Rock Mechanics Rock Mechanics

and Rock Engineering 16, 173--180 (1983) and Rock Engineering

9 by Springer-Verlag 1983

Back Analysis of Measured Displacements of Tunnels

By
S. Sakurai and K. Takeuchi
Kobe University, Kobe, Japan

Summary

The present paper deals with a method of back analysis to be utilized for the
interpretation of field measurements in monitoring the stability of tunnels. The
method belongs to an inverse approach based on the finite element formulation,
assuming the ground media in which tunnels are excavated to be linear, isotropic
and elastic. Assuming Poisson's ratio and the vertical initial stress, the method de-
rives the complete initial state of stress and Young's modulus from a set of relative
displacements measured between adjacent measuring points. In order to verify the
applicability of the proposed method of back analysis to practical engineering
problems, a case study is presented.

1. Introduction

The knowledge of the initial state of stress in the ground and of the
material properties is of fundamental importance both for understanding the
structural behaviour and modelling the reality numerically. The initial stres-
ses, as well as the material constants, may be determined by various types
of in-situ tests. The stress relief method by overcoring is the most popular
for determining the initial stresses. Plate-bearing tests and direct shear tests
are extensively used, particularly in dam construction sites, for determining
the deformability and strength of rock masses. It is well known, however,
that the results of these tests in general indicate a certain amount of scatter
because of non-homogeneous characteristics of the ground, though the tests
are conducted at nearly the same geological formations. Besides this, the
mechanical properties of the ground must be evaluated in relation to the
size and shape of the structures. This means that the values obtained from
the in-situ tests cannot be used directly as input data for an analysis, and
so great care should be taken when interpreting the results of the tests.
In order to overcome these difficulties, field observation of the defor-
mational behaviour of tunnels during construction seems to be most pro-
mising. Observations by means of displacement measurements such as tunnel
convergence and borehole extensometer measurements are desirable because
i3 Rock Mechanics, Vol. 16/3

0723-2632/83/0016/0173/$ 01.60
174 S. Sakurai and K.Takeuchi:

of the simplicity of instrumentation and reliability of the data obtained,

while direct stress measurements are always less reliable. Such displacement
measurements may then be used to back-calculate the initial stresses and
the material properties.
Back analysis problems may be solved in two different ways defined
as inverse and direct approaches (Cividini et at., 1981). In the inverse
approach the formulation is just the reverse of that in the ordinary stress
analysis, although the governing equations are identical. However, the num-
ber of the measured values should be greater than the number of unknown
parameters, so that optimization techniques can be used to determine the
unknowns. For simple geometrical configurations and material models,
closed form solutions may be found in the literature. For structures with
arbitrary shape under more complex geological conditions the Finite Ele-
ment Method (FEM) seems to be most promising. For example K a v a n a g h
(1973) proposed a back analysis formulation based on the FEM which may
make it possible to get the material constants, not only for isotropic ma-
terials, but also for inhomogeneous and anisotropic materials from both
measured displacements and strains. However, it is likely that if the input
data of displacements or strains indicate a certain scatter, Kavanagh's meth-
od never converges. Therefore it is questionable if the method may be ap-
plied to geotechnical engineering problems in which the measured values
always contain a scattering. The direct approach to back analysis is based
on an iterative procedure correcting the trial values of unknown parameters
by minimizing error functions. The advantage is that this method may be
applied to non-linear back analysis problems without having to rely on a
complex mathematical background ( G i o d a and Maier, 1980) and, in addi-
tion, that standard algorithms of mathematical programming may be adopted
for the numerical solution (Cividini et al., 1981). On the other hand, the
iterative solution requires quite time-consuming computations.
In this paper an inverse approach to back analysis is proposed, which
is based on a finite element formulation.

2. Objectives and Assumptions

For analysing the stability of tunnels, S a k u r a i (1981, 1982) proposed
the Direct Strain Evaluation Technique, which allows a quantitative inter-
pretation of the results of displacement measurements. The technique is
based on the concept of strain rather than stress. Since strain is kine-
matically related to displacement, strain distributions in the ground can be
obtained directly from the measured displacements without performing any
stress analyses. The stability of tunnels is then assessed by comparing the
derived strains with critical failure strains of the soil or rock. The main
objective of the back analysis technique outlined here is to determine the
initial stresses and material constants from measured displacements. This
data may then be used as input data to compute the strain distribution in
the vicinity of a tunnel.
 Back Analysis of Measured Displacements of Tunnels 175

The following assumptions are made:

- - The mechanical behaviour of the ground is idealized by an isotropic
linear-elastic model, so that the material constants reduce to Young's
modulus and Poisson's ratio. Since Poisson's ratio has less influence
on the results of an analysis, an appropriate value will be assumed.
- - The elastic constants of the lining are assumed to be known.
- - The initial state of stress is assumed to be constant all over the region
under consideration.

