You are on page 1of 58

7.

Seismic Design of

Underground Structures
G. Bouckovalas

Professor N.T.U.A.

G. Kouretzis

Civil Engineer, Ph.D. N.T.U.A.

October 2010

GEORGE BOUCKOVALAS, National Technical University of Athes, Greece, 2011

7.1

Preface
This lecture deals with the seismic design of
infinitely long cylindrical underground
structures i.e. tunnels, pipelines etc. The
presented methodology can be easily adapted
for the design of structures that feature a noncircular cross-section. However, this (and
other relevant) methodologies are not valid for
other types of underground structures such as
metro stations, storage facilities or shafts.
A common characteristic of the infinitely
long structures under consideration is their
high flexibility and small mass, compared to
the surrounding soil. Thus, unlike most
common above ground structures,
underground structures response to the
imposed displacements from the surrounding
soil is not dominated by inertia effects. As a
result, a static analysis (not a pseudo-static) is
sufficient for their design, given that the
surrounding soil displacements are determined
a-priori.

The following presentation focuses initially on the


failure patterns attributed to transient and
permanent seismic ground displacements. Next, a
methodology for the stress analysis of flexible
underground structures due to transient
displacements is presented in detail, a problem
that can be effectively treated by analytical means.
The methodology for the stress analysis of
underground structures due to permanent ground
displacements is more cumbersome, as it requires
the implementation of non-linear numerical tools,
at least for real-world design purposes. However,
the basic principles of this numerical methodology
are outlined, with the aid of a case study: the
stress analysis of the Thessaloniki to Skopje crude
oil steel pipeline at two active fault crossings: a
normal fault crossing and a strike-slip fault
crossing.

GEORGE BOUCKOVALAS, National Technical University of Athes, Greece, 2011

7.2

. Earthquakes and Underground


Structures
. Seismic design against transient
displacements
C. Seismic design against permanent
displacements

earthquake-induced transient ground displacements...


... are attributed to seismic wave propagation

Love wave (R)


Rayleigh wave (R)

compression wave
(P)

shear wave (S)


GEORGE BOUCKOVALAS, National Technical University of Athes, Greece, 2011

7.3

peripheral cracks

Rayleigh

concrete spalling
(compression)

Rayleig

P
GEORGE BOUCKOVALAS, National Technical University of Athes, Greece, 2011

7.4

longitudinal cracks
Rayleigh

shear cracks

Rayleigh

GEORGE BOUCKOVALAS, National Technical University of Athes, Greece, 2011

7.5

10
7

f ai

0.50g

11
2

24

lur

ep
ote
n
pro tial
pag due
ati to
on
s
eff eism
ect ic w
s.. av
e
.

0.25g

1 failure sites

23 30N

121 30E

121E

120 30E

epicenter
rupture

30 km

Locations of underground structures (concrete tunnels) that failed during the


C-C (1999) earthquake, plotted against recorded maximum ground
acceleration contours at ground surface.

Note that all tunnels that were damaged lie within the 0.25g contour

120cm/sec

10
7

8
9

11

60cm/sec

24

6
1 failure sites

23 30N

121 30E

121E

120 30E

epicenter
rupture

30 km

Locations of underground structures (concrete tunnels) that failed during the


C-C (1999) earthquake, plotted against recorded maximum ground
velocity contours at ground surface.
GEORGE BOUCKOVALAS, National Technical University of Athes, Greece, 2011
Note that all tunnels
that were damaged lie within the 60cm/sec contour

7.6

Chi-Chi 1999

Chi-Chi 1999

Correlation of strong motion levels with underground structure failures

earthquake-induced permanent ground displacements...

fa
ult

ru
pt

ur
e

...are due to ground failure attributed to seismic effects. Typical examples


include fault rupture propagation to the ground surface, landslides, steep
slope failures, and mild slope failures due to liquefaction (lateral spreading).

GEORGE BOUCKOVALAS, National Technical University of Athes, Greece, 2011

7.7

Tunnel collapse due to fault rupture propagation


(Shih-Gang dam tunnel, Chi-Chi, 1999)

Empirical relations for the determination of the MEAN anticipated


ground displacement due to fault rupture (Wells & Coppersmith, 1994).
The MAXIMUM anticipated displacement is about twice the mean value

GEORGE BOUCKOVALAS, National Technical University of Athes, Greece, 2011

7.8

Permanent
displacements
due to
SLOPE
FAILURE

and due to
liquefaction-induced
lateral spreading
GEORGE BOUCKOVALAS, National Technical University of Athes, Greece, 2011

7.9

(a)

(b)

sub-sea
landslides at

Eratini-Tolofonas beach
Aigio (1995) earthquake

Eratini port
()
GEORGE BOUCKOVALAS, National Technical University of Athes, Greece, 2011

7.10

river mouth
(b)

L=50-200m

GEORGE BOUCKOVALAS, National Technical University of Athes, Greece, 2011

W=50-300m

Manythousands
thousandscubic
cubicmeters
metersof
of
Many
soilare
aremobilized
mobilizedduring
duringslope
slope
soil
failures,and
andmove
movefor
fordistances
distances
failures,
ranging
from
a
few
centimeters
ranging from a few centimeters
toseveral
severalmeters...
meters...
to
(...whensoil
soilstrength
strength
(...when
deteriorates
duringshaking,
shaking,as
as
deteriorates during
forexample
exampleininthe
thecase
caseof
of
for
liquefaction).
liquefaction).
We will
will dwell
dwell more
more into
into the
the
We
estimation of
of slope
slope failurefailureestimation
induced
displacements
one
induced displacements inin one
of the
the following
following lectures.
lectures.
of
However,displacement
displacement
However,
estimates
dueto
toslope
slopefailure
failure
estimates due
arenot
notas
asstraightforward
straightforward(?)
(?)
are
asfault-induced
fault-induceddisplacements...
displacements...
as
several
seismologicalseveral seismologicalgeotechnical-topographical
geotechnical-topographical
factorsmust
mustbe
betaken
takeninto
into
factors
account.
account.

