ground surface
Boolean variables
Boolean variables
Probability
Peak factor
Earthquake duration
time
Code displacements
Shear wave velocity
Surface wave velocity
Ground acceleration recorded at time t in point P
(A is also a ground type, when stated explicitly)
Amplitude of the k-th cosine component of the ground acceleration in point P
(B is also a ground type, when stated explicitly)
Amplitude of the k-th sine component of the ground acceleration in point P
(C is also a ground type, when stated explicitly)
Ground type
Dummy variable
(also a ground type, when stated explicitly)
Power spectrum
Ground shear modulus
Transfer function
Index varying from 1 to i-1
Point located on top of a s.d.o.f. structure
Point located on top of a s.d.o.f. structure
Modulus of the transfer function from ground acceleration to structural total displacements
Index of the soil type
Point on the ground surface
Point on the ground surface
Dummy variable
Response spectrum
Period
Code displacements
Velocity
Dummy variable
Distance between points P and Q
Displacement of point P, between points M and Q
Total and relative displacements among points P, Q, L, M
Incoherency factor
Damping of the structure and in the Kanai-Tajimi power spectrum
Parameter in the proposed correction function applied to the Kanai-Tajimi power spectrum
Soil coefficient
Angle of the transfer function from ground acceleration to structural total displacements
Correction functions applied to the Kanai-Tajimi power spectrum
Correlation coefficient between BP and BQ and CP and CQ
Soil density
Standard deviation
Difference in arrival time of the earthquake waves between points P and Q
Dummy variable
Structural circular frequency
k-th circular frequency of the ground acceleration
Non dimensional distance
Angle between the vector of surface wave propagation and the vector that goes from P to Q
Inertia radius

24

Center of mass
Point on the ground surface where a soil discontinuity is located
Covariance matrix of B and C at the l points on the ground surface
Modal correlation in the complete quadratic combination, CQC, formula

25

PART 2
INFLUENCE OF EARTHQUAKE SPATIAL
VARIABILITY ON THE DIFFERENTIAL
DISPLACEMENTS OF SOIL AND SINGLE DEGREE OF
FREEDOM STRUCTURES: ANALYSES AND RESULTS

26

ABSTRACT
We aim at quantifying the ground and statically determined structures differential displacements caused
by earthquakes that vary in space. Our final goal is to propose an interpretation for the problem simple enough to
be used in the design codes.
We depart from the theoretical model set up in part 1 and first assess the sensitivity of ground
differential response to the many parameters influencing it; then we repeat the same process for the structural
response and are thus finally able to build up interpretations of results, one for the ground and the other for the
structures, simple enough to be used as design rules.
Comparison with the European, Italian Civil Protection and draft Italian bridges codes provisions show
that the first two may be improved on this aspect; and, for this reason, the Italian draft code for bridges has been
mainly drafted following the results of this study.

1. INTRODUCTION AND SCOPE


In part 1 we have set up the model to compute the differential displacements response of
points located on the ground and on top of linear elastic single degree of freedom (s.d.o.f.) structures,
due to spatially varying earthquakes. The spatial variability has been modelled following Vanmarcke
[1], for what concerns physical interpretation of the random field, coupled with the traditional Luco [2]
coherency function and a new, corrected, form of the Kanai-Tajimi-Clough-Penzien power spectrum
[3, 4, 5]. The final model reads:
AP ( t ) = BPK cos (K t ) + CPK sin (K t )
k

AQ ( t ) = BQK cos K ( t PQ ) + CQK sin K ( t PQ )

k
cos ( )
= X PQ
vapp
V

2
2

2

PQk = exp k X PQ

Gmodified-KT-CP = GKT-CP 0

PQ =

GKT-CP = G0

X PQ

4f + 4 f2 4f 2

2
f

2 2

+ 4 f2 4f 2

0 = 0.5 +

(1.1)

2
g

2 2

+ 4 g2 g4 2

2 E F
atan

E F

In equation (1.1) A is the measured acceleration in points P and Q at time t, k is an index varying from
1 to the number of circular frequencies k that we are considering, BPk and CPk are the amplitudes of
the k-th cosine and sine functions, XPQ is the distance from P to Q, vapp is the surface wave velocity,
is the angle between the vector of surface wave propagation and the vector that goes from P to Q, v is
the shear wave velocity, is the incoherency parameter, G0, f, g, f, g,E, F, are the parameters
of the modified Kanai-Tajimi-Clough-Penzien power spectrum. The terms in equation (1.1) are
discussed in part 1.
We have also determined the expression of the fractiles of the maximum differential
displacements between two points located on the ground surface, P and Q, or on top of linear elastic
s.d.o.f. structures, L and M, with circular frequencies and damping respectively equal to P, P, Q, Q,
as:

27

Z s*, p = Z * rs , p
2
2
Z2 = PK
+ QK
PQK
*

2
PK

=N

2
PK

2
2
GPPK ; QK
= N QK
GQQK

PQK = 2 PK QK PQK cos ( PK QK K PQ )

NP =

m F
F
m2 F2 P 2 + F32P
; P = atan 2 2 P atan 2 P
F12P + F22P
F
F1P
3P

F1P = P2 2 ; F2 P = 2 P P ; F3 P = m1 P2 + m2 F1P

NQ =

n2 F2Q 2 + F32Q
F +F
2
1Q

2
2Q

n F
F
; Q = atan 2 2Q atan 2Q
F3Q
F1Q

(1.2)

F1Q = ; F2Q = 2 Q Q ; F3Q = n1 + n2 F1Q


2
Q

2
Q

where rsp is the Vanmarcke peak-factor [6]. The problem geometry is depicted in Figure 1.

M
X

Structure 1

Soil
column

2 X

1 X

Linear elastic
s.d.o.f. system

Structure 2
Linear elastic
s.d.o.f. system

Q
Soil
column
earthquake
motion

earthquake
motion
BEDROCK

Figure 1: geometry of the problem

We have also examined what the European [7], Italian Civil Protection [8] and draft Italian [9]
code provisions are for this problem. We will indicate the three codes respectively with the acronyms
EC8, ICPC, DICB from now on.
We now want to assess the sensitivity of differential displacements computed with equation
(1.2) based on the model in (1.1) to the different variables. We will look at both the ground and the
structural displacements. We will finally compare the correct results, meaning by this word what we
obtain from the theoretical model, with what prescribed by the codes to finally present code rules on
which the draft Italian code provisions for bridges are based.

