Vous êtes sur la page 1sur 55

Are We Ready for . . . "El Centro"?

by Bobby Motwani
03/03/10
( http://bobbyness.net/NerdyStuff/structures/El%20Centro%20in
%20Malaga/ElCentroInMalaga.html )
NB. Images on this page are dynamically resized within your browser window. If need be
right-click and save or open in separate window for better viewing.
 Objectives
 Software
 Part I : El Centro
o Accelerogram
o Time Response of SDOF systems
o Response Spectrum
 Part II : Design Spectra
o Response Spectra Characteristics
 Time integration and Response Spectrum Calculation Revisited
o Design Spectra based on Response Spectra
o Seismic Hazard Analysis
o Uniform Hazard Spectra
o Design Codes
 EC8
 NCSR-02
 Are we ready for El Centro?
 References

Objectives:
The end goal of this study is fairly straightforward: I wish to determine how the buildings
designed to withstand seismic action in Malaga City would cope if the area were struck by
the historical earthquake that hit the city of El Centro, in the Imperial Valley in California, in
1940. More precisely, to compare the design-spectra specified by the Spanish Seismic
Building Design Code (currently the NCSE-02) with the response-spectra of the El Centro
Earthquake. I shall restrict this study to the effects of horizontal ground motion, and work
only with the N-S component of the El Centro ground motion by way of example (you could
carry out a similar study with the E-W component). Thus, the study seems fairly simple: all
I need to do is superimpose the N-S response-spectrum of the earthquake to the horizontal-
motion design-spectrum specified by the design code, and see whether the former lies
'within' the latter.
And that is precisely what I shall do. However, my ulterior motives for carrying out this
study are: a) to show how to compute response spectra of ground motion from data
collected by seismographs in the form of accelerograms, and b) to understand how design
spectra are constructed and to familiarize myself a bit with the Spanish Seismic Design
Code and the respective European Standards from which it is derived (Eurocode 8 ) and
with the design criteria adopted in these codes. In the process we shall also learn about the
nature of response spectra, the historical design spectra by Newmark and Hall, and a
primer on Seismic Hazard Analysis from which the concept of Uniform Hazard Spectra
derives.
Software:
To carry out the required mathematical computations and plot data, I have used Matlab. I
have deliberately used a general purpose numerical computing environment because:

a) part of the objective of this study is to learn to construct the required diagrams and not
rely on specialized software that does the job for you at the click of a button, which is not a
bad thing as long as you understand what is is the programme is doing to be able to
interpret the outputs correctly, but as I said, it's instructive to programme the process
yourself at least once! and,

b) for the flexibility this offers in handling data and carrying out operations to measure
rather than adhere yourself to the limited options a specialized programme provides.
I might also add that while the Matlab package is not available free of charge, you could just
as well do the same using other numerical computing environments such as Octave, Scilab,
or Rlab, which are available at no cost at all, or use a general purpose programming
language such as C/C++ or VB.NET (although with these I assume you'd have to develop
your own routines for solving differential equations and such, but those are topics
concerning numerical methods which are not the object of the present work!)

Part I : El Centro.
1. The accelerogram
The accelerogram for El Centro is provided by the NISEE (National Information Service for
Earthquake Engineering) supported by the University of California, Bekerley. The data is
provided as an ASCII-file (IMPVAL1.ACC) as a discrete series of readings. Ground
acceleration, velocity, and displacement are each provided in the same file. To simplify
reading from file into a MATLAB variable, I have separated each of the three sets of data
into separate text files (IMPVAL1-acc.txt, IMPVAL1-vel.txt, IMPVAL1-disp.txt), though I
shall primarily use the acceleration data.
To read the acceleration data from MATLAB, you can use the File I/O function "fscanf",
which reads and converts data from a text file into an array in column order. The
instructions are in the m-file ElCentroAccelerogram.m :
% El Centro Earthquake, Imperial Valley, May 18 1940, 270 degrees

% plot accelerogram

t=0:0.02:53.76;

fid = fopen('C:\Users\u\Desktop\Accelerograms\Imperial Valley (El Centro)\matlab\IMPVAL1-


acc.txt');
a = fscanf(fid, '%g'); % a is in mm/s/s
a = a./9800; % acceleration in g's
a = [0 ; a] % add data point for t=0 (a=0)

plot(t,a,'color','b');grid;
Title('IMPERIAL VALLEY EARTHQUAKE - EL CENTRO, 270degrees')
ylabel('acceleration (g)')
xlabel('time (s)')

ElCentroAccelerogram.m

The earthquake duration is 53.76 seconds (2688 acceleration data points spaced evenly every 0.02
seconds - note that the first data point is for t=0.02s; the acceleration for t=0 is 0). Exectuting the
previous m-file gives us the plot of the acceleration data:
N-S horizontal ground acceleration of El Centro earthquake

2. Time Response of SDOF systems


The time response of a linear SDOF system to a ground acceleration is governed by the
differential equation:

where üg is the ground acceleration, and u the displacement relative to the ground. We can
write this as:

or in terms of the natural period of vibration as:

