Tunnelling and Underground Space Technology 88 (2019) 87–97

Contents lists available at ScienceDirect

Tunnelling and Underground Space Technology

journal homepage: www.elsevier.com/locate/tust

A closed-form solution for the seismic racking and rocking behavior of T

rectangular tunnels
⁎
Carlos Gordo-Monsó , Jesús González-Galindo, Claudio Olalla-Marañón
E. T. S. de Ingenieros de Caminos, C. y P., Universidad Politécnica de Madrid, 28040 Madrid, Spain

A R T I C LE I N FO A B S T R A C T

Keywords: Current design methodologies for underground structures subjected to earthquake action are based on closed-
Racking form solutions of the so-called racking coefficient, which relates the soil free-field distortion deformation to the
Rocking structure distortion deformation. Existent closed-form expressions for this racking coefficient are derived for
Underground circular tunnel geometries embedded in an elastic and homogeneous medium, while expressions for rectangular
Tunnel
geometries are based on analogies of the circular case. In this paper, a new and intuitive methodology derived
Seismic
Soil-structure interaction
from first principles to obtain a closed-form expression for the racking coefficient for rectangular-like tunnel
geometries buried at a sufficient depth is presented, and it is shown that it provides a good fit to finite element
analyses results. Similarly, a procedure to obtain closed-form expressions for the rocking rotation of the struc-
ture, and the racking deformation profile at different distances from the soil-tunnel interface is presented.

1. Introduction embedded underground structure must accommodate with limited da-

mage. The approach consists, at a first stage, in determining the soil
Underground structures behavior subjected to the earthquake action displacement profile, at a distance far enough of the structure (i.e. free-
differ notably from the case of above ground structures. In most cases field displacement), and at the depth comprised by the underground
for underground structures, the inertial component of the structure structure (i.e. depth between invert slab and roofing system). A second
mass displacement is negligible and does not govern the dynamic be- stage consists in imposing this free-field displacement to the structure,
havior during an earthquake (Bobet, 2010; Huo et al., 2006; Penzien, analyze the internal forces, and design structural elements accordingly.
2000; Pitilakis et al., 2007; Tsinidis et al., 2015a; Wang, 1993), whereas Given that this approach imposes a free-field displacement to the
for surface structures this is the most significant parameter. Moreover, structure, it does not consider that the soil-structure interaction may
buried structures are constrained by the surrounding soil, and hence locally change the deformation profile in the nearby soil surrounding
subjected largely to a displacement range similar to that undergone by the underground structure.
the soil (Bobet et al., 2008; Hashash et al., 2010; Huo et al., 2005). Recognizing that the embedded structure presence may alter the soil
To assess the displacements, and hence internal forces, of under- deformation pattern in its proximity, a more refined approach termed
ground structures during an earthquake, three different approaches are pseudo-static soil-structure interaction method, was developed during
followed (Bobet et al., 2008; Hung et al., 2009; LA Metro Authority, the 90s by several authors, for which a comprehensive review is pro-
2012; WSDOT, 2010), these are: pseudo-static free-field displacement vided by Hashash et al. (2001). Similar in concept to the free-field
approaches, pseudo-static soil-structure interaction displacement ap- displacement approach, once that the soil displacement profile is de-
proaches, and numerical soil-structure interaction approaches. termined, an amplification or reduction factor termed racking coefficient
Pseudo-static free-field displacement approaches (Hendron and is applied to the free-field deformation to obtain the structure de-
Fernandez, 1983, Merrit et al., 1985, Monsees, 1991) constitute the formation which must be accommodated with limited damage. This
early attempts to study the linked behavior of soil and underground racking coefficient, which encapsulates the soil-structure interaction
structure during an earthquake. The first proposed method, developed (SSI) effect, was obtained analytically based on the elasticity theory,
for the San Francisco Trans-Bay Tunnel (Kuesel, 1969), accurately de- and further verified by numerical analyses. This method has been
scribes the main features of the problem: regarding the soil, the substantiated in different closed-form expressions for the racking
earthquake consists of a set of passing deformation waves that the coefficient, each one developed under slightly different assumptions.

⁎
Corresponding author.
E-mail addresses: [email protected] (C. Gordo-Monsó), [email protected] (J. González-Galindo), [email protected] (C. Olalla-Marañón).

https://doi.org/10.1016/j.tust.2019.03.002
Received 29 September 2018; Received in revised form 5 February 2019; Accepted 2 March 2019
Available online 12 March 2019
0886-7798/ © 2019 Elsevier Ltd. All rights reserved.
C. Gordo-Monsó, et al. Tunnelling and Underground Space Technology 88 (2019) 87–97

Fig. 1. (a) Block diagram of an arbitrary portion of the soil-structure system. (b) Pure shear distortion strain scheme for an individual block.