28

2 ORGANIZATION OF ANALYSES
We have first investigated the maxima differential displacements of two points P and Q on the
ground and then on top of single degree of freedom oscillators, points L and M. The excitation is
specified via its acceleration/displacement response spectra. The maxima differential displacements are
expressed by equation (1.2). The number of variables governing the differential responses is large and
we examine them in what follows.
First, for what the excitation is concerned, we have retained throughout the analyses a value of
1 m/s2 for the peak ground acceleration, pga. The structural damping has been taken equal to 5%. The
response spectra shape has been taken, for each ground type, as specified by the EC8 [7]. We have
considered only three ground types, A, B and D, as specified in EC8; these soils are the most different
situations encountered. The ground types in the EC8 are equal to those of the ICPC. They also
essentially correspond to A, B and C of the DICB [9]: in fact, soils A and C of the DICB are equal to
soils A and B, C, E of the EC8 while the differences between soil C of the DICB and soil D of EC8 are
disregardable from an engineering viewpoint, see Figure 5 and Figure 6 of part 1. We have then
assumed that each of the points P and Q could be in each ground type. It is worth noticing that the
spectra specified in the codes are at 50% probability of exceedance, i.e. that one in two earthquakes
produces higher effects.
However, we need the excitation specified in terms of power spectra, not response spectra. We
have computed the power spectra corresponding to each response spectrum using the procedure
outlined in the last section of part 1 and that has led us to formulate the correction factor 0 for the
modified Kanai-Tajimi-Clough-Penzien power spectrum, see Table 3 of part 1.
With the earthquake power spectra known, we have computed the differential ground
response, points P and Q, keeping as parameters the soil types in P and Q, the value of the
incoeherence factor , the surface wave velocity vapp (expressed as a parameter times the surface
wave velocity specified by the EC8 code for each ground type, i.e. vapp =.vapp_EC8) and the distance
between the points. In case the soils in P and Q are of the same type, there is no ambiguity in choosing
the parameters v and vapp. In case there are two different soil types below P (soil type 1) and Q (soil
type 2), we have used the homogenization formulas justified in part 1:
vapp = 1 vapp1 + 2 vapp 2
2

= 1 1 + 2 2
v
v
v
1
2

1 + 2 = 1

(2.1)

with a value for 1 of 0.5, i.e. the soil discontinuity is at the same distance from P and from Q, see
Figure 1.
The duration of the the earthquake has been assumed equal to 25 seconds and the comparisons
with the codes has been made by keeping the probability value at 0.5, since the spectra specified in the
codes are at that probability level. However we have also investigated the dependence of the maxima
differential displacements on the probability level. The organization of analyses for the ground
differential displacement is shown in Figure 2

29

SOIL TYPES
A

A
DISTANCE
SOIL TYPES

Results @
probability=P1

DISTANCE
Results @
probability=PN

Figure 2: organization of analyses for the assessment of soil differential displacements

For each soil coupling, the circle in Figure 2, we have obtained the differential displacements
as a function of the distance between points, and the values of and , (a cube of results in Figure
2); finally we have seen the effect of the probability level (the different cubes in Figure 2).
When we have examined the differential displacements on top of the structures, for the case of
linear elastic oscillators, two more parameters must be considered, the modal periods of each structure.
Since, as we will see, the influence of and will have been assessed by then, and, on the other side,
we need to keep the variables within a reasonable number, in the structural analyses we will assume
=0.5 and =1.
The values of the variables used for the assessment of differential displacements are summed
up in Table 1 and the values of the shear and apparent wave velocities, v and vapp, are in Table 2.
Table 1: values of the variables used for the assessment of differential displacements
#

Variable
Values assigned to variable
Variables used to assess the differential ground displacement
1
Soil type in P
A, B and D as defined by EC8 (1)
2
Soil type in Q
A, B and D as defined by EC8 (1)
3
Distance between P and Q
From 0 to 10,000 m
4
From 0.02 to 0.5
Incoherency parameter
0.5, 1, 1.5
such that vapp = vapp_EC8
5
6
Probability levels
From 0.01 to 0.99
Further variables used to assess the differential structural displacement
(only for the case of =0.5 and =1)
7
Period of the first structure TP
From 0.2 to 2
8
Period of the second structure TQ
From 0.2 to 2
(1)
A, B and D as defined by EC8 are equal to A, B and D of ICPC and A, B, C of DICB.
Table 2: values of the shear and apparent wave velocities, v and vapp, assumed in the analysis
parameter / ground type
description
v (m/s)
vapp (m/s)

A
rock
800
3000

B
gravel, sand, clay
580
2000

30

D
gravel, sand, clay
90
1500

The number of governing variables is large, larger than one could reasonably think to be able
to control. However, we will see that some fortunate regularity in the results will permit simplifications
and thus make the whole analysis easy to interpret.

3. RESULTS FOR THE GROUND DIFFERENTIAL DISPLACEMENTS


For each soil coupling, the parameters governing the differential displacements of P and Q are
four: the distance between P and Q, XPQ, the probability level, and two soil parameters, the
incoherence factor and the apparent surface wave velocity which is proportional to . Experimental
measurements indicate that the latter two are extremely dispersed, even within homogeneous soils. On
the other hand, it is intuitive after examination of equation (1.1), that is a more critical variable than
.
For this reason we have first of all fixed the probability level at 50%, at 1, and have
examined the results keeping only the distance between P and Q, XPQ, and the incoherence factor as
variables. The results are presented as level curves of the maximum differential displacement, with XPQ
on the x-axis and v/ on the y-axis, in doubly logarithmic scales. The horizontal lines at =0.02 and
=0.5, the lower and upper bounds assumed in this analysis, are highlighted in each diagram.
Let us first examine one of the results, the one for soil coupling D-D, shown in Figure 3

O4

O3

O1

O2

Figure 3: level curves of differential ground displacement in (m.10-2) for soil coupling D-D