To integrate this numerically in the time domain, I used MATLAB's ode45 solver; the ode
solvers require the differential equation to be defined as a first order system of ode's in a
seperate m file as a function that returns the value of the derivative vector at a specified
time based on the value of the function calculated in the previous step of the integration.
Other parameters can be passed on to this function, and since I shall be obtaining the time
response for different SDOF systems, it makes sense to pass the damping ratio and natural
period of vibration to this function as parameters; it is also necessary to pass as a
paramater the forcing function (ground acceleration), which I have passed as the pair of
vectors (acceleration vector and corresponding time vector), so that the acceleration at the
required time instant can be computed as necessary (via a linear interpolation, for
example). The m-file containing the equations is available as f.m :
function ydot = f(t,y,p1,p2,p3,p4)

xi = p1;
Tn= p2;
t1 = p3; % time vector
a2 = p4; % ground acceleration
a = interp1(t1,a2,t); % Interpolate the data set (t1,a2) at time t

ydot = [y(2); -(4*pi*xi/Tn)*y(2)-(2*pi/Tn)^2*y(1)- a];

f.m

Using this function, we can, for example, simulate the free response of a SDOF system - eg, for a
SDOF system with a 5% damping ratio and natural period of vibration of 10s:
% Time Response of SDOF systems

% eg free response SDOF system with 5% damping, Tn=10s:


xi=0.05; % relative damping
Tn=10; % natural period of SDOF system
a0 = zeros(length(a),1); % forcing function is zero
y0 = [1;0]; % initial condition
[T,Y] = ode45(@f,t,y0,[],xi,Tn,t,a0); % Y(:,1) is in m and
Y(:,2) in m/s
plot(T,Y(:,1));grid
Title('SDOF (5% damping) free response to initial
displacement')
AXIS tight
ylabel('displacement (m)')
xlabel('time (s)')

TimeResponseSDOFSystems.m

free response of SDOF system with 5% damping and natural period of 10s

To simulate the time-response to the ground acceleration, I have first augmented the simulation
time by 50% of the earthquake duration, just in case the peak value of the response (which will be
the value of interest to us) occurs during the natural vibration of the system that will follow the
extinction of the ground motion. I have then used ode45 to obtain the steady-state response of the
system to the ground acceleration, and plotted the system deformation (displacement of the mass
relative to the ground) as a function of time:

% increasing simulation time by 50% of eartquake duration (will be free response after
earthquake)
t1 = 0:0.02:80.6;
a2 = a*9.8 ; % a2 in m/s/s
diff = length(t1)-length(t);
a2 = [a2 ; zeros(diff,1)];

% steady state response of system to ground acceleration


y0 = [0;0]; % zero initial conditions (interested only in steady-state response)
[T,Y] = ode45(@f,[0:0.02:80.6],y0,[],xi,Tn,t1,a2); % Y(:,1) is in m and Y(:,2) in m/s

%plot of ground acceleration


subplot(2,1,1)
plot(t1,a2,'b')
grid
Title('IMPERIAL VALLEY EARTHQUAKE - EL CENTRO, 270degrees')
axis tight
ylabel('ground acceleration (m/s/s)')
xlabel('time (s)')

%plot of system steady-state response to ground acceleration


subplot(2,1,2)
plot(T,Y(:,1));grid
Title('SDOF (5% damping) response to IMPERIAL VALLEY EARTHQUAKE')
ylabel('displacement (m)')
xlabel('time (s)')
axis tight

%maximum displacement = D
max(abs(Y(:,1)))% max. displacement

TimeResponseSDOFSystems.m
time response of SDOF system with 5% damping and natural period of 10s to El Centro ground motion

The maximum displacement of the mass relative to the ground is seen to be, in this case,
38.41 cm.

3. Response Spectrum
If we repeat the previous analysis for different SDOF systems, maintaining the damping
ratio and varying the natural period of vibration, and represent the maximum value of each
response against the period, we obtain the Displacement (or Deformation) Response
Spectrum of the El Centro Earthquake for a particular damping ratio (it's customary to use
a damping ratio of 5%, which is the normal damping ratio of most buildings):

ResponseSpectra.m
El Centro Deformation Response Spectrum for 5% damping ratio
In logarithmic axes:
El Centro Deformation Response Spectrum for 5% damping ratio

It is also of interest to plot the pseudo-velocity and pseudo-acceleration response spectra.


The pseudo-velocity and pseudo-acceleration are given by:

that is,

The respective plots are: (the MATLAB code for this section can be found in the m-file
ResponseSpectra.m)
Pseudo-velocity response spectrum for El Centro Earthquake, 5% damping ratio

Pseudo-Acceleration response spectrum for El Centro Earthquake, 5% damping ratio

It is also of interest to plot normalised versions of these spectra, that is, normalised w.r.t to
maximum ground displacement, velocity, and acceleration. I shall represent the normalised
displacement spectrum and normalised acceleration spectrum, as these are of particular interest (I
shall why explain in a moment); ie: D/ug0 and A/üg0 :
Relative Deformation response spectrum for El Centro Earthquake, 5% damping ratio

Relative Acceleration response spectrum for El Centro Earthquake, 5% damping ratio


Although the deformation, pseudo-velocity, and pseudo-acceleration response spectra
contain the same information, it is useful to plot all three spectra because each one directly
provides a meaningful physical quantity: the deformation spectrum provides the peak
deformation of a system, the pseudo-velocity is directly related to the peak strain energy
stored in the system during the earthquake, and the pseudo-acceleration times the mass is
equal to the peak equivalent static force and base shear (see Chopra).

Finally, it is also of interest to plot all three spectra (deformation, pseudo-velocity, and
pseudo-acceleration) on the same graph. This is called a tripartite plot, and is possible
because all three separate quantities contain the same information, no more and no less.
Basically, we start off by plotting V against the period (or frequency) on logarithmically
scaled axes, as we did earlier. Now, since:

we see that lines of constant D have a slope of -1 on logarithmically spaced axes, hence, the
axis of deformation has a slope of 1. Similarly, noting that:

that is, lines of constant A have a slope of +1 on logarithmically spaced axes, hence, the axis
of pseudo-acceleration has a slope of -1. Thus, we have the two mutually perpendicular
vertical and horizontal axes for V and T, and another two mutually perpendicular axes for D
and A which for angles of 45º w.r.t. the vertical and horizontal axes. That is why this type of
plot is known as a four way log plot.

