Twin tunnels excavated in mixed face conditions

M.A.A.P. Silva & F.L. Gonçalves

Andrade Gutierrez Engenharia S.A., São Paulo, Brazil
A.A. Ferreira & H.C. Rocha
Cia. do Metropolitano de São Paulo Metrô-SP, São Paulo, Brazil.

ABSTRACT: This paper presents a final analysis of the excavation of two parallel tunnels for the
Metro of São Paulo Line 5 Expansion, which were excavated by two 6.9m EPB TBM. This current
paper focuses on the stretch where mixed face conditions in Precambrian materials were exca-
vated, which had high influence on the TBMs performance, such as difficulties in keeping a stable
face pressure, ground losses (high settlements), sinkholes, slow advances, clogging, and very fre-
quent and long hyperbaric interventions. The purpose of this document is to present a technical
explanation for the events occurred in the excavations in this stretch, and for the impact in terms
of production.

1 INTRODUCTION

This paper presents an analysis of the excavation of two parallel tunnels for the Metro of São
Paulo Line 5 Expansion, which were excavated by two 6.9m EPB TBMs. Table 1 shows the main
characteristics of them.
Although these tunnels were already object of study by Silva et al. (2015, 2016) and by Co-
mulada et al. (2016), this current paper focuses on the stretch where mixed face conditions in
Precambrian materials were excavated, in order to evaluate the relationship between TBM oper-
ational factors and these materials, as well as its consequences. The authors above give more
information about TBM performance in both soft sandy soils and stiff clays. Figure 1 shows a
schematic tunnel alignment, being also highlighted the stretch where mixed face conditions were
observed. shows the cutter head (CH) design and its main cutter tools used.

Figure 1. (a) Twin tunnels alignment and mixed face conditions stretch. (b) Cutter head design and main
cutter tools used.
 In this stretch, mixed face conditions could be the main reason for the impacts on the TBMs
performance, such as difficulties in keeping a stable face pressure, ground losses (high settle-
ments), sinkholes, slow advances, clogging, and very frequent and long hyperbaric interventions.
The total length of this stretch was close to 450m, which were excavated in approximately 3
months (average of 4 rings per day).

Table 1. TBM characteristics.

Shield Characteristics
Type EPB
External diameter 6.89 m
Total CH opening 39%
Nominal Torque 3560 kNm
Exceptional Torque 4984 kNm
Breakout Torque 5340 kNm
Number of thrust cylinders 16x2
Total installed thrust force 60,801 kN
Maximum thrust force 50,000 kN
Cutters Rippers, discs, scrapers (#41)
Minimum curve radius 250 m
Ring thickness 30 cm
Ring length 1500 mm
Ring segmentation 5+1

2 GEOLOGICAL AND GEOTECHNICAL CONDITIONS

As described above, this stretch is strongly characterized by mixed face conditions in weathered
gneiss and saprolite, and saprolite and residual soil. Some characteristics of these materials are
presented below.
The residual soils are composed of sandy silt and clayey silt, with SPT values ranging from 2
to 30 blows. In these materials, water level ranged between 10 to 25 m above tunnel crown. Re-
sidual soils also contain some amount of cobbles and geological structures inherited from parent
rock. It is highlighted that these materials can have high permeability to water (between 10 -4 and
10-6 cm/s, according to Futai et al., 2012). Furthermore, in regard to air, it is also accepted that
this permeability can be 10 to 100 times greater, which can make these materials very permeable
and unstable when operating in transition mode (compressed air on top) and in hyperbaric condi-
tions. Finally, Futai et al. (2012) have shown that these materials are very heterogeneous, which
explains the difficulty in adjusting the foam parameters, as well as the high efforts necessary to
have a good conditioned muck and to homogenize the material in the excavation chamber.
The saprolite term used here corresponds to the weathered rock impenetrable to the SPT sam-
pler. Usually, saprolite has SPT values higher than 30 blows. The geomechanical behavior of this
material is hard to be modeled, because it is highly heterogeneous and its behavior can be gov-
erned by the joints and discontinuities of the original rock. It is known from previous conventional
excavations that saprolite does not have good response when excavated (very low geomechanical
parameters when relieved) and it is associated to high ground losses.
As described by Oliveira et al. (2017), this saprolite (and some residual soils) has usually stiff
to hard consistency and, thus, requires high amount of liquid addition (very liquid foam or water)
in order to give it some plasticity. However, according to Hollman & Thewes (2012) clogging
tendency methodology (Figure 2), adding liquids will take this material to the strong clogging
tendency field and, thus, it will increase the probability of needing a hyperbaric intervention to
clean the cutterhead.