Wang (1993) proposed that a single parameter FR termed flexibility ratio (Anderson et al., 2008; Penzien, 2000; Wang, 1993) rely on the as-
(measuring the relative shear stiffness of the soil substituted by the sumption that the stress and strain field in the surrounding soil is si-
underground structure, compared to the shear stiffness of the structure milar to the one developed for a circular cross-section. The expression
itself) governs the SSI effect (the more flexible the structure is re- by Huo et al. (2006) is effectively derived for a rectangular cross sec-
spective to the soil, the more the structure deforms and vice versa), and tion, nevertheless the underlying mathematical development is intricate
proposed a closed form expression for the rocking coefficient of circular and may hinder the understanding of the mechanical phenomena to the
tunnels based on previous work by Peck (1969). Penzien (2000), engineer designing the underground structure.
Penzien and Wu (1998) developed a background for the racking coef- In the following, we propose a closed-form solution developed for
ficient based on the theory of elasticity and provided a closed-form the rectangular geometry case that, while trying to be as accurate as
solution for the racking coefficient similar to that of Wang. Anderson possible, provides the engineer with an effective mean to intuitively
et al. (2008) disregard the effect that the Poisson coefficient may have understand the very nature of the racking effect (namely, a larger
in the SSI phenomenon, providing a closed-form expression similar to structure deformation than the one corresponding to the free-field if the
that of Wang and Penzien corresponding to a constant Poisson coeffi- structure is more flexible than the soil, and vice versa). This intuitive
cient ν = 0.5. Huo et al. (2006) acknowledge that the previously ob- understanding is an aspect that should not be disregarded: as explained
tained racking coefficient expressions for rectangular tunnels disregard before, the free field methods (Kuesel, 1969; Monsees, 1991) are based
the effect of the out-of-plane displacement of the walls and slabs that on the assumption that the soil close to the embedded structure deforms
react against the surrounding soil, and hence develops a closed-form similarly to the soil far from it and that no SSI effect is present, probably
expression for the racking coefficient by means of conformal mapping guided by an intuition that was corrected with later studies.
techniques of a rectangular tunnel and a circular tunnel.
Numerical soil-structure interaction approaches constitute the most
3. Proposed solution for the racking coefficient
detailed methodology to study the effect of an underground structure
subjected to the seismic action. These consist on the numerical analysis
Consider a rectangular underground structure (of width b, depth d,
of the problem by developing a finite elements or finite differences
and aspect ratio λ = b/d), and a portion of surrounding soil of arbitrary
model of the buried structure-soil system, and performing whether a
size (of width L and depth H) concentric with the underground struc-
full dynamic analysis or a pseudo-static analysis, both of these linear or
ture as in Fig. 1a. Consider that the soil exterior perimeter is subjected
non-linear, depending on the feature of interest to be studied (Abate
to a pure shear stress state, and consider that our purpose is to de-
and Massimino, 2017; Debiasi et al., 2013; Fabozzi et al., 2017;
termine the strain field in the soil and structure.
Guoxing et al., 2015; Hashash et al., 2001; Hung et al., 2009; LA Metro
In determining the strain field, the following 3 assumptions are
Authority, 2012; Wang et al., 2013; WSDOT, 2010). Although this ap-
made:
proach might appear the most thorough and precise, given that it may
appropriately consider different mechanical effects, in practice it is
▪ The soil-structure system can be divided in rectangular blocks,
seldom used during early design stages of an underground structure (LA
numbered 1–9 (Fig. 1a).
Metro Authority, 2012). Firstly, the design process covers conceptual,
▪ Each block can only deform in a pure shear pattern (Fig. 1b), with
preliminary, and detailed design phases, during which the functional
the shear stiffness of an elastic, homogenous, and isotropic medium,
and architectural planning (i.e. passenger circulations in subway sta-
defined by its shear modulus G and the elasticity equation γ = τ/G.
tions) may greatly and frequently alter the previous structural and
Note that this equation measures the total sheared angle γ between
geotechnical configuration. This rends a detailed finite element ap-
orthogonal boundaries, but that the sheared angle respective to the
proach highly impractical until later design stages, when many relevant
vertical and horizontal boundaries of the block is γ/2 each (Hudson
engineering decisions may have already been taken, and there is little
and Harrison, 2000).
place for changes but for verification. Secondly, the solely reliance in a
▪ There is displacement compatibility between corners of adjacent
complex numerical model for obtaining relevant structural results (i.e.
blocks, that is, corners shared by different blocks displace horizon-
bending moments, shear forces, displacements, etc.) may leave the
tally and vertically the same amount. This would translate to a non-
engineer without a practical and more immediate tool to intuitively
slip condition in terms of previous studies (Penzien, 2000; Wang,
guide her decisions towards a safe and economic design.
1993).

2. Purpose of the proposed solution Stemming from the above, it is assumed that within each block
shear stresses and strains are constant through the entire width or
As exposed in previous sections, from the closed-form expressions depth, in particular at its boundaries. This assumption means that no
available to date for the analysis of the SSI effect, and hence the racking stress or strain concentration will take place in the corners of the soil-
deformation of a rectangular underground structure, 3 of them structure, and that only an “average stress” and “average strain” will be

88
C. Gordo-Monsó, et al. Tunnelling and Underground Space Technology 88 (2019) 87–97

Fig. 2. (a) Block diagram with horizontal layers, and horizontal shear stress transfer between layers, horizontal shear stiffness of each block. (b) Block diagram with
vertical layers, vertical shear stress transfer between layers, and vertical shear stiffness of each block.

