Ground Response

Analysis
Asrat Worku (Dr-Ing)
Associate Professor of Civil Engineering, AAU
Introduction
• Ground response analyses are employed
• To predict ground surface motions as input to structural
models in form of
• Time histories in time domain (of accelerations, velocities and
displacements);
• Spectra in frequency domain (response spectra, Fourier spectra,
etc)
• To evaluate stresses and strains in soils to assess
liquefaction potential
• To predict seismic forces that can lead to instability of
natural and man-made earthen slopes
Introduction
• Bedrock motion is assumed to be known by other
means
• Attenuation relationships are used that account for
• Source conditions;
• Travel path geology;
• Distance
• Despite the significant attenuation of seismic waves
travelling tens of kilometers of distance from the
source, the site soil of thickness just few tens of
meters can be decisive in determining the
destructive potential of the seismic ground motion at
bases of structures
Introduction
• The influence of site soils on seismic ground
motions has been recognized since the 1920’s
• Significant progress was made since the Niigata
and Alaska earthquakes of 1964 that caused
significant failures attributed to site effect.
• A number of methods have been developed over
the years that followed based on
• Constitutive laws: Linear, equivalent linear and nonlinear
• Dimensionality: 1D, 2D and 3D
• Most popular are 1D equivalent-linear and nonlinear
techniques
1D Ground Response Analysis
• As waves travel away
from source faults deep
in the earth’s crust, they
reach the soil-rock
interface nearly
vertically due to multiple
refraction at interfaces
of layers of
progressively
decreasing stiffness.
1D Ground Response Analysis
• 1D ground response
analyses methods are
based on
• A vertically incident SH waves at
the soil-rock interface;
• Stratification within the soil is
assumed horizontal.
• Techniques based on these
assumptions give results in
good agreement with
measurements
• The definitions shown in the
figure are commonly used
1D Ground Response Analysis: Linear Analysis
Transfer Functions
• Transfer functions are used to express responses at the
surface in terms of the input motion at the bedrock
interface
• Responses could be acceleration, velocity,
displacement, stresses, strains etc.
• Procedures that use transfer functions are based on the
superposition principle – thus linear response is implied.
• Transfer functions for a serious of progressively more
complicated geotechnical conditions are considered
next
• Linear response is dealt with first; nonlinear problems
can be handled using iterative techniques and
equivalent linear properties.
1D Ground Response Analysis: Linear Analysis
Uniform, Undamped Soil layer on Rigid
Bedrock
• Consider the uniform, isotropic elastic layer shown
excited by a harmonic SH-wave entering from the rock
• The vertically traveling waves in the soil layer is given
by the 1D wave equation:

Note that no damping

was considered in
the derivation of the
wave equation and G
is the small strain
shear modulus.
1D Ground Response Analysis: Linear Analysis
Uniform, Undamped Soil layer on Rigid
Bedrock
• The solution is given by

 
u z, t   Aei t z   Bei t z   Aeiz  Be iz eit

• Where κ=ω/Vs is the wave number. The first term

represents waves travelling upwards and the second
represents waves travelling downwards
1D Ground Response Analysis: Linear Analysis
Uniform, Undamped Soil layer…
• The BC that the shear stress at z=0 vanishes yields

• This results in A=B so that

• This represents a standing wave of amplitude

• This is created by the constructive upward and
downward traveling waves resulting in a constant
shape of deformation of the soil layer
1D Ground Response Analysis: Linear Analysis
Uniform, Undamped Soil layer…
• The transfer function for the surface deformation w.r.t. the
bedrock motion is defined and obtained as