Figure 2. (a) Residual soil and saprolite clogging tendency (adapted from Oliveira et al., 2017). (b) Abra-
sivity Cerchar Index (Monteiro e Rocha, 2015b).

The biotite gneisses are highly foliated and present steeply dipping NE-SW trending orienta-
tion. Uniaxial compressive tests carried out in weathered gneiss showed that strength can easily
vary from 5 to 60 MPa depending on the content of biotite bands. Figure 3 shows the geological
profile from the stretch analyzed.
Recent studies presented by Monteiro & Rocha (2015) have revealed that the biotite gneisses
also exhibit high CERCHAR abrasivity index (Figure 2b).

Figure 3. Geological profile (Monteiro & Rocha, 2015a).

Because of the high heterogeneity and the presence of different materials in the excavation face
in a transitional ground, the specific weight of the excavated material can constantly change. It
usually fluctuated between 2.2 and 2.5 t/m³, which made the control of the excavation volume
with the scales quite difficult.
 3 THE IMPACT ON THE SHIELD PERFORMANCE
3.1 Conditioning
In the transitional zone, it often occurred three different materials in the excavation face – rock,
saprolite and residual soil. Besides, these materials are very heterogeneous and, thus, very diffi-
cult to be conditioned into a homogeneous paste.
It must be remembered that water, air and foam agents (tensides and polymers) are added in
the cutterhead front to condition the excavated material, in order to give to the mixture formed in
the working chamber the workability required, without loosing its capabilities to counterbalance
earth and water pressures. It must be also highlighted that air bubbles formed in these mixtures
have an important function, once they are the ones to provide the workability mentioned. And
thus, its structures must be stable enough to keep the air trapped.
The relationship between liquid (water plus tenside) and foam (water plus tenside plus air)
injection is given by the foam expansion rate (FER). For cohesive materials, wet foam is required
(FER < 6), while cohesionless materials require dryer and more viscous foam (FER > 8). Finally,
it is worth to be remembered that the foam injected in the cutterhead prevents clogging and, be-
cause of that, it is very important to have a straight and continuous monitoring during the exca-
vation, being constantly alert to any change in the geological conditions. In transitional ground,
that need is even more important because of the everchanging conditions at the tunnel face.
In a specific stretch of the Line 5 Precambrian mixed face conditions, residual soil with cohe-
sionless material occurred in the tunnel crown while there were weathered rock or a harder mate-
rial in the bottom. The cohesionless residual soil disaggregated easily in the presence of water, so
dryer foam was necessary to assure the stability of that layer. On the other hand, the harder ma-
terial demanded a greater water volume to control the temperature during mining. That mixed
face condition affected the balance between water and air injections in the excavation chamber.
Consequently, the cutterhead openings and the foam injection points were sometimes blocked,
as well as high temperatures in the excavation chamber often occurred, which led to the use of
“free” water injections in the working chamber (WI), with the aim to control temperature and give
some plasticity to the muck.
Sometimes, extremely wet foam (FER < 4) was used in the bulkhead. However, the total flow
injected through this line was not enough to give the plasticity needed for the muck and, thus,
more “free” water injection in the chamber was constantly necessary.
Other attempts consisted in increasing the foam injection rate (FIR) to try to achieve higher
workability, maintaining the dry foam to stabilize the cohesionless soil layer. In those cases, the
air accumulation in the top of the excavation chamber packed the material in the bottom, leading
again to the need of “free” water injection. However, the excess of “free” water injected in the
chamber damaged the foam structure, releasing the air bubbles to the top of the working chamber.
Although the machines were not designed as hybrid machines, some attempts to inject benton-
ite though the foam lines during the advances were made. The bentonite showed some results in
cooling the material and reducing the segregation. However, since the machine was not designed
to work with high injection rates of bentonite, its available flow was not sufficient to allow a
continuous advance. The injections were possible for only one or two rings.
However, the bentonite injection let at the foam lines is more susceptible to clogging. Thus, it
was not possible to evaluate if the addition of bentonite would modify the chances of clogging.
In summary, rock and saprolite were not well conditioned. Those materials need water for
cooling, while soft residual soils (SPT < 6 blows) need a dry and stable foam for a good muck
conditioning and to keep the face stable. Regardless of the change in the foam parameters, with
such different materials in the front, segregation always occurred in the chamber.
A higher cutterhead rotation speed could theoretically help to mix material in front and foam;
however, in mixed face high rpm would increase the risk of cutting tools damage, as the impact
of discs to the rock face at higher speeds increases, resulting in flat discs and thus creating the
need for hyperbaric interventions. Besides, higher rpm also increases temperature in the chamber,
which leads to stoppages, increase in the clogging possibility and also hyperbaric interventions.
Because of that, the machines could not work in EPB mode, but only in Transition Mode with
compressed air on the top (Maidl et al., 2012), which enhanced the probability of an overexcava-
tion, even maintaining a high face pressure (EP1 higher than 2.5 bars with the water pressure
below 2.0 bars). Controlling the air bubble in the chamber was the key to avoid any possible face
instability. Nevertheless, because of the high permeability in the transition zones, it was not rare
having air escaping to the ground and face pressure fluctuations, leading to soil instabilities. In
total, three sinkholes were associated with the operation in the mixed face stretch.
Since the geological conditions were always varying, foam parameters were also changing all
the time, showing the difficulties in forming and keeping a homogeneous muck in the chamber.
Figure 4(a) shows the relationship between average FIR, water injection rate (WIR) and velocity.
It is clear the use of more foam and water in the stretch where lower velocities were achieved. In
the same way, it is also clear that the faster the machine advanced, the lower FIR and WI were.
The FIR values ranged between 100 and 130%, while the FER was between 5 and 8.