considered for each block. Moreover, the total angle γ sheared by each as shown in Eqs. (6)–(9).
block will be the sum of the sheared angle that produces horizontal c
displacement γH, plus the sheared angle that produces vertical dis- K1, V = K3, V = K7, V = K 9, V = 2·G·
a (6)
placement γV.
If one follows the transfer of, say, horizontal shear force from the c
K2, V = K 8, V = 2·G·
top to the bottom boundary through the depth of the system, from b (7)
equilibrium considerations, it is easy to note that the total shear force at
any horizontal section must be constant (Fig. 2a) (even if the horizontal d
K 4, V = K 6, V = 2·G·
a (8)
shear stress at each block may be different). Similarly, stemming from
the above assumptions, and from displacement compatibility con- d
siderations between contiguous blocks, one can note that for each G· b
K5, V = K STRU , V = 2·
horizontally contiguous block (that is pertaining to the same horizontal FR (9)
layer, namely the top layer of blocks 1–3, the intermediate layer of
From the previous expressions (1)–(4), and owing to the shear strain
blocks 4–6, and the bottom layer of blocks 7–9) the horizontal shear
compatibility between horizontal adjacent blocks, the shear stiffness for
distortion strain γH must be of equal value.
each horizontal layer (KTOP for the top layer, KINT for the intermediate,
Therefore, the horizontal displacement stiffness for each horizontal
and KBOT for the bottom layer) can be obtained as an in-parallel sum of
layer of blocks must be the in parallel sum of the shear stiffness of each
stiffness (Eqs. (10) and (11)).
of the three blocks within the layer. In particular, and considering that
the horizontal displacement corresponds solely to the horizontal KTOP, H = KBOT , H = K1, H + K2, H + K3, H = K7, H + K 8, H + K 9, H
sheared angle γ/2, the stiffness Kn,H of each block n for a horizontal a+b+a L
force will be as shown in Eqs. (1)–(4). = 2·G· = 2·G·
c c (10)
a
K1, H = K3, H = K7, H = K 9, H = 2·G· 2a
c (1) KINT , H = K 4, H + K5, H + K 6, H = 2·G· + K STRU , H
d (11)
b
K2, H = K 8, H = 2·G· The shear stiffness for each vertical layer (KLEFT for the left layer,
c (2)
KCENT for the central layer, and KRIGHT for the right layer) can be ob-
a tained similarly (Eqs. (12) and (13)).
K 4, H = K 6, H = 2·G·
d (3)
KLEFT , V = KRIGHT , V = K1, V + K 4, V + K7, V = K3, V + K5, V + K 9, V
b
G· d c+d+c H
K5, H = K STRU , H = 2· = 2·G· = 2·G·
FR (4) a a (12)

where the term FR in Eq. (4) corresponds to the classical flexibility ratio 2c
K CENT , V = K2, V + K5, V + K 8, H = 2·G· + K STRU , V
as defined by Wang (1993), which in the case of linearly elastic a single b (13)
barrel rectangular tunnel of width b, depth d, and wall and slab bending
inertias Iw and Ib respectively is defined by Eq. (5). If one were to estimate the average horizontal shear stiffness KAVG,H
of the soil and structure ensemble (that is the 9 blocks), by measuring
b
K SOIL G· d the force required at the top face in order to produce a unit horizontal
FR = = 1 displacement, an appropriate method consistent with the above as-
K STRU
d2·b d3
+ sumptions would be to sum the in-series stiffness of each horizontal
24·E·Ib 24·E·Iw (5)
layer (Eq. (14)).
Note that this particular expression for FR is not relevant for the
1 1
procedure described in this paper, and can be easily changed to any KAVG, H = 1 1 1
= 1 1
other shear deformation stiffness for a given application (i.e. stiffness KTOP , H
+ KINT , H
+ KBOT , H
2· L + 2a
2·G· 2·G· + K STRU , H (14)
c d
with inner columns, inner horizontal slabs, etc.).
A similar reasoning would be valid for the vertical shear force Similarly for the average vertical shear stiffness KAVG,V of the soil
transferring right to left (Fig. 2b). Using appropriate geometrical di- and structure ensemble, adding the in-series stiffness of each vertical
mensions for the stiffness Kn,V of each block n for a vertical force will be layer (Eq. (15)).