• Its modulus is

• Since the denominator is less than or equal to 1, H and vs

being constant for a given soil layer, the surface displacement
amplitude is always larger or equal to the bedrock amplitude
depending on the frequency indicating that the soil layer
amplifies the rock motion
1D Ground Response Analysis: Linear Analysis
Uniform, Undamped Soil layer…
• Resonance occurs when
• The plot of the amplitude against ω (or kH) is as shown below
Note the infinite displacement at resonance
Caused due to omission of damping.
Displacements are, however, always finite due to
existence of energy dissipation
1D Ground Response Analysis:
Uniform, Undamped Soil layer…
• Example
1D Ground Response Analysis: Linear Analysis
Uniform, Damped Soil layer on Rigid Rock
• A more realistic approach to account for damping is to
use the Kelvin-Voigt visco-elastic constitutive model,
instead of an elastic one, that is given by
  G  
• Note that the additional second term depends on the
strain rate. Using this model, the elastic wave equation
derived earlier takes the form

• It can be shown that the solution to this equation is

similar in form to the undamped case and is given by:
1D Ground Response Analysis: Linear Analysis
Uniform, Damped Soil layer on Rigid Rock
u z , t   2 A cos  * z e it

• Where κ*=ω/vs* is the complex wave number

• Following similar steps, one obtains the transfer

function as

• Introducing the frequency-independent shear modulus

1D Ground Response Analysis: Linear Analysis
Uniform, Damped Soil layer on Rigid Rock
• For small values of damping,

• The transfer function then becomes

• With the identity of

1D Ground Response Analysis: Linear Analysis
Uniform, Damped Soil layer on Rigid Rock
• Again, for small damping, , and the modulus
takes finally the form

• This represents a frequency-dependent

amplification, which will never become infinite
because of the damping (see its plots below for
different values of damping ratio)
1D Ground Response Analysis: Linear Analysis
Uniform, Damped Soil layer on Rigid Rock

• Obviously, the influence

of damping is higher for
responses at higher
frequencies

• Peak values occur at

1D Ground Response Analysis: Linear Analysis
Uniform, Damped Soil layer on Rigid Rock
• Hence

• The maximum amplification occurs at the fundamental

frequency (n=0) so that

• The corresponding period is known as the characteristic

site period and is

• This provides an important indication of the period, at

which the most significant amplification occurs at a soil
site.
1D Ground Response Analysis: Linear Analysis
Uniform, Damped Soil layer on Rigid Rock
• At each frequency a standing wave develops. Mode shapes
corresponding to the first three lowest frequencies are
obtained from the expression in the brackets of the solution
without damping
u z , t   2 A cos z eit   z eit
1D Ground Response Analysis: Linear Analysis
Uniform, Damped Soil layer on Elastic Rock
• In this case downward-travelling waves are
partly reflected and partly transmitted at the
soil-rock interface.
• The latter helps in dissipating some energy
unlike in the rigid case, in which the energy that
entered from the rock remains trapped in the
soil layer.
• This provides one form of radiation damping
that further reduces the vibration amplitudes as
compared to the rigid-rock case.
1D Ground Response Analysis: Linear Analysis
Uniform, Damped Soil layer on Elastic Rock
• Consider the following soil layer overlying an
elastic rock half space
1D Ground Response Analysis: Linear Analysis
Uniform, Damped Soil layer on Elastic Rock
• The displacements due to the traveling waves
are (in the soil and rock, respectively)

• Where the subscript s stands for soil and r for rock

• The surface boundary condition requires As=Bs

1D Ground Response Analysis: Linear Analysis
Uniform, Damped Soil layer on Elastic Rock
• The continuity • where
conditions are

is the complex
• From these conditions
impedance ratio,
follow and v*ss and v*sr are
the shear-wave
velocities of the soil
and elastic rock
1D Ground Response Analysis: Linear Analysis
Uniform, Damped Soil layer on Elastic Rock
• Suppose a shear wave of amplitude, A, travels
through the rock. If the soil didn’t exist, the free-
surface effect results in a bedrock outcropping
motion of amplitude of 2A.
• With the presence of the soil the free-surface
motion amplitude becomes
1D Ground Response Analysis: Linear Analysis
Uniform, Damped Soil layer on Elastic Rock
• With F3 defined as the ratio of the free-surface
amplitude to the outcropping amplitude

