SEISMIC DISPLACEMENT ANALYES

The seismic response of

inifinitely long slopes:
From linear to non-linear
approaches

1 3.8
1. Dynamic elastic solution
2. Dynamic elastic-viscoelastic solution
3. Reduction to one degree of freedom
4. 2D and 1D SS elastic plastic solutions
5. Reliability of numerical solutions

3.8 2
Dynamic solution (1D geometry):
the dynamic amplification phenomenon
Balance of momentum

with

3.8 3
 1. 1D dynamic elastic solution: 1
the dynamic amplification phenomenon
In case of linear isotropic elasticity, the solution is
independent of slope inclination (n and t are substituted by
z and x)
By changing variables and by imposing the
acceleration along vertical direction to be nil:
∂ 2 ui Solution:
s ij , j + ρ =
0
∂t 2

∂τ xz ∂τ xy ∂σ x ∂ 2 ux Ut= Ut(z,t)
+ + +ρ 2 = 0
∂z ∂y ∂x ∂t Elastic constitutive relationship
∂τ xy ∂σ x

τ xz = Gγ xz
= = 0
∂y ∂x
∂τ xz ∂ 2 ux
+ρ 2 = 0
∂z ∂t 3.8 4
 ∂u x ∂u z ∂u x
γ xz =
−( + )=
− Compatibility equation
∂z ∂x ∂z

∂γ xz ∂ ux 2
G +ρ 2 = 0 D’Alambert equation
∂z ∂t

∂ ux ρ ∂ ux
2 2
− =
0
∂z 2
G ∂t 2
Uncoupled one dimensional equation
3.8 5
PARTIAL DIFFERENTIAL EQUATION OF HYPERBOLIC TYPE

∂ ux ρ ∂ ux
2 2
− =
0
∂z 2
G ∂t 2

∂ ux G ∂ ux G
2 ∂ ux
2 2 2
= = vs 2 vs =
∂t 2
ρ ∂z 2
∂z ρ

a2-b2 = (a-b) (a+b)

∂ ∂ ∂ ∂
( − vs )( + vs )u x ( z , t ) =
0
∂t ∂z ∂t ∂z
3.8 6
 1 1 z z
t =(η + ξ ) ; z =(η − ξ )
2 2 ξ=
t− ; η=
t+
vs vs
Variable
∂ 1 ∂ 1 ∂ substituttion ∂ 2 ux
= ( − );
∂ξ 2 ∂t vs ∂z =0
∂ξ∂η
∂ 1 ∂ 1 ∂
= ( +
∂η 2 ∂t vs ∂z
) ux F (ξ ) + G (η )
=

CHARACTERISTIC LINE DEFINITION

∂ux ∂ux ∂ux ∂ux

− vs = 0 + vs = 0
∂t ∂z ∂t ∂z
u x ( z , t=
) au x1 ( z − vs t ) + bu x 2 ( z + vs t )
3.8 7
To integrate INITIAL AND BOUNDARY CONDITIONS
must be imposed

z The inclination of these lines

coincides with wave velocity

3.8 8
 Harmonic progressive waves:
vs = ω/κ
Vs = ω/ κ =u x a cos(kz − ωt )

ωt =2π v

Κ λ = 2π
λ = 2π/k
λ wave length ω vs
fr = ω/2π = Αngular frequency
= k =
vs
; λ 2π
ω
3.8 9
 Stationary waves 1

A stationary wave can be obtained by superimposing fwo

harmonic waves with the same amplitude propagating in the
opposite direction:

u x a cos(kz − ωt ) + a cos(kz + ω=
= t ) 2a cos kxz cos ωt
since che
dato 2π

cos( =
p + q ) cos p cos q − senpsenq

ω
In general :
u x = a cos zx cos ωt
vs
3.8 10
 Free motion 1

By imposing lower
boundary conditions:
By substitutibg x with z
Ux(z =H, t)=0
=
π
+ nπ
Cos(kH)=0,
2

vs π
ω0 = Fundamental frequency
H 2 3.8 11
 Forced motion
If we excite the system harmonically:
ux (H,t) = cos kH cos ωt
we obtain again that the solution is stationary as
it was in case of free motion :
ω
Z = 0 ground surface u x = a cos xz cos ωt
vs

Amplification factor

The amplification factor depends on stratum

thickness, input frequency and material mechanical
poperties 3.8 12
 H

ω
u x = a cos xz cos ωt
vs λ/H
If ωΗ/ vs is small, that iis if is
sufficiently large, the motion is

λ = vs / fr
syncronous, otherwise it becomes
asyncronous as epresented in the
figure above
3.8 13
RESONANCE Natural frequencies of the stratum

Amplification function in case of homogeneous stratum placed on a

inifinitely long sub-stratum. Generalization for generic input.
3.8 14
Homogeneous elastic stratum on a deformable
infinite stratum
s
r
Impedance ratio
When impedence goes to 1, that is when
the stratum is homogeneous and inifinitely
thick the amplification factor goes to 1

3.8 15
3.8 16
 2

If we change the constitutive relationship?