Note that for T = 2 π, A = V = D, as can be readily seen from the relations

To represent the three response spectra on a single 4-way log plot, all we need to do is
represent V against T and then plot lines of constant D and lines of constant A. The file
ResponseSpectra.m contains the code I used to generate these lines and their labelling.
Tripartite plot of El Centro Response Spectrum. (Green lines: pseudo-acceleration lines;
maroon lines: constant displacement lines).

Part II: Design Spectra

1. Response Spectra Characteristics


The main reason for plotting all three response spectra on the same plot (tripartite plot)
rather than on seperate plots is because the shape of the spectrum can be approximated
more readily for design purposes with the aid of all three spectral quantities than with one
alone. Let's have another look at the response spectrum for El Centro, for a damping ratio of
5%. This time I plot a larger range of natural vibration periods (Tn=0.01:0.1:50.1) :

Tripartite plot of El Centro Response Spectrum(blue), maximum ground motions(red)

I have also plotted in red the peak ground motion values: the peak ground displacement as
constant displacement line, the peak ground velocity as a constant pseudo-velocity line, and
the peak ground acceleration as a constant pseudo-acceleration line. This is to compare the
three spectral response quantities (displacement, pseudo-velocity, and pseudo-
acceleration) with the respective peak ground motion maxima (note though that peak
pseudo-velocity and pseudo-acceleration are not the peak relative velocity and acceleration
of the system, so strictly speaking it doesn't make sense to plot peak ground velocity and
peak ground acceleration on the same scales as system response peak pseudo-velocity and
peak pseudo-acceleration respectively, at least not on their own, - however, I can only
assume that this is done to normalize the pseudoquantities with respect to ground maxima
(which is okay as we are really only plotting normalised pseudoquantities), which gives us
a means to construct smooth design spectra from peak ground motion parameters based on
the trends in the response spectra which are observed, as we shall see in what follows).
When we observe response spectra such as that plotted for El Centro in the above plot, for
different site readings of earthquake motion or different damping ratios, we notice the
following trends:
 for systems with very short natural periods of vibration, the peak pseudo-
acceleration (A) approaches üg0 and the relative displacement (D) is very small.
Why? Systems with very short periods are very rigid with relatively very little
inertia, thus the mass moves rigidly with the ground. As a consequence,
o D is very small, since D is deformation or relative displacement w.r.t. ground
displacement, and
o relative acceleration (ü(t)) will be nearly zero, which implies that üg(t) ≈ üt(t).
For lightly damped systems, it can be shown that üt(t) ≈ -A(t) (see Chopra),
therefore A ≈ üg0.

Response spectra of El Centro for various damping ratios, figure borrowed from Chopra.

for systems with very long natural periods of vibration (usually above 15s or so), D
approaches ug0, and A is very small (since the forces in the structura are directly related to
mA, these are small too). The reasoning is as follows: systems with very long natural
periods are very flexible with relatively large inertia (inertia >> rigidity), therefore, the
mass remains almost stationary as the ground below it moves. Hence:
 üt(t) ≈ 0 → A(t) ≈ 0, and
 u(t) ≈ - ug(t) → D ≈ ug0.

Response spectra of El Centro for various damping ratios, figure borrowed from Chopra.
To be honest, my plot of the response spectrum for a damping ratio of 5% doesn't
converge to the peak ground displacement, at least not for natural periods of 50s or
less. Initially I thought that constructing the response spectrum for even longer
periods (of little practical interest as far as civil structural design goes) would
eventually lead to this convergence, but the results for the same response spectrum
presented by other authors (such as the one I borrowed from Chopra to illustrate
the trends, as shown above), show that the maximum relative displacement D, (for
different damping ratios too) tends to the maximum ground displacement long
before 50s (clearly not a fault of the different units being used). My results seem to
be OK though for smaller periods, as I have contrasted them with those provided by
other sources - I can only assume that the ode solver I have used of MATLAB is not
producing accurate results for large periods, maybe because the equations for such
periods require other specialized solvers to avoid growth of numerical errors and
such, but that's a topic I shall not go into. I shall just pretend, for this study, that my
plot follows, approximately, the trends I have described above, a summary of which I
show in the following figure:
Trends in Response Spectra relative to ground motion parameters, borrowed from Chopra.

From the previous plot it can be observed that response spectra tend to be characterized by
three regions:
 a region of approximately constant acceleration, in the high-frequency (short
period) portion of the spectra: this region is called the acceleration sensitive region
because structural response seems to be most directly related to ground
acceleration;
 constant displacement, at low frequencies (long-period region): this region is called
the displacement sensitive region as structural response seems to be related most
directly to ground displacement;
 constant velocity, at intermediate frequencies: this region is called the velocity-
sensitive region, because structural response seems to be better related to ground
velocity than to other ground motion parameters.
In each region, the corresponding ground motion is amplified the most (although for very
low and very high periods, the amplification tends to unity, as we have reasoned). (NB,
although I used the term 'amplified', you should bear in mind that it is the pseudo-spectral
quantities that we are viewing as amplifications of ground maxima. Remember that pseduo-
acceleration and pseudo-velocity are not the peak acceleration and peak velocity of the
system response respectively, thus, to be strict, the amplification factors are not showing
how peak ground velocity and ground acceleration are amplified. However, it seems that
the psedo-velocity and pseudo-acceleration are good approximations of relative velocity
and absolute acceleration in certain cases).

Regions of the response spectra, graph borrowed from Chopra.