31

For an easier understanding of the figure, let us momentarily rename x, y and z the quantities
on the axes, i.e. respectively the distance, v/ and the differential displacements. The level curves
shown are projection of the surface z(x,y) on the (x,y) plane, at constant values of z, exactly as it is done
on a topographic or marine chart. Let us concentrate on an important characteristic of this diagram: (i),
from bottom left to top center, the level curves are first oriented at 45 degrees (in doubly logarithmic
scales) with respect to the x-axis and then they get vertical, through a transition branch. Imagine now
that we cut the surface z(x,y) with a plane parallel to the (x,z) plane, i.e. at constant v/. What we get is
the function z(x) at constant y. Let us imagine to do this the first time at y= v/ =100 m/s and the
second time at y= v/ =105 m/s. We anticipate that with the former value, for any ground coupling, we
always are in the area with level curves at 45 degrees, while with the latter value we always are in the
area with vertical level curves. For this reason, we assume these values, 100 and 105 m/s, as reference
ones and we will refer to the function z(x,y=100 m/s) as z100(x) and to the function z(x,y=105 m/s) as
z1e5(x).
The value of z for a point of coordinates (x,y) can be simply read from Figure 3. But also
notice that it can be obtained from z100(x) and z1e5(x) by using property (i) depending on where it is. If it
is in the area with level curves at 45 degrees, e.g. point O1 in Figure 3, then the differential
displacement z can be found projecting (x,y) at 45 degrees (i.e. parallely to the level curves) on the line
y= v/ =100 m/s, thus obtaining the coordinates of a new point, O2 in Figure 3. If it is in the area with
vertical level curves, e.g. point O3 in Figure 3, the point can be projected parallely to the y-axis up to
the line y= v/ =105 m/s, thus obtaining the coordinates of a new point, O4 in Figure 3. We disregard
the possibility that the point is on the transition branch, because it is small and, in this case, we would
have results contained within the two previous ones.
So the value of z(x,y) for any point, depending on where it is on the (x,y) plane, can also be
computed as z100=z(x1,y=100 m/s) or as z1e5=z(x,y=105 m/s). Notice that, in any case, its correct value
will be the higher between the two.
Also notice that for a diagram in doubly natural scale, two points (xa,ya), (xb,yb) on a line at 45
degrees have the property that the difference in the x coordinate is equal to the difference in the y
coordinate, i.e. [x= xb - xa]=[y= yb - ya]. For a diagram in doubly logarithmic scale, this property
translates in the fact that the ratio of coordinates is equal, i.e. [xb/ xa]=[ yb/ ya].
Hence we can set:
z PQ ( X PQ , v ) = max ( z100 , z1e5 )

100 m/s

z100 = z X = X PQ
, v = (100 m / s )
v /

5
z1e5 = z X = ( X PQ ) , v = (10 m / s )

(3.1)

Equation (3.1) is important. It shows the influence of the incoherence factor on the ground
differential displacement for a given distance between points XPQ and of soil shear wave velocity v: for
low values of the incoherence factor (i.e. top part of Figure 3 and third of equations (3.1)), the
influence of is nothing and the differential displacement depends solely on the distance between
points. For higher values of incoherency (i.e. bottom part of Figure 3 and second of equations (3.1)) its
effect is exactly like the distance doubling the incoherence has the same effect as doubling the
distance between points.
Now that the influence of the incoherence factor is assessed, let us examine the influence of
a change in the surface wave velocity, vapp. Figure 4 and Figure 5 show the results obtained with the
same parameters as Figure 3, the only difference being in the value of , equal to 0.5 and 1.5
respectively.

32

Figure 5: level curves of differential ground


displacement in (m.10-2) for soil coupling D-D
and =1.5

Figure 4: level curves of differential ground


displacement in (m.10-2) for soil coupling D-D
and =0.5

The only difference with respect to before is in the position of the upper (i.e. vertical level
curves) part of the diagram. A decrease of makes the upper part move left, an increase makes it
move right. The only change to equation (3.1) needed to quantify its influence is then the expression of
z1e5.
From a physical point of view, this means that an increase of the surface wave velocity, i.e. a
decrease of the time lag PQ between the accelerations in P and Q, equation (1.1), has the effect of
broadening the range of influence of . Broadly speaking, the shear wave velocity v and the surface
wave velocity vapp have similar effects on the differential ground displacements: all other things equal,
differential ground displacements increase with decreasing v and vapp.
One point still needs to be clarified. Let us look at Figure 3 again. Why does the differential
displacement show that weird oscillatory pattern in the upper-right part of the diagram? After all, this
does not occur in different parts of the diagram. Besides, this happens also for different values of , see
Figure 4 and Figure 5. Well, the explanation is within equations (1.1) and (1.2). The maximum ground
differential displacement can be written as:
Z s*, p = Z * rs , p

(3.2)

2
2
Z2 = PK
+ QK
PQK
*

PQK

X

2
= 2 PK QK exp K2 X PQ
cos K PQ

v
vapp

The term within the square brackets in PQK is responsible for this behavior. To get convinced, we have
plotted exp[-(k.XPQ./v)2] and exp[-(k.XPQ./v)2].cos(-k.XPQ/vapp) separately in Figure 6 and Figure 7
for k =2, XPQ from 0 to 10,000 m, vapp =1,500 m/s (as in soil D), and two values of v/, 1,000 and
60,000 m/s.

33

Figure 6: diagram of exp[-(k.XPQ./v)2]

Figure 7: diagram of exp[-(k.XPQ./v)2] .cos(-k.XPQ/vapp)


For the lower value of v/, from Figure 6 one can see that the term exp[-(k.XPQ./v)2], which is the
correlation between amplitudes of the sine and cosine terms, goes from 1 to 0 within the first 400 m.
Correspondingly, the term exp[-(k.XPQ./v)2] .cos(-k.XPQ/vapp), Figure 7, goes also to 0 within the first
400 m, without oscillating as the cos(-k.XPQ/vapp) would love to do. For the higher value of v/, the
opposite happens. This is what we see in Figure 3: apart from the upper right part of the diagram, with
fixed v/ the level curves are monotonic with distance up to the maximum value, equal to 0.065m. In
the upper part of the diagram, the oscillatory behavior springs up and the level curves, with fixed v/,
are monotonic with distance up to 0.065m and then they start oscillating around this value because of
the cosine term within equation (3.2).
All that has been said up to now, making reference to one soil coupling, is true for all soil
coupling.

34

The last variable whose influence has to be assessed is the probability level. The results shown
above are relative to 0.5 probability of exceedance: what happens if we make the same diagrams with a
different probability? For the sake of conciseness we do not show the level curves at different
probability levels but we simply give the final result: the level curves at different probability levels are
identical to those shown for 0.5 probability apart from a scale factor which is a very smooth function of
probability and is very well interpolated by a third degree polynomial:
z PQ ( p ) = z PQ ( p = 0.5 ) f ( p )
f ( p ) = 2 p 3 2.7 p 2 + 1.38 p + 0.735

(3.3)

The scale factor f(p) is plotted in Figure 8 for p between 0.01 and 0.99.