What about resonance?
Let try to interpret the experimental results
in the light of linearity, by inttroducing
dissipation:

Viscous and hysteretic models:

From the laboratory expeerimental
evidence to the theoretical interpretation of
results
3.8 17
2. Dynamic elastic solution (1D geometry):
the dynamic amplification phenomenon

3.8 18
 Experimental evidence:
laboratory devices

Cyclic triaxial test Cyclic torsional shear test

solid cylinder

If the applied frequency is sufficiently low, the inertia effects can be

neglected
3.8
19
Micro structural evolution in granular
media: [ISH 82]
self adaptive microstructure
very small strains (γ < 10-5)
small strains (10-5 < γ < 10-3)
large strains (γ > 10-3) microstructure evolution
coupling between volumetric and
shear strains, liquefaction

3.8
Stress chains 20
 Large numbers of cycles
The axial load is cycled

shakedown

Cyclic response of subballast material (load

frequency = 5 Hz): a) evolution of total The conditioning phase
deviatoric strain (solid line) and permanent
deviatoric strain (dashed line) with load
cycles; b) evolution of total volumetric ε p =Aβ m N b
strain (solid line) and permanent β= σ d σs
volumetric strain(dashed line) with load
cycles [SUI 02].
ηcyc
n3 = 0,96, n9 =0,98 n=
η stat ,max
asymmetric cycles
3.8Dingqing and Selig [DIN 96] 21
Viscous and hysteretic damping

Spring-viscous
damping model τ Gγ + cγ
=
c =cost
D∝ω
γ = γ 0 sin ω t
=τ Gγ 0 sin ω t + ω cγ 0 cos ω t
t0 + 2π ω
∂γ
= WD = ∫t ∂t τ π ωγ 2
Spring-hysteretic dt c 0
damping model 0

c ∝ ω –1 1
D = cost W S = G γ 2
0
h
2
(LIN 1996) WD 1 π cωγ 02 cω h
=
D = = =
h=cω 4π Ws 4π 2 Gγ 0 1 2
2G 2G
3.8 22
 Forced vibration with viscous damping

F0sinωt F0 sin ω t
mz + cz + Gz =

=z Z 0 sin(ω t − φ )
−1
Z0 =
F0 cω
φ = tan
G
( G − mω ) + ( cω )
2 2 2
G − mω 2

c=1
G=1
F0=1
m≅0

D→0 for ω=0

with c=cost

3.8 23
 Determination of parameters G and D

F0
Z0 =
( G − mω ) + ( cω )
2 2 2

WD 1 π cωγ 02 cω h
=
D = = =
4π Ws 4π 2 Gγ 0
1 2
2G 2G
F0
G=
⇒ for ω → 0 : G is obtained by means of static tests Z0

⇒ once G is known, D is evaluated= WD h

D =
4π Ws 2G
⇒ if the G dependency on ω is taken into
consideration , two coupled equations must be
solved
3.8 24
 Torsional shear tests on Toyoura sand:
a) stress-strain records at 10th cycle of I to VII stages,
b) shear moduli and damping ratios [IWA 78].
3.8 25
 vs depends on the void ratio
and the effective confining
pressure

Variation of shear wave velocity with σ0 for round-

grained dry Ottawa sand obtained by resonant column
tests, with peak-to-peak shear strain amplitude of 10-3
rad (Hardin and3.8 Richart 1963). 26
 Resonant column apparatus
Longitudinal vibration of short=
bar u ( x, t ) U ( x )( A1 sin ω nt + A2 cos ω nt )
∂ 2 u ( x, t ) ρ ∂ 2 u ( x, t ) ∂ 2U ( x ) ρ 2
− =
0 ⇒ + (ω n ) U ( x ) =
0
∂x 2
E ∂t 2
∂x 2 E
U ( x) B1 sin (ω n x vc ) + B2 cos (ω n x vc )
End condition: Fixed - Free
At x = 0 (fixed end) U(x) = 0 ⇒ B2 = 0
dU ( x ) B1ω n  ωn L 
At x = L (free end) dU(x)/dx = 0 ⇒ = 0= cos  
dx vc  vc 
( 2n − 1) π 2 n =
⇒ ω n L vc = 1, 2,3..
for n=1
1
ω n 2=
= π fn π vc L =
, vc E ρ
2
Torsional vibration
1
ω n 2=
= π fn π vs L =
, vs G ρ
2 3.8 27
 Determination of internal damping
from resonant column tests
Free vibration with viscous damping mx + cx + kx =
0
x (
X cos ωn 1 − D 2 t − α )
v +Dω x 
2
=
D c=
cc c 2 km ( )
− D ωn t
X e x + 0
2 n 0