89
C. Gordo-Monsó, et al. Tunnelling and Underground Space Technology 88 (2019) 87–97

1 1 (15) with the average horizontal and vertical stiffness of the more in-
KAVG, V = 1 1 1
= 1 1
KLEFT , V
+ KCENT , V
+ KRIGHT , V
2· H + 2c ternal perimeter i − 1, KAVG,H,i-1 and KAVG,V,i-1, and choose geometrical
2·G· 2·G· + K STRU , V (15)
a b dimensions ai, bi, ci, di, Li and Hi corresponding to the given control
With the previous results at hand, one can compute an approximate perimeter i, as shown in Eq. (17).
racking coefficient for the shear distortion of the embedded structure Li
+ b ·K i
H
~ γAVG, H , i − 1 + γAVG, V , i − 1 di·KINT , H , i − 1 i CENT , V , i − 1
and the average shear distortion of the soil and structure ensemble, that Ri = = Li H
is to say, the ratio of shear distortions between the inner structure γAVG, H , i + γAVG, V , i
Hi·KAVG, H , i
+ L ·K i (17)
i AVG, V , i
γINN = γINT,H + γCENT,V and the outer perimeter γAVG = γAVG,H + γAVG,V,
when the outer perimeter is subjected to a pure shear stress state of If we follow the reasoning, we may think that the total racking
value τ. This racking coefficient would be the ratio of the sum of strains coefficient relating the embedded structure shear distortion γSTRU, and
~
as shown in Eq. (16), where the hat in R denotes that this racking ex- the free-field distortion γFF, can be approximated by a sufficiently large
pression is approximate and conditional to the block behavior as- portion of soil surrounding the structure with a fine discretization of
sumptions stated before. successive control perimeters, and by the product of each individual
racking coefficient as in the multiplicative Eq. (18).
τ ·L / KINT , H τ ·H / KCENT , V L H
γINT , H + γCENT , V + + b·K
~ d b d·KINT , H CENT , V γSTRU i=n
γi − 1 i=n
R=
γAVG, H + γAVG, V
= τ ·L / KAVG, H
+
τ ·H / KAVG, V
= L
+ L·K
H R=
γFF
≅ ∏ γi
= ∏ R~i
H L H ·KAVG, H AVG, V i=1 i=1 (18)
(16) ~
In Eq. (18) the Ri expression is given by Eq. (17), and in Eq. (17) for
From the previous expressions, and for the sake of simplicity con- i = 1 corresponding to the innermost control perimeter, the KAVG,H,i-1=0
sidering only the horizontal shear strain contribution (Eqs. (10), (11), and KAVG,V,i-1=0 are the structure stiffness given by Eqs. (4) and (9).
(14), and (16)), one can readily obtain some important information From the exposition above, we must note that the proposed pure
regarding the behavior of the soil and structure ensemble. As it would shear block model cannot consider appropriately the solution depen-
be expected, if the intermediate layer were stiffer than the soil-structure dence on the Poisson coefficient demonstrated by other researchers
ensemble, the intermediate layer would deform less than the ensemble (Huo et al., 2006; Penzien, 2000; Wang, 1993). This is because, being
(Fig. 3a). On the contrary, if the intermediate layer were softer, the each block subjected to a pure shear state, there is no direct dependence
intermediate layer would deform more than the soil and structure en- for the shear strain on the Poisson coefficient once that the G modulus is
semble (Fig. 3b). defined. Given that in a pure shear state there is no change in volume,
Notice that the distortion stiffness of the intermediate layer is one might expect that the proposed multiplicative closed-form solution
governed by the distortion stiffness of the embedded structure: a small should best replicate the behavior for those cases in which the Poisson
structure stiffness when compared with the substituted soil (FR > 1) coefficient is close to ν = 0.5. This lack of dependence of the proposed
would yield an intermediate layer softer than the ensemble, and vice model on the Poisson coefficient can be adjusted introducing a mod-
versa. A similar reasoning can be made regarding the vertical compo- ification factor which will be presented later in this paper.
nent of the shear strain, for which two cases one with a stiffer structure In order to verify the accuracy of the proposed closed-form solution
than soil (FR < 1), and another with a softer structure than soil for the racking coefficient, a series of finite element models (FEM) have
(FR > 1) are depicted in (Fig. 4a) and (Fig. 4b) respectively. been carried out in OpenSEES (McKenna et al., 2010). The FEM ana-
Until this point in the discussion, the dimensions L and H for the soil lyses developed represent the problem of a sufficiently large elastic,
perimeter have been chosen arbitrarily. If we think for which dimen- isotropic, and homogenous soil medium with an embedded rectangular
sions the block assumptions may be approximately valid, we may structure at its center. The soil has been meshed with 4 nodes quad-
consider that, for a soil perimeter close enough to the structure, the rilateral elements and plane-strain behavior, whereas the lining
average shear distortion can be supposed constant in each layer. meshing has been performed with beam elements whose nodes are
Therefore, we may think that we can compute approximately a racking linked in vertical and horizontal displacement to the adjacent soil
coefficient between a soil perimeter close enough to the structure and nodes, as to represent a fully bonded concrete-soil behavior. The soil
the structure itself by means of Eq. (16). external boundaries do not present displacement restrictions, and are
Moreover, we may assume that if we consider successive soil control subjected to a pure shear stress state of value τ. Only the bottom left-
perimeters (Fig. 5), close enough one to each other (say perimeters i − most soil node has horizontal ux and vertical uy displacements re-
1 and i), we may compute an approximate individual racking coefficient stricted, and the bottom rightmost soil node has vertical displacement
~
Ri , relating the shear distortion of successive perimeters, using Eqs. uy restricted, as to achieve numerical equilibrium. The considered soil
(14)–(16). shear wave velocity is Vs = 360 m/s and the soil specific weight is
For doing so in any given control perimeter i, it will suffice to ρ = 20 kN/m3, yielding a shear modulus G = 259.20 MPa. Regarding
substitute the structure stiffness KSTRU,H and KSTRU,V in Eqs. (14) and the tunnel, a concrete structure has been considered with Young

Fig. 3. (a) Horizontal shear profile for a structure stiffer than the substituted soil (FR < 1), with intermediate layer shear strain γINT,H smaller than the average strain
γAVG,H. (b) Horizontal shear deformation profile for structure softer than the substituted soil (FR > 1), with intermediate layer shear strain γINT,H larger than the
average strain γAVG,H.

90
C. Gordo-Monsó, et al. Tunnelling and Underground Space Technology 88 (2019) 87–97

Fig. 4. (a) Vertical shear profile for a structure stiffer than the substituted soil (FR < 1), with central layer shear strain γCENT,V smaller than the average shear strain
γAVG,V. (b) Vertical shear deformation profile for a structure softer than the substituted soil (FR > 1), with central layer shear strain γINT,V larger than the average
shear strain γAVG,V.

intermediate perimeter distorts respective to the free-field distortion.

Regarding the multiplicative closed-form method presented in the
previous section, an expression similar to Eq. (18) to compute such
profile for a given perimeter j would be Eq. (19).
γPERIMj i=n i=n
γi − 1 ~
RPERIMj =
γFF
≅ ∏ γi
= ∏ Ri
i=j+1 i=j+1 (19)

As for the racking coefficient, the same set of FEM analyses has been
used to obtain numerical results for the racking profile. The average
shear distortion γFEM,PERIM j at each control perimeter j in the FEM
analyses has been computed as the horizontal and vertical displacement
difference for the bottom leftmost corner and the top rightmost corner
of each perimeter with Eq. (20) (Fig. 7a), and then normalized by the
free-field distortion γFF to obtain the racking profile at each perimeter j.
ΔHORIZ , TOP, j − ΔHORIZ , BOTTOM , j ΔVERT , RIGHT , j − ΔVERT , LEFT , j
γFEM , PERIMj = +
Hj Lj
(20)
Results are shown in (Fig. 8) where the horizontal abscissa has been
Fig. 5. Successive series of control perimeters 1 to n. Each individual racking
~ normalized among models as the ratio of the distance of the control
coefficient Ri measures the ratio of shear distortion between control perimeter
~ γ perimeter to the structure X , divided by the structure width b (XPERIM-
i-1 and i, being Ri = AVG ,i−1
.
ETER/bSTRUCTURE = 0 means a perimeter adjacent to the structure, and
γAVG, i