7.11

transverse permanent displacement


high bending and tensile
strains

longitudinal permanent
displacement

tensile strain

compressive strain

Ching-Shue tunnel before and after Chi-Chi earthquake

GEORGE BOUCKOVALAS, National Technical University of Athes, Greece, 2011

7.12

Failures at the portals of Ling-Leng tunnel (left)


and Maa-Ling tunnel (right)

Slope failure
above the western portal of the Malakassi C Tunnel.
GEORGE BOUCKOVALAS, National Technical University of Athes, Greece, 2011
7.13
(E. Hoek & P. Marinos, 4th Report on Egnatia Highway Project, March 1999).

Afterword
Compared to permanent displacements, transient displacements effects are less
detrimental regarding underground structure response, since transient displacements
are not only... transient, but also related to significantly smaller magnitudes.
For example, a very strong earthquake with predominant period 0.70sec and maximum
ground acceleration amax = 0.80g, will result in transient displacements with a magnitude
in the order of few centimeters only.
Smax= amax T2/(2)2 = 10 cm only
In comparison, the permanent displacement due to the fault rupture will well exceed
1.00m
On the other hand ...
Transient displacements due to wave propagation affect the whole length of the
underground structure (possibly several km), and not only the part of the structure
located at the vicinity of the fault trace (+ 50m). Moreover, transient displacements
due to wave propagation have a considerably smaller return period (100-500 years),
compared to the return period of fault activation (10,000-100,000 years).
Both these factors highlight the importance of transient displacement effects for
the seismic design of underground structures.

GEORGE BOUCKOVALAS, National Technical University of Athes, Greece, 2011

7.14

. Earthquakes and Underground


Structures
. Seismic design against transient
displacements
C. Seismic design against permanent
displacements

Preface
Compared to permanent displacements, transient displacements effects are less
detrimental regarding underground structure response, since transient displacements
are not only... transient, but are also related to significantly smaller magnitudes.
For example, a very strong earthquake with predominant period 0.70sec and maximum
ground acceleration amax = 0.80g, will result in transient displacements with a magnitude
in the order of few centimeters only.
Smax= amax T2/(2)2 = 10 cm only
In comparison, the permanent displacement due to the fault rupture will well exceed
1.00m
On the other hand ...
Transient displacements due to wave propagation affect the whole length of the
underground structure (possibly several km), and not only the part of the structure
located at the vicinity of the fault trace (+ 50m). Moreover, transient displacements
due to wave propagation have a considerably smaller return period (100-500 years),
compared to the return period of fault activation (10,000-100,000 years).
Both these factors highlight the importance of transient displacement effects for
the seismic design of underground structures.

GEORGE BOUCKOVALAS, National Technical University of Athes, Greece, 2011

7.15

Basic Assumptions
Harmonic seismic waves

No sliding at the soil-structure interface


kinematic soil-structure interaction effects can be ignored
(flexibility factor F)
As a result of the last two assumptions,
underground structure displacements are equal and in phase
with the surrounding soil displacements

the 3rd assumption is valid when

F=

2 Em (1 vl2 )

( 2)
D

El (1 + vm )t

> 20

m=grounds Youngs modulus


l=structures Youngs modulus
m=grounds Poisson ratio
l= structures Poisson ratio
D=cross-section diameter
t=cross-section thickness

Example: steel pipeline (e.g. natural gas pipeline)


m=1GPa (Cs=400m/sec)
l=210GPa
m=0.33
l= 0.2
D=1m
t=0.01m

F850>>20

GEORGE BOUCKOVALAS, National Technical University of Athes, Greece, 2011

7.16

the 3rd assumption is valid when

F=

2 Em (1 vl2 )

( 2)
D

El (1 + vm )t

m=grounds Youngs modulus


l=structures Youngs modulus
m=grounds Poisson ratio
l= structures Poisson ratio
D=cross-section diameter
t=cross-section thickness

> 20

Example: concrete tunnel (e.g. Metro tunnel)


m=1GPa (Cs=400m/sec)
l=30GPa
m=0.33
l= 0.2
D=10m
t=0.25m

F385>>20

the 3rd assumption is valid when

F=

2 Em (1 vl2 )

( 2)
D

El (1 + vm )t

> 20

m=grounds Youngs modulus


l=structures Youngs modulus
m=grounds Poisson ratio
l= structures Poisson ratio
D=cross-section diameter
t=cross-section thickness

Example: concrete sewage pipeline


m=1GPa (Cs=400m/sec)
l=30GPa
m=0.33
l= 0.2
D=2.5m
t=0.1m

F95>>20

GEORGE BOUCKOVALAS, National Technical University of Athes, Greece, 2011

7.17

Basic Assumptions
Harmonic seismic waves

No sliding at the soil-structure interface


kinematic soil-structure interaction effects can be ignored
(flexibility factor F)

All
Allanalytical
analyticalexpressions
expressionshereinafter
hereinafterrefer
referto
tostrains
strainsrather
ratherthan
thanstresses.
stresses.
That
Thatisisdue
dueto
tothe
thefact
factthat
thatwave
wavepropagation
propagationimposes
imposestransient
transient
on
underground
structures,
and
not
inertial
deformations
forces,as
asinin
deformations on underground structures, and not inertialforces,
common
commonaboveground
abovegroundstructures.
structures.Design
Designof
ofunderground
undergroundstructures
structuresaims
aimsat
at
ensuring
that
the
structure
can
sustain
those
deformations
without
failure.
ensuring that the structure can sustain those deformations without failure.

