EARTHQUAKE ENGINEERING AND STRUCTURAL DYNAMICS, VOL.

27, 283—300 (1998)

STRESSES IN LININGS OF BORED TUNNELS

JOSEPH PENZIEN1* AND CHING L. WU2

1 International Civil Engineering Consultants, Inc., Berkeley, CA 94704, U.S.A.
2 Bechtel Corporation, San Francisco, CA 94104, U.S.A.

SUMMARY
An analytical procedure is presented for evaluating the stresses in linings of bored tunnels caused by kinematic soil—lining
interaction. Three cases of plane—strain interaction are treated as produced by (1) relaxation of in situ soil stresses near
a lining following its installation, (2) overburden pressure at the soil surface, and (3) two-dimensional free-field soil
response normal to the lining axis as produced by an earthquake. The procedure is applied separately to a steel lining and
a concrete lining for a site located at the lower end of Market Street in San Francisco through which a BART (Bay Area
Rapid Transit) tunnel passes. ( 1998 John Wiley & Sons, Ltd.

Earthquake Engng. Struct. Dyn., 27, 283—300 (1998)

KEY WORDS: tunnel linings; tunnel/soil interaction

INTRODUCTION
When boring a cylindrical tunnel through relatively soft soil, a rigid open-face shield (Figure 1) is pressed
forward longitudinally (z direction) into the soil, much like the action of a cookie cutter, transferring the
in situ soil stresses to the outer surface of the shield. As the shield moves forward, a narrow segmental lining
ring is installed directly behind the shield and grout is injected between the outer surface of the ring and the
inner surface of the cylindrical cavity. It is intended that this grout fill any voids or gaps that may be present.
This ring-installation operation is continually repeated as the shield moves forward with the adjacent lining
rings being bolted together to secure them in place within the tunnel.
As the shield passes by a fixed point along the longitudinal axis of the tunnel, the soil pressure distribution
is transferred to the lining. However, since the lining is quite flexible compared to the shield, the lining will
deform under this distribution which, in turn, will modify the pressure distribution. This coupled behaviour
of cause and effect is referred to herein as kinematic soil—lining interaction. This same type of interaction will
be assumed also for the other two interaction cases mentioned above, namely that produced by overburden
pressure at the soil surface and that produced by free-field soil response during an earthquake.
In 1964, Burns and Richard1 published a paper entitled ‘Attenuation of stresses for buried cylinders’ which
presented a procedure for evaluating the soil—structure interaction behaviour produced by an overburden
pressure at the soil surface. The procedure treated the soil surrounding the cylinder as a continuum and
derived two-dimensional plane-strain solutions using the Michell series form of the Airy stress function.
Later in 1972, Peck et al.2 used these solutions to approximate the stresses in a tunnel lining produced by the
relaxation of in situ soil stresses near the lining following its installation by the shield tunnelling method. The
Michell series form of the Airy stress function was also used by Einstein and Schwartz3 in 1979 to analyse the
linings of tunnels having various support conditions.

* Correspondence to: Joseph Penzien, International Civil Engineering Consultants, Inc., 1995 University Avenue, Suite 119, Berkeley,
CA 94704, U.S.A.

CCC 0098—8847/98/030283—18$17)50 Received 27 August 1996

( 1998 John Wiley & Sons, Ltd. Revised 12 August 1997
284 J. PENZIEN AND C. L. WU

Figure 1. Open face tunnel shield (MUNI Metro Turnback Project, Drawing No. DL-10089)

Because the soil—lining interaction behaviour produced by an applied overburden pressure at the soil
surface differs from that behaviour produced by relaxation of the in situ soil stresses near a lining following its
installation, different solutions should be used to determine these two behaviours. Therefore, this paper
presents separate solutions for these two interaction cases using the classical theory of elasticity. In addition,
a solution for the case of soil/lining interaction resulting from free-field soil response during an earthquake is
presented. Various aspects of global seismic response of linings have been published by Merrit et al.4 in 1985
and by Wang5 in 1993, which, in particular, evaluate seismic deformation of cross-sections using the
formulations of Peck et al.2 with the lateral soil coefficient (K ) set equal to !1)0.
0
In each of the three soil/lining interaction cases mentioned above, subsets of solutions are obtained for two
different conditions at the interface between the soil and the lining, namely (1) full continuity of the soil and
lining displacements, and (2) full slippage, without normal separation, resulting in no tangential shear forces.
The effect of pore pressure due to the presence of ground water is included in the solutions.
The subsequent development of analytical procedures for treating the three cases of soil—lining interaction
is separated into four parts (1) in situ stress state of the soil through which the tunnel is bored, (2) generalized
force/displacement relations of the soil, (3) generalized force/displacement relations of the lining, and (4)
soil—lining interactions.

IN SI¹º STRESS STATE OF SOIL

To assess kinematic soil—lining interaction, the in situ stress state of the soil must be known. Consider an
infinitesimal element of soil at Pt. O as shown in Figure 2, which is located at the origin of coordinate axes x,
y and z. The z-axis coincides with the intended longitudinal axis of the tunnel to be bored into the soil as
indicated by the dashed circle of diameter D in this figure. The x and y axes will become the transverse and
vertical axes of the tunnel cross-section, respectively.