1b. Time integration and Response Spectrum Calculation Revisited
I've decided to try and improve on two aspects of my procedure so far that I'm not too
happy with. The first, to use another method of integrating the system response in the time
domain, in view of my uncertainty of MATLAB's ode45's performance for high periods. This
will give me a means to contrast the results obtained up till now.
The proposed method of integration of the dynamic equation is the "Piecewise exact
method" (see FEMA notes). Basically, the loading function is appromiated by a series of
straight-line segments, and then, within each of these intervals, the dynamic equation is
solved (rather, the analytical expression of the solution for a ramp input, which is easily
obtainable, is evaluated) as a function of the values of the response at the start of the
interval (initial conditions). This gives us the initial conditions for the next segment. The
deduction of the analytical equations of both the deformation and velocity to a ramp
loading function are shown in the document piecewiseMethod.jpg. The MATLAB file in
which I've implemented the method is piecewiseIntegration.m.
Analytical solution of dynamic equation to ramp loading.
piecewiseIntegration.m.

It's advantageous to use this method because even though it is an approximate method due
to the fact that the loading function is approximated by a series of straight-line segments, in
our case, the loading function is given by a series of discrete points, which essentially is a
series of straight-line segments (it is the most precise data we have of the load anyway). So
barring the precision of the accelerogram data, the method is exact, because we are directly
evaluating the analytical expressions solution to the differential eqution. And it's far less
costly from a compatitonal point of view than using ode45, which is based on a four and
five stage Runge-Kutta method for numerical integration of odes.
The response spectrum, constructed using piecewise integration to solve the dynamic
equation for different natural periods, is shown in the next figure (the code details can be
seen in the m-file ResponseSpectrumUsingPiecewiseIntegration.m
Deformation Response Spectrum for El Centro, 5% damping, obtained using piecewise integration.

A scaled version, w.r.t. maximum ground motion, is shown next:


Scaled Deformation Response Spectrum for El Centro, 5% damping, obtained using piecewise
integration.

The result is interesting, because even for a natural period of up to 50s, the deformation of
the system doesn't taper down to that of the ground's, as I was half-expecting it to. The
amplification of the displacement is roughly the same as that obtained using the ode solver.
I shall take this behaviour to be correct then, despite my earlier scepticism, though I don't
know why it is so. Perhaps one needs to simulate extremely long periods for that trend to
be observed, in the particular case of El Centro.
Even though I have accepted the trend for long periods computed using ode45, as the
piecewise method has predicted the same trend, it is still interesting to compare the results
obtained using both methods. The following figure shows the response spectrum computed
using ode45 in red, and the response spectrum computed using the piecewise method, in
blue. For both, computations of maximum relative displacement were carried out every
0.02 seconds from T=0.01 to 50s.
Despite the piecewise method of integration not having thrown any light as to the trend
issue for long periods, I'm glad I decided to implement it, as the previous graph shows that
the results obtained by the ode solver seem to become somewhat haphazard especially for
longer periods. The piecewise method, which evaluates the system response from the
analytical expression of the solution to the dynamic equation, is 'exact' in this respect (as
far as the loading can be approximated by a series of straight-line segments, which is
actually all the data we have considering the accelerogram is given as a series of discrete
points), and it does not show this haphazard behaviour. I shall henceforth use the piecewise
method, which by my previous arguments, I'd reason provides the more precise results.
Despite the piecewise method of integration not having thrown any light as to the trend
issue for long periods, I'm glad I decided to implement it, as the previous graph shows that
the results obtained by the ode solver seem to become somewhat haphazard especially for
longer periods. The piecewise method, which evaluates the system response from the
analytical expression of the solution to the dynamic equation, is 'exact' in this respect (as
far as the loading can be approximated by a series of straight-line segments, which is
actually all the data we have considering the accelerogram is given as a series of discrete
points), and it does not show this haphazard behaviour. I shall henceforth use the piecewise
method, which by my previous arguments, I'd reason provides the more precise results.

The second aspect I wanted to improve was on the selection of natural period times for the
construction of log-plots of the response spectrum. Notice that if the periods are evenly
spaced, the corresponding data points are not evenly spaced on a log plot, but 'bunched'
towards the right of the graph. Eg, if the previous plot of the El Centro Response Spectrum,
with data points plotted every 0.02 seconds, is plotted on logarithmic axes, we see that the
actual data points (in red) are not evenly spaced, and the left side of the graph is made up
mostly of few straight lines interpolating these data points which are far apart.

For plots of spectra on logarithmic axes, whose object is to magnify the information relative to
shorter periods as against the very large periods (note that for civil engineering purposes, the
natural periods of interest usually are shorter than a couple of seconds at the very most), it makes
much more sense to carry out the calculations at logarithmically spaced points as well. Thus, in
MATLAB if we define Tn as Tn = logspace(-2,1.7,500), we obtain 500 points logarithmically spaced
between 10-2=0.01 and 101.7 ≈ 50s, which will appear linearly spaced on a logarithmic axis:
and the pseudo-velocity response spectrum:
where the red dots are actual data points.
And the tripartite:
where the red dashed-lines indicate peak ground motion values.

2. Design Spectra based on Response Spectra


The jaggedness of a response spectrum, such as the one of El Centro, is a typical feature of
response spectra because they are for individual earthquakes recorded at particular sites
with particular soil characteristics and what not. We cannot use such a spectrum for design
purposes, because small variations in system frequency would produce very different
design demands. And the design would only be valid (optimised) for that particular
earthquake at that particular location.
Jaggedness of Response Spectra, figure borrowed from FEMA.

