Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
k
xx
k
xy
k
xz
k
yx
k
yy
k
yz
k
zx
k
zy
k
zz
2
4
3
5
Thus, as long as the tensor is symmetric, eigenvalues and
eigenvectors can be calculated in order to nd the directions and
magnitudes of principal permeability:
K
k1
k2
k3
2
4
3
5
The magnitude of tensor components are strictly related to the
distributionandphysical properties of discontinuities of the rock mass.
Tensor K can be estimated from physical properties of the
discontinuities, but some assumptions must be made (Snow, 1968,
Kiraly, 1969):
The global permeability of the intact rock must be much less, almost
null, compared to the permeability of the discontinuities.
Discontinuities must be persistent in a representative element of
volume of the rock mass.
The permeability must be isotropic on the plane of discontinuities.
The mean velocity vector V
m
of the ow must have a linear variation,
withthe projectionof the gradient vector J
g
12m
d
N
i1
f
i
d d
3
i
d I
!
n
i
!
n
i
h i
1
where:
g gravity acceleration (9.81 m/s
2
)
v kinematic viscosity of water (3.20 10
6
m
2
/s)
N Total number of discontinuities sets
f average frequency of the i-set of discontinuities (m
1
)
d average aperture of the i-set of discontinuities (m)
I identity matrix
!
n n
1
; n
2
; n
3
dimensionless unitary vector normal to the average
plane of the discontinuity set
The term [I n
n
2
1
n
1
n
2
n
1
n
3
n
2
n
1
n
2
2
n
2
n
3
n
3
n
1
n
3
n
2
n
2
3
2
4
3
5
Tensor Kcan be referred to a Cartesian coordinate systemas well as
direction cosines or azimuthal (Dip/Dip-Direction) spherical coordi-
nate system. The parameter d used in Eq. (1) refers to ssures
characterized by smooth and parallel surfaces.
Taking into consideration a set of perfectly parallel and smooth
ssures in a non-permeable matrix, the permeability of the ssures
can be easily expressed by:
k
gd d
3
d f
12m
2
where
g gravity acceleration (9.81 m/s
2
)
v kinematic viscosity of water (3.20 10
6
m
2
/s)
d average aperture of planar ssures (m)
f average frequency of ssures (m
1
)
The average physical aperture E of ssures measured in the eld
cannot be considered as a representative value for the d parameter
due to the fact that natural discontinuities are characterized by wavy-
Fig. 5. K
m
variability with depth in MML.
Fig. 6. Graphical explanation of the effective hydraulic opening (from Barton, 2004a).
178 N. Coli et al. / Engineering Geology 101 (2008) 174184
rough surfaces which lead to a great lowering of the magnitude of k
(Barton, 2004a,b).
The effect of roughness can be taken into account by means of an
effective hydraulic opening (e) (Fig. 6) dened as follows (Barton,
2004a,b):
e
E
2
JRC
2:5
0
3
where:
E average physical aperture of ssures (mm)
JRC
0
7J
r
3 (where J
r
=Barton Q-System roughness index [Barton
et al., 1974]) in this study JRC
0
was assumed equal to Barton
JRC index (Barton and Choubey, 1977).
Fig. 7 shows the effect of varying JRC on the value of k using the
effective hydraulic opening in Eq. (2).
Since roughness has a big impact on permeability it must be taken
into account in the main K formulation.
Therefore Eq. (1) is replaced with:
K
g
12m
d
N
i1
f
i
d e
3
i
d I
!
n
i
!
n
i
h i
4
Where e (m) is the effective hydraulic opening.
Eq. (4) permits the calculation of K tensor for a given geostructural
survey station once discontinuity properties have been mapped.
Eq. (4) has been implemented in an automated script developed for
the software Mathematica (Wolfram). For each geostructural survey
station the script inputs are in this form:
Dip/Dip-Direction () of each set of discontinuities.
Average frequency f (m
1
) for each of the i-set of discontinuities.
Average effective hydraulic opening e (m) for each of the i-set of
discontinuities.