( 1998 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn., 27, 283—300 (1998)
 STRESSES IN LININGS OF BORED TUNNELS 285

Figure 2. Soil profile and soil element at Pt. O

Figure 3. In situ soil loading over tunnel height D

Planes xy, yx, and zy are principal planes of the element upon which principal normal stresses act as
follows:

p "!c H "!(p6 #u), p "p "!(K pN #u) (1)

y 4 4 1 x z 0 1
in which c is the average unit weight of the soil over height H , pN is the effective vertical compressive soil
4 4 1
stress, u is the pore pressure given by

u"c H (2)
8 8
and K is the effective lateral soil coefficient, representing the horizontal component of the initial in situ
0
stress, which can have a wide range of values depending upon the site conditions (e.g. if the tunnel is bored
into overconsolidated clay, the value of K can exceed unity considerably).
0
For the purpose of analysis, it will be assumed that the tunnel is to be located sufficiently below ground
surface so that approximate solutions can be obtained by treating the in situ stress state of the soil as being
uniform over the vertical-diameter dimension D as indicated in Figure 3. In doing so, the loading represented
in this figure will be separated into its dilatational and shear-type loadings as shown in Figures 4(a) and 4(b),
respectively.

( 1998 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn., 27, 283—300 (1998)
286 J. PENZIEN AND C. L. WU

GENERALIZED FORCE/DISPLACEMENT RELATIONS OF THE SOIL OUTSIDE

THE TUNNEL

Due to generalized forces on cavity wall

Consider boring a cylindrical tunnel of diameter D into a soil region which is in the in situ stress-state
condition shown in Figure 3 and assume no lining rings are installed following passage of the shield. As the
shield moves forward, the surface of the cylindrical cavity at a particular cross-section will displace inward
unless the full in situ soil loadings transferred to the shield can somehow be maintained. Since these loadings
are not maintained following passage of the shield, the cavity surface will displace inward due to full release of
the in situ loadings shown in Figure 3 which act in opposite directions on the cavity surface. To calculate the
inward displacement due to the release of these loadings, which corresponds to applying the same loadings
shown in Figure 3 to the cavity surface, they will be separated into their dilatational and shear-type
components. Further, for convenience in analysis, these separate cavity-boundary loadings will be expressed
in cylindrical co-ordinates. Doing so for the loadings shown in Figures 4(a) and 4(b), but ignoring the pore
pressure u (since its effects will be treated separately later), one obtains,
pN
p (R, h)" 1 (1#K ), q (R, h)"0 (3)
r 2 0 rh
and
pN pN
p (R, h)"! 1 (1!K )cos 2h, q (R, h)" 1 (1!K )sin 2h (4)
r 2 0 rh 2 0
respectively, which follow the standard Timoshenko sign convention.
Using the cavity loadings expressed by equations (3) and (4), elasticity theory6 can be used to obtain the
stress fields for the region R)r(R and 0)h)2n. The stress field produced by the dilatational loading
of equation (3) is
R2pN !R2pN
p (r, h)" 1 (1#K ), p (r, h)" 1 (1#K )
r 2r2 0 h 2r2 0
(5)
q (r, h)"0
r,h
To obtain the stress fields produced by the cavity-boundary loadings expressed by equations (4), the
compatibility equation, which also satisfies equilibrium, namely

A BA B
L2 1 L 1 L2 L2/ 1 L/ 1 L2/
# # # # "0 (6)
Lr2 r Lr r2 Lh2 Lr2 r Lr r2 Lh2

Figure 4. Separation of in situ soil loading into its dilatational (a) and shear-type (b) components

( 1998 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn., 27, 283—300 (1998)
 STRESSES IN LININGS OF BORED TUNNELS 287

must be used, in which the Airy stress function / is of the following form:
/(r, h)"f (r) cos 2h (7)
which separates the variables r and h. Substituting this function into equation (6) yields the ordinary
differential equation

A BA B
d2 1 d 4 d2f 1 df 4
# ! # ! "0 (8)
dr2 r dr r2 dr2 r dr r2
which has the solution
1
f (r)"Ar2#Br4#C #D (9)
r2
Equation (7) therefore becomes

A B
1
/(r, h)" Ar2#Br4#C #D cos 2h (10)
r2
Having evaluated the Airy stress function, the stress fields in the region R)r(R and 0)h(2n can be
obtained using
1 L/ 1 L2/ L2/
p (r, h)" # , p (r, h)"
r r Lr r2 Lh2 h Lr2
(11)
1 L/ 1 L2/
q (r, h)" !
rh r2 Lh r Lr Lh
Upon substitution of equation (10) into these relations, one obtains

A B A B
6C 4D 6C
p (r, h)"! 2A# # cos 2h, p (r, h)" 2A#12Br2# cos 2h
r r4 r2 h r4
(12)

A B
6C 2D
q (r, h)" 2A#6Br2! # sin 2h
rh r4 r2
The constants A, B, C, and D in these equations can be obtained by satisfying the stress boundary conditions
at r"R as expressed by equations (4) and at r"R as given by p (R, h)"0 and q (R, h)"0. To facilitate
r rh
investigation of normal and tangential interaction between the lining and surrounding soil medium sepa-
rately, equations (4) are divided into the two sets of two separate cavity-boundary loadings given by
pN
p (R, h)"! 1 (1!K ) cos 2h; q (R, h)"0
r 2 0 rh
and (13)
pN
p (R, h)"0; q (R, h)" 1 (1!K ) sin 2h
r rh 2 0
The corresponding stress fields given by equations (12) become

