Tunnelling and Underground Space Technology 94 (2019) 103110

Contents lists available at ScienceDirect

Tunnelling and Underground Space Technology

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

Clogging and flow assessment of cohesive soils for EPB tunnelling: Proposed T
laboratory tests for soil characterisation
Daniela G.G. de Oliveiraa, , Markus Thewesb, Mark S. Diederichsc
⁎

a
Tunnelsoft – BabEng, Lübeck, Germany
b
Institute for Tunnelling and Construction Management, Ruhr-University Bochum, Bochum 44801, Germany
c
Department of Geological Sciences and Geological Engineering, Queen’s University, Kingston, ON K7L3N6, Canada

ARTICLE INFO ABSTRACT

Keywords: For earth pressure balance (EPB) machines, it is fundamental to characterise the excavated ground in its natural
EPB machine and conditioned states. The characteristics of the excavated material have a direct influence on machine op-
Soil conditioning eration, impacting the functions as a support medium maintaining the support pressure ahead of the tunnel face,
Mixed soils besides its transportation and disposal. The current available methods to assess the clogging potential focus
Clogging
mainly on sedimentary pure clayey soils, not including mixed soils. Mixed soils exist widely in nature, including
Flow behaviour
Fluidity
a wide variety of residual soils, which are by-products of intense tropical weathering. This paper presents a
combined laboratory routine to characterise and evaluate the clogging and fluidity of soils, including mixed soils
by considering different clay fractions. This routine still requires tests conducted directly on-site, where the
results should be compared with machine operation parameters, such as torque of cutterhead and screw con-
veyor, temperature, pressure, as well as occurrence of clogging. This testing procedure provides on-site as-
sessment that can also be conducted prior the excavation using borehole material. The intention of this routine is
to provide a geological-geotechnical qualitative characterisation of the material to be excavated, and of the
effectiveness of the additives that could be added, as a general guidance for EPB soil conditioning, especially of
cohesive soils. The advantage of this routine is its simplicity and low cost to be reproducible in different la-
boratories and jobsites, making feasible the comparison between different soils by different operators, before and
during excavation.

1. Introduction can be used as soil conditioner in EPB machines, growing the require-
ments of optimising the excavation process towards a sustainable ap-
The characteristics and behaviour of a conditioned ground play a plication of EPB additives.
fundamental role in the entire operation of earth pressure balance Even though cohesive soils lie within the ideal operation field of
(EPB) machines. This conditioned material, the muck, can be char- EPB machines (BTS, 2005; Maidl et al., 2012), some particularities of
acterised by the excavated ground itself combined with any form of this soil type cause certain challenges for tunnel operation, clogging
additives, such as water, foam, polymers, and fines. The muck con- being one of them (Thewes, 1999, 2004; Sass and Burbaum, 2009;
tributes to sustain the pressure ahead of the machine, combined with Feinendegen et al., 2011; Hollman and Thewes, 2012, 2013; Peila et al.,
the advance and retraction mechanisms of the hydraulic jacks and the 2016; Basmenj et al., 2016; Khabbazi et al., 2017; Avunduk and Copur,
rotation speed of the screw conveyor. Additionally, it needs to hold 2019; Oliveira et al., 2019). Moreover, the flow behaviour of soils
certain characteristics to be successfully transported and disposed. should be known for any EPB excavation, as it will have a great influ-
Several characteristics of the muck (e.g. permeability, compressibility, ence on the operation, affecting the support pressure, machine para-
flowability) that might be beneficial for face support or flow inside the meters, such as the torque of machine and screw conveyor, as well as
chamber and screw conveyor can result in challenges for its transpor- transportation and disposal (Galli and Thewes, 2014, 2018; Galli,
tation and disposal (EFNARC, 2005; Herrenknecht et al., 2011; Maidl 2016).
et al., 2012; Thewes et al., 2012; Galli and Thewes, 2014). Besides that, Both characteristics, i.e. clogging potential and flow behaviour,
there is an increasing environmental concern about the products that should be examined in combination, as they could influence each other.