The script performs the calculation of tensor K and the determina-
tion of the principal K=[k1,k2,k3] directions by evaluating eigenvec-
tors and eigenvalues of the matrix. This process involves two
coordinate transformations.
Fig. 8 illustrates the script workow.
The model described above has been validated by means of simple
tests. Discontinuity properties assumed in tests are described in Table 1.
The rst test (Test A) was performed using a single discontinuity set
(Set1) inorder to compare k numerical values withthosegivenby Eq. (2).
In the second test (Test B) two sets of perpendicular discontinuities
were assumed with different physical properties in order to achieve
very different permeability values. It is clear that the maximum
hydraulic conductivity will be in the plane of Set1 which has a greater
effective hydraulic opening and the minimum hydraulic conductivity
will be in the plane of Set2. Therefore we expect that simulation will
produce compatible results.
Fig. 9. shows stereonets of the average Set1 and Set1 planes and
computed principal K max and min directions.
It can be seen that computed principal K directions perfectly agree
with expected results in both Tests A and B. Moreover k values given
by Test A agree with values given by Eq. (2) (Table 2).
Other similar tests were carried out and all gave expected results,
thus the model can be considered valid and functional.
Therefore for each of the 32 geostructural survey stations tensor
K=[k1,k2,k3] was calculated (Table 3) and plotted in stereographic
stereonets (Fig. 10).
Some stations are not included because they were discarded after a
geological consistency verication. These stations in fact refer to very
local geological structures or accidents and are not representative of
the dominant rock-mass structure.
6. Finite element 2D seepage analysis
2D nite element seepage analysis (Phase
2
, Rocscience Inc.) was
used in order to evaluate water ow into tunnels, once principal K
directions had been calculated for each geostructural survey station. The
model provides tunnel inowthrough unitary sections of rock mass (1 m).
Tunnels were rst subdividedintogeologicallyhomogeneous lengths,
as manyas neededfor anexhaustive coverage of the geological variability
of each tunnel. Then geological sections of each homogeneous length,
transverse to tunnel paths (-plane), were developed for a total number
of 38 sections.
Initial phreatic levels utilized in the nite-element models, referred
to the water table prole developed by SPEA in the Geotechnical
Longitudinal Prole of the project. Where more than one phreatic level
were present, only the highest one was taken into account in order to
simulate the worst inow scenario.
Geological sections (-plane) were then associated to the nearest
andmost representative gestructural stations andthus tothe relative K
tensors.
Then principal K values projected to the -plane were calculated.
In fact, Phase
2
( Rocscience Inc.) seepage module requires only two
components of hydraulic conductivity as inputs (k1,k2), mutually
Fig. 8. Exemplicative scheme of the script workow.
Table 1
Discontinuities properties assumed for validation tests
Set 1 Set 2
Dip-Direction/Dip () 190/89 100/89
f (m
1
) 3.30 4.00
e (m) 0.05 0.00000028
Fig. 7. Effect of JRC on the nal value of k for a perfectly parallel set of ssures with
f =3.30 and E=10 mm. k
E
is calculated using the physical aperture of ssures (E) while
k
e
is calculated using the effective hydraulic opening (e).
179 N. Coli et al. / Engineering Geology 101 (2008) 174184
perpendicular, lying on the 2D model plane (-plane) and allowed to
rotate by an angle from the +x axes of the model datum.
Transformation from the absolute principal K components to the
-plane principal K components involves these steps:
each of the three principal K components (k1,k2,k3) were projected
to the corresponding -plane (k1
, k2
,
k3
) through trigonometric
transformations.
Once k1
, k2
,
k3
) was discarded.
k1
, k2
, k2
values
have a mean value of (k1
, +k2
/k2
) =(k1
/k2
).
FE seepage simulations were then performed: geological sections
were reproduced into nite element software and specic materials
were assigned to different lithologies according to computed k1
, k2
vectors.