CA B A B D
R 2 1 R 4
p (r, h)"!pN (1!K ) ! cos 2h
r 1 0 r 2 r

AB
pN R 4
p (r, h)"! 1 (1!K ) cos 2h (14)
r 2 0 r

CA B A B D
pN R 2 R 4
q (r, h)"! 1 (1!K ) ! sin 2h
rh 2 0 r r

( 1998 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn., 27, 283—300 (1998)
288 J. PENZIEN AND C. L. WU

and

CA B A B D
R 2 R 4
p (r, h)"!pN (1!K ) ! cos 2h
r 1 0 r r

AB
R 4
p (r, h)"!pN (1!K ) cos 2h (15)
h 1 0 r

C A B A BD
pN R 4 R 2
q " 1 (1!K ) 2 ! sin 2h
rh 2 0 r r
respectively.
Of interest in the subsequent analytical development is the radial-strain relation e (r, h) for each of the
r
above loading cases, since it is needed to determine the corresponding diameter-change distributions with
respect to h. This relation will now be determined under the condition of plane stress as given by Hooke’s law
1
e (r, h)" [p (r, h)!l p (r, h)] (16)
r E r 4 h
4
in which l and E represent Poisson’s ratio and Young’s modulus, respectively, for the soil in its in situ state.
4 4
The resulting diameter-change distributions can then be easily converted to the plane-strain condition
(e "0).
z
Substituting separately the first two of equations (5), (14), and (15) into equation (16) yields
R2pN (1#K )
e$ (r, h)" 1 0 (1#l )
r 2E r2 4
4

CA B A BD
pN (1!K ) R 2 1 R 4
e4/ (r, h)"! 1 0 ! (1#l ) cos 2h (17)
r E r 2 4 r
4

CA B A BD
pN (1!K ) R 2 R 4
e45 (r, h)"! 1 0 !(1#l ) cos 2h
r E r 4 r
4
respectively. The superscript d has been added to the strain term e (r, h) above to denote the dilatational
r
loading given by equations (3), while the superscripts sn and st have been added to denote the normal and
tangential components of the shear-type loading as expressed by the first and second of equations (13),
respectively.
Having the radial strain functions of equations (17), the corresponding diameter-change distributions
around the cavity wall can be obtained using
=
*D(h)"!2
PR er (r, h) dr (18)

Substituting separately each of equations (17) into this expression yields

RpN
*1 D (h)"! 1 (1#K )(1#l )
$4 E 0 4
4
RpN
*1 D4/ (h)" 1 (1!K )(5!l ) cos 2h (19)
44 3E 0 4
4
2RpN
*1 D45 (h)" 1 (1!K )(2!l ) cos 2h
44 3E 0 4
4
( 1998 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn., 27, 283—300 (1998)
 STRESSES IN LININGS OF BORED TUNNELS 289

Adding the last two of these three equations gives the diameter-change distribution resulting from both
components, sn and st, of the shear-type loading acting together. Doing this yields

RpN
*1 D (h)"*1 D4/ (h)#*1 D45 (h)" 1 (1!K )(3!l ) cos 2h (20)
44 44 44 E 0 4
4
The bar has been placed above * in equations (19) and (20) to denote the plane-stress condition. These
equations can be converted to represent the plane-strain condition by substituting E /(1!l2 ) for E and
4 4 4
l /(1!l ) for l giving
4 4 4
RpN
*D (h)"! 1 (1#K )(1#l )
$4 E 0 4
4
RpN
*D4/ (h)" 1 (1!K )(1#l )(5!6l ) cos 2h
44 3E 0 4 4
4
(21)
2RpN
*D45 (h)" 1 (1!K )(1#l )(2!3l ) cos 2h
44 3E 0 4 4
4
RpN
*D (h)" 1 (1!K )(1#l )(3!4l ) cos 2h
44 E 0 4 4
4
Note that *D (h)"*1 D (h) because under the dilatational plane-stress loading condition of equation (3),
$4 $4
p (r, h)"!p (r, h) as shown in equation (5); thus e (r, h)"0 which corresponds identically to the plane-
r h z
strain condition.
Using equations (21), four different generalized stiffness coefficients for the soil can be obtained as follows:
(1) a dilatational coefficient k defined as that value of loading pN (1#K )/2 which will produce the
$4 1 0
diameter-change distribution *D "!1, (2) a shear-component coefficient k4/ defined as that value of
$4 44
pN (1!K )/2 which will produce the diameter-change distribution *D4/ (h)"cos 2h, (3) a shear-component
1 0 44
coefficient k45 defined as that value of pN (1!K )/2 which will produce the diameter-change distribution
44 1 0
*D45 (h)"cos 2h, and (4) a combined shear-component coefficient k defined as that value of pN (1!K )/2
44 44 1 0
which will produce the diameter-change distribution *D (h)"cos 2h. Using these definitions and satisfying
44
equations (21), one obtains

E 3E
k " 4 ; k4/" 4 (22)
$4 2R(1#l ) 44 2R(1#l )(5!6l )
4 4 4
3E E
k45 " 4 ; k " 4 (23)
44 4R(1#l )(2!3l ) 44 2R(1#l )(3!4l )
4 4 4 4
It can be shown that
k4/ k45
k " 44 44 (24)
44 (k4/#k45 )
44 44
corresponding to the two stiffnesses k4/ and k45 being in series.
44 44

Due to an overburden pressure on the soil surface

Consider now the application of an overburden pressure p to the surface of a uniform soil half-space
without a tunnel being present. The principal normal stresses produced at Pt. O in Figure 2 by this loading