In contrast to response spectra, design spectra need to be smooth, since the natural period of a civil
engineering structure cannot be determined with that much accuracy, and design specification
should not be very sensitive to a small change in natural period. Based on our previous
observations on the characteristic shapes of response spectra when represented on four-way log
plots (well not our observations, theirs, but you know what I mean), Newmark and Hall decided that
the design spectrum should consist of a series of straight lines in each of the three 'amplification'
zones, and then two connecting lines to the A = üg0 and D = ug0 regions that we saw are characteristic
for very short and very long periods respectively:
Creating a smooth spectrum for design.

Newmark and Hall plotted the response spectra of dozens of earthquakes that were recorded on
firm soil sites for the western United States. Each ground motion was normalised so that all had the
same peak acceleration, eg. üg0 = 1g.
Response Spectra of different earthquakes (for the same damping ratio).

At each period, the mean value of spectral ordinates and the standard deviation were calculated,
and the corresponding spectra were obtained:

Mean and mean + 1 σ spectra for damping ratio = 5%.

he peak ground velocity and displacement are average values over the set of ground
motions (remember that each individual spectrum was normalised w.r.t. the same peak
ground acceleration). The values are (for peak ground acceleration of 1g):

Using these results, and based upong the shape-trends discussed earlier, Newmark and Hall
proposed a simplified spectrum made up of straight lines, based on the previous ground
motion parameters:
Idealised Newmark spectrum (shown for the mean case), for systems with 5% damping ratio.

The design spectra shown in red on the previous plot is for systems with a 5% damping ratio, and
for a 50% probability of exceedance. For different damping ratios, and different probabilities of
exceedance, different idealised spectra are obtained. However, the recommended period values are
Ta = 1/33s, Tb=1/8s, Te=10s, and Tf=33s for all design spectra, whereas the periods T c and Td are
obtained by intersection of the lines bc, cd and de (which are the lines of constant acceleration α Aüg0,
αBu'g0, and αDug0, although plotted as scaled values on the previous graph) for each particular
idealised spectrum. The amplification factors, obtained carrying out the previous construction for
different damping ratios and for both the mean and mean-plus-one-standard-deviation are found to
be:
Amplification factors for Newmark and Hall's design response spectra(from Newmark and Hall, 1982).

The Newmark-Hall idealised response spectrum is an empirically derived elastic design


spectrum (as it's derived from elastic response spectra). To construct the Newmark-Hall
design spectrum for a given ground motion, we start off by plotting the peak ground motion
values (if only the peak ground acceleration is available, the peak ground velocity and
displacement obtained from the average values of the records analysed by Newmark and
Hall, given by

can be used. Then all you have to do is look up the amplification factors from the table for
the corresponding damping ratio and level of exceedance probability desired, and plot the
constant pseudo-acceleration, pseudo-velocity and relative displacement lines between
Tb=1/8s and Te=10s (the intersections of these lines give you Tc and Td). For T 33s, we
plot the lines A/üg0 = 1 and D/ug0 = 1, respectively, and finally connect points a and b, and e
and f with transition lines.
Carrying out this process in MATLAB, using the El Centro Earthquake peak ground motion
values, for a 50% probability of exceedance, and systems with a 5% damping ratio, the
result is: (the code is avaliable in the m-file: NewmarkDesignSpectrum.m).
Newmark Design Spectrum for El Centro Ground Motion, 50% exceedance probability, 5% damping
systems.

The values of Tc and Td (although obtained graphically in the previous plot), can easily be
determined analytically:
The elastic design spectrum by Newmark and Hall provides a basis for calculating the
design force and deformation for SDOF systems to be designed to remain elastic. This basic
form (with some modifications) is now the basis for structural design in seismic regions
throughout the world (typically plotted against structural "period", the inverse of
frequency). A nominal level of damping is assumed (5% of critical damping).

Let's plot the Newmark Design Spectrum for El Centro ground motion parameters, for 5%
damping ratio, that we obtained in the last section, but instead as pseudo-accelertaion
against system period, on arithmetic scales. For each straight-line segment of the graph, the
pseudoacceleration is given by:
 for
 for

 
  for

  for

  for
  for
 
  for

  for

  for
  for

 For
If we plot this in MATLAB: (see file NewmarkDesignSpectrum.m)

Newmark's Design Spectrum for El Centro Earthquake, 50% exceedance probability, 5% damping
ratio.

where I have only plotted to 2.5 seconds, so not all curves are shown. We can compare the El Centro
response spectrum with the design spectrum for El Centro - we did this earlier on the triparte
representation, but let's see how it looks on the pseudoacceleration-T arithmetic axes:
El Centro Response Spectrum (blue), Newmark Design Spectrum for El Centro for 50% exceedance
probability. Damping ratio 5%.