Figure 8: scale factor f(p) as a function of the probability of exceedance

This nice regularity holds for all tested soils, distances and coherency parameters and could
not have been easily predicted from the expression of the Vanmarcke peak factor.
All we need at this point to completely define the soil differential displacement, given the soil
types in P and Q, as a function of all the other governing variables, namely distance, incoherence
parameter , and probability level, is a reasonably simple closed form expression for z100 and z1e5.
More precisely, given a soil coupling, we will have one expression for z100 (which is independent on )
and three expressions for z1e5 (at the three considered values for =0.5, 1, 1.5). Two such curves, z100
and z1e5 relative to soil coupling D-D and =1 are shown in Figure 9 and Figure 10, together with the
closed form approximating function.

35

Figure 9: z100 for soils D-D and =1

Figure 10: z1e5 for soils D-D and =1

The approximating functions have the form:


(
z PQ

appr )

(X ) = q
PQ

+ q2 log ( X PQ ) z Pmax
Q
q3

(3.4)

and its parameters have been fitted to the numeric values of zPQ(XPQ) up to 90% the maximum value, in
order to have small errors, from an engineering viewpoint. So we propose to compute zPQ(XPQ) using
equation (3.4):
(
z PQ ( X PQ ) z PQ

appr )

(X ) = q
PQ

36

max
+ q2 log ( X PQ ) z PQ
q3

(3.5)

In equation (3.5) we are disregarding the oscillatory pattern of z1e5 at large distances that we
have discussed before. In the tables that follow, we list the values of q1, q2, q3, zmaxPQ for all soil
coupling and for three values for =0.5, 1, 1.5. In the tables, the dot symbol by a matrix component
indicates that it must be computed by symmetry. Equation (3.5) must be used with meter units.
Table 3: parameters q1, q2, q3, zmaxPQ in equation (3.5) to compute z100
Parameter
Soil P/Q
A
B
D

A
0
.
.

100.q1
B
0.7
0
.

D
3.3
2.0
0

A
141.2
.
.

105.q2
B
86.6
226.4
.

D
42
93.8
396.9

A
1.99
.
.

D
2.60
2.37
1.85

100.zmax
A
B
D
2.9
3.5
5.7
.
3.9
5.5
.
.
7.0

q3
B

100.zmax
A
B

7.33
4.96
.

8.43
8.11
5.19

4.35
.
.

4.8
6
.

7.5
7.4
10.5

7.85
5.99
.

8.83
8.75
5.65

4
.
.

4.7
5.7
.

7.4
7.3
10.4

8.47
6.32
.

9.78
8.93
6.36

3.7
.
.

4.4
5.5
.

7.1
7.3
9.6

q3
B
2.26
1.86
.

Table 4: parameters q1, q2, q3, zmaxPQ in equation (3.5) to compute z1e5
A

100.q1
B

A
B
D

0
.
.

0.7
0
.

3.3
2
0

A
B
D

0
.
.

0.7
0
.

3.3
2.0
0

A
B
D

0
.
.

0.7
0
.

3.3
2
0

Parameter
Soil P/Q

105.q2
B

=0.5

2.5e-2 8.7e-4 9.1e-5 5.64


.
1.8e-1 3.4e-4
.
.
.
2.5e-1
.
=1.0
3.6e-3 1.5e-4 1.9e-5 6.34
.
1.4e-2 4.4e-5
.
.
.
6e-2
.
=1.5
9.6e-4 2.8e-5 1.6e-6 6.82
.
5.1e-3 1.9e-5
.
.
.
1.0e-2
.

Finally, an interesting thing to see for application in the codes, is, for each soil coupling, the
worst possible situation. This is given by the case with the highest incoherence factor , 0.5,
irrespective of the value of . The differential displacements for this case, identified as zv/0.5, are shown
in Figure 11 and can be computed via use of equation (3.5) with the parameters listed in Table 5.

37

Figure 11: differential displacements with =0.5

Table 5: parameters q1, q2, q3, zmaxPQ in equation (3.5) to compute zv/0.5
Parameter
Soil P/Q
A
B
D

A
0
.
.

q1.100
B
0.7
0
.

D
3.3
2.0
0

A
16.97
.
.

q2.105
B
D
1.36
17.05
14.28 28.91
.
265.54

A
2.52
.
.

q3
B
3.87
2.90
.

zmax.100
A
B
D
2.9
3.4
5.6
.
3.9
5.5
.
.
6.9

D
2.80
2.81
1.91

One may notice that the maxima differential displacements for this case are in good agreement
with what predicted by EC8/ICPC [7, 8], equation (3.1) of part 1. However, with reference to soil
coupling DD, one may notice from Figure 3 that the maximum differential displacement occurs for the
lowest value of the incoherence factor , i.e. at v/(=0.02), and is equal to 0.0725 m, which is slightly
more than 0.069 m, the maximum for v/(=0.5). Furthermore, from Figure 4, one may also notice that
the maximum differential displacement for =0.5 and v/(=0.02) is still a bit larger and is equal to
0.08 m, i.e. an increase of 16% with respect to 0.069 m. This happens for all soil couplings. To
facilitate the comparison, we have shown the maxima differential displacements for these three cases,
EC8/ICPC [7, 8], (zv/0.5 and =1) and (zv/0.02 and =0.5), in Table 6.

Table 6: comparison between the maxima differential displacements in (m.10-2) predicted by


EC8/ICPC [7, 8] and those relative to (zv/0.5 and =1) and (zv/0.02 and =0.5)
Soil coupling
EC8/ICPC
zv/0.5 and =1
zv/0.02 and =0.5

AA
2.9
2.9
3.4

AB
3.7
3.4
4.3

AD
5.7
5.6
7

BB
4.3
3.9
5.5

BD
6.2
5.5
7.8

DD
7.6
6.9
8

The differences between the values in Table 6 predicted by (zv/0.5 and =1) and (zv/0.02 and
=0.5) range from +16% (coupling DD) to +42% (coupling BD) while the differences between the
values predicted by (EC8/ICPC) and (zv/0.02 and =0.5) range from +5% (coupling DD) to +25%
(coupling BD).
For this reason, and because the maxima occur at large distances, we find it both reasonable
and safe and physically sound to compute the maximum differential displacements by using the
EC8/ICPC formulation with an increase of 25%:

MAX
z PQ
= 1.25 0.025 pga

( P TPC TPD ) + ( Q TQC TQD )


2

(3.6)

For the variation of the differential displacement with distance, we find it reasonable again to keep the
formulation of equation (3.5) and the parameters in Table 5, but with zmax computed as in equation
(3.6). This is what has been implemented in the DICB