( 1998 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn., 27, 283—300 (1998)
290 J. PENZIEN AND C. L. WU

will be
p "!p; p "p "!KM p (25)
y x z 0
in which
KM "l /(1!l ) (26)
0 4 4
In this case, the lateral pressure coefficient KM is a function of Poisson’s ratio, l , only. It does not depend on
0 4
the degree of consolidation within the soil.
The corresponding loading on the soil within the imaginary dashed circle of diameter D in Figure 2 can be
separated into its dilatational and shear-type components in the same way the in situ loading of Figure 3 was
separated as shown in Figure 4. It is easily determined that the diameter-change distributions of the circle
caused by these dilatational and shear-type components are
pR
*D (h)"! (1#KM )(1#l )(1!2l )
$# E 0 4 4
4 (27)
pR
*D (h)" (1!KM )(1#l ) cos 2h
4# E 0 4
4
respectively. If the soil within the imaginary circle is now removed bringing the boundary normal and shear
stresses on the tunnel wall to zero, additional diameter changes will take place. These changes are obtained
by dividing the dilatational and shear-type components of loading eliminated by removal of the soil by the
corresponding generalized stiffness coefficients given by the first of equations (22) and the second of
equations (23), respectively, resulting the diameter-change distributions
pR
*D (h)"! (1#KM )(1#l )
$5 E 0 4
4 (28)
pR
*D (h)" (1!KM )(1#l )(3!4l ) cos 2h
45 E 0 4 4
4
The total dilatational diameter-change distribution of the tunnel wall caused by the overburden pressure p is
obtained by adding the first of equations (27) to the first of equations (28); and, the total shear-type
diameter-change distribution is obtained by adding the second of these two sets of equations.
Doing so, yields
2pR
*1 D (h)"! (1#KM )(1!l2 )
$ E 0 4
4 (29)
4pR
*1 D (h)" (1!KM )(1!l2 ) cos 2h
4 E 0 4
4
in which a bar has been added above the terms *D (h) and *D (h) to denote the loading case of an
$ 4
overburden pressure p applied to the soil surface.

Due to seismic ground response

Assume the free-field ground response produced during an earthquake is in the transverse direction to the
tunnel. From a standard site-response analysis, using the vertically propagating shear-wave model, the
distribution of horizontal free-field ground displacement with depth y, u( y, t), is obtained. In this analysis,
one finds the critical distribution u( y, t ) at time t which produces the maximum shear-type deformation of
# #
the soil over the depth D of the intended tunnel. The function u( y, t ) will usually be reasonably linear over
#
( 1998 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn., 27, 283—300 (1998)
 STRESSES IN LININGS OF BORED TUNNELS 291

Figure 5. State of stress due to seismic soil response

the range !D/2(y(D/2. In this case, the average free-field shear strain of the soil in the xy plane over
this depth is given by the approximate relation
u(!D/2, t )!u(D/2, t )
c" # # (30)
# D
The corresponding state of stress over the depth D is of the shear-type shown in Figure 5(a) which is
equivalent to that state of stress shown in Figure 5(b). The individual stresses in this case are given by
c EM
q"p "!p " # 4 (31)
1 2 2(1#l6 )
4
in which the shear modus GM is equal to EM /2(1#l6 ) and should be assigned a value compatible with the
4 4 4
critical shear-strain level c . These stresses will produce an ovalling of the imaginary circle of diameter
#
D shown in Figure 2 with its major and minor axes in the $45° directions as indicated in Figure 5(b). If
a tunnel of diameter D is present when the earthquake occurs, its diameter-change distribution in the ovalling
mode is obtained by simply substituting c EM /2(1#l6 ) for p(1!KM )/2 and cos 2(h#n/4) for cos 2h in the
# 4 4 0
second of equations (29) resulting in the relation
*2 D (h)"4Rc (1!l6 ) cos 2(h#n/4) (32)
4 # 4
in which the double bar over the term *D (h) denotes the seismic free-field ground response causing the
4
diameter changes.

GENERALIZED FORCE/DISPLACEMENT RELATIONS OF THE LINING

The generalized dilatational stiffness of the lining, defined as that intensity of uniform pressure required on its
outer surface to produce an inward diameter-change distribution equal to unity under the plane-strain
condition e "0, is given by
z
k "E A /2R2(1!l2 ) (33)
$- - - -
For the purpose of evaluating generalized stiffnesses of the lining due to shear-type loading as shown in
Figure 4(b) of intensity p, this loading will be converted to its equivalent normal and tangential components
acting on the outside of the lining as expressed by
p (R, h)"p cos 2h, p (R, h)"p sin 2h (34)
/ 5
which will allow separate investigations of the normal and tangential interactions between the lining and the
surrounding soil medium.

( 1998 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn., 27, 283—300 (1998)
292 J. PENZIEN AND C. L. WU

Figure 6. Sign convention for force components in lining

Using the sign convention shown in Figure 6, it can be shown that the distributions of internal axial force,
shear, and bending moment per unit of longitudinal (z direction) dimension are given by
P (h)"!1 pR cos 2h, » (h)"!2 pR sin 2h, M (h)"!1 pR2 cos 2h (35)
/ 3 / 3 / 3
when only the loading p (R, h) is acting on the lining, and by
/
P (h)"!2 pR cos 2h, » (h)"!1 pR sin 2h, M (h)"!1 pR2 cos 2h (36)
5 3 5 3 5 6
when only the loading p (R, h) is acting on the lining. When both loadings p (R, h) and p (R, h) are acting
5 / 5
together, the distributions are
P(h)"!pR cos 2h, »(h)"!pR sin 2h, M(h)"!1 pR2 cos 2h (37)
2
Using the method of virtual work and assuming the plane-strain condition, the diameter changes at h"0
and h"n/2 due to loading p (R, h) and the diameter changes at these same values of h due to loading p (R, h)
/ 5
are given by the relations
2pR4
*D4/ (R, 0)"!*D4/ (R, n/2)" (1!l2 )
4- 4- 9E I -
- -
and (38)
pR4
*D45 (R, 0)"!*D45 (R, n/2)" (1!l2 )
4- 4- 9E I -
- -
respectively, in which l , E , and I represent the lining’s Poisson’s ratio, Young’s modulus, and its
- - -
circumferential cross-section moment of inertia per unit of longitudinal dimension. The contributions of axial
and shear stresses in the lining to these shear-mode diameter changes have been ignored as they are
negligible. Adding equations (38) yield the diameter changes
pR4
*D (R, 0)"!*D (R, n/2)" (1!l2 ) (39)
4- 4- 3E I -
- -
due to both loadings p (R, h) and p (R, h) acting together.
/ 5
( 1998 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn., 27, 283—300 (1998)
 STRESSES IN LININGS OF BORED TUNNELS 293