• After Euler’s expansion,

• The modulus becomes

1
F3  ,   
cos 2  s* H   z2 sin 2  s* H
1D Ground Response Analysis: Linear Analysis
Uniform, Damped Soil layer on Elastic Rock
• In order to observe the influence of the elastic rock, we
intentionally drop the damping. Thus for the undamped
soil, one obtains

• Note that unlike the rigid-rock case, there would be no

resonance even if damping is zero, as the denominator is
always positive.
• The larger the impedance ratio, the smaller the free-
surface amplitude. This is as a result of waves
transmitted back into the elastic rock.
1D Ground Response Analysis: Linear Analysis
Uniform, Damped Soil layer on Elastic Rock
• The influence of the bedrock stiffness is clearly seen in the plots
shown for selected impedance ratios
• This finding is important in accounting for the influence of rock
stiffness on soil amplification (Note: Design spectra for rock sites
of East and West USA are different for a related reason)
1D Ground Response Analysis: Linear Analysis
Layered, Damped Soil layer on Elastic Rock
• Soils are often layered with varying stiffness
• Consider the soil with N layers on an elastic rock as
shown
1D Ground Response Analysis: Linear Analysis
Layered, Damped Soil layer on Elastic Rock
• Representing each layer as a Kelvin-Voigt solid, the
solution to the wave equation is as usual

• Introducing the local coordinate, z, compatibility

requires for the deformation at each interface

• But
1D Ground Response Analysis: Linear Analysis
Layered, Damped Soil layer on Elastic Rock
• Then, the deformation compatibility condition
yields

• Similarly, compatibility for shear stress at each

interface demands

• But
1D Ground Response Analysis: Linear Analysis
Layered, Damped Soil layer on Elastic Rock
• And

• Thus, the stress compatibility yields

• From the two conditions follow

1D Ground Response Analysis: Linear Analysis
Layered, Damped Soil layer on Elastic Rock
• In the above relations, the complex impedance
ratio is defined as

• Since the shear stress is zero at the ground

surface, it follows from the above relations
1D Ground Response Analysis: Linear Analysis
Layered, Damped Soil layer on Elastic Rock
• By applying the above equations for A and B
successively, frequency functions relating
amplitudes in any layer, m, to that of the first
layer can be obtained as

• The transfer function relating displacements at

any two layers can be given in form of
1D Ground Response Analysis: Linear Analysis
Layered, Damped Soil layer on Elastic Rock
• Because of the harmonic nature of the motion,
the same function is applicable to the velocities
and accelerations.
• Hence, if the motion at a certain point is known,
the motion at any other point can be computed
– (a result useful for deconvolution)
1D Ground Response Analysis: Linear Analysis
Use of the Transfer Functions
• Once the transfer functions are known the
complete analysis involves the following steps:
• The known bedrock time history is transformed into Fourier
series
• Each term in the series is multiplied by the appropriate transfer
function for the response required
• The ground response is obtained as the sum of the individual
products
• The time history of the ground response is obtained through
inverse Fourier transformation of the sum
• Thus, the transfer function determines which
frequencies are amplified most and which ones are
deamplified
1D Ground Response Analysis: Nonlinear
Response – Equivalent Linear Approximation
• The explanation in the preceding slides assumes
linear parameters – thus, it is directly applicable for
layers modelled as a linear elastic material.
• As already noted, the ELA is a modification of the
linear approach available for a better estimate of the
response, especially in problems involving
moderately large strains such as caused by moderate
to strong earthquakes.
• Analysis is done in the frequency domain on a 1D soil
column idealized as a continuum of infinite lateral
extent.
1D Ground Response Analysis: Nonlinear
Response – Equivalent Linear Approximation
• Actual hysteretic stress-strain behavior is
approximated by equivalent soil
properties: hysteretic damping ratio:
1 Aloop 1 Aloop
 
4 Atriangle 2 Gsec c 2
• The two parameters are strain
dependent. There is thus a need to deal
with them iteratively
• An objective strain level is needed to
select these parameters