Figure 4. (a) Foam and water injection data in mixed face conditions. (b) Advancing TBM parameters.

3.2 Digging
In Precambrian mixed face conditions, Shield advances are usually limited by the cutterhead ro-
tation speed and the penetration rate, that limit the maximal forces applied in the discs in order to
avoid damage. Ferber (2013) recommended the values shown in Table 2.

Table 2 – Recommended TBM advance parameters (Ferber, 2013)

Material Netto-penetration (mm/min) Netto-rotation speed (rpm)

Sand 30-40 1.2-2.0
Gravel and gravel sand 15-30 1.2-2.0
Compact clay 30-40 1.2-2.0
Stiff clay 40-50 1.2-2.5
Soft clay 40-50 2.0-3.0
- additionally boulders/stones 15-20 1.2-2.0

For the Line 5 Precambrian mixed face conditions, penetration rate was hardly higher than 10
mm/rev and rotation speed was around 2.0 rpm. Frequently, penetration values bellow 5 mm/rev
were observed. With such low velocity, sedimentation in the chamber often occurred, making
conditioning harder and, thus, having more difficult conditions during advance.
In critical stretches, penetration was limited by Torque or Thrust. In Figure 4(b), it can be seen
an increase in discs forces, denoting a harder material in the front, which demands limiting the
penetration in order to avoid disc damages. In this figure it also can be seen that the machine was
advancing close to its torque limit.
In mixed face conditions, high cutting tool wear was observed, incurring in a high number of
hyperbaric interventions to change them. The passage of the cutting tools through material of
different resistance, promotes the impact of the tools, damaging them. That is one of the reasons
to limit the cutterhead revolutions. Some damaged discs are shown in Figure 5(a). In this picture,
it can be seem that the discs stopped rotating around their axis, leading to a high wear on them.
The increase of steering forces was one indication that something was not working properly in
the front and the discs were probably damaged.