Generalized stiffness coefficients for the lining in the shear-type modes under the plane-strain condition
will now be defined as follows: (1) a shear-component coefficient k4/ defined as that value of p which will
4-
produce the diameter-change *D4/ (R, 0) equal to unity, (2) a shear-component coefficient k45 defined as that
4- 4-
value of p which will produce the diameter change *D45 (R, 0) equal to unity, and (3) a shear-coefficient
4-
k defined as that value of p which will produce the diameter change *D (R, 0) equal to unity when both
4- 4-
p (R, h) and p (R, h) are acting together. Using these definitions and satisfying equations (38) and (39), one
/ 5
obtains the coefficients
9E I 9E I 3E I
k4/" - - ; k45" - - ; k " - - (40)
4- 2R4(1!l2 ) 4- R4(1!l2 ) 4- R4(1!l2 )
- - -
which satisfy the relation
k "k4/ k45/(k4/#k45 ) (41)
4- 4- 4- 4- 4-
All of the above generalized stiffness coefficients of the lining will be used in the subsequent soil—lining
interaction analyses.

SOIL—LINING INTERACTIONS
For each of the three loading conditions of the soil treated above, namely (a) relaxation of the in situ soil
stresses at the tunnel wall location, (2) application of an overburden pressure p on the soil surface, and (3)
exposure to free-field seismic ground response, dilatational and shear-type diameter changes of the tunnel
wall with no lining present have been evaluated for generalized displacement patterns of the soil and
generalized stiffness coefficients have been defined for the corresponding generalized soil and lining displace-
ments. This information is sufficient for determining the lining diameter changes which take place due to
soil—lining interaction in each loading case.

Due to relaxation of soil stresses following installation of lining

Consider first the case of relaxing the dilatational soil-stresses in an unlined tunnel following passage of the
shield. After relaxation of the stresses, the cylindrical cavity will have a diameter equal to D#*D ; thus, in
$4
the case when unstressed lining rings of diameter D are installed, the return of pore pressure and the
development of soil—lining interaction following construction will impose a uniform pressure at the
soil/lining interface causing the lining to deform inward in a dilatational mode with diameter change * and
$-
the surface of the soil cavity to deform outward in a dilatational mode with diameter change * . The
$4
diameter change * will consist of one part *u caused by return of the pore pressure u and another part
$- $-
*s resulting from soil—lining interaction. That part caused by the pore pressure is given by
$-
*u "u/k (42)
$- $-
To satisfy compatibility of the soil lining diameter changes, it is necessary that
*u #*s #* "!*D (43)
$- $- $4 $4
in which *D is given by the first of equations (21). To satisfy equilibrium at the soil/lining interface, it is
$4
required that
k * #u"k (*u #*s ) (44)
$4 $4 $- $- $-
Solving equations (42)—(44) simultaneously for *s yields
$-

C D
k u
*s " $4 *D # (45)
$- (k k ) $4 k
$-` $4 $-
( 1998 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn., 27, 283—300 (1998)
294 J. PENZIEN AND C. L. WU

Substituting the first of equations (21), the first of equations (22), and equation (33) into this equation gives,
after some rearrangement of terms, the relation
pN R(1#K )(1#l ) 2uR2(1!l2)
*s " 1 0 4! - (46)
$- E (1#a ) E A (1#a )
4 $ - - $
in which a is a dimensionless parameter defined by
$
a ,E A (1#l )/RE (1!l2 ) (47)
$ - - 4 4 -
Substituting equation (33) into equation (42) yields
2uR2(1!l2)
*u " - (48)
$- EA
- -
Adding equations (46) and (48) expresses the total inward diameter change of the lining, which can be
simplified to the form
pN R(1#K )(1#l ) 2uR2(1!l2)a
* " 1 0 4# - $ (49)
$- E (1#a ) E A (1#a )
4 $ - - $
Equations identical in form to equation (45), but with its pore pressure term u/k removed, can be used to
$-
obtain the diameter changes of the lining due to shear-type soil—lining interactions. If this type of interaction
takes place with full continuity of the soil and lining displacements at the soil/lining interface, then by
substituting the fourth of equations (21), the second of equations (23), and the third of equations (40) into
equation (45), with the term u/k removed, one obtains
$-
pN R(1!K )(1#l )(3!4l )
* " 1 0 4 4 cos 2h (50)
4- E (1#a )
4 4
in which
6E I (1#l )(3!4l )
a" - - 4 4 (51)
4 R3E (1!l2 )
4 -
If the interaction takes place with full slippage at the interface, without normal separation, resulting in no
tangential shear forces, then by substituting the fourth of equations (21), the second of equations (22), and the
first of equations (40) into equation (45), again with the term u/k removed, yields
$-
pN R(1!K )(1#l )(3!4l )
*4/" 1 0 4 4 cos 2h (52)
4- E (1#a4/)
4 4
in which
3E I (1#l )(5!6l )
a4/" - - 4 4 (53)
4 R3E (1!l2 )
4 -
Due to an overburden pressure on the soil surface
Equation (45), again modified with the term u/k removed, can be used to obtain the diameter changes of
$-
the lining due to an overburden pressure p being applied to the soil surface. Substituting the first of
equations (22), the first of equations (29), and equation (33) into this modified form of equation (45), one
obtains the dilatational inward diameter change
2pR(1#KM )(1!l2 )
*1 " 0 4 (54)
$- E (1#a )
4 $
( 1998 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn., 27, 283—300 (1998)
 STRESSES IN LININGS OF BORED TUNNELS 295