0
 ω 1− D 
2
 n 
tan −1 ( v0 + D ωn x0 ) ωn x0 1 − D 2
α=
X n +1 exp ( − D ωntn +1 )
= exp  − D ωn ( tn +1 − tn ) 
=
Xn exp ( − D ωntn )

2π
for t =
ωn 1 − D 2

δ ln ( X=
n X n +1 ) 2π D 1 − D 23.8 28
Dynamic tests
torque excitation on uniform dry
dense sand specimens Too large
(Lin 1996) frequencies

very small strains (γ < 10-5)

WD 1 π cωγ 02 cω h
=
D = = = h
4π Ws 4π 123.8Gγ 02 2G 2G 29
Cyclic torsional shear tests
on Ottawa sand and I-lan sand (Lin 1996) 30

3.8
Relationship between the damping and the frequency

 Af 
=
D Af + B f f =c G π + Bf  (Lin 1996)
3.8  f  31
Frequency effect on material damping ratio,
small strain amplitude Meng, 2003

Clayey sand, drained tests

3.8 32
[SHI 95], torsional shear and resonant column

Elastic-plastic or
Hysteretic models

Visco-Elastic models

3.8 33
VISCOELASTIC MATERIAL: reduction of the amplification factors

Equation under stationary conditions

Solution for small values of D

Amplification function
3.8 34
 2

3.8 35
3. Reduction to one degree of freedom

3.8 36
REDUCTION TO A ONE DEGREE OF FREEDOM SYSTEM m

free motion By recalling: vs = ω/κ k

vs π
u x = b cos kz cos ωt ω0 =
H 2
3

H ∂ u xx
2
ρ ∂ ux
H
2

∫0 ∂z 2
dz = ∫
G 0 ∂t 2
dz

H ρ ∂ u x (t )
H
2
−bk ∫ cos kzdz u (t ) =
2
b ∫ cos kzdzdz
0 G 0 ∂t 2

u x+=2 2
k v su x 0 mu=
x+ Ku x 0
vsπ π G 2ρ H Gπ
kv= ω= = = m = ;K
2H 2H ρ π
s 0
2H 3.8 37
 3
mu + K (u − ur ) =
0
mu + Ku =
Kur
Where the term on he right is known and can be considered as
a forcing load F(t)

Case 1: F(t) harmonic load

ORDINARY DIFFERENTIAL
Case 2: F(t) Impulsive load
EQUATION: the solution is obtained
by adding the general integral to
F0 sin ωt
mx + kx = the particular written here below

F0 / k
X0 =
F0 = ku0  ω  2
2
x = X 0 sin ωt
1 − 2 
 ω 3.80  38
 x = X 0 sin ωt + A sin ω0t + B cos ω0t
Constants A and B are obtained by imposing the initial conditions
For instance for x= x= 0

 ω 
x X 0  sin ωt − sin ω0t 
 ω0 
x

beats
In this case ω is fixed

3.8 39
4. 2D and 1D SS elastic plastic solutions

3.8 40
2D and 1D SS ELASTIC PLASTIC NUMERICAL RESULTS 4

SS to to 1D reduction

H=10
H=10 °m m α=30°
a=30 °
Residual displacement Final profiles and
Permanent displacement on the free surface Irreversible shear strains

Sovraconsolidazione
dinamica

41 3.8
 The site efffect: straucrtures placed
along natural slopes
Horizontal stratum (15 m)
Seismic Input
(Kobe ’95 earthquake 80%
reduced )

α = 0°: horizontal
stratum

Acceleration
amplification
SS numerical analysis
42 3.8
Acceleration at the top layer
Comparison with the
elastic solution