Strain state of thin-walled underground structures


z

y
x

1
A 3-D shell modeling the underground structure, when subjected to imposed
displacements from the surrounding soil, will develop...
axial strains
in-plane shear strains
hoop strains h
The out-of-plane normal strains, and the shear strains at the inside and the outside
face of the shell can
be customarily
ignored...
(why??)
GEORGE
BOUCKOVALAS, National
Technical University
of Athes, Greece, 2011

7.18

harmonic excitation:

2
(x Ct )
u = Amax sin

displacement (u)

Amax

time (t)

period (T)

=wavelength C=propagation velocity t=time

P-wave propagating along the axis


(xy plane)
diameter
(D)
y

2
x
2

u x = Amax sin
x C pt

plane view

strain amplitude
ground strain=
structure strain:

thus

a = x =

u x 2 max
2

=
cos
x C pt
x

2 max

Vmax

C p 2011
GEORGE BOUCKOVALAS, National Technical University of Athes, Greece,

a ,max =

maximum
seismic
velocity

7.19

S-wave propagating along the axis


(xy plane)
No-slip assumption at the soilstructure interface shear beam model
No slip suggests zero relative
displacements between crosssections, thus zero bending strains.

A full-slip assumption at the soilstructure interface would suggest that


only bending strains develop, and that
shear strains are zero
(bending beam model)

In the real world, some slippage will always occur, especially during strong excitations
The conservative no-slip assumption is adopted for design purposes

adopting the shear beam model

y
x

2
(x C s t )
u y = Amax sin

plane view

ground strain=
structure strain:

u y
x

Vmax
2
(x C s t )
cos
Cs

Vmax
C SUniversity of Athes, Greece, 2011
GEORGE BOUCKOVALAS, National Technical
max =

7.20

strain distribution along the cross-section

strains on a 3-D shell


(except axial strain)
do not retain a constant value along
the cross-section.
Strain distribution is a function
of the polar angle and, of course, time

max
direction
of motion

=maxcos

max

propagation axis

1.5

/|max|

t=to

cos

0.5

0
-0.5

t=to+T/2

-1
-1.5

strong motion axis

y
0

45

90

135

180

225

270

315

360

polar angle (degrees)

angle is measured clockwise,


starting from z-axis

Transverse P-wave (yz plane)


L/2

2
u y = Amax sin

y
direction of
propagation
y C pt

section view

max=L/L
L

12

h,max=max

ground strain:
y =

u y
y

11

h = y =

structure strain at
points & :
Vmax
2
( y C p t )
cos
Cp

GEORGE BOUCKOVALAS, National Technical University of Athes, Greece, 2011

7.21

strain distribution along the cross-section


MX

h,max
direction
of motion

h=0
h,max

1.5

t=to

h/|h,max|

|cos|

0.5

-0.5
t=to+T/2

-1

-1.5
0

45

90 135 180 225 270 315 360


polar angle (degrees)

axis of propagation
& axis of strong motion

angle is measured clockwise,


starting from z-axis

Transverse S-wave (yz-plane)


z

section view

y
direction of
2
propagation
( y C s t )
u z = Amax sin

max

ground strain:
yz =

u z
y

12

h,max=max/2

max/2

11

h,&'

structure strain:
V

2
= yz / 2 = max cos ( y Cs t )
2Cs

Vmax
2
( y C s t )
cos
2C s


GEORGE BOUCKOVALAS, National Technical University of Athes, Greece, 2011
7.22

h , & ' = yz / 2 =

strain distribution along the cross-section


compression
max/2

Principal strains (tensile and compressive)


are located at a plane inclined by 45
12

(why??)

11

extension
max/2

strong motion axis

1.5
sin2

t=to

h/|h,max |

h,max=max/2

max/2

0.5

0
-0.5
-1

t=to+T/2

-1.5

propagation axis
y

45

90 135 180 225 270 315 360


polar angle (degrees)

Transverse S-wave (z-plane)

side view
direction of
propagation
z

2
(z C s t )
u x = Amax sin

ground strain=
structure strain:

u x Vmax
2
(z C s t )
=
cos
z
Cs

GEORGE BOUCKOVALAS, National Technical University of Athes, Greece, 2011

7.23

strain distribution along the cross-section

strong motion axis

1.5

/|max|

t=to

0.5

sin2

-0.5
-1

t=to+T/2

-1.5
0

45

90 135 180 225 270 315 360


polar angle (degrees)

propagation axis

Transverse S-wave propagation results in shear structure strains


(as in the case of S-wave propagation along the axis, at xy-plane)
but with different strain distribution along the cross-section

Summarizing
Case
P-wave along the axis
(xy)
S-wave along the axis
(xy)

,max
Vmax

max

Cp
Vmax

Cs

Vmax

Transverse P-wave (yz)

Vmax

Transverse S-wave (yz)


Transverse S-wave (xz)

h,max

Vmax

Cp
2C s

Cs

GEORGE BOUCKOVALAS, National Technical University of Athes, Greece, 2011

7.24

However, in the real world


a seismic wave will cross the axis of the structure under a
random angle , measured at the plane defined by the
structure axis and the wave propagation axis.
Orientation of this plane at the 3-D space is random (there

is no practical way to predict it or define it)

direction of
wave propagation

structure axis

plane of
wave motion

wave-structure
plane

GEORGE BOUCKOVALAS, National Technical University of Athes, Greece, 2011

7.25

A SV-wave (the direction of motion coincides with the


plane of propagation xy) can be analyzed into 4
apparent waves:

Amax

transverse P (xy plane)


& transverse S (xy plane)
-wavelength /sin
-propagation velocity Cs/sin

/sin

x'
Amax

Amaxcos

-Amaxsin

direction of
wave propagation

P, S along the axis


(xy plane)
-wavelength /cos
-propagation velocity Cs/cos
(the wave period is constant)
x

Amax

structure axis

/cos

In order to estimate maximum design strains for this


generic case, we must consider:
a) superposition of strains (distribution along the cross-section)
that result from each apparent wave propagation.
b) maximization of the resulting expressions for strains, to
eliminate the unknown angles &

(angle is the angle formed by the plane of strong motion and


the plane of propagation, and is defined for S-waves only)

GEORGE BOUCKOVALAS, National Technical University of Athes, Greece, 2011

7.26

Implementing this procedure for S- and P-waves propagating at a random


angle relatively to the structure axis, results in the following maximum design strains...

maximum strains

S-wave
(normalized over the
Vmax/Cs ratio)

P-wave
(normalized over the
Vmax/Cp ratio)

axial

0.50

1.00

shear

1.00

1.00

hoop h

0.50

1.00

von Misses vM

0.87/(1+ l)

1.00/(1+ l)

major principal 1

0.71

1.00

minor principal 3

-0.71

-1.00

ATTENTION: Strain components do not attain these maximum values at the


same position along the cross-section. Thus, simply adding them to derive von
Misses and principal strains is an over-conservative approach.