XPERIMETER/bSTRUCTURE = 2 means a perimeter at two widths from the

modulus Ec = 30 GPa, and constant wall and slab thickness of structure).
hw = hb = 1.50 m providing the gross bending inertia. A range of flex- In Fig. 8 the multiplicative expression Eq. (19) shows a good
ibility ratios FR has been studied by varying the width b and depth d agreement with the finite element results, and indicates that at a dis-
dimensions of the structure from 1 m to 20 m. Different structure aspect tance of 1.5–2 structure widths the shear distortion value remains close
ratios λ = b/d ranging from 1 to 3 have been analyzed to study its in- to the free-field distortion (R ≈ 1), that is to say, at farther distances
fluence on the solution. The soil boundaries, both for the FEM analyses than two widths the influence of the embedded structure becomes al-
and the closed form solutions proposed in this paper have been chosen most negligible.
50 m away horizontally and vertically from the structure. For the eva-
luation of the closed-form solution, a value of 1000 control perimeters
5. Proposed solution for the rocking coefficient
(n = 1000 terms in Eq. (18)) has been employed.
Fig. 6a shows a very good agreement of the proposed multiplicative
Several researchers (Cilingir and Madabhushi, 2011; Debiasi et al.,
solution Eq. (18) with the FEM analyses for all the studied aspect ratios
2013; Iwatate et al., 2000; Pitilakis and Tsinidis, 2016; Tsinidis, 2017;
λ, and for the case in which the Poisson coefficient is ν = 0.5 as ex-
Tsinidis et al., 2015b; Tsinidis and Pitilakis, 2018; Ulgen et al., 2015)
pected. Similarly, this expression closely matches results from other
have pointed an interesting rocking behavior shown by buried struc-
closed-form expressions for circular geometries (Anderson et al., 2008;
tures presenting relative stiffness departing from the substituted soil
Penzien, 2000; Wang, 1993) when ν = 0.5. For other Poisson coeffi-
stiffness. Stiff structures (FR < 1) tend to present a rigid-body rocking
cient ν values, and when the modification factor introduced later in this
displacement pattern, in the sense that the structure tends to rotate
paper is applied, similar good agreement with the FEM analyses, and
rigidly in the same direction as the global shear distortion. On the
similar trends as those in other closed-form expressions can be appre-
contrary, flexible structures (FR > 1) tend to present a rocking pattern
ciated.
in the opposite direction to the global shear distortion. The rocking
rotation coefficient is defined by Tsinidis (2017) as the rotated angle θ
4. Proposed solution for the racking profile by the bottom or top slab, and further normalized by the free field
distortion angle as the magnitude θ/γFF. A close-up of the tunnel region
We might also consider to analyze and compare the profile of the and corresponding displacement patterns resulting from FEM analyses
racking coefficient at different distances (close and far from the struc- is shown in (Fig. 9), similar to the ones by Tsinidis and Pitilakis
ture) along a section as shown in (Fig. 7b), that is, how much an (Pitilakis and Tsinidis, 2016; Tsinidis, 2017; Tsinidis et al., 2016;

91
C. Gordo-Monsó, et al. Tunnelling and Underground Space Technology 88 (2019) 87–97

Fig. 6. Comparison of closed-form solutions by Wang (1993), Penzien (2000), Anderson et al. (2008), and the ones proposed in this study, plotted along with the
finite element results developed in this study, and the ones provided by Wang (1993). For Poisson coefficients of (a) ν = 0.5, (b) ν = 0.4, (c) ν = 0.3, (d) ν = 0.2.

Fig. 7. (a) Displacement control points for an intermediate control perimeter j of width Lj and depth Hj, to compute intermediate racking coefficients. (b) Location of
the racking coefficient profile.

92
C. Gordo-Monsó, et al. Tunnelling and Underground Space Technology 88 (2019) 87–97

Fig. 8. Racking coefficient profiles computed with closed-form (solid lines), and FEM (dotted lines) for different aspect ratios λ, FR values, and Poisson coefficients.
Horizontal distances range from adjacent to the structure (XPERIM/bSTRU = 0) to a distance of 2 structure widths (XPERIM/bSTRU = 2).

Tsinidis and Pitilakis, 2018), where color shades represent vertical presented before, an explanation to this rocking behavior comes natu-
displacements normalized to the maximum in each figure (not among rally. Referring to (Fig. 10), and considering for instance the case of a
figures) as to highlight the vertical displacement pattern in each figure. very stiff structure (FR ≈ 0), the racking of the structure will be of a
When considering the displacement pattern of the block mechanism lesser value than the racking corresponding to the outer boundaries of a

Fig. 9. Displacement patterns resulting from FEM analyses as a function of the flexibility ratio FR and aspect ratio λ .