⁎
Corresponding author at: Einsiedelstr. 28, 23554 Lübeck, Germany.
E-mail address: [email protected] (D.G.G.d. Oliveira).

https://doi.org/10.1016/j.tust.2019.103110
Received 11 August 2018; Received in revised form 5 September 2019; Accepted 5 September 2019
0886-7798/ © 2019 Elsevier Ltd. All rights reserved.
D.G.G.d. Oliveira, et al. Tunnelling and Underground Space Technology 94 (2019) 103110

An adopted solution to decrease clogging should not turn into an issue potential for clogging and the flow behaviour of any cohesive soil, with
for the flow of a certain excavated material, for example, an extremely or without conditioning, offering a possibility for laboratory and on-site
liquid muck causing transportation problems, such as being con- investigation prior and during an EPB tunnel excavation. The main
tinuously spilled out of the conveyor belt. advantage of the testing routine is its simplicity and the relative low
Several laboratory tests and methodologies have been proposed for cost of the testing devices. The low quantity of material required for
evaluating, directly or indirectly, the clogging potential (Thewes, 1999, testing makes feasible to test the material obtained from boreholes, as
2004; Sass and Burbaum, 2009; Spagnoli et al., 2010; Feinendegen well as the possibility to conduct tests on-site. Most of the time it is
et al., 2011; Hollman and Thewes, 2012, 2013; Zumsteg and Puzrin, challenging to transport soil, especially oversees, due to high costs and
2012, 2013, 2016; Hollmann, 2014; Feinendegen and Ziegler, 2016; regulations, which is one of the main limitations for larger laboratory
Khabbazi et al., 2017; Hernandez et al., 2018; Milatz et al., 2019; devices that cannot be transported to jobsites. This work shall be con-
Spagnoli et al., 2019) and flow behaviour by means of slump and tinued reaching its next steps: testing on-site excavated material and
rheometer devices (Quebaud et al., 1998; Leinala et al., 2000; EFNARC, comparing it with excavation parameters.
2005; Peila et al., 2008; 2013, 2016; Borio, 2010; Budach, 2011; Considering that some of the tests included foam, the main foam
Thewes et al., 2012; Galli and Thewes, 2014, 2018; Galli, 2016). parameters, namely foam concentration (Cf), foam expansion ratio
Most of the research, as pointed out by Oliveira et al. (2018a, b, (FER), and foam injection ratio (FIR), are defined here according to
2019), focused on either sand or clayey soils mainly originating from EFNARC (2005).
sedimentary deposition. However, mixed soils exist frequently in mSurfactant
nature, such as the tropical residual ground, which can be found at Cf (%) = 100 ×
mFoamSolution (1)
great depths in tropical areas (Vargas, 1980; Rahardjo et al., 2004;
Viana da Fonseca and Topa Gomes, 2010; Huat et al., 2012; Oliveira where
et al., 2017a,b). EPB excavation along this type of ground, especially
when including mixed transitional ground, as defined by Oliveira and mSurfactant = mass of surfactant in foaming solution
Diederichs (2016), can lead into major excavation issues, as pointed out mFoam Solution = mass of foaming solution
by several authors (Della Valle, 2001; Thewes, 2004; Zhao et al., 2007; VFoam
Shirlaw, 2015, 2016; Comulada et al., 2016; Gong et al., 2016; Peixoto FER =
VFoamSolution (2)
da Silva et al., 2017; Oliveira and Diederichs, 2016; Oliveira et al.,
2017a, 2017b). where
Oliveira et al. (2018a, b, 2019), after testing assembled soils by
mixing clay and sand in different proportions to simulate the basis of a VFoam = volume of foam
residual soil in the laboratory, proposed tests to evaluate the clogging VFoam Solution = volume of foaming solution
potential and flow behaviour of those mixed soils. The clogging po- VFoam
tential was a modification from the original method suggested by FIR (%) = 100 ×
VSoil (3)
Zumsteg and Puzrin (2012). In the original methodology, Zumsteg and
Puzrin (2012) used a HOBART® mixer and defined a clogging para- where
meter (λ) by comparing the amount of soil stuck in the mixing tool
(MT) and the total amount of soil mass. Oliveira et al. (2018a) de- VFoam = volume of foam
monstrated that this method, as originally proposed, would lead into an VSoil = volume of tested soil
overestimation of the clogging potential, especially for the case of
mixed soils. In response to this issue, the authors included an additional 2. Material and methods
step done by dropping the MT from a certain height, to a certain sur-
face, by means of a newly assembled device, ATUR.1 In this section, the authors describe the assembled soils and the
Subsequently, a flow table was used to evaluate the flow behaviour laboratory methodology sequence. Some testing limitations are also
of mixed cohesive soils, resulting in the proposed method described by pointed out.
Oliveira et al. (2018b). Results obtained from the flow table, slump
tests, and rheometer were compared and some assumptions about flow 2.1. Tested material
behaviour of soils, depending on the clay percentage and mineral,
water content, grain size, and additives, were made. The biggest con- Preparing all the soils in the laboratory assured reproducibility of
tribution was offering a testing routine to evaluate the flow of cohesive the tests and control of all variables, including three main mineral
soils. Another advantage is the required small quantities of sample for products, starting as powdered components: quartz sand, kaolinite, and
testing (less than 1 kg), when compared to the slump test (more than bentonite. Table 1 details all the main characteristics of these three
8 kg), being feasible to easily conduct tests on-site and with borehole elements.
samples. There were a couple of reasons for choosing kaolinite and bentonite
This present study combines the flow behaviour and clogging as- as tested clay minerals. First, they represent the extreme spectra of
sessment methods, adding more stages to it, such as the CBFactor. The plasticity index for clay minerals, kaolinite being a considerably low
procedure of soil assemblage has been also modified, where enough plasticity clay, and bentonite, a highly plastic one. Second, kaolinite is
time was given to allow homogenous moisturisation of the soil samples, the main clay mineral present in tropical residual ground (Vargas,
after water was added. This procedure, together with more frequent 1980; Huat et al., 2012), justifying the choice of this mineral and a
mixing of the samples when being prepared, decreased the variability of slightly higher number of tests with it. Bentonite, because of its swelling
the results. This paper also includes insights obtained when conducting properties, is expected to be a bigger challenge for shield tunnel drives.
the same routine with natural soils from Singapore, comparing with A total of 9 sample combinations were assembled, with different
previous experience when excavating those soils with shield machines. water contents and, for some of the tests, with different conditioning
This laboratory routine proposes a method of evaluating the products. Table 2 presents all the sample combinations with propor-
tions of minerals and their Atterberg values. The sand portion of the
samples were assembled with two different sand gradations, as illu-
1
ATUR - Adhäsive Tone Untersuchung RUB-Queens (RUB-Queens Clay strated in Fig. 1.
Adhesion Tool) Regarding the conditioning additives, it was chosen to keep the