However, in the real world


Underground structures are located relatively close to the ground
surface, and are subjected to surface Rayleigh wave effects too (e.g.

at valleys, near slopes, at large distances from rupture)


A Rayleigh waves is equivalent to a
P-wave and a SV-wave, propagating
simultaneously and featuring a
phase lag equal to /2

S-component

x'

ATTENTION: We should not just...

6
P-component

Rayleigh wave

structure axis
x

add strains for a P-wave and a SVwave propagating at the same


angle in order to find the
maximum strains, as maxima do
not occur at the same time.

GEORGE BOUCKOVALAS, National Technical University of Athes, Greece, 2011

7.27

Maximum (after proper superposition) design strains for


a Rayleigh wave propagating at a random angle relatively
to the structure axis:
maximum strain

R-wave
(normalized over the
Vmax,V/CR ratio)

axial

0.68

shear

hoop h

0.68

von Misses vM

0.86/(1+ l)

major principal 1

0.68

minor principal 3

-0.68

Note:
Rayleigh wave propagation velocity CR0.94 Cs is estimated at a depth z=1.0R

It is worth trying to verify the above expressions for Rayleigh waves at home!

GEORGE BOUCKOVALAS, National Technical University of Athes, Greece, 2011

7.28

Homework
Leukada 16/8/03 earthquake
0.4

Vmax=0.317m/sec

homogeneous soil
Cs=200m/sec =1/3

V (m/s)

0.2
0
-0.2
-0.4
0

10

Time (s)

15

20

Verify the structural integrity of the following structures:


steel pipeline
concrete tunnel
- D=1.5m, t=0.015m
- allcompression=40t/d (%)<5% (EC8)
- allextension =2% for the main body
=0.5% for peripheral
butt welds

- D=10m, t=0.20m
- allcompression=0.35%
- allextension=2%

Justify the expressions that you apply for the estimation of seismic strains!

GEORGE BOUCKOVALAS, National Technical University of Athes, Greece, 2011

7.29

However, in the real world

Consider the case of an underground structure located in


a soft soil layer, overlying the bedrock
Seismic waves propagate to the
soil-bedrock interface at a random
angle rock, refract, and continue to
the soft soil layer

soil
Csoil

according to Snells law:


cosasoil = cosarock

rock
Crock

soil

rock

Csoil
Crock

the predominant period of


the refracted wave
is not altered, thus...

soil = rock

Csoil
Crock

The refracted wave, propagating at an angle soil relatively to the


structure, can be analyzed (as before) into apparent waves...
A. a horizontal apparent wave with a propagation velocity Csoil/cosasoil
trans=soil/sinsoil

soil

structure axis

(projection)

axial=soil/cossoil

soil
rock

rock/cosrock

and

wavelength: horiz = soil cosa

soil

rock

Csoil
Crock
cosarock

Csoil
Crock

horiz =

rock

The horizontal apparent wave is equivalent to the apparent wave


propagating alongGEORGE
the BOUCKOVALAS,
soil-bedrock
interface.
National Technical
University of Athes, Greece, 2011

cosarock

7.30

B. a vertical apparent wave with a propagation velocity Csoil/sinasoil


C
on the other
sinsoil = 1 cos 2 soil = 1 cos 2 rock soil
hand:
Crock
1

0.125
0.25
0.333

0.96

sinsoil

0.92

Csoil/Crock=0.5

0.88
0.84
0

10 20 30

40 50

rock

60 70 80

90

So, the apparent propagation velocity of the vertical apparent wave is


practically equal to the wave propagation velocity in soft soil Csoil
(and trans=soil=Csoil T)
An extra, unknown parameter must be considered here, compared to the
homogeneous rockmass case- the angle rock
Design strains are estimated via the superposition of strains along the crosssection, and the subsequent maximization of the resulting expressions for the
unknown angles , rock
normalized
strains

S-wave
(C=CS)

P-wave
(C=CP)

axial

0.50Csoil/Crock

0.3Csoil/Crock

shear

0.43Csoil/Crock+0.98

2.0Csoil/Crock

hoop h

0.36Csoil/Crock+0.50

0.5Csoil/Crock+1.0

von Misses vM

(0.38Csoil/Crock+0.85)/(1+l)

(0.58Csoil/Crock+1.0)/(1+l)

major principal 1

0.5Csoil/Crock+0. 5

0.63Csoil/Crock+1.0

minor principal 3

-0.5Csoil/Crock-0. 5

-0.63Csoil/Crock-1.0

Vmax

C soil

The above expressions are valid for Csoil/Crock ratio values less than 0.35. For Csoil/Crock>0.35 , results will
be over-conservative, and the use of the corresponding expressions for homogeneous rockmass conditions is
proposed.
GEORGE BOUCKOVALAS, National Technical University of Athes, Greece, 2011
7.31

Comparison to ALA-ASCE (2001) recommendations


ALA-ASCE 2001 a =

Vmax

aC

Vmax= peak ground velocity generated by ground


shaking
C=

apparent propagation velocity for seismic waves,


conservatively (?) assumed to be equal to
2000 m/s.

a=

1 for P and R waves, 2 for S waves


Strains

S-wave
(C=CS)

P-wave
(C=CP)

axial

0.50Vmax/Crock

0.3Vmax/Crock

shear

(0.98+0.43Csoil/Crock) Vmax/Csoil

2.00Vmax/Crock

hoop h

(0.50+0.36Csoil/Crock) Vmax/Csoil

(1.00+0.50Csoil/Crock) Vmax/Csoil

Comparison to ALA-ASCE (2001) recommendations


input data:

Csoil=200m/sec
Crock=2000m/sec (bedrock)
Cp=2Cs
Vmax=75cm/sec

Homogeneous rock

Soft soil

Strains

Rayleigh wave
S-wave

P-wave

S-wave

P-wave

axial

0.019%

0.019%

0.019%

0.005%

0.255%

shear

0.037%

0.019%

0.383%

0.037%

0.375%

hoop h

0.019%

0.019%

0.201%

0.196%

0.255%

a =

Vmax

ALA-ASCE 2001

aC =0.037% (conservatively a=1)

major discrepancy...!