in which a is defined by equation (47); substituting the second of equations (23), the second of equations (29),
$
and the third of equations (40), into the modified form of equation (45) yields, for the case of full continuity of
the soil and lining displacements at the soil/lining interface,
2pR(1!KM )(1!l2 )
*1 " 0 4 cos 2h (55)
4- E (1#a )
4 4
in which a is defined by equation (51); and finally, substituting the second of equations (22), the second of
4
equations (29), and the first of equations (40) into the modified form of equation (45) gives, for the case of full
range slippage at the soil/lining interface,
4pR(1!KM )(1!l2 )
*1 4/" 0 4 cos 2h (56)
4- E (1#a4/ )
4 4
in which a4/ is defined by equation (53). A bar has been placed above the terms * , * , and *4/ in
4 $- 4- 4-
equations (54), (55), and (56), respectively, to denote the loading case of an overburden pressure p applied to
the soil surface.

Due to seismic ground response

Again using the modified form of equation (45), in the same way equations (55) and (56) were obtained but
using the unlined tunnel diameter-change distribution given by equation (32), one obtains

A B
4Rc (1!l6 ) n
*2 " # 4 cos 2 h#
4- (1#a ) 4
4 (57)

A B
4Rc (1!l6 ) n
*24/" # 4 cos 2 h#
4- (1#a4/ ) 4
4
The double bars above the terms * and *4/ denote seismic ground motion response as the source of loading
4- 4-
on the lining.

FORCE COMPONENTS IN TUNNEL LINING

Adopting the sign convention shown in Figure 6 for the internal force components P(h), M(h), and »(h), one
can establish for the dilatational displacement mode:
EA*
P "! - - $- , M (h)"»(h)"0 (58)
$ 2R(1!l2 ) $
-
Substituting into the first of equations (58) the values of * given by equations (49) and (54) yields the values
$-
of P for the stress-relaxation and overburden-pressure cases of loading, respectively.
$
For the shear-type displacement modes, one can use equations (37) and (39) to obtain
3E I * 3E I * 3E I *
P (h)"! - - 4- cos 2h, M (h)"! - - 4- cos 2h, » (h)"! - - 4- sin 2h (59)
4 R3(1!l2) 4 2R2(1!l2) 4 R3(1!l2)
- - -
and equations (35) and the first of equations (38) to obtain
3E I *4/ 3E I *4/ 3E I *4/
P (h)"! - - 4- cos 2h, M (h)"! - - 4- cos 2h, » (h)"! - - 4- sin 2h (60)
4/ 2R3(1!l2) 4/ 2R2(1!l2) 4/ R3(1!l2)
- - -
which represent the case of full continuity of the soil and lining displacements at the soil/lining interface and
the case of full slippage at the soil/lining interface, respectively. Substitution of the values of * and *4/ given
4- 4-
( 1998 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn., 27, 283—300 (1998)
296 J. PENZIEN AND C. L. WU

by equations (50) and (52) into the above equations (59) and (60), respectively, will yield results for the
stress-relaxation case of loading and substitution of the values *1 and *1 4/ given by equations (55) and (56)
4- 4-
will yield the corresponding results for the overburden-pressure case of loading. Equations (59) and (60) can
also be used for the seismic loading case by substituting the double-bar values of equations (57) for * and
4-
*4/ and by substituting (h#n/4) for all values of h.
4-
Direct superposition can be used to find the combined internal force components for any two, or all three,
of the loading cases treated above. In doing so, it is helpful to recognize that, when combining the internal
force components due to seismic loading with those of the other two loadings, the following identity can be
used:

A cos 2h#B cos 2h"JA2#B2 cos 2(h#j) (61)

in which

j"1 tan~1(B/A) (62)

2

APPLICATION OF THEORY
The previously developed theory will now be applied to two cases involving a steel lining and a concrete
lining. Both case studies have been carried out for a site located at the lower end of Market Street in San
Francisco where a BART (Bay Area Rapid Transit) tunnel passes through soft soil material (Recent Bay
Mud). The conditions at this site as represented in Figure 1 are the following:

E "14)36]106 N/m2 (300 k/ft2); K "0)65; t "0)394

4 0 4
c "17)28]103 N/m3 (0)110 k/ft3); H "16)31 m (53)5 ft) (63)
4 4
c "9)80]103 N/m3 (0)0624 k/ft3); H "13)26 m; (43)5 ft)
8 8
The soil properties vary considerably at this site so the above numerical values for E , K , and t represent
4 0 4
mean levels.