2
D.G.G.d. Oliveira, et al. Tunnelling and Underground Space Technology 94 (2019) 103110

Table 1
Main characteristics of each component of assembled soils.
Product specification Supplier Spec. Chemical Comp. Specific Surface Cation Exc. Capacity pH Moisture (%)
Grav. (m2/g) (meq./100 g)

Silica sand Euroquarz 2.65 SiO2 – – 6.5–7 –

Kaolinite GHL KS 80*2 Georg H. Luh GMBH 2.62 Al2O3, SiO2, Fe2O3 18.2 – 4.5–9.5 0.6–2
Active Bentonite IBECO B1 Imerys Civil 2.65 Al2O3, SiO2, Fe2O3, MgO, CaO, 600–800 70 ± 10 10 11 ± 3
Engineering K2O, Na2O

Table 2
Tested samples with proportions of minerals and Atterberg limits values.
Sample ID Type of sand Sand (%) Kaolinite (%) Bentonite (%) Liquid limit Plastic limit Plasticity Index
WL (%) WP (%) PI (%)

Ksfine30 Sfine 70 30 – 22 16 6
Ksfine50 Sfine 50 50 – 32 21 11
Ksfine70 Sfine 30 70 – 44 29 16
K100 – 0 100 – 62 40 22
Kmedium50 Medium 50 50 – 31 23 8
Bsfine20 Sfine 80 – 20 86 27 59
Bsfine50 Sfine 50 – 50 171 31 140
Bsfine70 Sfine 30 – 70 245 34 211
B100 – 0 – 100 486 53 433

Fig. 1. Grain size distribution for sand portion of soil mixture presented in Table 2.