3. F o r m u l a t i o n of B a c k A n a l y s i s

Genuine rock pressure may be taken into account in an FE-computa-

tion by applying equivalent nodal forces {p0} at the excavation surface,
which correspond to the initial state of stress at this place. These forces
may be determined by
{po} = S [B] T {~o} dV (1)
V

where [B]: matrix depending on geometry only,

initial stresses,
V: excavation volume.
For the sake of simplicity, in this study a two-dimensional formulation
is presented, whereby one principal axis (z-axis) is orientated parallel to
the tunnel axis. With the components of the initial stress being defined in a
Cartesian coordinate system x, y, z by
= (ozo

the vector {po} may be expressed according to

(po) =o-zO(B 1} +ayO {B 2} +rxyO {B ~) (2)
where the vectors
{B~} = {B1~ B2~)~' (i--1, 2, 3)
are obtained from (1) by integration over the excavation volume, i. e. (Bq
depend on geometry only. The relation between the nodal forces {P} of the
system and the nodal displacements {u} may be expressed by the well known
relationship
{P} = [K] {u} (3)
where [K] denotes the stiffness matrix of the assembled finite element system.
For a lined tunnel with the Young's moduli E It and EL of the rock and the
lining respectively the stiffness matrix may be split into

[K]=E ~ ([K ~] + - /~L

~ [KL]) = E~ [K*] (4)

13"
176 S. Sakurai and K.Takeuchi:

where [Ke] represents the stiffness matrix for the ground for E R = I and
[KL] that for the lining for E L= 1. If the FE-mesh is chosen in such a way
that the measuring points coincide with the nodes of the mesh, the displace-
ments may be grouped in a vector {u m} which contains measured points
only and a vector {u u} which represents the unknown displacements at
the rest of the nodal points
{~} = {u~) + {u%
If this relation together with the stiffness matrix [K*] expressed by (4) is
substituted in (3) for the load vector {P} = {p0} according to (2) one obtains

B~a j Ke2""J
Solving for {u u} and substitution leads to
ax ~ {B.} + av ~ {By} +zzy ~ {Bxy} = E R [KN*] {u m}
with
{gx} = {B11} - [K12;'] [K22"] -1 {B21},
{By}= {B12}-[K12;:][ K 2 2 * ] - 1 { B 2 2 } ,
{Bxy} = {B, a} - [K,2':] [K22"]-1 {B23},
[KN*]= [ K l l ' : ] - [K12':] [ K 2 2 ' : ] - 1 [K21':'].

This equation may be rewritten as

{.~} = ~ 1 [A] {o0} = [A] {8~ (5)

where the matrix [A]
[A] = [ [KN:']-I {Bx} [KN::']-I (By} [KN*]-I {Bxy} ]
depends on geometry and the Poisson's ratios for both the ground and the
lining. The ratio {80) of the initial stress to Young's modulus of the rock
is called normalized initial stress. Thus (5) contains three unknown values

72

~\~_0 .....
] Ul

--•
Fig. 1. Displacementcomponents at two adjacent measuringpoints

of normalized initial stress and consists of as many equations as measured

displacement components. If the number of measured displacements is greater
than three, the normalized initial stresses can be determined with an opti-
mization procedure.
 Back Analysis of Measured Displacements of Tunnels 177

It should be noted, however, that in practice it is easier to measure

relative displacements between two different measuring points than absolute
displacements, so that {u m} in Eq. (5) should be transformed to relative
displacements. The relative displacements between adjacent measuring points
shown in Fig. 1 can be expressed in terms of absolute displacements as
follows:

{Z~U}= V2' Vl'f = sinO --cosO - - s i n O c o s O ] [;ssj

Hence the absolute displacements {u m} are related to the measured rela-

tive displacements by
{A u m} = [T] {u'~}. (7)

The transformation matrix [T] is obtained with the consideration of (6).

Substituting (5) into (7) yields
{a u ~} = [n*] {~0} (8)
with
[A*] = IT] [A].
If the least squares method is adopted in (8) the normalized initial stresses
can be determined uniquely from the measured relative displacements by
{8~ = [ [a*] T [a*] ] -1 [a*] T {A u~}. (9)

4. Calculation Procedure

4.1 Unlined Tunnels

Since the ratio EL/E ~ is zero for unlined tunnels, the matrix [A*] in (9)
is defined uniquely for a given Poisson's ratio ~i~ of the rock. Assuming vR
and substitution of {A u m} by a set of measured relative displacements leads
therefore immediately to the normalized initial stress. If the value of the
initial stress itself is required, a relation between this stress and Young's
modulus E R would be necessary. Because such a relation does not exist
one component of the initial stress or E R must be determined in some
other way. We propose to assume the vertical component (ry0 proportional
to the overburden H and the specific weight y e of the ground
~yo = 7R H. (10)
Hence E n is determined by
ER - )'R H
~yo (tl)
and the other two components of the initial stress by
Gx 0 = ~R ~x 0,
v x y o = E R Txy o.
178 S. Sakurai and K.Takeuchi:

It should be noted, however, that if only the evaluation of the strain distri-
bution is required, there is no need to determine the initial stresses and
Young's modulus separately. Instead, only the ratio of these quantities is
important.
4.2 L i n e d T u n n e l s

In the case of lined tunnels, the matrix [A*] is not known a priori even
when the Poisson's ratio ~r of the lining and E L are assumed, because it de-
pends on the ratio EL/E R. Therefore an iterative solution must be adopted. For
this an adequate value of ER is assumed, the procedure described for unlined
tunnels is carried out and the resulting value for E R defined by (11) is
compared with the assumed one. If the two values differ more than a pre-
scribed limit, this procedure is repeated with an adjusted value of E R until
satisfying the limit.