Acceleration at the top layer

Comparison with the

viscoelstic solution

43 3.8
 α = 30°
Acceleration of the top layer

Asymmetry due to the

initial state of stress:
inclined stratum

Acceleration of the top layer Comparison with the

Newmark solution

44 3.8
5. Reliability of numerical
elastic-plastic/viscoplastic solutions

3.8 45
Discussion about the reliability of the numerical solution
1D approach

1. λ/H ratio
2. Effects related to the material fragility
3. Mesh dependency
4. Viscoplastic regularization
5. Strain localization and signal
amplification
6. Influence of the cyclic Hysteresis
3.8 46
Three anthagonistic factors dominate the ssytem
response:

1. The elastic amplification of the signal

2. The asyncronous response in case λ/H reduces
3. The progressive energy dissipation due to
yielding

In any case here the mechanical response is

ductile and localization is absent. In all these
cases the Newmark hypothesis of rigid stratum
seems not be acceptable 3.8 47
 48
H=10 m α=30°
Effecto of λ/H Hardening 1DSh model
Ormsby wavelet
Localization is absent
Varying frequency mean values =2,5,9
λ/H=7.5, 3, 1.67
Hz
= 1/cos(H/λ)

3.8 48
Effecto of λ/H
49
H=10, 5, 2 m α=30°
Ormsby wavelet Hardening 1DSh model

Varying H, fmean =9 Hz λ/H=1.67, 3.33, 8.33

Comparison with the elastic-viscoelstic

solution

3.8 49
 Effects related to the material fragility 50
di Prisco & Pisanò 2010, V Int. Conf. Earth. Eng. & Soil Dyn. (San Diego)

η =∞
Solution objectivity
Mesh dependence of the elasto-plastic solution

Refining the mesh the strain localization phenomenon

seems to disappear: exactly the opposite with respect to
what obtained in case of quasi-static conditions
3.8 50
VISCO-PLASTICITY CREEP TESTS
INCREMENTAL CONSTITUTIVE
RELATIONSHIP

 primary CREEP  Secondary creep  Tertiary creep

4.1
Comparison between a standard strain controlled triaxial
test ( solid line)
and a load controlled test (dashed line)
4.1
STRESS RELAXATION TESTS

4.1
 ELASTIC VISCO-PLASTIC MODELS

VISCOUS NUCLEUS DEFINITION

φ( f )

f TWO ALTERNATIVE CHOICES

4.1
 P1
f >0
P0

f <0 f =0

4.1
Viscoplastic Reduced efficiency for low
regularization viscosity materials

(Needleman ’88, Loret &

Characterisic time and length
Prévost ’90 e ’91, Sluys ’92)

η =20 s-1

3.8 56
 Effects related to the material fragility 57
di Prisco & Pisanò 2010, V Int. Conf. Earth. Eng. & Soil Dyn. (San Diego)

1DS viscoplastic non-local model (Pijaudier-Cabot & Bazant 1987, di Prisco et al. 2002)
( y − s )2
∂γxyvp ( y )
−
H  2 L2

= sign (τ xy )ηΦ  ∫ W ( y − s ) f ( s ) ds 
e
W ( y − s) = L micro-structural
( y − s )2
∂t
H
− parameter
0  ∫e 2 L2
ds
0

η =50 s-1 L=1 cm

The efficent use of non-local approaches for seismic analyses requires:

Implementation of adaptive remeshing techniques
3.8 57
Influence of the microstructure on the structural
response

Interaction between viscous and non-local

characteristic lengths

w = max {wVP , wNL }

3.8 58
 The role of fragility and of
viscoplasticity
1D SS numerical results

Strain localization and

signal amplification

59 3.8
 60
Influence of the cyclic Hysteresis
Modelling of the cyclic behaviour
Kelvin-Voigt non-linear viscoelasticity

1 30

25
0.8

Damping
η

Damping Ratio (%)

sec/G0
G 0.6 Shear Modulus
20

max
15

GG/G
Damping Ratio
0.4
10

0.2
5

0 0
The viscoelasticity does not capture the Elasto-(visco)-plasticity for 0.0001 0.001 0.01 0.1 1 10
irreversbility mechanical behaviour of the cyclic loading γ
Shear Strain (%)

material
Multi-surface Models with kinematic hardening (Mroz 1967, Prévost 1985,
Elgamal et al. 2003)
τ τ
Boundary Hysteresis for loading- Mechanical behaviour for
surface unloading τ
symmetric cyclles

σ γ

γ
3.8 60
 Influence of the cyclic hysterisis 61

A numerical example

experimental
Experimental Perfectly plastic
multi-surface (5) Isotropic hardening

The additional dissipation reduces the

Comparison of seismic responses displacements
H = 5m, α = 20°
Isotropic hardening

multi-surface(3,5,7)

Perfectly plastic

Ruolo essenziale della rigidezza pre-rottura 3.8 61