Table 3 is a regular EPB foam and products B and C contain anti-clay polymer.
General description of additives used for tests with conditioned The products B and C are from two different suppliers. For all tests, Cf
samples. was always 3%, the FER target was always 9.5 ± 0.3, and FIR con-
Product Name Description sidered the moisture content necessary to bring the consistency of a soil
down from a certain consistency index to another for its calculation.
Product A Foam Table 4 presents the initial consistency (initial) of the soil only with
Product B Foam with anti-clay polymer
water and the consistency achieved (final) with the additive.
Product C Foam with anti-clay polymer
For producing the foam, was used the same foam generator as de-
scribed by Galli (2016) as “TLB Mini Foam Generator”. This foam
details of them undisclosed, replacing the names of products and their generator is designed to supply pressure air/liquid up to 4 bar, having
respective suppliers by letters (A, B, and C), specifying if it is only foam, glass pearls of 5 mm in diameter inside the tube where the liquid/air is
or foam with anti-clay polymer, as indicated in Table 3. The product A mixed. The flow regulation is done manually by means of a flowmeter
(Krohne) and was calibrated to generate foams as in the EPB machines.

3
D.G.G.d. Oliveira, et al. Tunnelling and Underground Space Technology 94 (2019) 103110

Table 4
Variables for tests with soil additives, including main foam parameters.
Soil Sample Test ID IC Initial (only soil + water) IC Finala (soil + water + additives) FER FIR (%)

Ksfine50 1 1.33 0.81 9.6 91.5

2 1.16 0.68 9.3 77.7
3 1.05 0.54 9.4 84.5
4 0.98 0.43 9.5 90.0
5 0.50 0.02 9.6 86.3
Bsfine50 1 1.03 0.99 9.4 88.0
2 0.94 0.88 9.4 114.5
3 0.77 0.73 9.3 103.5
4 0.71 0.68 9.3 80.5
5 0.54 0.44 9.2 118.3

a
Values are average from all tests. The standard deviations are lower than 0.1.

Fig. 2. Illustration of parts of testing sequence: (a) soils were mixed with required mineral and water contents and stored in closed containers; (b) a reasonable
quantity of sample was placed as lumps in the HOBART® mixing bowl; (c) the HOBART® mixer with soil specimen; (d) B-flat beater with stuck soil on it before any
drop (λ0); (e) B-flat beater placed in the ATUR device after one drop; (f) detail of the dimensions of the ATUR device; (g) flow table with soil specimen after 40
joltings of the table; (h) squeezing of soil specimen to assess muck consistency and flow.

2.2. Proposed methodology after 24 h of sample curing2. For highly plastic samples (IP > 40%) the
samples were mixed a second time, waiting another 24 h to check the
The methodology combines a sequence of testing procedures, moisture content. If the target IC was achieved, with a difference
starting with the preparation of samples, verification of the target smaller than ± 0.04, the tests were conducted. Otherwise, either pow-
consistency index, and test with the HOBART® mixing methodology. dered soil or water was added to achieve the target consistency. The
This is followed by the sequential dropping of the beater with the ATUR samples were mixed again, and we waited another 24 h before checking
device, then cleaning and washing the beater and bowl, the flow table the moisture content. Some of the samples, such as the ones with higher
test, and finally, the assessment of samples by squeezing them. Fig. 2 content of bentonite, took longer than 72 h to be fully homogenised.
illustrates the stages of this methodology and Fig. 3 provides a flow- The fact that the soil had enough time to cure, as well as the use of a
chart of the methodology sequence. high speed mixer, homogeneously moisturising, made it possible to
All samples were assembled initially as dried specimens, starting achieve the target consistency, which was tested at least twice, before
with mixing powder. The target water content was added (deionised and after the test.
water) and after the mixing stage, soils were stored in closed containers Equation (4) was applied for calculating the consistency index,
(Fig. 2a). Most of the time, a high amount of soil was prepared (ap- considering the values of plastic and liquid limits (WP and WL, re-
proximately 2 kg of wet soil), using the same specimen for several tests, spectively), together with the water content (W).
which guaranteed testing the same consistency always. Each sample
combination, for each specific water content, was tested for a minimum
of three times. Water was never added during the test, only during the
soil assemblage procedure.
For low plasticity specimens, the moisture content was checked 2
Sample curing means that the sample was given enough time for moisture
homogenisation by simply waiting for a certain period.