Figure 5. (a) Some damaged disc. (b) Temperatures observed in mixed face conditions

A similar analysis can be done through Specific Energy (SE) evaluation, as shown in Equation
1. It can be understood as the necessary effort to advance the machine.
𝐸∙𝑉+2𝜋∙𝑅∙𝑇
SE = A∙V
, (1)

where E = total thrust; V = velocity; R = rotation speed; T = torque; and A = cross section area.

As illustrated in Figure 6, in saprolite full face condition, SE ranged around 100 MJ/m³. In
mixed face with rock and saprolite, values higher than 150 MJ/m³ occurred, denoting a very hard
material in the front and high efforts necessary to advance. In soft residual soils, SE ranged be-
tween 20 and 70 MJ/m³. Thus, these high values can explain the high number of hyperbaric in-
terventions in this stretch and also the high temperatures reached, as explained below in this paper.
Furthermore, it was clear that high SE values observed in Sao Paulo mixed face had close
relationship with values registered in other geographical areas also characterized by tropical rock
weathering, as presented for Singapore and Hong Kong mixed face conditions in Shirlaw (2015)
and Shirlaw (2016). In Figure 6, it is clear the relationship between material “hardness” and ve-
locity. The softer the material is (lower SE), the faster the machine advances.

Figure 6. Differences in TBM’s performances in term of SE, velocities and geology conditions.
 On the other hand, the machine has more difficulty in advance the harder the material is (higher
SE). These problems were more evident in mixed face with Saprolite and rock.
It is also important to report that the highest velocities were achieved after hyperbaric interven-
tions, with new cutting tools and clean cutterhead. However, in some cases, velocities fell sharply
after few rings, due to clogging or even cutting tools early damage.
Finally, once penetration was limited in order to decrease the chances of disc damages, heating
was often observed. In critical stretches, the machine had to stop several times due to the high
temperatures in the chamber, endangering the main bearing sealing and aggravating the muck
sedimentation problem in the chamber. In some cases, temperatures in the chamber achieved val-
ues higher than 50º (Figure 5(b)), even when very wet foam and “free water” at the maximum
flow rate were injected in the chamber. That high temperature caused vitrification of the material
in the chamber, increasing the chances of clogging and resulting in the need of more hyperbaric
interventions.

3.3 Different performances and hyperbaric interventions

In the mixed face stretch, the machines presented different performances, which can be related to
a geological variance in the excavation conditions, as showed by Comulada et al. (2016) and Silva
et al. (2016). Figure 6 also shows the specific energy difference between the machines and its
relationship with the geological conditions. Variations in weathering and fracturing of the exca-
vated material had important relevance in both TBM performances.
Other indicative of the performance difference is the number of hyperbaric interventions in
each machine – 4 in Machine 1 and 14 in Machine 2. The total schedule impact was 56 days in
Machine 2, besides two sinkholes and high ground losses (as shown in item 3.4).
Among the 14 hyperbaric interventions in Machine 2, only in three of them was not observed
any ground instability. Four of them were aborted due to ground instabilities and 5 were possible
only after the execution of ground treatment from the surface with chemical injections, as related
ahead. Two of the interventions were due to cutterhead blocking. In Machine 1, there was no
impact in the work schedule, but there was one sinkhole and also some high ground losses.
It is worth mentioning that the difficulties in carrying out the hyperbaric interventions can be
associated with the saprolite heterogeneity. In some passages this material can be highly compact
and with very low permeability, which results in a very thin filter cake layer. However, the exist-
ence of fissures and faults in some stretches makes it difficult to completely seal the front, causing
compressed air leaks, which lead to pressure drops and destabilization of the front.
In the passages with residual soil in the crown, some interventions were aborted due to face
instabilities. Despite the bentonite filter cake in the excavation front, long hyperbaric interven-
tions led to difficulties in keeping hyperbaric pressures constants.
Soil treatments based on the use of colloidal silica (Figure 7 – on the left) were carried on both
tracks to recompose the cavities opened by instabilities and, later, to allow shield advancing. This
procedure had the purpose of reducing permeability and giving some apparent cohesion to the
material in the front, notably the residual soil. Its success was directly associated with the total
amount of product injected.