4. RESULTS FOR THE STRUCTURAL DIFFERENTIAL DISPLACEMENTS


In this section we examine the structural differential displacements. In order to keep the
problem variables within a number which may be governed, and because by now we know what the
influence of and is, at least for the ground differential displacements, we keep them at fixed value,
respectively 0.5 and 1. Initially, we also keep the probability level at 0.5, but we anticipate that its
influence will be investigated in what follows. We cast the problem as:

38

z LM = ( z PQ + u LM ) D

(4.1)

Equation (4.1) means that we are looking for the value of the variable D that makes the correct
structural differential displacement, zLM, equal to the differential structural displacement computed
code-style, i.e. summing up the contribution of the ground and of the structures considered separately.
The contribution of the ground is computed in the correct way as in the previous section, and the
contribution of the structures is computed with a modal analysis with CQC modal combination at fixed
base, see equation (3.3). The reason why the problem is cast as in equation (4.1), and not in a different
functional form, is that we have seen that D is more regular with this formulation as compared to
alternative formulation, e.g. z LM = z PQ + u LM D . One should notice that D takes into account both the
correlation between structural differential displacements and between the differential displacements of
soil and structures above it.
For each soil coupling, D depends on the distance and on the natural periods of the structures,
i.e. D=D(XPQ, TP, TQ). We have noticed that the shape of D(XPQ) has two distinct behaviour depending
on the distance between periods |TP -TQ|: for values of the distance between periods lower or equal than
0.1 sec, D grows with the space distance up to its asymptotic value; for values of |TP -TQ| larger than 0.1
sec, D is practically constant.
We have discretized the natural periods at 0.1 sec step, ranging from 0.2 to 2 sec, we have
divided the results in two groups, period distance lower or equal and larger than 0.1 sec, and, within
each group, we have computed the quantities:
Dm ( X PQ ) =
De ( X PQ ) =

max D ( X PQ , TP , TQ ) + min D ( X PQ , TP , TQ )
TP ,TQ

TP ,TQ

(4.2)

max D ( X PQ , TP , TQ ) min D ( X PQ , TP , TQ )
TP ,TQ

TP ,TQ

For a given distance XPQ , Dm(XPQ) is halfway between the minimum and maximum of
D(XPQ,TP,TQ) for any natural period; De(XPQ) is the distance of the maximum and the minimum from
Dm(XPQ). By plotting the two quantities one can see both the trend of D and the error committed if one
assumes D(XPQ, TP, TQ) Dm(XPQ). The results are shown in Figure 12 and Figure 13 for |TP-TQ|0.1 sec
and Figure 14 and Figure 15 for larger differences in natural periods, respectively for Dm and De. On
the x-axis there are the values of the distance between structures, XPQ=XLM.

39

Figure 12: value of Dm for all soil coupling and periods TP, TQ such that |TP -TQ|0.1 sec

Figure 13: value of De for all soil coupling and periods TP, TQ such that |TP -TQ|0.1 sec

Figure 14: value of Dm for all soil coupling and periods TP, TQ such that |TP -TQ|>0.1 sec

40

Figure 15: value of De for all soil coupling and periods TP, TQ such that |TP -TQ|>0.1 sec

Let us examine the figures. For periods TP, TQ such that |TP -TQ|>0.1 sec, the interpretation is
simple: no matter what the distance and the soil coupling is, the value of 1 is a reasonable upper bound
for D. The correct value of D depends on the soil coupling, the distance and TP and TQ and ranges from
about 0.6 to 0.98. While there is a discernible pattern as a function of the distance and the soil coupling,
we have been unable to find regularities depending on the natural periods. In order to make things
simple, from an engineering perspective, we think that the assumption of 1 for D is acceptable.
For periods TP, TQ such that |TP -TQ|0.1 sec, the interpretation is less straightforward:
depending on the soil coupling, the asymptotic value of Dm, from about 0.65 to 0.75, is reached at
distances between 200 (DD coupling) and 1,000 (AA soil coupling) meters; before that, Dm grows with
the distance. Within the first 200 meters, De is lower or equal than about half the corresponding value
of Dm. For larger distances it decreases to about 20% the corresponding value of Dm. For this group of
structures, with |TP -TQ|0.1 sec, we have also approximated the shape of Dm(XPQ) with the power of
log function already used before, in equation (3.4). One such approximation is shown in Figure 12 for
soil coupling AA. The approximations appear to work well for all the soil couplings. So, for this case,
we can set:
D ( X PQ ) Dm ( X PQ ) + De ( X PQ ) 1.5 Dm ( X PQ ) 0.90
Dm ( X PQ ) = q4 + q5 log ( X PQ )

(4.3)

q6

In the same fashion as in the previous section, we have tabled the values of q4, q5, q6, for each
soil coupling in Table 7.
Table 7: parameters q4, q5, q6 in equation (4.3) to compute Dm with TP, TQ such that |TP -TQ|0.1 sec
and CQC modal combination
Parameter
Soil P/Q
A
B
D

A
0
.
.

q4.100
B
D
34.8 61.3
0
52.1
.
0

41

A
7.76
.
.

q5.103
B
D
0.25
1.5
13.54 3.98
.
70.65

A
2.31
.
.

q6
B
3.76
2.1
.

D
2.66
2.36
1.46

The reason why we have singled out this case from the general and simple D=1 result, is
because on one side this is what the mathematical model suggests and on the other it is a very common
situation, at least in Italy. In fact the great majority of the Italian railway and highway bridges [10]
have simply supported prestressed concrete decks, with piers that either are, or, when in the linear
elastic range, can be reliably modeled as, s.d.o.f systems; and when the highway or railway crosses a
plain the piers are typically equal structures at short distances, say 20-30 meters. Hence, the use of the
general D=1 rule would be far too prudential. However, we stress again that the use of equation (4.3)
requires that the soils in P and Q be homogeneous and that |TP -TQ|0.1 sec; very seldom one can rely
on these assumptions. For the soil homogeneity assumption, in fact, a detailed ground knowledge is
required, a condition which is rarely satisfied. And the assumption that |TP -TQ|0.1 sec is further very
arguable because, even for simple structures, the natural periods depend on soil-structure interaction,
earthquake intensity, correct functioning of the supports, etc. In sum, one should be very cautious when
using equation (4.3) and this is the reason why in the DICB we have always retained a value of 1 for D.
As an example, with reference to Figure 1, assume that structures 1 and 2 are designed as
equal structures on equal soils. Assume that due to inherent randomness in the soil properties and the
construction phases, their real natural periods can be modelled as independent gaussian random
variables, with mean value equal to the design natural period, Td, and coefficient of variation equal to
0.4. The difference in the natural periods, T=T1-T2, is then a gaussian random variable, with mean
equal to 0 and standard deviation equal to (0.4).(Td).(2)0.5. The probability that | T | is lower or equal
than 0.1, i.e. motions are correlated and equation (4.3) is applicable, is then equal to:

Prob ( T1 T2 0.1) = 1 2 0.1, ( = 0 ) , = Td 0.4 2

(4.4)

where (x,,) is the cumulative distribution function, evaluated at x, of the gaussian


distribution with mean and standard deviation . Equation (4.4) is plotted in the following figure for
Td from 0.2 to 2 seconds.