Phase
2
seepage module can only evaluate steady-state ow
conditions and it cannot directly simulate tunnel inow immediately
after excavation. Initial inow was then calculated by means of a
specic conguration of the model: two vertical xed piezometric
levels at the same level as the undisturbed ones were placed close to
tunnel boundaries. With this simulation the FE model reaches the
Fig. 9. Stereonets of Test A and Test B. Straight lines represent ciclographs of average
Set1 and Set2 planes, circles and triangles are the traces of the three principal k1, k2, k3
directions. Calculated k directions perfectly agree with expected results.
Table 2
Values of k calculated by the model (Test A) and using Eq. (2)
Simulation Eq. (2)
k(m/s) 2.30 10
2
1.50 10
2
Fig. 10. Example stereonet for station S11. Principal Kdirections are plotted with relative
magnitudes.
Table 3
Values of principal k directions calculated for all geostructural survey stations
Station k1 k2 k3
m/s DD/Dip m/s DD/Dip m/s DD/Dip
S3 1.00 10
4
35/13 1.00 10
4
300/21 1.00 10
6
333/64
S4 3.30 10
4
55/02 3.30 10
4
324/20 4.40 10
13
331/70
S5 1.20 10
6
331/47 8.40 10
7
134/71 3.80 10
7
57/17
S6 1.61 10
6
228/78 1.22 10
6
346/6 6.60 10
7
257/10
S7 1.78 10
5
77/07 1.17 10
5
339/50 1.80 10
8
353/38
S8 2.04 10
6
250/73 2.01 10
6
287/14 2.99 10
8
194/10
S11 5.30 10
4
261/8 5.30 10
4
169/10 1.00 10
9
29/77
S12 1.16 10
4
181/71 1.16 10
4
306/11 3.60 10
8
219/15
S13 5.84 10
8
207/90 5.79 10
8
24/31 6.15 10
10
291/5
S14 8.20 10
8
123/40 8.12 10
8
65/38 8.50 10
10
357/26
S15 1.16 10
4
350/15 1.16 10
4
252/27 2.38 10
8
287/58
S19 4.00 10
6
327/10 3.80 10
6
240/10 2.00 10
7
207/71
S23 1.74 10
7
109/75 1.47 10
7
83/13 2.83 10
8
355/6
S25 6.56 10
5
293/6 6.56 10
5
212/55 4.70 10
7
199/34
S26 1.16 10
4
352/15 1.15 10
4
254/28 4.61 10
7
287/58
S27 7.24 10
6
277/25 7.24 10
6
120/63 6.28 10
6
192/09
S28 2.04 10
6
250/73 2.01 10
6
287/14 2.99 10
8
194/10
S29 1.40 10
8
325/32 1.30 10
8
321/57 4.80 10
10
234/02
S30 5.10 10
4
2/78 5.00 10
4
335/09 8.13 10
8
245/05
S31 4.25 10
7
59/64 4.25 10
7
276/20 5.22 10
12
181/14
S32 3.76 10
7
4/35 3.75 10
7
74/26 1.34 10
7
317/43
Values below 10
9
must be considered as non-permeable.
180 N. Coli et al. / Engineering Geology 101 (2008) 174184
steady-state ow condition with a piezometric level over tunnel head
very close to the undisturbed one. The two xed vertical piezometric
levels were placed symmetrically from the boundary of the tunnel
section at a distance equal to the tunnel radius, considering that
during excavation the most relaxed part of rock mass, where water
ow is more relevant, is about one diameter all around tunnel
boundaries. Obviously during excavation the piezometric level above
the tunnel head will not be equal to the undisturbed phreatic level of
the water table, due to the lowering of the phreatic level caused by
tunnel overall drainage. Thus initial inow evaluated in the above
method is referable to maximumtheoretical inowat the beginning of
excavation.
Fig. 11. Finite elements seepage analysis performed on a single section. A) Initial inow immediately after tunnel excavation, B) Steady-state inow with impermeable invert.
181 N. Coli et al. / Engineering Geology 101 (2008) 174184
The following FE seepage simulations were performed:
Initial inow at excavation.
Long-term steady-state inow with impermeable invert: real long-
term steady-state inow condition.
Hypothetical long-term steady-state inow with fully-permeable
tunnel boundaries: this simulation was performed in order to have a
comparison with other standard analytical inow computational
methods.