Figure 7. Soil treatment with colloidal silica in front.

 However, in the presence of ground discontinuities, the injected material flowed to these and
sometimes did not seal the permeable residual soil, which can explain some instabilities verified
during previously treated hyperbaric interventions. Figure 7 (on the right) reveals some injected
silica colloidal in apparent discontinuity, verified during a hyperbaric intervention.

3.4 Settlement Analysis

Most of the ground losses in Precambrian materials were related to the operation mode (tran-
sition mode with applied compressed air), which might have led to some overexcavation. How-
ever, some settlements were also related to insufficient grouting and to the inherent material rhe-
ological properties, once settlements increase were observed even after a long distance after the
TBM passage (Figure 8). Figure 9 shows values obtained from monitoring data.
Settlement analyses were made in order to evaluate remaining voids in the ground. Gaussian
curve (Peck, 1969) and Yield Density Curve (Celestino & Ruiz, 1988) methodology, given by
Equations (2) and (3) respectively, were used for this approach to estimate the ground loss per-
centage (Vs, in Table 3). Analyses were made in sections affected by face instability during hy-
perbaric interventions (M104-1 and M106-2), and with overexcavation observed during advance
(M104-2; M105-both; and M106-1). Apparently, there were no operational problems in the exca-
vation under the section M121.
−x
 = max e2 i2 (2)

max
=
x b (3)
1 + (a)
where  = settlement; x = front distance; i = inflection point; and a,b = calibration parameters.
0

Via 1 - 104
-10
Via 2 - 104
Recalque (mm)

-20 Via 1 - 105

Via 2 - 105
-30 Via 1 - 106
Via 2 - 106
-40 Via 1 - 121
Via 2 - 121
-50
Front Shield
-70 -50 -30 -10 10 30
Tail Shield
Distância da frente (m)
Figure 8.Observed longitudinal observed settlements

Figure 9. Some observed transversal settlements.

 Table 3 gives the obtained results. As can be seen, the ground losses calculated with the Yield
Density Curve were higher. Further, it seems that this curve also gives a better settlement trough
estimation, once the ground losses inferred from the data acquired in the machine [(inferred ex-
cavated volumes minus grouting volumes)/ Theoretical excavated volume] in the critical stretch
were ~7% in Track-1 and ~5% in Track-2. Thus, it seems that the Gaussian curve can underesti-
mate ground losses estimations. Those high settlement values reported justified the need of soil
treatment from the surface in the following sections with the same geological characteristic, as
written before.

Table 3.Gaussian and e approaches.

Gaussian Curve (Peck, 1969)
max Vs
Section i (m) R²
(mm) (%)
104-1 25.2 14.54 2.46 1.0
104-2 33.7 17.75 4.01 0.89
105-1 21.1 16.83 2.38 1.0
105-2 30.7 18.45 3.80 1.0
106-1 34.7 16.73 3.89 0.91
106-2 24.7 13.76 2.28 0.85
121-1 12.7 16.65 1.42 0.81
121-2 7.58 12.66 0.64 1.0
Yield Density Curve (Celestino and Ruiz, 1998)
max a Vs
Section b R²
(mm) (m) (%)
104-1 25.1 16.5 2.98 2.69 1.0
104-2 30.2 22.9 6.0 3.88 0.97
105-1 21.1 19.2 2.66 2.77 1.0
105-2 30.5 21.4 2.71 4.43 1.0
106-1 32.2 21.1 5.8 3.82 1.0
106-2 26.6 12 1.30 6.22 0.99
121-1 12.0 18.5 5.5 1.26 0.98
121-2 7.58 14.5 2.65 0.75 1.0