GEORGE BOUCKOVALAS, National Technical University of Athes, Greece, 2011

7.32

Homework
Repeat the previous homework assignment for the case where the
the structure is constructed near the surface of a homogeneous soft
soil layer (CS,SOIL=200m/s), overlying the marl bedrock
(CS,ROCK=700m/s).

Case studies of underground structures failures during


Kobe & Chi-Chi earthquakes
10
7

120cm/sec

8
9

11
2

24

60cm/sec

6
1 failure sites

23 30N

121 30E

121E

120 30E

epicenter
rupture

GEORGE BOUCKOVALAS, National Technical University of Athes, Greece, 2011

30 km
7.33

Case studies of underground structures failures during


Kobe & Chi-Chi earthquakes
10
8

8
24

2
4

1 failure sites

Holocene alluvium
Pleistocene deposits
Miocene deposits
Oligocene deposits
Neocene deposits

Metamorphic rocks
23 30N

121 30E

121E

120 30E

bedrock
epicenter
rupture
0

30 km

Soil classification
NEHRP (1997)

Hard rock

CS (m/sec)

This study

CS (m/sec)

CS>1500

Granite bedrock

2000

B1

760<CS<1500

Rocks

Cretaceous rocks (igneous and


metamorphic rocks, limestone, solid
volcanic deposits)

B2

Stiff soils and soft rocks (PleioPleistocence sandstones,


conglomerates, schists, marls)

Stiff soils

Soft soils

360< CS <760

Pleistocene clayey deposits, sands and


gravels

180< CS <360

Holocence alluvium (sand, gravels,


clay) and man-made deposits

850

550

250

soft soil

Very stiff soils/soft


rocks

1200

CS<180

GEORGE BOUCKOVALAS, National Technical University of Athes, Greece, 2011

7.34

Frequency of failure types


40

30

20

18

18
15

10

8
2

2
pl
ift
flo
or
u

cr
ac
ks

ot
he
rc
ra
ck
s
sh
ea
rc
ci
ra
rc
ck
um
s
fe
re
st e
nt
el
ia
lc
re
in
ra
fo
ck
rc
s
em
en
tb
uc
kl
in
g

lo
ng
i

tu
di
na
l

0
sp
al
lin

number of recorded failures

34

Failure criterion
development of cracks wider than w>0.2mm
crack width is related to the stress applied on the steel reinforcement bars
(Gergely and Lutz, 1968, ACI Committee 224, 1995)

w = 0.076 f s 3 dc A
for a typical tunnel... fs,lim=140MPa
failure when:
1*steel >fs,lim

stresses higher than fs,lim suggest larger crack widths, or concrete spalling
GEORGE BOUCKOVALAS, National Technical University of Athes, Greece, 2011

7.35

a-posteriori failure prediction


600
this study
ASCE-ALA/EC8
fs,lim for crack width=0.2mm

550
500
450

proposed: 64%
ASCE&EC8: 8%

s (Pa)

400
350
300
250
200
150
100
50
0

soil type:A

B1
14

B2

19
29
Number of failures

D
47

38

Prediction of actual structure response


600
550
500

structures that failed


structures that survived
fs,lim for crack width 0.2mm

450

s (Pa)

400
350
300

proposed: 74%
ASCE&EC8: 57%

+ all non-successful predictions of ASCE & C8


are non-conservative estimates

250
200
150
100
50
0
soil type:A

B1
34

B2
52

C
78

GEORGE BOUCKOVALAS, National


Technicalof
University
Number
failuresof Athes, Greece, 2011

D
92

101
7.36

Afterword
The presented analytical expressions are valid for flexible (F>20), and infinitely long
underground structures. The stress state becomes more complicated in areas of bends and
-ees. However, analytical methodologies have been proposed for such cases too, and
provide relatively accurate results. When the underground structure is constructed by
discrete, jointed pieces, the flexibility of the joints must also be taken into account in the
assessment of its response.
Generally speaking, when the in-situ conditions diverge significantly from the discussed
assumptions, more elaborate numerical analysis tools must be applied for the seismic
design of the underground structure. The structure is modeled as a beam or a shell, soilstructure interaction is simulated via elasto-plastic springs, and the seismic excitation is
applied at the base of the springs that are used to model soil response.
An extensive presentation of such numerical methodologies is beyond the scope of this
lecture. Their basic principles are however presented in the following case study,
regarding the numerical stress analysis of a crude oil steel pipeline, due to the possible
activation of a normal and a strike slip fault crossing its route (permanent ground
displacements)

GEORGE BOUCKOVALAS, National Technical University of Athes, Greece, 2011

7.37

. Earthquakes and Underground


Structures
. Seismic design against transient
displacements
C. Seismic design against permanent
displacements

GEORGE BOUCKOVALAS, National Technical University of Athes, Greece, 2011

7.38

C.J. Gantes, G.D. Boukovalas


Department of Civil Engineering
National Technical University of Athens

ANALYSIS
OF BURIED PIPELINES
AT FAULT CROSSINGS

National
Technical
University
of Athens

Contents of presentation

Pipeline Geometry, Material Properties & Construction Techniques


Failure Criteria
Downthrows at Seismic Faults
Pipeline Modeling - Fault Movement Modeling
Beam & Mixed Model Results - Influence of Internal Pressure
Conclusions

GEORGE BOUCKOVALAS, National Technical University of Athes, Greece, 2011


Gantes & Bouckovalas ANALYSIS OF BURIED PIPELINES AT FAULT CROSSINGS

7.39

National
Technical
University
of Athens

Geometry, material properties and


construction techniques

At the specific locations of active faults crossings, heavy wall NPS 16 line
pipes with a wall thickness of 8.74mm will be installed, made of API5LX60 steel.