Case 1: Steel lining

The steel lining consists of segmental rings (six segments per ring) having the cross-section shown in
Figure 7. The properties of this lining are

D"5)49 m (18 ft); E "2)04]1011 N/m2 (29)6]106 psi); t "0)3

- -
(64)
I "9)63]10~5 m4/m (5)87 in4/in); A "0)0314 m2/m (1)236 in2/in)
- -
First, let us assume the case of no slippage between the tunnel lining and its surrounding soil. Then, using the
above lining properties along with those given for the soil previously and introducing their values appro-
priately into equations (47), (51), (49), (50), (58), and (59), one obtains

a "249; a "0)886; * "0)54 mm (0)0213 in); * "10)8 mm (0)425 in)

$ 4 $- 4-
P (h)"!69)8]104 N/m (!47)8 k/ft); P (h"0)"!3)39]104 N/m (!2)32 k/ft) (65)
$ 4
M (h"0)"!46)5]103 Nm/m (!10)4 k ft/ft)
4
As indicated directly below, the maximum deformation and forces in the tunnel lining, particularly the
bending moment, due to relaxation of the in situ soil stresses are significantly smaller than those obtained by

( 1998 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn., 27, 283—300 (1998)
 STRESSES IN LININGS OF BORED TUNNELS 297

Figure 7. Cross-section of steel lining ring

the formulations provided in Reference 2 (with correction for the effects of pore pressure on the hoop stress
and using the condition of no slippage between the lining and soil):

By stress relaxation By overburden pressure (Reference 2)

Max. bending moment 46)5]103 Nm/m (10)4 k ft/ft) 81)82]103 Nm/m (18)39 k ft/ft)
Max. thrust 73)2]104 N/m (50)10 k/ft) 83)10 N/m (56)94 k/ft)
Max. diameter change 11)4 mm (0)446 in) 19)53 mm (0)769 in)

At the above-mentioned BART tunnel site, it has been estimated that the average free-field soil shear strain
c could reach the level 0)0028 under maximum credible earthquake conditions (Reference 8). Making use of
#
this value, equations (57) and (59) yield

* "9)98 mm (0)392 in); P (h"!n/4)"!3)13]104 N/m (!2)14 k/ft)

4- 4
(66)
M (h"!n/4)"42)9]103 Nm/m (!9)64 k ft/ft)
4
The maximum combined force components in the ovalling mode, when combining the above stress-
relaxation and seismic values in accordance with equation (61), are the following:

P (0°#j)"![(3)39)2#(3)13)2]1@2104"!4)61]104 N/m (!3)16 k/ft)

4
(67)
M (0°#j)"![(46)5)2#(42)9)2]1@2103"!63)3]103 Nm/m (!14)2 k/ft)
4
The corresponding value of j is (1/2) tan~1(3)13/3)39)"21)4°. Adding P (!21)4°) to P (h) gives the total
4 $
combined axial force at h"!21)4° equal to !74)4]104 N/m (!51)0 k/ft) where the moment equals
!63)3]103 Nm/m (!14)2 k ft/ft). Combining the resulting circumferential normal stresses produced by
these components, one obtains for the cross-section properties and dimensions shown in Figure 7

p (h"!21)4°)"#10)5]106 N/m2 (#1508 psi)

0
(68)
p (h"!21)4°)"!106)4]106 N/m2 (!15 433 psi)
*
at the outer and inner faces of the lining, respectively.

( 1998 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn., 27, 283—300 (1998)
298 J. PENZIEN AND C. L. WU

At h"(!90°!21)4°), P (!111)4°)"#4)61]104 N/m (#3)16 k/ft) and M (!111)4°)"#63)3]103

4 4
Nm/m (#14)3 k/ft). Combining the normal stresses produced by [P (!111)4°)#P (h)] and M (!111)4°),
4 $ 4
one obtains

p (h"!111)4°)"!54)9]106 N/m2 (!7951 psi)

0
(69)
p (h"!111)4°)"#61)0]106 N/m2 (#8991 psi)
*
at the outer and inner faces, respectively. The maximum compressive stress (!106)4]106 N/m2) and the
maximum tensile stress (#61)0]106 N/m2) are approximately 43 and 25 per cent of the linings yield stress
(#249]106 N/m2), respectively.
Let us now consider the case of full slippage between the tunnel and surrounding soil. Using the
appropriate formulas for the full-slippage case contained in this paper and repeating the same calculations
as for the no-slippage case, the combined normal stresses at the outer and inner fibres of the lining are found
to be:

p (h"!21)4°)"#12)3]106 N/m2 (#1787 psi)

0
p (h"!21)4°)"!108)8]106 N/m2 (!15 760 psi)
*
p (h"!111)4°)"!56)8]106 N/m2 (!8230 psi)
0
p (h"!111)4°)"#64)3]106 N/m2 (#9318 psi)
*
The effect of full slippage increases the bending moment and diameter change by approximately 4 per cent
and reduces the thrust by approximately 4 per cent. The maximum tensile and compressive stresses in the
lining are increased by approximately 4 and 2 per cent, respectively.
From the above results, it is quite clear that the BART steel lining would perform quite satisfactory under
the above combined static and seismic loading conditions.