93
C. Gordo-Monsó, et al. Tunnelling and Underground Space Technology 88 (2019) 87–97

Fig. 10. Conceptual rocking patterns as a function of the structure stiffness.

given soil control perimeter. Therefore, when we consider the boundary been compared to a series of FEM analyses, where the rotation of the
conditions for the displacement field (i.e. the horizontal bottom bottom slab has been computed as in Eq. (23), where ΔVERT,RIGHT and
boundary of the outermost soil perimeter must remain horizontal be- ΔVERT,LEFT stand for the vertical displacements of the top-right and
cause of the contact with the bedrock), we must apply a rigid body bottom-left embedded structure corners, and b stands for the structure
rotation to the block ensemble to compatibilize prescribed displace- width.
ments. For the particular case depicted of FR ≈ 0, this yields a clockwise
ΔVERT , RIGHT − ΔVERT , LEFT
rigid body rotation of the embedded structure if the outer soil distortion θFEM =
b (23)
is rightwards.
If the structure is as stiff as the substituted soil (FR = 1), the racking The results of the proposed closed-form expression, and numerical
of the structure will be of the same magnitude as the outer soil peri- results of the FEM analyses show a reasonably good agreement as
meter, and when applying the rigid body rotation to comply with the shown in (Fig. 11).
boundary conditions, the structure slab will not show any rotation.
Finally, if the structure is softer than the surrounding soil (FR > 1), 6. Modification of the proposed solutions for consideration of the
given that the racking of the structure will be larger than the racking Poisson coefficient
corresponding to the outer boundaries of the soil block, the rigid body
rotation applied to comply with the boundary conditions will not be As explained previously, the underlying mechanical model of the
enough to cancel the rotation of, say, the horizontal slabs. Therefore, a proposed solutions, namely the block behavior assumption, is unable to
counterclockwise rotation will be apparent if the shearing is rightwards. consider the dependence on the Poisson coefficient. In order to cater for
Considering the successive perimeters conceptual model proposed this effect, a simple adjustment factor α can be introduced for the
in this paper, the cumulative horizontal boundary rotation of each inner multiplicative closed-form racking expression given by Eq. (18), for the
perimeter, relative to each adjacent outer perimeter would yield a total multiplicative closed-form racking profile expression given by Eq. (19),
structure rotation in a direction depending on the FR value. and for the summative closed-form rocking expression given by Eq.
Conceptually, we can think of this rocking behavior as a matter of (22). This modification factor should be able to adjust the closed-form
perspective, which depends on the boundary that we are choosing as a solution results to the finite element model results.
reference and on which direction respective to this the structure de- The chosen functional form for the modification factor α is shown in
forms. Quantitatively, considering the block model depicted in previous Eq. (24), being a logarithmic expression depending both on the Poisson
sections, and following the notation in (Tsinidis, 2017), we can think coefficient ν and the flexibility ratio FR. The corresponding coefficients
that the relative rotation of the horizontal boundaries within two computed by a best-fit approach are shown in Table 1.
consecutive perimeters i − 1 (inner) and i (outer), normalized to the i
(outer) shearing deformation is given by Eq. (21). 1, FR < 1
α=⎧
⎨
⎩ (α1· ν + α2 )· Ln (FR) + α3, FR ≥ 1 (24)
~ γV , i − 1 − γV , i
θi − 1, i =
γH , i + γV , i (21) After application of the modification factor α, the closed form ex-
pression for the racking coefficient becomes Eq. (25), and where the
~
Given that, when the surrounding soil at a far distance is strained in term Ri is given by Eq. (17).
pure shear a magnitude γFF, the i perimeter will deform its racking i=n
γSTRU ~
coefficient RPERIMi times the free field strain γFF, the total angle rotated R= ≅ α· ∏ Ri
by the structure slab can be computed by adding all the relative angles γFF i=1 (25)
rotated between consecutive perimeters (i − 1, and i) normalized to the
Similarly, after application of the modification factor α, the closed
shear distortion of the outer perimeter i, weighting each relative angle
form expression for the racking coefficient profile, at control perimeter
by the racking coefficient of that outer perimeter i, as in Eq. (22).
j, becomes Eq. (26). In the computation of the αj factor, the value of the
i=n n j=n γj − 1 ⎞ flexibility ratio FR should be that of the j perimeter being analyzed.
θ ~ ⎛ γV , i − 1 − γV , i
γFF
≅ ∑ θi −1,i·RPERIMi = ∑ ⎜ γ + γV , i
· ∏ γj ⎟ γPERIMj i=n
i=1 i=1 ⎝ H ,i j=i+1 (22) ~
⎠ RPERIMj =
γFF
≅ αj · ∏ Ri
i=j+1 (26)
To assess the accuracy of the proposed expression, the results have

94
C. Gordo-Monsó, et al. Tunnelling and Underground Space Technology 88 (2019) 87–97

Fig. 11. Comparison of closed-form solutions for the rocking coefficient for three aspect ratios λ = 1, 2, 3, and FEM results developed in this study, and those in
(Tsinidis and Pitilakis, 2018), where available. For Poisson coefficients of (a) ν = 0.5, (b) ν = 0.4, (c) ν = 0.3, (d) ν = 0.2.