The nominal backfill cover varies from 0.60m in rocky areas to 0.90m in
cross-country areas and 1.20m under major roads.

The pipeline steel is modeled with a typical API-5LX60 tri-linear stressstrain curve based on the provisions of ASCE.

Gantes & Bouckovalas ANALYSIS OF BURIED PIPELINES AT FAULT CROSSINGS

National
Technical
University
of Athens

Pipeline material properties

600

600

(2,2)

(1,1)

500
400

(minimum)
Ultimate Tensile Strength

(minimum)
Yield Strength

200
stress (MPa)

STRESS (MPa)

400

300

E=210.000 KPa

0
-100
-200

200

100

-300
-400
-500

Tri-linear Idealization
Ramberg-Osgood
0

STRAIN (%)

-600
-0.06 -0.05 -0.04 -0.03 -0.02 -0.01 0
0.01 0.02 0.03 0.04 0.05 0.06
strain

GEORGE BOUCKOVALAS, National Technical University of Athes, Greece, 2011


Gantes & Bouckovalas ANALYSIS OF BURIED PIPELINES AT FAULT CROSSINGS

7.40

National
Technical
University
of Athens

Failure criteria

Pipelines resist the imposed displacements mainly through axial (tensile or


compressive) strains, thus it is more meaningful to talk in terms of strains
than stresses.
Considered Failure Criteria
1. Limiting compressive strain to avoid elastic or plastic buckle, associated
with local wrinkling.

e*c = 0.84 0.0035

D
t

For an NPS 16 in. x 8.74mm pipe the former expression yields 0.677%.

Gantes & Bouckovalas ANALYSIS OF BURIED PIPELINES AT FAULT CROSSINGS

National
Technical
University
of Athens

Failure criteria

2. Limiting the tensile strain for girth welds due to metallurgical


alterations induced to the heat-affected zone during the welding process.
The allowable tensile strain for butt (peripheral) welding is
conservatively taken as 5. The latter criterion is used throughout
the analysis.

Section Conclusions
Relative ground movements caused by fault rupture
are displacement-controlled
The limiting strain is taken equal to 5
GEORGE BOUCKOVALAS, National Technical University of Athes, Greece, 2011
Gantes & Bouckovalas ANALYSIS OF BURIED PIPELINES AT FAULT CROSSINGS

7.41

Downthrows
at seismic faults

National
Technical
University
of Athens

Two fault types are considered:


Normal fault
Strike slip fault
The critical crossings are identified with reference to:
Anticipated ground movements
Geology at the area of crossing
Angle of intersection between fault trace and pipeline axis

Gantes & Bouckovalas ANALYSIS OF BURIED PIPELINES AT FAULT CROSSINGS

National
Technical
University
of Athens


Wells & Coppersmith,
1994).
.

GEORGE BOUCKOVALAS, National Technical University of Athes, Greece, 2011


Gantes & Bouckovalas ANALYSIS OF BURIED PIPELINES AT FAULT CROSSINGS

7.42

National
Technical
University
of Athens

Normal ffault
ault
Fault geometry

Y
Pipeline axis

Pipeline axis

Fault trace

Fault trace

Anticipated downthrow DZ = 30 cm

DX = DZ cot( ) cos( ) DY = DZ cot( ) sin( )


For =45 and = 63 DX=DY=0.36DZ
Gantes & Bouckovalas ANALYSIS OF BURIED PIPELINES AT FAULT CROSSINGS

National
Technical
University
of Athens

Strike slip fault


Fault geometry

X
Y
Pipeline axis

Fault trace
Anticipated left lateral Slip S = 30 cm, =80

DX = S cos( ) = 0.17 S

DY = S sin ( ) = 0.98S

GEORGE BOUCKOVALAS, National Technical University of Athes, Greece, 2011


Gantes & Bouckovalas ANALYSIS OF BURIED PIPELINES AT FAULT CROSSINGS

7.43

10

National
Technical
University
of Athens

Pipeline modeling

Non linear FE Analysis for material behavior and large deformations are
considered using NASTRAN.
Two models, a beam (BM) and a mixed beam-shell model (MM) are used for
the analysis.
A straight pipeline segment of length 1200m is considered for both models
and the fault rupture is applied in the middle.
The Beam Model (BM) implements 3D beam elements, having the
mechanical properties of a tube with 16 diameter and 8.74mm thickness
made of API-5LX60 steel.
The Mixed Model (MM) combines shell elements near the expected fault
to capture stress concentrations, and 3D beam elements further away
from the fault, where low stresses are expected. Coupling of shell and the
beam part of the model is done with the use of rigid elements.
Gantes & Bouckovalas ANALYSIS OF BURIED PIPELINES AT FAULT CROSSINGS

National
Technical
University
of Athens

11

Pipeline modeling

V1
L1
C1
G14

Z
Y
X

Coupling of shell and beam elements

GEORGE BOUCKOVALAS, National Technical University of Athes, Greece, 2011


Gantes & Bouckovalas ANALYSIS OF BURIED PIPELINES AT FAULT CROSSINGS

7.44

12

National
Technical
University
of Athens

Modeling of soil
around buried pipelines

The soil is modeled with four sets of inelastic springs, two in


the local X-Y plane and two in the global vertical Z direction
where different upward and downward reactions occur

The soil springs are computed


assuming cohesionless materials
(sands) such as the backfill soil
used along the pipeline route

Properties of the springs are calculated according to


ALA-ASCE (2005) guidelines for the seismic design of buried pipelines
(for sand trench backfill)
Gantes & Bouckovalas ANALYSIS OF BURIED PIPELINES AT FAULT CROSSINGS