5. Illustrative Example

An example from engineering practice demonstrates the applicability of

the proposed method for practical problems. For a tunnel with an over-
burden of approximately 400 m convergence and extensometer measure-
ments were started right after excavation. The arrangement of the measure-
ments together with the readings at the final stage of construction are
represented in Fig. 2.
Disregarding the effects of the lining the back analysis was carried out
as outlined in ch. 4.1. Assumption of ~R=0.3 led immediately to the nor-
malized initial stresses

{~o}= 8yo = 0.0609 9

~xy ~ 0.0003

Using in a second analysis the data of the convergence measurements only,

the results were
0.0490 /
= 0.0587

- 0.0001

Comparing these values it may be seen that the difference between the
two sets is very small. This means that for this case the convergence mea-
surements alone would be sufficient for a reliable back analysis.
The normalized initial stresses obtained may now be used to determine
the initial stresses. Assumption of the vertical component

ay ~ = 0.0245 9400 = 9.80 MPa

according to (10) leads to

E R = 160.9 MPa,
ax ~ = 7.87 MPa,
~zy ~ = 0,04 MPa.
 Back Analysis of Measured Displacements of Tunnels 179

The displacement and strain distributions in the vicinity of the tunnel are
now computed by means of an ordinary finite element analysis utilizing as
input data the initial stresses and Young's modulus obtained here. The
corresponding results are illustrated in Fig. 2. It may be seen that there is a
fairly good agreement between the computed and measured displacements.

e-----~ Measurement /

- - Back-Analysts /

~SJ" /t-- 4__

%
So

CZ) 0

SCALE

0 I 2 3 rn
cra

50

~" z.O [-:.:i

c)

r 3O

Measuremer~t

1
~::::!
2O I I Back-Analysis
;:.:;
o
o 10 IN
r.,.1

9 @ @ 9 @ @ | @
Measurement Line Number b
Fig. 2. Comparison between the measured and back-analyzed displacements
a) Extensometer measurements; b) Convergence measurements
180 S. Sakurai and K.Takeuchi: Measured Displacements of Tunnels

6. C o n c l u s i o n s

In this paper a back analysis procedure based on the FEM has been
proposed for determining the initial stresses {o0} and Young's modulus ER
of the rock mass from displacement measurements. The main conclusions
may be summarized as follows:
1
- - If only the normalized stresses _~_{~0) are required, assumption of the
Poisson's ratio vR leads to the desired results. If E R and {(~0} should be
determined separately, one component of {~r0} or E e must be assumed.
- - In the proposed method relative displacements between two measuring
points can be taken into account. This is an advantage because in
practice measuring relative displacements is much easier than absolute
ones.
- - In the case of lined tunnels a small number of iterations is necessary
for the calculation of the initial stresses, while in unlined tunnels a
single calculation is good enough.
- - Compared with similar methods the proposed procedure is numerically
very stable. Thus it may be applied also for measured data with a
large scatter.

Acknowledgements

The authors are much obliged to the Nihon Doro Kodan (Japan High-
way Public Corporation) for the permission to publish some of the field
measurement data. The authors wish to express their gratitude to Dr.
P. Fritz, Federal Institute of Technology, Zurich, for his useful suggestions
on the preparation of the manuscript. Thanks are also due to Mr. S. M o r i ,
Graduate Student of Kobe University, for his considerable assistance with
the case study computation.

References
Cividini, A., Jurina, L., Gioda, G. (1981): Some Aspects of Characteriza-
tion Problems in Geomechanics. Int. J. Rock Mech. Min. Sci. & Geomech. Abstr.
18, 487--503.
Gioda, G., Maier, G. (1980): Direct Search Solution of an Inverse Problem
in Elastoplasticity: Identification of Cohesion, Friction Angle and In-situ Stress by
Pressure Tunnel Tests. Int. J. Num. Meth. Engng. 15, 1823--1848.
Kavanagh, K. (1973): Experiment Versus Analysis: Computational Techni-
ques for the Description of Static Material Response. Int. J. Num. Meth. Engng. 5,
503--515.
Sakurai, S. (1981): Direct Strain Evaluation Technique in Construction of
Underground Opening. Proc. 22nd U. S. Rock Mech. Syrup., MIT, 278--282.
Sakurai, S. (1982): Monitoring of Caverns During Construction Period.
Symp. Caverns and Pressure Shafts, ISRM, Aachen, 433--441.