Table 1 form expressions), and accuracy respective to the results obtained with
Coefficients for the modification factor α. the FEM analyses. A convergence ratio dependent on the number of
α1 α2 α3
control perimeters involved in the closed-form solution can be com-
puted as the ratio of the closed-form racking coefficient (evaluated by
−0.342 0.176 1.0 means of Eq. (25)) to the racking coefficient obtained with the FEM
analyses for the case of ν = 0.5 and λ = 1. Corresponding results for a
range of flexibility ratios FR are shown in (Fig. 12a), where it can be
Lastly, after the application of the modification factor α, the closed appreciated that for approximately 200 perimeters an accuracy of
form expression for the rocking coefficient becomes Eq. (27). As in the ± 10% is reached. A further increase in the number of perimeters
previous case, in the computation of the αi factor, the value of the provides stability to the predicted racking coefficient value.
flexibility ratio FR should be that of the i perimeter being evaluated. The accuracy of the proposed racking and rocking coefficients so-
n n j=n lutions, computed as the ratio of the values obtained with the closed-
θ ~ ⎛ γV , i − 1 − γV , i ⎞ ⎛ γj − 1 ⎞⎞
γFF
≅ ∑ (αi·θi −1,i)·(αi·RPERIMi) = ∑ ⎜⎛⎜αi· γH , i + γV , i ⎠ ⎜
⎟ · αi· ∏ γj ⎟⎟
forms to the FEM analyses, has been evaluated and shown in (Fig. 12b)
i=1 i=1
⎝⎝ ⎝ j=i+1 ⎠⎠ for a range of flexibility ratios FR, aspect ratios λ, and Poisson coeffi-
(27) cients ν. As it can be appreciated, the racking coefficient provides a
good accuracy in the order of ± 10%, while the rocking coefficient
shows a larger dispersion.
7. Performance of the proposed closed-form solutions

It is insightful to examine the performance of the closed-form so- 8. Summary and conclusions
lutions, both in terms of efficiency respective to the fineness of the
discretization (i.e. number of control perimeters involved in the closed- A new and intuitive method has been presented to analyze the shear

95
C. Gordo-Monsó, et al. Tunnelling and Underground Space Technology 88 (2019) 87–97

Fig. 12. (a) Convergence ratio for the rocking coefficient as a function of the number of control perimeters used in the discretization. (b) Error in the value of the
racking and rocking coefficients computed with the closed-form solutions respective to the value computed with the FEM analyses.

distortion of underground rectangular structures subjected to far field Undergr. Sp. 2, 88–105. https://doi.org/10.1016/j.undsp.2017.03.005.
shear strains. The method is applicable for the case of a homogeneous, Guoxing, C., Su, C., Xi, Z., Xiuli, D., Chengzhi, Q.I., Zhihua, W., 2015. Shaking-table tests
and numerical simulations on a subway structure in soft soil. Soil Dyn. Earthq. Eng.
isotropic, elastic medium, and elastic structure, buried at a sufficient 76, 13–28. https://doi.org/10.1016/j.soildyn.2014.12.012.
depth as to be able to assume that the medium is subjected to a pure Hashash, Y.M.A., Hook, J.J., Schmidt, B., I-Chiang Yao, J., 2001. Seismic design and
shear state. analysis of underground structures. Tunn. Undergr. Sp. Technol. 16, 247–293.
https://doi.org/10.1016/S0886-7798(01)00051-7.
The method is based on first principles of mechanics, namely on the Hashash, Y.M.A., Karina, K., Koutsoftas, D., O’Riordan, N., 2010. Seismic design con-
simple behavior of elastic blocks subjected to pure shear, and on the siderations for underground box structures. In: Proceedings of the 2010 Earth
discretization of the soil-structure ensemble in successive control peri- Retention Conference, pp. 620–637. https://doi.org/10.1061/41128(384)64.
Hendron, A.J., Fernandez, G., 1983. Dynamic and static design considerations for un-
meters to which recursively apply the simple concept of a shear block. derground chambers. In: Howard, T.R. (Ed.), Seismic Design of Embankments and
This method is able to accurately predict three magnitudes: the Caverns. ASCE, New York, pp. 157–197.
racking distortion of the structure, the rocking rotation of the structure, Hudson, J.A., Harrison, J.P., 2000. Engineering Rock Mechanics: An Introduction to the
Principles. Elsevier.
and the soil shear distortion profile near and far from the structure
Hung, C.J., Monsees, J.E., Munfah, N., Wisniewski, J., 2009. Technical Manual for Design
boundaries. Closed-form expressions are provided for these three and Construction of Road Tunnels – Civil Elements. Washington, D.C.
magnitudes. This method has been verified by comparing its predictions Huo, H., Bobet, A., Fernández, G., Ramírez, J., 2005. Load transfer mechanisms between
with those obtained with numerical analyses by finite element models, underground structure and surrounding ground: evaluation of the failure of the
Daikai station. J. Geotech. Geoenviron. Eng. 131, 1522–1534.
and the numerical results and closed form solutions of other re- Huo, H., Bobet, A., Fernández, G., Ramírez, J., 2006. Analytical solution for deep rec-
searchers. tangular structures subjected to far-field shear stresses. Tunn. Undergr. Sp. Technol.
The method can be easily implemented in a spreadsheet, but more 21, 613–625. https://doi.org/10.1016/j.tust.2005.12.135.
Iwatate, T., Kobayashi, Y., Kusu, H., Rin, K., 2000. Investigation and shaking table tests of
importantly, it provides a simple conceptual framework to understand subway structures of the Hyogoken-Nanbu earthquake. In: Proceedings 12th World
the mechanic behavior of a rectangular structure embedded in a sur- Conference on Earthquake Engineering. Auckland, New Zealand, pp. 1–8.
rounding medium subjected to a far field pure shear deformation. Kuesel, T.B., 1969. Earthquake design criteria for subways. J. Struct. Div. ASCE 95,
6616–6626.
LA Metro Authority, 2012. Metro Rail Design Criteria Section 5 Structural/Geotechnical.
Appendix A. Supplementary material Los Angeles, CA.
McKenna, F., Scott, M.H., Fenves, G.L., 2010. Nonlinear finite-element analysis software
architecture using object composition. J. Comput. Civ. Eng. 24, 95–107. https://doi.
Supplementary data to this article can be found online at https://
org/10.1061/(ASCE)CP.1943-5487.0000002.
doi.org/10.1016/j.tust.2019.03.002. Merrit, J.L., Monsees, J.E., Hendron, A.J., 1985. Seismic design of underground struc-
tures. In: In: Mann, C.D., Kelley, M.N. (Eds.), Proc. of the 1985 Rapid Excavation
Tunnelling Conference, Society of Mining, Metallurgical and Petroleum Engineers 1.
References
New York, pp. 104–131.
Monsees, J.E., 1991. Underground seismic design. In: Fall Lecture Series of the Boston
Abate, G., Massimino, M.R., 2017. Parametric analysis of the seismic response of coupled Society of Civil Engineers Section. Structural Group Lecture Series, Boston, MA, pp.
tunnel–soil–aboveground building systems by numerical modelling. Bull. Earthq. 1–41.
Eng. 15, 443–467. https://doi.org/10.1007/s10518-016-9975-7. Peck, R.B., 1969. Deep excavations and tunneling in soft ground. In: 7th International
Anderson, D.G., Martin, G.R., Lam, I., Wang, J.N., 2008. NCHRP 611 – Seismic Analysis Conference on Soil Mechanics and Foundation Engineering, pp. 225–290.
and design of Retaining Walls, Buried Structures, Slopes, and Embankments. Penzien, J., 2000. Seismically induced racking of tunnel linings. Earthq. Eng. Struct. Dyn.
Washington, DC. 29, 683–691. https://doi.org/10.1002/(SICI)1096-9845(200005)29:5<683::AID-
Bobet, A., 2010. Drained and undrained response of deep tunnels subjected to far-field EQE932>3.0.CO;2-1.
shear loading. Tunn. Undergr. Sp. Technol. 25, 21–31. https://doi.org/10.1016/j. Penzien, J., Wu, C.L., 1998. Stresses in linings of bored tunnels. Earthq. Eng. Struct. Dyn.
tust.2009.08.001. 27, 283–300. https://doi.org/10.1002/(SICI)1096-9845(199803)27:3<283::AID-
Bobet, A., Fernandez, G., Huo, H., Ramirez, J., 2008. A practical iterative procedure to EQE732>3.0.CO;2-T.
estimate seismic-induced deformations of shallow rectangular structures. Can. Pitilakis, K., Anastasiadis, A., Raptakis, D., Boussolas, N., Papageorgiou, E., 2007. Seismic
Geotech. J. 45, 923–938. https://doi.org/10.1139/T08-026. design loads for metropolitan subway tunnels: the case of the thessaloniki Metro. In:
Cilingir, U., Madabhushi, S.P.G., 2011. A model study on the effects of input motion on Proceedings of the 4th International Conference on Earthquake Geotechnical
the seismic behaviour of tunnels. Soil Dyn. Earthq. Eng. 31, 452–462. https://doi. Engineering. Thessaloniki, pp. 1–15.
org/10.1016/j.soildyn.2010.10.004. Pitilakis, K., Tsinidis, G., 2016. Recent advances on the seismic behaviour and design of
Debiasi, E., Gajo, A., Zonta, D., 2013. On the seismic response of shallow-buried rec- tunnels. In: Conference in Honour of Michele Maugeri. Catania, Italy, pp. 1–46.
tangular structures. Tunn. Undergr. Sp. Technol. 38, 99–113. Tsinidis, G., 2017. Response characteristics of rectangular tunnels in soft soil subjected to
Fabozzi, S., Licata, V., Autuori, S., Bilotta, E., Russo, G., Silvestri, F., 2017. Prediction of transversal ground shaking. Tunn. Undergr. Sp. Technol. 62, 1–22. https://doi.org/
the seismic behavior of an underground railway station and a tunnel in Napoli (Italy). 10.1016/j.tust.2016.11.003.