Figure 16: probability that |T1-T2|0.1 sec


From Figure 16 one can see that, for periods larger that 0.4 seconds, the probability that |T1T2|0.1 sec is already below 0.35, i.e. far too low to rely on the decrease of differential displacements
due to correlation.
Two more points are worth discussing before finishing this section: the first concerns the
possibility to use the square root of sum of squares, SRSS, instead of CQC, modal combination, to

42

compute the structural differential displacement uLM in equation (4.1). This would be desirable because
we have the aim of simplifying computations as much as possible and SRSS allows simple hand
computation. This is actually possible, both for the case |TP -TQ|>0.1 sec and for the case |TP -TQ|0.1
sec. For the former case, one can assume again D=1. This is on one side intuitive (CQC gives the same
results as SRSS when the periods are separated) and on the other confirmed by the plots of Dm and De,
similar to Figure 14 and Figure 15 and not shown here for the sake of brevity. For the case |TP -TQ|0.1
sec, equation (4.3) may be used, with the values of q4, q5, q6 for each soil coupling in the following
table.
Table 8: parameters q4, q5, q6 in equation (4.3) to compute Dm with TP, TQ such that |TP -TQ|0.1 sec
and SRSS modal combination
Parameter
Soil P/Q
A
B
D

A
0
.
.

q4.100
B
43
0
.

D
69
60
0

A
12.6
.
.

q5.103
B
0.24
17.9
.

D
1.5
4.2
81

A
2.16
.
.

q6
B
3.9
2.0
.

D
2.70
2.39
1.47

The last point concerns the influence of the probability level on the structural differential
displacement zLM. All that has been said in this section up to now is relative to the median value of the
responses. What happens with different fractiles? The fact that we have expressed the problem as in
equation (4.1) makes us hope that we will be able to move from one probability value to the other via a
simple multiplication of the median response value. Actually this is the case, and we have tested it on
all soil couplings and distances and periods shown in this study. Again, for the sake of brevity, we do
not show here different fractiles of zLM but only give the final result: different fractiles of the structural
differential displacements may be obtained from the median value by multiplying it by the same scale
factor f(p) as in equation (3.3). Explicitly we can write:
zLM ( p ) = zLM ( p = 0.5) f ( p )

(4.5)

which is again a nice formulation and definitely solves the problem. The reason why in the DICB, for
the case of fragile failures, it is advised to multiply the final differential displacement by 1.25 (after
equation (3.4) of part 1) is that by doing so, one obtains both for the soil and the structural differential
displacements a probability of 90% of them not being exceeded.

5. NUMERICAL EXAMPLE
The object of the numerical example is the Lenze - Penze di Valle viaduct already studied in
[10] and [11]. It is a reinforced concrete highway viaduct, built in 1966, located on the A16 NapoliCanosa Italian highway. The bridge is composed of eight piers plus abutments, with total heights
varying between 18 and 27 m. The structural scheme is statically determined since the prestressed
concrete slabs, which have a mass of 20,000 kg/m, are simply supported. The two extreme piers are
single columns with the same height (18m) and cross section (rectangular and hollow, 3m x 2.5m x
0.4m); the remaining piers are portal frames with the same cross sections. The piers concrete has
strength fck 30 Mpa, with steel rebars classified in Italy as Feb38k (fyk=380 Mpa), for a total of 150 cm2.

43

32 m

32 m

18 m

Pier 1

32 m

Pier 2

32 m

32 m

Pier 3

32 m

32 m

Pier 6

Pier 5

Pier 4

32 m

32 m

Pier 7

Pier 8

18 m

PIERS CROSS SECTIONS

2.5 m

0.4 m

3m

PIERS CROSS SECTION

Figure 17 : structural scheme of the Lenze Penze di Valle viaduct

The intensity of the 500 years return period earthquake is 9.4 IMM [11], corresponding to about 0.4g.
The correspondence is obtained with the correlation law between Mercalli intensity and peak ground
acceleration in [12], which is specific for southern Italy:

pga[m/s 2 ] =

100.237IMM 0.594
10

(5.1)

We assume that the viaduct is in soil type D. The soil differential displacements, in the longitudinal
bridge direction, for the EC8/ICPC and DICB are respectively equal to:

I
uPQ
= X PQ pga

TC
vapp

= 14 mm EC8/ICPC

(5.2)

q
II
uPQ
= pga q1 + q2 log ( X ) 3 = 112 mm DICB

To compute the differential displacements we have used the codes provisions of equation (3.1) of part
1. The coefficients qi are equal to those of Table 5. The structural natural periods, differential
displacements with synchronous motion and CQC modal combination and total differential
displacements are shown in Table 9
Table 9: periods, response spectrum and differential displacement between the top of the piers. Sd is
the spectral displacement, uLM is the CQC differential displacement, zLM is the total differential
displacement
Pier #
1
2
3
4
5
6
7
8
Period (sec)
0.38
0.53
0.61
0.61
0.61
0.55
0.41
0.38
Sd (mm)
51
96
128
128
128
103
53
51
109
160
256
256
165
116
74
uLM (mm)
zLM (mm) EC8/ICPC
123
174
270
270
179
130
88
zLM (mm) DICB
221
272
369
369
277
228
185
1.25zLM (mm) DICB
276
340
461
461
346
285
231
From equation (5.2) one can notice that the difference in the soil differential displacement is almost one
order of magnitude and is equal to about 100 mm. This difference is then maintained in Table 9, where

44

the differential displacements between the top of the piers are shown. The last row of Table 9 shows
the differential displacement with 10% probability of exceedance in 500 years, and is the value with
which one should verify the bridge supports.
Piers 3, 4 and 5 are equal, both from the material and geometric point of view. If one could
rely on the soil below them being homogeneous, which would not, in our opinion, be the case this time
since the viaduct is in a valley, the structural differential displacement between pier 3 and 4 and
between pier 4 and 5 would be computed as:
z PQ D = 112 mm D 74 mm
uLM D = 256 mm D 169 mm

(5.3)

z LM = ( z PQ + uLM ) D 243 mm

q
1.46
D Dm + De 1.5 Dm = 1.5 q4 + q5 log ( X ) 6 = 1.5 70.65 103 log ( 32 ) = 0.66

In this case the correlation between motions would decrease the differential displacement by a
minimum of 1-0.66=34%.