National
Technical
University
of Athens

13

Pipeline modeling

Transverse soil springs in the beam part of the model


GEORGE BOUCKOVALAS, National Technical University of Athes, Greece, 2011

Gantes & Bouckovalas ANALYSIS OF BURIED PIPELINES AT FAULT CROSSINGS

7.45

14

National
Technical
University
of Athens

Pipeline modeling

Transverse soil springs in the shell part of the model


Gantes & Bouckovalas ANALYSIS OF BURIED PIPELINES AT FAULT CROSSINGS

National
Technical
University
of Athens

15

Pipeline modeling

Transverse soil springs in the shell part of the model


GEORGE BOUCKOVALAS, National Technical University of Athes, Greece, 2011

Gantes & Bouckovalas ANALYSIS OF BURIED PIPELINES AT FAULT CROSSINGS

7.46

16

National
Technical
University
of Athens

Axial soil springs

Maximum axial soil force


per unit length of
the pipeline

+1
Tu = DH 0 tan
2
2 3 for a steel pipeline
t= 3mm (dense sand)
5mm (loose sand)
[sum of interface shear (friction) forces along the perimeter of the pipeline]
Gantes & Bouckovalas ANALYSIS OF BURIED PIPELINES AT FAULT CROSSINGS

National
Technical
University
of Athens

17

Transverse horizontal
soil springs

These springs simulate the resistance from the surrounding soils to any
horizontal translation of the pipeline. Thus, the mechanisms of soilpipeline interaction are similar to those of vertical anchor plates or
footings moving horizontally relative to the surrounding soils and thus
mobilizing a passive type of earth pressure.
Relationship between force p per unit
length and horizontal displacement y

p=

y
A +By

A = 0.15 yu/pu
B = 0.85/pu
pu = H Nqh D
Nqh=horizontal bearing capacity factor
for loose sand
yu = 0.07 to 0.10 (H+D/2)
0.02 to 0.03 (H+D/2)
for dense sand
GEORGE BOUCKOVALAS, National Technical University of Athes, Greece, 2011
Gantes & Bouckovalas ANALYSIS OF BURIED PIPELINES AT FAULT CROSSINGS

7.47

18

Transverse horizontal
soil springs

National
Technical
University
of Athens

maximum lateral soil


force per unit length

Pu = N qh HD

[passive earth pressure on a


horizontally moving shallow
footing]

bearing capacity factor


for sand (Hanshen, 1961)

Gantes & Bouckovalas ANALYSIS OF BURIED PIPELINES AT FAULT CROSSINGS

19

Transverse vertical
soil springs

National
Technical
University
of Athens

Downward Motions: the pipeline is assumed to act as a


cylindrically-shaped strip footing and the ultimate soil
resistance qu is given by conventional bearing capacity
theory. For cohesionless soils the force per unit length
is:

qu = H Nq D+ 0.5 D 2 N

Nq, N = bearing capacity factors for horizontal


strip footings, vertically loaded in the downward
direction
=

effective unit weight of soil


GEORGE BOUCKOVALAS, National Technical University of Athes, Greece, 2011
Gantes & Bouckovalas ANALYSIS OF BURIED PIPELINES AT FAULT CROSSINGS

7.48

20

Transverse vertical
soil springs

National
Technical
University
of Athens

Downward Motions: the pipeline is assumed to act as a


cylindrically-shaped strip footing and the ultimate soil
resistance qu is given by conventional bearing capacity
theory. For cohesionless soils the force per unit length
is:

Meyerhof, 1965

qu = H Nq D+ 0.5 D 2 N

Nq, N = bearing capacity factors for horizontal


strip footings, vertically loaded in the downward
direction
=

effective unit weight of soil


Gantes & Bouckovalas ANALYSIS OF BURIED PIPELINES AT FAULT CROSSINGS

21

Transverse vertical
soil springs

National
Technical
University
of Athens

Upward motions: Based on tests performed with


pipes buried in dry uniform sand, the
relationship between the force q and the vertical
upward displacement z, has been shown to vary
according to the following hyperbolic relation

q=
A = 0.07 zu /qu
B = 0.93/qu
ultimate uplift resistance

z
A +Bz
qu = H Nqv D
Trautmann & O Rourke, 1983

The vertical uplift factor Nqv is a function of the depth to diameter ratio H/D
and the friction angle of the soil
GEORGE BOUCKOVALAS, National Technical University of Athes, Greece, 2011

Gantes & Bouckovalas ANALYSIS OF BURIED PIPELINES AT FAULT CROSSINGS

7.49

22

National
Technical
University
of Athens

Load modeling

Fault activation is modeled as an imposed displacement


at the base of the soil springs attached on one side of the fault line

Fault
movement

Deformed shape

Gantes & Bouckovalas ANALYSIS OF BURIED PIPELINES AT FAULT CROSSINGS

National
Technical
University
of Athens

23

Load modeling

For the Mixed Model the internal pressure is modeled as a uniform load
normal to the internal face of the shell elements. The nominal pressure is
10.2 MPa according to the specification of ASME.

The displacement field is imposed in a number of steps to capture eventual


non-linearities in the pipes response with regard to fault rapture
GEORGE BOUCKOVALAS, National Technical University of Athes, Greece, 2011
Gantes & Bouckovalas ANALYSIS OF BURIED PIPELINES AT FAULT CROSSINGS

7.50

24

Normal ffault
ault
rresults
esults

National
Technical
University
of Athens

Beam Model Deflected shape


Gantes & Bouckovalas ANALYSIS OF BURIED PIPELINES AT FAULT CROSSINGS

National
Technical
University
of Athens

25

Normal ffault
ault
rresults
esults

Beam Model Axial spring stresses


GEORGE BOUCKOVALAS, National Technical University of Athes, Greece, 2011
Gantes & Bouckovalas ANALYSIS OF BURIED PIPELINES AT FAULT CROSSINGS

7.51

26

Normal ffault
ault
rresults
esults

National
Technical
University
of Athens

Beam Model Stresses

Beam Model Strains


Gantes & Bouckovalas ANALYSIS OF BURIED PIPELINES AT FAULT CROSSINGS

National
Technical
University
of Athens

27

Normal ffault
ault
rresults
esults

Mixed Model Deflected shape


GEORGE BOUCKOVALAS, National Technical University of Athes, Greece, 2011
Gantes & Bouckovalas ANALYSIS OF BURIED PIPELINES AT FAULT CROSSINGS