• Invariably, computer codes are necessary

to conduct the ELA
1D Ground Response Analysis: Nonlinear
Response – Equivalent Linear Approximation
• A typical soil shear stress-strain response to cyclic loading
is shown for two level of strains (Kottke, 2010).
1D Ground Response Analysis: Nonlinear
Response – Equivalent Linear Approximation
• The corresponding difference in shear modulus and
damping are shown on the G/Gmax and damping curves
(Kottke, 2010).
1D Ground Response Analysis: Nonlinear
Response – Equivalent Linear Approximation
• Since computed strain values depend on the equivalent
linear properties, the iterative approach takes place as
shown schematically. This is illustrated below:
1D Ground Response Analysis: Nonlinear
Response – Equivalent Linear Approximation
• The G/Gmax and ξ curves generated from lab test results are based on peak shear
strains of harmonic loads that occur regularly until the excitation is stopped.

• On the other hand, a typical time history of an earthquake ground motion is highly
irregular with few spikes occurring within a short period of time as shown in
comparison with a harmonic load of same peaks.

• This difference needs to be accounted for in the analysis programs

• Obviously, the harmonic loading is more severe; it is thus logical to employ a

reduced peak of the transient record, commonly 50% to 70%, often 65%, called
effective strain
1D Ground Response Analysis: Nonlinear
Response – Equivalent Linear Approximation
• The iteration procedure:
• Estimate initial G and ξ for each layer (usually small-strain values see Point (1)
above);
• Use these to compute time histories including that of shear strain for each layer as in
the linear case
• Compute effective/reduced shear strain for each layer, j, as

• The factor may be taken as the factor suggested above (=0.65) or as the magnitude
dependent one proposed by Idriss and Sun (1992):

• Use the effective shear strain to obtain new equivalent linear properties, and (see
(2) and (3) above)
• Repeat these steps until G and ξ converge sufficiently.
1D Ground Response Analysis: Nonlinear
Response – Equivalent Linear Approximation
• Since the Fourier-series computation of strains in each step is a
linear method, this procedure is an approximation to the actual
nonlinear behavior of soils

• Though the soil properties, G and ξ, are made strain-compatible

through iteration, they are constant throughout the earthquake
duration regardless of the varying strains at particular time
points in the duration

• The procedure was first coded in the widely-used software

SHAKE (Schnabel et al, 1972) and in its subsequent variants
and other more recent codes like DEEPSOIL developed in the
decades that follow.
1D Ground Response Analysis: Nonlinear
Response – Nonlinear Approach
• Here the soil column is idealized as a system of lumped
masses and non-linear springs. The usual equations of
motion of an MDF oscillator is found.

• Direct numerical integration of the equations of motion in the

time domain in small time steps is undertaken.

• The approach accommodates for any nonlinear stress-strain

relation or advanced constitutive model.

• Most models attempt to directly trace the actual stress-strain

hysteresis loops.
1D Ground Response Analysis: Nonlinear
Response – Nonlinear Approach
• In order to define the hysteretic rule, the method relies on a
mathematical relationship for the backbone curve and the
loading-unloading curves as explained in the previous chapter.

• The most famous first generation non-linear models use the

explicit stress-strain Kondner-Zelasko (KZ) model with a
single-parameter hyperbolic rule for the backbone curve given
by (Matasovic and Vucetic, 1993)

• The single parameter is .

1D Ground Response Analysis: Nonlinear
Response – Nonlinear Approach
• A more common improved backbone curve is the multi-
parameter Matasovic-Kondner-Zelasko (MKZ) model,
which is a modified version of the original KZ hyperbolic
model:

• Here, α and s are additional parameters introduced with

the aim to better fit actual cyclic stress-strain curves.
1D Ground Response Analysis: Nonlinear
Response – Nonlinear Approach
• Some more advanced models include Ramber-Osgood model,
Hardin-Drnevich-Cundall-Pyke (HDCP) model, Martin-Davidenkov
model and Iwan-type model
• Also, a number of techniques can be used to integrate the equations
of motions in short time steps
• Some computer programs that use non-linear models are listed below
END
Thank You