6. CONCLUSIONS
We have made a fairly wide study of the effect of spatial variability of seismic motion in
terms of the ground differential displacements for different soil types and incoherency parameters. As a
natural extension, we have also investigated what would happen in terms of differential displacements
to linear elastic statically determined structures founded on such soils. We have then compared these
results with the Eurocode8 [7], which coincides with the Italian Code from Civil protection [8], and
draft Italian code provisions for bridges [9].
In part 1 we have shown the basic theoretical developments. In this part the main numerical
results are given.
The most important points that have clearly arisen are:
The soil differential displacements vary smoothly to their asymptotic value, attained at distances
between 600 (soil coupling D-D) and 4,000 meters (soil coupling A-A), depending on the soil
types. These values are conservative and take into account the most unfavourable combinations of
soil properties.
When the soil is homogeneous, the differential displacement at zero distance is zero; when two
different soil types are contiguous, the differential displacement at zero distance has a finite, nondisregardable value
The variation of soil differential displacement with the distance can be accurately approximated
with a power of log function.
The surface wave velocity is much less influential than the shear wave velocity and, in view of
necessary code simplification, may be disregarded altogether
The incoherence parameter , within the expression v/, is, together with the distance, the most
influential variable. Its assessment for each soil type appears, from our viewpoint, at best difficult
and questionable and significant more research effort is needed on this point. To account for this
problem, which is probably due more to lack of basic comprehension of physical behavior than to
intrinsic randomness, we think it reasonable to keep on the safe side, as has been done in the Draft
Italian Code for bridges [9]
The maximum differential soil displacements at large distances can be obtained using the
expression given by Eurocode 8 [7] increased by 25%.
To compute the total structural differential displacements, one may sum up the soil differential
displacement and structural differential displacement. The latter can be obtained with the simpler,
as compared to CQC, SRSS rule (root of the sum of the squares of top displacements for
synchronous motion). Then one needs to correct this result by a coefficient that accounts for the
correlation of motions. This coefficient can be safely taken equal to 1. For structures with natural
periods less distant than 0.1 seconds, founded on homogeneous soils, the correlation coefficient
could be reduced. However, due to period uncertainties, this assumption could be easily
unrealistic, and therefore the evaluation should be the same also for structures having the same
period.
The above has been said with reference to the median value of differential displacements of both
the soil and the structures. To compute them with varying probability of exceedance, one may

45

multiply the median value by the probability factor f(p). If the median value is multiplied by 1.25,
one has the differential displacements with 10% probability of exceedance.
It has been shown that the probability factor f(p) to go from the median response to higher fractiles
can be obtained by a simple polynomial expression independent on the soil type and on the
structural periods.
For what concerns differential displacement for bridges, Eurocode 8 [7] and the Italian Civil
Protection Code [8] appear inaccurate and unconservative. This is especially true in the range of
distances where most civil engineering structures are, below 100 m. For instance, in the example
application of the previous section, with a pier distance of 32 m, Eurocode 8 and the Italian Civil
Protection Code estimation of soil differential displacements is 14 mm while with the draft Italian
Code for bridges [9] the estimation is 112 mm. The latter value is consistent with the model
predictions.

As a final remark, we would like to highlight that earthquake spatial variability does appear to be a
significant problem for failure modes governed by differential displacements, also for structures of
minor importance like small bridges. Since its inclusion in the design phase brings about small or no
extra cost for most situations, we stress the importance of a rapid code update on this subject.

7. REFERENCES
1. Vanmarcke, E. H., Fenton, G. A., Conditioned simulation of local fields of earthquake ground
motion, Structural Safety, 10, 1-3, May 1991, pages 247-264
2. Luco, J. E., Wong, H. L., Response of a rigid foundation to a spatially random ground motion,
Earthquake Engineering & Structural Dynamics, 14, 6, Nov.-Dec. 1986, pages 891-908
3. Kanai, K., Semi-empirical formula for the seismic characteristics of the ground, University of
Tokio Bulletin of Earthquake Research Institute, vol. 35, pp. 309-325, 1957
4. Tajimi, H., A statistical method of determining the maximum response of a building structure
during an earthquake, Proceedings of the Second World Conference on Earthquake Engineering,
Tokyo, Vol. II, 1960, pages 781-797
5. Clough, R. W., Penzien, J., Dynamics of structures, McGraw-Hill, Inc., New York, 1975, 634
pages
6. Vanmarcke, E.H., Fenton, G.A., Heredia-Zavoni, E., SIMQKE-II, conditioned earthquake ground
motion simulator : user's manual, version 2.1, Princeton University, [Princeton, N.J.], 1999, 25
pages
7. Comit Europen de Normalisation, CEN, Eurocode 8: design of structures for earthquake
resistance, Draft n. 2, doc cen/tc250/sc8/n320, May 2002
8. Presidenza del Consiglio dei Ministri, Primi elementi in materia di criteri generali per la
classificazione sismica del territorio nazionale e di normative tecniche per le costruzioni in zona
sismica, Ordinanza, March 2003, (in Italian).
9. Consiglio Superiore dei Lavori Pubblici, Aggiornamento delle Norme Tecniche per le Costruzioni
in zone sismiche, draft version of October 2003, 2003 (in Italian, available at
http://host.uniroma3.it/master/mica/mica.asp
>
normativa
and
at
http://www.
infrastrutturetrasporti.it/main/facciamo/consup/consup.html > normativa)
10. Donferri, M., Giannini, R., Nuti, C., Pinto, P.E., [1998] Analysis of seismic risk on the bridges of
Autostrade network, Autostrade, vol. 2, pp. 7-15
11. Nuti, C., Vanzi, I., To retrofit or not to retrofit, Engineering structures, vol.25 (6), pp. 701-711,
Elsevier, UK, 2003
12. Decanini, L.D., Gavarini, C., Mollaioli, F., Algunas consideraciones sobre la correlaciones entre
intensidad macrosimica y parametros del movimento del suelo, IX international seminar on
earthquake prognostics, San Jos, Costa rica, 1994