7.52

28

Normal ffault
ault
rresults
esults

National
Technical
University
of Athens

Percentage Difference of Total Translation


8%

Disp 0.15 m
Difference

4%

Disp 0.30 m
0%

-4%

-8%
580

585

590

595

600

605

610

615

620

Pipe Length (m)

Mixed Model % difference of total translation due to internal pressure


29

Gantes & Bouckovalas ANALYSIS OF BURIED PIPELINES AT FAULT CROSSINGS

Normal ffault
ault
rresults
esults

National
Technical
University
of Athens

No pressure
Major Stress at Bottom Plane
500

Mixed Model Stresses

disp = 0,15 m
disp = 0,30 m

300
200

Major Strain at Bottom Plane

100
0
580

0.30%

585

590

595

600

605

610

615

disp = 0,15 m

0.25%

620

Pipeline length (m)

disp = 0,30 m

0.20%

Strain

Stress (MPa)

400

0.15%
0.10%

Mixed Model Strains

0.05%
0.00%
580

585

590

595

600

605

Pipeline length (m)


GEORGE BOUCKOVALAS, National Technical University of Athes, Greece, 2011
Gantes & Bouckovalas ANALYSIS OF BURIED PIPELINES AT FAULT CROSSINGS

610

615

620

7.53

30

Normal ffault
ault
rresults
esults

National
Technical
University
of Athens

Pressure 10.2 MPa


Major Stress at Bottom Plane

Mixed Model Stresses

500

disp = 0,15 m
disp = 0,30 m

300
200

Major Strain at Bottom Plane


100
0
580

0.30%

disp = 0,15 m

0.25%
585

590

595

600

605

610

615

620

disp = 0,30 m

0.20%

Pipeline length (m)


Strain

Stress (MPa)

400

0.15%
0.10%
0.05%

Mixed Model Strains

0.00%
580

585

590

595

600

605

610

615

620

Pipeline length (m)


Gantes & Bouckovalas ANALYSIS OF BURIED PIPELINES AT FAULT CROSSINGS

31

Normal ffault
ault
rresults
esults

National
Technical
University
of Athens

Mixed Model Strains


With pressure
Displacement 0.30m

Mixed Model Strains


With pressure
Displacement 0.15m

GEORGE BOUCKOVALAS, National Technical University of Athes, Greece, 2011


Gantes & Bouckovalas ANALYSIS OF BURIED PIPELINES AT FAULT CROSSINGS

7.54

32

National
Technical
University
of Athens

Strike sslip
lip ffault
ault
rresults
esults

Beam Model Deflected shape


Gantes & Bouckovalas ANALYSIS OF BURIED PIPELINES AT FAULT CROSSINGS

National
Technical
University
of Athens

33

Strike sslip
lip ffault
ault
rresults
esults

Beam Model Axial spring stresses


GEORGE BOUCKOVALAS, National Technical University of Athes, Greece, 2011
Gantes & Bouckovalas ANALYSIS OF BURIED PIPELINES AT FAULT CROSSINGS

7.55

34

Strike sslip
lip ffault
ault
rresults
esults

National
Technical
University
of Athens

Beam Model Stresses

Beam Model Strains


35

Gantes & Bouckovalas ANALYSIS OF BURIED PIPELINES AT FAULT CROSSINGS

Strike sslip
lip ffault
ault
rresults
esults

National
Technical
University
of Athens

Major Strain at Bottom Plane

No Pressure
Mixed Model Stresses

0.40%

0.20%

disp = 0.15 m
disp = 0.30 m

0.10%

0.00%
580

585

590

595

600

605

610

615

Major Strain at Bottom Plane

620

Pipeline length (m)

0.40%

0.30%

disp = 0.15 m
Strain

Strain

0.30%

0.20%

disp = 0.30 m

0.10%

Mixed Model Strains


Pressure 10.2 MPa

0.00%
580

585

590

595

600

605

Pipeline length (m)


GEORGE BOUCKOVALAS, National Technical University of Athes, Greece, 2011
Gantes & Bouckovalas ANALYSIS OF BURIED PIPELINES AT FAULT CROSSINGS

610

615

620

7.56

36

Strike sslip
lip ffault
ault
rresults
esults

National
Technical
University
of Athens

Mixed Model Strains


With pressure
Displacement 0.30m

Mixed Model Strains


With pressure
Displacement 0.15m
Gantes & Bouckovalas ANALYSIS OF BURIED PIPELINES AT FAULT CROSSINGS

National
Technical
University
of Athens

37

Analysis conclusions

No failure of the pipeline at zones of active faults is expected.


For the strike-slip fault case yielding of the pipeline at places near the
fault is anticipated.
Good agreement of the results from the two models for both studies is
observed.

GEORGE BOUCKOVALAS, National Technical University of Athes, Greece, 2011


Gantes & Bouckovalas ANALYSIS OF BURIED PIPELINES AT FAULT CROSSINGS

7.57

38

National
Technical
University
of Athens

Construction
countermeasures

Heavy wall sections near active faultsThis measure will increase the pipeline
stiffness relative to that of the backfill and will lead to smaller curvatures and
smaller internal strains
Arrangement of loose backfill around the pipe, extending beyond the anticipated
displacement along the critical zone

Wrap the pipeline with a proper, friction reducing geotextile which will provide a
lower friction coefficient

Enclose the pipeline within a casing, so that the pipeline can move freely along the
intensely distorted length, on both sides of the fault trace

Gantes & Bouckovalas ANALYSIS OF BURIED PIPELINES AT FAULT CROSSINGS

National
Technical
University
of Athens

39

Construction
countermeasures

GEORGE BOUCKOVALAS, National Technical University of Athes, Greece, 2011


Gantes & Bouckovalas ANALYSIS OF BURIED PIPELINES AT FAULT CROSSINGS

7.58

40