Fig. 11 shows an example of a FE seepage analysis.
The unitary inow values obtained for each section were then
extended for the length of the tunnel with the same geological setting
of the section. The total inow value for each tunnel was obtained by
the sum of inow values calculated for each geologically homo-
geneous lengths of the tunnel path.
This calculation was performed taking into account water table
variability within every geologically homogeneous length. When a
phreatic level within a geologically homogeneous length suffered
important variability, its inuence in the water inow calculation was
taken into account by varying initial phreatic levels in the correspond-
ing section during FE analysis.
Total inow values for tunnels were estimated, for both the initial
and the long-term steady-state conditions.
7. Discussion
Every mathematical model obtains results whose reliability is
directly linked to input data reliability. For hydrogeological models the
most important parameter is the hydraulic conductivity K.
K values evaluated with the procedure presented in this paper are
more reliable than those derived from Lugeon tests, because they are
calculated from discontinuities properties and they have a dened
spatial orientation, therefore these data were used for the nite
element seepage analysis models.
The tunnel water inow values derived from the nite element
seepage models have been compared to those derived from other
standard analytical methods: The Goodman formula (Goodman et al.,
1965) and the Heuer abacus (Heuer, 1995). The Goodman formula is a
very simple mathematical relationship deriving from the Darcy's law:
Q
2d Kd L
ln
2L
r
0
where:
Q unitary ow at steady-state ow (for tunnel meter)
K permeability of homogeneous and isotropic rock mass
L water head above tunnel
r
0
tunnel radius
The value of Q refers to a theoretical inow in an idealized circular
tube intoa perfectly homogeneous andisotropic rock mass. This formula
is very simple and many authors warn about using it in fractured rock
masses because this kind of rocks are neither homogeneous nor
isotropic. In fact the formula usually provides higher Q-values than the
real ones, even by some greater order of magnitude.
Due to the poor reliability of the Goodman formula in fractured
rock masses the Heuer abacus was developed (Fig. 12). It is an
empirical relationship between water inow and equivalent K values
obtained from eld Lugeon tests. It derives from a long experience in
tunnelling even if in a geological framework different from the object
of the present study.
Fig. 12. The Heuer abacus.
182 N. Coli et al. / Engineering Geology 101 (2008) 174184
The values of tunnel inows obtained for six sections (-plane)
with FE analysis were compared in Table 4 with those calculated with
Goodman formula and Heuer abacus.
Of course these methods are very different in their formulation and
they shall not give the same values, anyway some interesting
considerations can be made.
Looking at Figs. 13 and 14 Q-values distributions have the same
trend for all of the three methods. This is an important result because,
excluding the relative magnitudes, it means that there is a global
coherence and therefore methods can be compared even if they are
methodologically very different.
Inparticular the two simple methods of GoodmanFormula andHeuer
abacus have exactly the same trend (Fig. 14) while the FE analysis, which
is much more complicated than the others involving a greater number of
variables, is only a bit different. This means that the workowdeveloped
toperformFE seepage simulations was methodologicallycorrect andthat
results are reliable and consistent with the overall behaviour given by
traditional and widely used inow evaluation methods.
As expected, values derived from the Goodman formula are always
higher than the others, eventen times higher, while values derived from
the Heuer abacus are of the same magnitude as those in the FE analysis.
Values computed with FE analysis can be therefore considered reliable
and not outscaled.
It must be considered that inows derived from F.E analysis are
probably a bit higher than the real inows that will occur in tunnels
because:
The FE model assumes a constant inow head in the upstream
boundary of the model. Even if the boundary was kept at the
maximum allowable distance according to the hydrogeological
framework, it is possible that the actual recharge of the acquifer
might not be enough to permanently provide the constant inow
estimated by the model.
In the choice of input data, the higher values were always assumed
in order to be as conservative as possible.
8. Conclusions
In this paper a new practical approach to the evaluation of rock-
mass permeability tensor and the prediction of tunnel inow was
presented.