8. APPENDIX A: SUMMARY OF FORMULAS TO COMPUTE THE ASYNCHROUS


EARTHQUAKE RANDOM FIELD AND THE DIFFERENTIAL DISPLACEMENTS

46

To make applications easier, we think it worth to summarize the most important formulas
presented in parts 1 and 2.
Power spectrum representation of the earthquake (equation (4.6) of part 1)
Gmodified-KT-CP = GKT-CP 0
GKT-CP = G0

+ 4 f2 4f 2
4
f

2
f

2 2

+ 4 f2 4f 2
G0 =

0 = 0.5 +

2
g

2 2

+ 4 g2 g4 2

pga
g1 log ( s ) + g 2

2 E F
atan

E F

How to go from the response spectrum representation of the earthquake to the power spectrum
representation
If the response spectrum is relative to the ultimate limit state and is defined as in the Eurocode 8 [7] or
the Civil Protection Italian Code [8] or in the draft Italian Code for bridges [9] then the parameters are
(Table 3 of part 1):
Soil / Parameter
Firm
Medium
Soft
A of EC8
B of EC8
D of EC8

g1
g (%)
f
f (%)
Kanai-Tajimi-Clough-Penzien, equation (4.3)
1.5
60
15
60
184
1
60
10
40
125.5
0.5
60
5
20
90.2
Modified Kanai-Tajimi-Clough-Penzien, equation (4.6)
2.8
97
23
43
142.8
2.8
94
18.5
53
65.5
2.8
87
11.1
61
30

g2

557.2
286.3
95.8

N.A.
N.A.
N.A.

225.8
98.2
39.3

10.0
10.0
10.0

Otherwise, use the iterative procedure outlined in section 4 of part 1.


Earthquake representation and coherency function for different soils (equation (1.1) and (2.12),
(2.9) of part 1)
AP ( t ) = BPK cos ( K t ) + CPK sin ( K t )
k

AQ ( t ) = BQK cos K ( t PQ ) + CQK sin K ( t PQ )

k
cos ( )
= X PQ

v
V
app
2


2
= exp k2 X PQ

PQ =
PQk

X PQ

= 1 1 + 2 2
v
v
v
1
2

vapp = 1 vapp1 + 2 vapp 2

1 + 2 = 1

Statistics of the differential response between two points on the soil or two structures (equation
(2.27), (2.28) of part 1 and equation (3.3))

47

Z s*, p = Z * rs , p

Z2 = GZ K
*

2
2
GZ * K = PK
+ QK
PQK
2
2
PK
= N PK
GPPK
2
2
QK
= N QK
GQQK

PQK = 2 PK QK PQK cos ( PK QK K PQ )


Also notice that:
Z s*, p = Z s*, p = 0.5 f ( p )
f ( p ) = 2 p 3 2.7 p 2 + 1.38 p + 0.735

Differential displacements of two points on the ground (equation (3.1) of part 1)


MAX
uPQ = pga q1 + q2 log ( X ) 3 u PQ

MAX
1.25 u P2 + uQ2
uPQ

uP = 0.025 pga P TPC TPD


uQ = 0.025 pga Q TQC TQD

If the response spectrum is relative to the ultimate limit state and is defined as in the Eurocode 8 [7] or
the Civil Protection Italian Code [8] or in the draft Italian Code for bridges [9] then the parameters are
(Table 2 of part 1):
Soil P/Q
Parameter
A
B
C

A
0
.
.

B
100.q1
0.7
0
.

3.3
2.0
0

17
.
.

B
105.q2
1.4
14.3
.

17
29
266

2.5
.
.

B
q3
3.9
2.9
.

C
2.9
2.9
1.1

Differential displacements of the top of two s.d.o.f. structures (equation (3.4) of part 1)
U LM = uPQ + uLM
uLM = S d21 + S d22

9. APPENDIX B: UNITS AND LIST OF SYMBOLS


Unless stated otherwise, all units are within the MKS (meter, mass kilogram, second) system.
Log is the natural (base e) logarithm.
Symbol
f
k
p
pga
q1, q2, q3
r
s
t
u

Meaning
Factor to compute the fractiles of the response as a function of the exceedance probability
1-p
Index of the ground acceleration and displacements circular frequencies
Probability
Peak ground acceleration
Constants in the proposed approximating functions
Peak factor
Earthquake duration
time
Code displacements

48

v
vapp
x
y
z
AP(t)
BPK
CPK
D
E
G
H
L
M
N
P
Q
S
T
U
XPQ
ZP, ZMQ
Z*

,f,g,

PQ

PQ

k
i

Shear wave velocity


Surface wave velocity
Distance
v/
Displacement
Ground acceleration recorded at time t in point P
(A is also a ground type, when stated explicitly)
Amplitude of the k-th cosine component of the ground acceleration in point P
(B is also a ground type, when stated explicitly)
Amplitude of the k-th sine component of the ground acceleration in point P
(C is also a ground type, when stated explicitly)
Factor to compute the differential displacements
(D is also a ground type, when stated explicitly)
Dummy variable
(also a ground type, when stated explicitly)
Power spectrum
Transfer function
Point located on top of a s.d.o.f. structure
Point located on top of a s.d.o.f. structure
Modulus of the transfer function from ground acceleration to structural total displacements
Point on the ground surface
Point on the ground surface
Response spectrum
Period
Code displacements
Distance between points P and Q
Displacement of point P, between points M and Q
Total and relative displacements among points P, Q, L, M
Incoherency factor
Damping of the structure and in the Kanai-Tajimi power spectrum
Parameter in the proposed correction function applied to the Kanai-Tajimi power spectrum
Soil coefficient
Error in the estimation of D, factor to compute the differential displacements
Angle of the transfer function from ground acceleration to structural total displacements
Correction functions applied to the Kanai-Tajimi power spectrum
Multiplier of the surface wave velocity given by the codes
Correlation coefficient between BP and BQ and CP and CQ
Standard deviation
Difference in arrival time of the earthquake waves between points P and Q
Structural circular frequency
k-th circular frequency of the ground acceleration
Non dimensional distance
Angle between the vector of surface wave propagation and the vector that goes from P to Q
Point on the ground surface where a soil discontinuity is located
Covariance matrix of B and C at the l points on the ground surface
Modal correlation in the complete quadratic combination, CQC, formula

49

S-ar putea să vă placă și