4 CONCLUSION

Analyzing soil conditioning and both TBM performances, it can be realized that the conditioning
was the main challenge of excavating in Mixed Face Conditions. High efforts made in condition-
ing had little or no reflex in order to make TBM’s advance easier. With some probability of suc-
cess, maybe a hybrid machine could be more suitable for the excavation in such conditions. Some
advances were tried with bentonite injection, but once the machines were not designed for that, a
detailed analysis and its conclusions could not be made. Besides conditioning, high tool wear was
observed, resulting in a high number of hyperbaric interventions. Furthermore, in terms of ma-
chine configuration, although a powerful machine would definitely help facing the problems ex-
perienced, it is not clear that it would avoid all of them under mixed face conditions. Difficulties
in advancing are therefore related to the intrinsic characteristic of the excavated material, which
can be considered as a unique material that cannot be treated as a soil, but not also as a rock.
Finally, it must be highlighted that differences were observed in both TBM performances,
which can also be attributed to geological differences between each tunnel alignment, besides
showing the difficulty in foreseeing TBM performance under these situations. Ground instabilities
and high ground losses were consequences of boring in such difficult conditions.
5 ACKNOWLEDGMENTS

The authors would like to thank Metrô-SP and the Andrade Gutierrez – Camargo Correia Join
Venture for the opportunity to publish this unique experience.

REFERENCES

Celestino, T.B. & Ruiz, A.P.T. 1998. Shape of settlement troughs due to tunnelling through different types
of soft ground. Feslbau 16(2): 118-121.
Comulada et al. 2016. Experiences gained in heterogeneous ground conditions at the twin tube EPB shield
tunnels in Sao Paulo Metro Line 5. In Proceedings of the World Tunnel Congress. United States: San
Francisco.
Ferber, S. 2013. São Paulo – Metro Linha 5 Lote 3: Ground Conditioning and TBM Parameters. In Her-
renknecht Training Course, Herrenknecht AG. Brazil: São Paulo.
Futai et al. 2012. Resistência ao Cisalhamento e Deformabilidade de Solos Residuais da Região Metropo-
litana de São Paulo. In Negro et al. (ed.), Twin Cities: solos das regiões metropolitanas de São Paulo e
Curitiba, São Paulo, ABMS: 155-187.
Hollmann, F. & Thewes, M. 2012. Evaluation of the Tendency of Clogging and Separation of Fines on
Shield Drives. Geomech. & Tunnelling 5: 574-580.
Maidl, B.et al. 2012. Mechanized shield tunnelling. In 2nd Edition. Berlin: Ernst & Sohn Verlag.
Monteiro, M.M. & Rocha, H.C. 2015a. Personal Comunnication.
Monteiro, M. M. & Rocha, H. C. 2015b. A experiência do Metrô de São Paulo nos estudos de abrasividade
de rochas: técnicas para previsão do consumo de ferramentas de corte em escavações subterrâneas. Re-
vista Engenharia 625: 119-124.
Oliveira, D.G.G. et al. 2017. EPB Conditioning of Mixed Transitional Ground: Investigating Preliminary
Aspects. In WTC 2017. Norway: Bergen.
Peck, R.B. 1969. Deep excavations and tunneling in soft ground. In 7th International Conference on Soil
Mechanics and Foundation Engineering, Mexico City, State-of-the-Art volume: 225–290.
Shirlaw, N. 2015. Pressurized TBM tunnelling in mixed face conditions resulting from tropical weathering
of igneous rock. In International Conference on Tunnel Boring Machines in Difficult Grounds (TBM
DiGs), Singapore, 18–20 November 2015.
Shirlaw, N. 2016. Mixed ground tunnelling in Hong Kong and Singapore. Presentation. In Symposium on
Underground Development and Technology, Singapore, November 2016.
Silva et al. 2015. Twin tunnels excavated in sandy soils in a density urban area. In Proceedings of the World
Tunnel Congress. Croatia: Dubrovnik.
Silva et al. 2016. Lições Aprendidas na Escavação de Túneis Paralelos da Linha 5 – Lilás do Metrô de São
Paulo. In Proceedings of the XVIII Congresso Brasileiro de Mecânica dos Solos e Engenharia Geotéc-
nica, COBRAMSEG, 2016. Brazil: Belo Horizonte.