For the realization of the newhighway fromFirenze to Barberino di
Mugello (Tuscany), twelve tunnels are planned to be excavated in the
shaly Sillano Formation (SIL) and in the calcareous Monte Morello
Formation (MML), our target was to predict the water ow into those
tunnels.
The determination of hydraulic conductivity is very difcult for
those kinds of geological formations, because the water ow into the
rock mass is controlled by the discontinuity network, therefore
permeability is high anisotropic and it changes with the variation of
discontinuity properties and the geological structure of the formation.
Measured K values resulting from Lugeon tests have shown to be
not fully representative of the mass permeability in fractured and
structurally complex rocks, because Lugeon K
m
values can be referred
only to a very local volume and they do not agree with the real
hydrogeological behaviour of rocks. Moreover K
m
values measured
with Lugeon tests are strongly affected by the lithology and the degree
of fracturing of rock masses where tests are performed. Shaly (i.e. SIL)
or highly fractured rock masses (i.e. fault zones in appenninic ysh
formations) cause the water owbetween borehole walls and packers,
with the consequence of an overestimation of the measured K
m
values.
To evaluate the hydraulic conductivity tensor K of the studied rock
masses, the approach proposed by Kiraly (1969, 1978) was used and
improved with the introduction of the effective hydraulic opening (e)
(Barton, 2004a,b). Discontinuities properties for K tensor calculation
were collected by means of a eld survey in 32 structural survey
stations located according to tunnel paths and geological framework.
Once structural data were collected, principal K directions were
calculated for each geostructural survey station.
Tunnels were then subdivided into geologically homogeneous
lengths, and geological sections of each homogeneous length,
transverse to tunnel paths (-plane) were developed for a total
number of 38 sections. Geological sections were associated to the
nearest and most representative gestructural stations and to the
relative K values, which were then projected to the -plane.
In order to evaluate water ow into tunnels, 2D nite element
seepage analysis (Phase
2
, Rocscience Inc.) were performed on the 38
Table 4
Inow values obtained by means of FE analysis, Goodman formula and Heuer abacus
Q (l/min/m)
FE analysis Goodman formula Heuer abacus
Section 1 0.23 1.29 0.11
Section 2 0.86 12.09 1.12
Section 3 36.00 307.92 9.04
Section 4 1.60 2.86 0.55
Section 5 0.18 0.86 0.14
Section 6 0.42 1.35 0.17
Fig. 13. Distribution of Q-values from Table 4. The left side small graph is a enlargement
for a better view of FE analysis and Heuer abacus trends.
Fig. 14. Distribution of Q-values from Table 4, logarithmic y-axis scale. Trends of the
three different methods are almost the same.
183 N. Coli et al. / Engineering Geology 101 (2008) 174184
geological sections (-plane), using the K values calculated for each
-plane.
Three kinds of simulation were carried out: initial inow
immediately after the excavation, long-term steady-state inow
with impermeable invert, long-term steady-state inow with fully-
permeable tunnel boundaries.
Inow values given by the FE model were then compared and
validated with inow values given by two classical inow prediction
methods: the Goodman formula (Goodman et al., 1965) and the Heuer
abacus (Heuer, 1995).
In conclusion, the workowpresented in this paper resulted to be a
valid approach in the determination of hydraulic conductivity values
in fractured rock masses and in evaluating water ow into tunnels,
because it takes into account the geological variability of the rock
mass, the properties of the discontinuities and the hydrogeological
context.
Acknowledgements
The study presented in this paper was supported by SPEA S.p.A..
The authors are thankful to Prof. F. Rosso for the support in the
mathematical aspects of the study, especially for the development of
the automated script of the software Mathematica ( Wolfram).
References
Abbate, E., 1992. Introduzione all'escursine pre-congresso (Introduction to the pre-
conference eld-trip). Guida all'attraversata dell'Appennino Settentrionale. Soc.
Geol. It., pp. 114.
Barton, N., 2004a. The theory behind high pressure groutingPart 1. Tunnels &
Tunnelling International, September, pp. 2830.
Barton, N., 2004b. The theory behind high pressure groutingPart 2. Tunnels &
Tunnelling International, October, pp. 3335.