3. Seismic Hazard Analysis


The Newmark design spectrum was developed from records on rock site. However, one
would expect that the site conditions influence the frequency content in the ground motion,
and therefore the spectral amplification factors would depend on them. Although studies
have been carried out and site dependent spectra have been obtained (amplification factors
for different soil types), the Newmark design spectrum was developed for a specific
collection of ground motion records. Where no ground motion records are available, the
procedure developed by Newmark for constructing design spectra would be extremely
limited.
What has been done is that Seismic hazard models of sites are used to develop another kind
of design spectrum known as a Uniform Hazard Spectrum that is used today for design
purposes. I will briefly introduce the concepts of Seismic Hazard Analysis.
Seismic Hazard Analysis describes the potential for dangerous, earthquake-related natural
phenomena such as ground shaking, fault rupture, or soil liquefaction. It should not be
confused with Seismic risk analysis, which assesses the probability of occurrence of losses
(human, social, economic) associated with the seismic hazards. For example, a hazard
associated with earthquakes is ground shaking. The risk is structural collapse and, possibly,
loss of life.
Seismic Hazard Analysis includes the quantitative assesment of expected levels of ground
shaking at a given location due to future earthquakes in the surrounding region. It is based
on the development of a seismicity model for the location and size of future earthquakes in
the region, and a ground motion model by which the level of shaking can be predicted at a
given site as a result of the earthquake scenarios predicted by the seismicity model. The
integration of the two models provide us with a model for the expected levels of shaking at
the site of interest.
A seismicity model, or more precisely, a seismic-hazard source model, is a description of
the magnitude, expected locations, and frequency of all earthquakes that affect a given site.
It is based on geological studies and records of seismic sources (active fault lines of the
region), regional earthquake catalogues, instrumental recordings, and such. I will not go
into details, as it is an extremely complex area of study, but basically the model must
specify a list of scenarios (seismic sources that would cause seismic hazard at the site of
study), quantifying the magnitude, location, and rate of each one of them.

Consideration of the different source types that could affect a given site
in the development of a seismicity model for the site.

The ground motion model must give estimations of the ground motions caused by
earthquakes. The most basic ground motion models give the ground motion level (which
may be any parameter that characterises the shaking, such as peak ground acceleration
(PGA), spectral acceleration (for a certain period), etc) as a function of magnitude and
distance of the earthquake, but more complex models include other parameters to allow for
different site types (eg, rock vs soil). Equations are developed by fitting analytical
expression to observations (or to synthetic data where observations are lacking).

Example of an empirical attenuation relationship.


Illustration of attenuation with distance.
The attenuation relationships, which form part of the ground motion model, are derived for
a particular region of interest. It depends a lot on the geological features of the region.

Probabilistic Seismic Hazard Alalysis (PSHA).


A deterministic seismic hazard analysis (DSHA) would simply consider the main sources
that affect a site, determine their magnitudes, and determine the ground motion that each
one would provoke at the site (using the attenuation relationships):

Deterministic seismic hazard analysis.


In a probabilistic analysis, a recurrence relationship (eg, frequency of earthquakes above a
certain magnitude) is introduced as is the uncertainty in each step of the process. Each of
the uncertainties are included in a probabilistic analysis, and the result is a seismic hazard
curve that relates the design motion parameter (eg PGA) to the probability of exceedance:

Example of Hazard Curve: annual frequency of exceedence vs Peak Ground Acceleration.


The probabilistic seismic hazard provides the annual exceedence probability for given ground
motions in hazard curves, and this methodology has today practically substituted the classical
deterministic methods that determine ground shaking from mapped faults assuming a certain
rupture
Once a hazard curve has been developed, we can obtain a design ground motion. The
hazard curve represents values of the average annual rate of exceedance for any given
ground-motion value. Assuming that ground motions may be described by a Poisson
distribution over time, then, if we define N(t) as the number of events that take place in
time interval [0,t] (random variable), the probability of there being k number of ground
motion events in a time t is given by:

where λ is the average number of events that occur in one time unit ( λt is the average
number of events that occur in a time t). Note that λ (which defines the probability density
function), is provided by the hazard curves as a function of the ground motion parameter.
The probability of no events in a time period 0 to t is thus:

and the probability of there being at least one event in the time 0 to t is thus:
It follows that the average annual rate of exceedance corresponding to a probability P of at
least once exceedance within a time period t is given by:

For example, if a designer wished to design a dam for the ground motion that has only a
10% probability of being exceeded in a 50 year period, first they would compute the
corresponding average rate of exceedance that defines the poisson distribution which
would give this probability for at least one event to occur in t=50yrs:

(this corresponds to a return-period of 1/λ ≅ 474.6 yrs):

The ground motion parameter (eg PGA) would be taken from the seismic hazard curve for
the site of the dam:
In other words, there is a 10% percent chance that an earthquake that would provoke a
PGA of more than 0.33g at the site of the dam will occur in 50 years. (The hazard curve tells
us that a PGA that exceeds 0.33g occurs on average 0.021 times per year, or once every 475
yrs (return period)). (Likewise, for example, there would be a 63.21% probability for at
least one ground motion of PGA above 0.33g to occur at the site in a period of time of 475
years:

(note that a 10% probability of exceedance in 50yrs has the same return period (475 years)
(or equivalently, a frequency of exceedance of 1/475) as a 63.21% probability of
exceedance in 475 years. There are infinte combinations of P and t for a given λ. The hazard
curve tells us that at the site in question, a frequency of exceedance of 1/475 events/year
corresponds to the frequency of there being a ground motion with PGA above 0.33g at that
site). For building design, a 50-year base line is used as this is estimated as the service life
of a typical building. For structures such as dams, a longer return period may be more
appropriate as the required service life might be much longer.
Once a ground motion level has been defined, it is a question of engineering techniques to
design a structure that can withstand the shaking.
Seismic Hazard Maps.
Seismic maps are a representation of seismic hazard levels on territorial maps. The maps
are typically expressed in terms of probability of exceeding a certain ground motion.
Hazard maps commonly specify a 10% chance of exceedance (90% chance of non-
exceedance) of some ground motion parameter for an exposure time of 50 years,
corresponding to a return period of 475 years (this return period is selected rather
arbitrarily). The SESAME project has developed a continental-scale hazard map for Europe:
The Seismic Hazard Map of the European-Mediterranean region, in terms of peak ground
acceleration at a 10% probability of exceedance in 50 years for stiff soil conditions, published
in 2003.
This map depicts Peak Ground Acceleration (PGA) with a 10% chance of exceedance in 50
years for a firm soil condition. PGA, a short-period ground motion parameter that is
proportional to force, is the most commonly mapped ground motion parameter because
current building codes that include seismic provisions specify the horizontal force a
building should be able to withstand during an earthquake.
The 10% probability of exceedance in 50 years maps depict an annual probability of 1 in
475 of being exceeded each year. This level of ground shaking has been used for designing
buildings in high seismic areas. The maps for 10% probability of exceedance in 50 years
show ground motions that we do not think will be exceeded in the next 50 years. In fact,
there is a 90% chance that these ground motions will NOT be exceeded. This probability
level allows engineers to design buildings for larger ground motions than what we think
will occur during a 50-year interval, which will make buildings safer than if they were only
designed for the ground motions that we expect to occur in the next 50 years.
The purpose of representing earthquake actions in a seismic design code such as EC8 is to
circumvent the necessity of carrying out a site-specific seismic hazard analysis for every
engineering project in seismically active regions. For non-critical structures it is generally
considered sufficient to provide a zonation map indicating the levels of expected ground
motions throughout the region of applicability of the code and then to use the parameters
represented in these zonations, together with a classification of the near-surface geology, in
order to construct the elastic design response spectrum at any given site.