Case 2: Concrete lining

Consider a concrete lining of diameter D"5)49 m (18 ft) and of thickness t"0)305 m (1 ft) having
a compressive strength f @"34)47]106 N/m2 (5000 psi), a Young’s modulus E "27)77]109 N/m2
# -
(0)58]106 k/ft2), and a Possion’s ratio l "0)17.
-
Carrying out the same evaluations described above for discrete values of thickness t, the stresses in the
inner fibres at the critical locations h"!j and h"(!90°!j) can be obtained as functions of lining
thickness. These functions are shown in Figure 8 over the thickness range 0(t(0)61 m (2 ft). For
t"0)305 m (1 ft), the maximum combined compressive stress at h"!j equals !8)96]106 N/m2
(!1298 psi), of which !6)43]106 N/m2 (!932 psi) and !2)53]106 N/m2 (!367 psi) are produced by
deformations in the ovalling and dilatational modes, respectively. For this same thickness, the maximum
combined tensile stress at h"(!j!90°) equals #4)37]106 N/m2 (#634 psi), of which #6)43]106 N/m2
(#932 psi) and !2)05]106 N/m2 (!298 psi) are produced by deformations in the ovalling and dilatational
modes, respectively. The maximum compressive stress (!8)96]106 N/m2 (!1298 psi)) is well below
f @"34)47]106 N/m2 (!5000 psi); however, the maximum tensile stress (#4)37]106 N/m2 (#634 psi))
#
exceeds the tensile strength of the concrete by about 30 per cent. As shown by Pt. A on the upper curve,
a thickness of 0)366 m (1)2 ft) is required to reduce the maximum tensile stress to 3)45]106 N/m2 (500 psi)
corresponding to the tensile strength of the concrete. As seen in Figure 8 for a given concrete lining thickness,
the maximum normal stresses at the outer and inner fibres of the lining are up to 12 per cent higher for the
full-slippage case than for the no-slippage case.
For a thickness in the approximate range 0)1 m (0)328 ft)(t(0)214 m (0)7 ft), a small increase in its value
results in increasing the critical tension stress; while, for a thickness t'0)214 m (0)7 ft), an increase in its

( 1998 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn., 27, 283—300 (1998)
 STRESSES IN LININGS OF BORED TUNNELS 299

Figure 8. Stresses at inner fibres of concrete lining

value results in decreasing this critical stress. This latter decreasing of the critical stress is due to the
increasing strong soil—lining interaction which takes place.

CONCLUDING REMARKS
The unified treatment of kinematic soil—lining interactions produced by (1) relaxation of in situ soil stresses
near a lining following its installation, (2) overburden pressure applied to the soil surface, and (3) free-field soil
response resulting from an earthquake, as presented herein, provides simple formulas for the calculation of
lining deformations, forces, and corresponding stresses. These response quantities as produced by the
relaxation of in situ soil stresses are significantly reduced from those produced by the ‘equivalent’ overburden
pressure loading, p"c H thus, distinctly different solutions should be used for these two loading cases to
4 4
avoid unnecessary conservatism in design.
The separate solutions obtained for the cases of full slippage and no slippage at the soil/lining interface
during the occurrence of shear-type deformations indicated that full slippage has only a small effect on the
shear and bending moments in the lining, but does reduce the maximum axial force by approximately 50 per
cent. However, since the axial force produced by the dilatational-type deformation is much greater than the
maximum axial force produced by the shear-type deformation, this 50 per cent reduction has little effect on
the combined axial force due to both types of deformation. Consistent with this observation, slippage has
little effect on the maximum combined normal stresses in the lining as produced by in situ stress relaxation
and seismic free-field soil response.

( 1998 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn., 27, 283—300 (1998)
300 J. PENZIEN AND C. L. WU

It should be emphasized that the analytical procedure developed herein for determining lining stresses
following installation is based on the assumption of perfect construction methods, i.e. the lining is placed in
position in a circular stress-free state and grouting outside the lining completely fills all voids present. Since
a perfect installation of this type cannot be achieved, the stress-state of linings as actually installed and the
deficiencies which can occur in grouting should be taken into consideration in the design process. As pointed
out by Kuesel,7 the steel rings installed in the BART tunnels were subjected to considerable ovalling due to
elastic flexure and working of the joints under their own dead weight and additional distortions were
introduced as the linings were bolted together. Even though uncertainties exist regarding the stress state of
linings during and following construction, evaluation of stress states under assumed ideal conditions, using
the procedures developed herein, provides a sound starting point, the results of which can then be adjusted
using engineering judgement as deemed appropriate.
Under extreme seismic conditions, the lining cross-section may deform significantly into the inelastic
range. Such deformations should be controlled so as to force them to occur in the interior regions of the
lining segments, not in the circumferential joints. Under moderate seismic conditions, the lining should be
designed to remain elastic; in which case, the analytical procedures developed herein can be used to assess
performance.8,9

REFERENCES
1. J. Q. Burns and R. M. Richard, ‘Attenuation of stresses for buried cylinders’, Proc. Symp. on Soil—Structure Interaction, ºniversity of
Arizona, Engineering Research Laboratory, Tucson, AZ, September, 1964.
2. R. B. Peck, A. J. Hendron Jr. and B. Mohraz, ‘State of the art of soft-ground tunneling’, Proc. RE¹C, Vol. 1, 1972.
3. H. H. Einstein and C. W. Schwartz, ‘Simplified analysis for tunnel supports’, J. Geotech. Engineering Div. ASCE, GT4, 499—518 (April
1979).
4. J. L. Merritt, J. E. Monsees and A. J. Hendron, ‘Seismic design of underground structures’, Proc. RE¹C, Vol. 1, 1985.
5. J.-N. Wang, Seismic Design of ¹unnels, Monograph 7, Parsons Brinkerhoff Quade & Douglas, June 1993.
6. S. Timoshenko and J. N. Goodier, ¹heory of Elasticity, 2nd edn, McGraw-Hill, New York, 1951.
7. T. R. Kuesel, ‘Soft ground tunnels for the BART project’, Proc. RE¹C, Vol. 1, 1972.
8. C. L. Wu and J. Penzien, ‘Seismic design of MUNI metro turnback project’, Proc. 5th º.S. National Conf. on Earthquake Engineering,
Vol. IV, 1994.
9. C. L. Wu and J. Penzien, ‘Stress analysis and design of tunnel linings’. Proc. RE¹C, Las Vegas, NV, June 1997.

( 1998 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn., 27, 283—300 (1998)