Barton, N., Choubey, L., 1977. The shear strength of rock joints in theory and practice.
Rock Mech. 10, 154.
Barton, N., Lien, R., Linde, J., 1974. Engineering classication of rock masses for the
design of tunnel support. Rock Mech. 6 (4), 189239.
Boccaletti, M., Coli, M., Decandia, F.A., Giannini, E., Lazzaretto, A., 1980. Evoluzione
dell'Appennino Settentrionale secondo un nuovo modello strutturale (The evolu-
tion of the Northern Apennine orogenic belt according to a new structural model).
Mem. Soc. Geol. It. 21, 359373.
Bortolotti, V., 1962. Domini paleogeograci precoci (Earlypalaeogeographic domains). Guide
Geologiche RegionaliVolume 4: Appennino Tosco-Emiliano. Soc. Geol. It., pp. 1618.
Briganti, R., Ciufegni, S., Coli, M., Polimeni, S., Pranzini, G., 2003. Il sottosuolo di Firenze:
evoluzione geologica Plio-Quaternaria dell'area orentina (Florence underground:
Plio-Quaternary geological evolution of the Florentine area). Boll. Soc. Geol. It. 112 (3),
435445.
Cicali, F., Pranzini, G., 1987. Il rapporto portata delle sorgenti-deusso di base dei
torrenti nel bilancio idrogeologico di un rilievo carbonatico (Monti della Calvana
Firenze) (Hydrogeological studies of a carbonatic relief: the Monti della Calvana,
Florence). Geol. Appl. Idrogeol. XXI, 155172 1986.
Coli, M., Fazzuoli, M., 1983. Assetto strutturale della Formazione di Monte Morello nei
dintorni di Firenze (Structural setting of the Monte Morello Formation in the
environs of Florence). Mem. Soc. Geol. It. 26, 543551.
Goodman, R., Moye, D., Shalkwyk, A., Javandel, I., 1965. Groundwater inow during
tunnel driving. Eng. Geol. 2.
Heuer, R., 1995. Approche quantitative, thorique et empirique, del Ronald E.Heuer, sur le
venues d'eu en tunnel (A quantitative, empirical and theoretical approach on water
ow into tunnels). Rapid Excavation and Tunneling Conference, San Francisco, Cal.,
June 1821, 1995.
Hudson, J.A., Harrison, J.P., 1997. Engineering rock mechanics. Pergamon Press. 440 pp.
La Pointe, P.R., Hudson, J.A., 1985. Characterization and Interpretation of Rock Mass Joint
Patterns. Geol. Soc. of America, Special Paper, vol. 199. 37 pp.
Kiraly, L., 1969. Anisotropie et htrognit de la permabilit dans les calcaires
ssurs (Anisotropy and heterogeneity of permeability in fractured limestones).
Eclogae Geol. Helv. 62/2, 613619.
Kiraly, L., 1978. La notion d'unit hydrogologique. Essai de denition (Denition of the
hydrogeological unit). Bull. Cent. Hydrogol. 2, 83216.
Kiraly, L., 2002. Karstication and groundwater ow. In: Gabrovek, F. (Ed.), Evolution of
Karst: From Prekarst to Cessation. Zalozba ZRC, Postojna-Ljubljana, pp. 155190.
Pranzini, G., 2002. Le venute d'acqua nelle gallerie appenniniche della linea ferroviaria
ad alta velocit (Water ow into the high-speed-railway tunnels of the Northern
Apennines). Acque Sotterranee 6 (80), 912 December.
Sestini, G., Curcio, M., 1965. Aspetti quantitativi delle impronte di fondo da corrente
nelle torbiditi dell'Appennino Tosco-Emiliano. (Quantitative features of bedding
sole structures in turbiditic formations of the Northern Apennines). Boll. Soc. Geol.
It. 84 (3), 143168.
Snow, D.T., 1968. Rock fracture spacings, openings and porosities. J. Soil Mech. Found.
Div. 94 na. SM1, January 1968.
184 N. Coli et al. / Engineering Geology 101 (2008) 174184