4. Uniform Hazard Spectra


If a series of seismic hazard curves is developed for a range of different parameters (e.g.,
PGA, 0.1 sec spectral acceleration (fo a given damping ratio), 0.2 sec spectral acceleration,
etc.) and values are extracted for a certain constant probability (say a 10% in 50 year
probability of exceedance, which determines a return period, in this case of 475 years),
then a design response spectrum may be created by plotting the parameter magnitudes vs
period. The following plot shows the first point on such a spectrum -- in this case, the PGA
with a 10% probability of being exceeded in 50 years:
PGA (= üg0) = 0.33g is the peak ground acceleration that has a probability of 10% of being
surpassed in 50 years at the site.
If we now repeat the process with the hazard curve that gives us, for different values of
spectral acceleration of SDOF systems with a natural period of 0.2s, the average frequency
of exceedance, at the site, of each of these spectral accelerations, :

PSA = 0.55g is the peak spectral acceleration of a T=0.2s (5% damped) system, that has a
probability of 10% of being surpassed in 50 years at the site.
If the process is continued, the result is a complete elastic response spectrum. It is called a
uniform hazard spectrum (UHS) because each point on the spectrum has the same
probability of being exceeded in the given period. These spectra could then be used in a
response spectrum analysis of the structure.
Uniform Hazard Spectrum, 5% damped systems, 10% prob of exceedance in 50 years.
The UHS is considered an appropriate probabilistic representation of the basic earthquake
actions at a particular location. The UHS will often be an envelope of the spectra associated
with different sources of seismicity, with short-period ordinates controlled by nearby
moderate-magnitude earthquakes and the longer-period part of the spectrum dominated
by larger and more distant events. As a consequence, the motion represented by the UHS
may not be appropriate for artificial ground motion generation:

It should be possible to generate a realistic earthquake ground motion that matches the small
or large earthquake. However, a generated earthquake that matches the UHS would be
unrealistic (and possibly too demanding).
However, if the only parameter of interest to the engineer is the maximum acceleration that
the structure will experience in its fundamental mode of vibration, regardless of the origin
of this motion or any other of its features (such as duration), then the UHS is a perfectly
acceptable format for the representation of the earthquake actions.
5. Design Codes
Until the late 1980s, seismic design codes invariably presented a single zonation map,
usually for a return period of 475 years, showing values of a parameter that in essence was
the PGA. This value was used to anchor a spectral shape specified for the type of site,
usually defined by the nature of the surface geology, and thus obtain the elastic design
spectrum. In many codes, the ordinates could also be multiplied by an importance factor,
which would increase the spectral ordinates (and thereby the effective return period) for
the design of structures required to perform to a higher level under the expected
earthquake actions, either because of the consequences of damage (e.g. large occupancy or
toxic materials) or because the facility would need to remain operational in a post-
earthquake situation (e.g. fire station or hospital). A code spectrum constructed in this way
would almost never be a UHS. Even at zero period, where the spectral acceleration is equal
to PGA, the associated return period would often not be the target value of 475 years since
the hazard contours were simplified into zones with a single representative PGA value over
the entire area. More importantly, this spectral construction technique did not allow the
specification of seismic loads to account for the fact that the shape of response spectrum
varies with earthquake magnitude as well as with site classification, with the result that
even if the PGA anchor value was associated with the exact design return period, it is very
unlikely indeed that the spectral ordinates at different periods would have the same return
period (McGuire, 1977). Consequently, the objective of a UHS is not met by anchoring
spectral shapes to the zero-period acceleration.
Within the drafting committee for EC8 there were extensive discussions about how the
elastic design spectra should be constructed, with the final decision being an inelegant and
almost anachronistic compromise to remain with spectral shapes anchored only to PGA. In
order to reduce the divergence from the target UHS, however, the code introduced two
different sets of spectral shapes (for different site classes), one for the higher seismicity
areas of southern Europe (Type 1) and the other for adoption in the less active areas of
northern Europe (Type 2). The Type 1 spectrum is in effect anchored to earthquakes of
magnitude close to Ms ~7 whereas the Type 2 spectrum is appropriate to events of Ms 5.5.
At any location where the dominant earthquake event underlying the hazard is different
from one or other of these magnitudes, the spectrum will tend to diverge from the target
475-year UHS, especially at longer periods.
5.1 Eurocode 8 Elastic Design Spectra.
"For the purpose of this Eurocode, national territories shall be subdivided by the
National Authorities into seismic zones, depending on the local hazard. By definition,
the hazard within each zone can be assumed to be constant.