96
C. Gordo-Monsó, et al. Tunnelling and Underground Space Technology 88 (2019) 87–97

Tsinidis, G., Pitilakis, K., 2018. Improved R-F relations for the transversal seismic analysis International Conference on Natural Hazards & Infrastructure. Chania, Greece,
of rectangular tunnels. Soil Dyn. Earthq. Eng. 107, 48–65. https://doi.org/10.1016/j. pp. 1–9.
soildyn.2018.01.004. Ulgen, D., Saglam, S., Ozkan, M.Y., 2015. Dynamic response of a flexible rectangular
Tsinidis, G., Pitilakis, K., Anagnostopoulos, C., Madabhushi, G., 2015a. Seismic response underground structure in sand: centrifuge modeling. Bull. Earthq. Eng. 13,
and design of rectangular tunnels. In: Proceedings of the SECED 2015 Conference: 2547–2566. https://doi.org/10.1007/s10518-015-9736-z.
Earthquake Risk and Engineering Towards a Resilient World. World, Cambridge, UK, Wang, H.F., Lou, M.L., Chen, X., Zhai, Y.M., 2013. Structure-soil-structure interaction
pp. 1–10. between underground structure and ground structure. Soil Dyn. Earthq. Eng. 54,
Tsinidis, G., Pitilakis, K., Heron, C., 2015b. Dynamic response of flexible square tunnels: 31–38. https://doi.org/10.1016/j.soildyn.2013.07.015.
centrifuge testing and validation of existing design methodologies. Geotechnique 65, Wang, J.N., 1993. Seismic Design of Tunnels. A Simple State of the Art Approach. New
401–417. https://doi.org/10.1680/geot.SIP.15.P.004. York, NY.
Tsinidis, G., Pitilakis, K., Mantikas, G., 2016. Deformation patterns of rectangular tunnels WSDOT, 2010. Appendix B8 Seismic Design Criteria SR 99 Bored Tunnel Alternative
in soft soils subjected to ground shaking. In: Proceedings of the 1st ICONHIC Design-Build Project. Seattle, WA.

97