"For most of the applications of this Eurocode, the hazard is described in terms of a
single parameter, i.e. the value of the reference peak ground acceleration on rock or
firm soil agR. Additional parameters required for specific types of structures are given
in the relevant Parts of Eurocode 8.

"The reference peak ground acceleration, chosen by the National Authorities for each
seismic zone, corresponds to the reference return period chosen by National
Authorities. To this reference return period an importance factor γ1 equal to 1.0 is
assigned. For return periods other than the reference the design ground acceleration
ag is equal to agR times the importance factor γ1.

"Within the scope of EN 1998 the earthquake motion at a given point of the surface is
generally represented by an elastic ground acceleration response spectrum, henceforth
called “elastic response spectrum”

"Two different shapes of response spectra, Type 1 and Type 2 may be adopted. The
National Authority must decide which elastic response spectrum, Type 1 or/and Type
2, to adopt for their national territory or part thereof. In selecting the appropriate
spectrum, consideration should be given to the magnitude of earthquakes that affect
the national territory or part thereof. If the largest earthquake that is expected within
the national territory has a surface-wave magnitude Ms not greater than 5½, then it is
recommended that the Type 2 spectrum should be adopted..

"The elastic response spectrum Se(T) for the reference return, period is defined by the
following expressions:

design spectrum parameters, EC8


EC8 5% damped, elastic spectra. Type 1 spectra are enriched in long period and are suggested
for high seismicity regions. Conversely, Type 2 spectra are proposed for low to moderate
seismicity areas (like France), and exhibit both a larger amplification at short period, and a
much smaller long period contents, with respect to Type 1 spectra. These propositions,
however, were constrained using few events mostly recorded on analogical instruments.

the parameters S, TB, TC and TD are given in tables for the different subsoil types for each of
the two types of spectra.
The influence of near-surface geology on response spectra is taken into account via the soil
parameter S. The fact that locations underlain by soil deposits generally experience
stronger shaking than rock sites during earthquakes has been recognised for many years,
both from field studies of earthquake effects and from recordings of ground motions. The
influence of surface geology on ground motions is now routinely included in predictive
equations. Code specifications of spectral shapes for different site classes generally reflect
the amplifying effect of softer soil layers, resulting in increased spectral ordinates for such
sites, and the effect on the frequency content, which leads to a wider constant acceleration
plateau and higher ordinates at intermediate and long response periods.
EC8 is unique amongst seismic design codes in that it is actually a template for a code
rather than a complete set of definitions of earthquake actions for engineering design. Each
member state of the European Union will have to produce its own National Application
Document, including a seismic hazard map showing PGA values for the 475-year return
period, select either the Type 1 or Type 2 spectrum and, if considered appropriate, adapt
details of the specification of site classes and spectral parameters.
The basic mechanism for defining the horizontal elastic design spectrum in EC8 is outdated
and significantly behind innovations in recent codes from other parts of the world, most
notably the US. It is to be hoped that the first major revision of EC8, which should be carried
out 5 years after its initial introduction, will modify the spectral construction technique,
incorporating at least one more anchoring parameter in addition to PGA.
In the Luso-Iberian peninsula, seismic hazard is the result of moderate-magnitude local
earthquakes and large-magnitude earthquakes offshore in the Atlantic. The Spanish seismic
code handles their relative influence by anchoring the response spectrum to PGA but then
introducing a second set of contours, of a factor called the ‘contribution coefficient’, K, that
controls the relative amplitude of the longer-period spectral ordinates; high values of K
occur to the west, reflecting the stronger influence of the large offshore events.

5.2 NCSR-02 (Spanish Seismic Design Code) Elastic Response Spectrum.


parameters for determining elastic response spectrum, NCSR-02.
elastic response spectrum, NCSR-02.

Are We Ready for El Centro?


I shall firstly construct the design spectrum for Malaga City, according to NCSR-02.
According to the zonation map (hazard map) (see NCSR-02), we have that the values of a b
and K are respectively 0.11g and 1. Assuming a construction of normal importance, we take
the importance factor equal to unity ( ρ = 1). Also, assuming the site is located on firm soil
(hard rock), C = 1, and the soil coefficient is found to be S = 0.8067.
Elastic Design Spectrum for Malaga according to NCSR-02.

Comparison of El Centro Design Spectrum with NCSR-02 Design Code for Malaga
I shall plot this design spectrum in MATLAB, as well as the Newmark-Hall pseudo-
acceleration design spectrum for El Centro, for a 50% prob. of exceedance, for systems with
a 5% damping ratio. The NCSR spectrum corresponds to a 10% of exceedance in a period of
50 years (return period 475 yrs). The code for generating the plot can be found in the m-
file: ComparisonNewmarkNCSR02.m.
Comparison of Newmark Design Spectrum for El Centro (50% prob exceedance) with
Elastic Design Spectrum for Malaga given by NCSR-02, for stiff soil and a return period of 475
yrs.
References.
Anil K. Chopra "Dynamics of Structures" .
FEMA notes
Farzad Naeim The Seismic Design Handbook.
Response Of Structures to Earthquake GroundMotions.
Lecture 18 - Earthquake Response Spectra..
Ahmed Y. Elghazouli (edited by), Seismic Design of Buildings to Eurocode 8
European Seismological Commission. UNESCO-IUGS International Geological Correlation
Program Project no. 382 SESAME
Eurocode 8
NCSR-02
El Centro Accelerogram Data;
(individual files I used for reading data to MATLAB):
 Acceleration Data
 Velocity Data
 Displacement Data

Vous aimerez peut-être aussi