4
D.G.G.d. Oliveira, et al. Tunnelling and Underground Space Technology 94 (2019) 103110

sample not too liquid), the soil was placed in the mixing bowl as lumps,
which is closer to the reality of an EPB excavation that scrapes off the
soil ahead of the cutterhead in sizes of lumps (Peila et al., 2016). The
total soil mass was recorded, which represents the parameter GTOT, as
suggested by Zumsteg and Puzrin (2012).
In terms of the suggested total mass for the tested soil, it will ba-
sically depend on the shear resistance of each specimen, also related to
its consistency index. A stiff soil with IC > 1.0 can easily break the
mixing machine with only 700 g of total wet mass. For instance, tests
with pure kaolinite, with IC = 1.04, could only proceed with maximum
of 400 g of wet soil. For lower consistencies the total wet mass was
approximately 800–1000 g. It was also noticed that changing the mass
around these values did not affect the results, as several tests were done
varying the soil amount between 400 and 1200 g.
The B-flat beater (MT) was then placed in the machine, mixing with
the HOBART® in speed 1 (100 rpm) for 3 min. The beater was carefully
removed and the mass of the stuck soil in the MT was weighed. The
mass of the beater was subtracted, resulting in the parameter GMT0,
following the procedure as suggested by Zumsteg and Puzrin (2012)
(illustrated in Fig. 2d).
The next stage was to place the beater in the ATUR device and let
the beater drop, as described by Oliveira et al. (2018a) (shown in
Fig. 2e and f). The mass of the soil stuck at the beater was weighed
again, obtaining GMT1. The same procedure was performed until the
third drop of the beater, always completely cleaning the surface base of
ATUR and avoiding influence of the already dropped material sticki-
ness, and then, obtaining GMT2 and GMT3. The MT was then dropped
four consecutive times and weighed to obtain GMT7.
Surely, the number of drops could change in the future, after
comparing data obtained in tunnel drives. In a later testing campaign
conducted in Singapore, testing natural soils that have already been
excavated by TBMs, it was tried different dropping options. It was
chosen, for instance, to drop the beater until the first mass of soil would
be dropped. For the samples called old alluvium clay, for example, it
took 40 drops until a small mass of soil was dropped. It is known, from
Fig. 3. Flowchart of combined testing sequence suggested to assess clogging
and conditioning of cohesive soils for EPB machine operation. previous excavation experience, that major clogging issues occurred
while excavating old alluvium clay with both shield types (Mixshield
and EPB). For the other samples, such as the grey clays from the Kallang
WL W
IC = Formation, or the residual soil, a by-product from the Bukih Timah
WL WP (4)
granite, the opposite has happened. These samples would lose soil mass
within the first three drops of the beater. Many times, however, these
The target consistency was chosen based on the main limits of the
soils showed significant stuck material without any drop of the beater
clogging range (e.g. low, medium, high) according to the diagram of
(equivalent to λ0). From previous experience, neither of these last-
Hollmann and Thewes (2012). This is the only diagram providing a
mentioned soils implied in clogging issues when excavated with EPB
correlation between clogging potential and consistency index based on
machines. It was clear then that soils with severe clogging tendencies,
a considerable amount of tunnel data. It was decided to use this graph
would remain stuck at the beater, even after several drops.
as a base to calibrate the test results, as explained in the Section 3.1.
The sequence of drops was chosen mainly to correlate with the
For the specific case of the tests with foam, the consistency was
diagram of Hollman and Thewes (2012), which is so far the only dia-
chosen in a way that the FER and FIR would be realistic, based on
gram based on a considerable amount of tunnel drive data. The pro-
tunnel drives common values. The initial consistency was reached by
cedure was a trial-and-error, where many options have been tried be-
mixing the soil with water, for its assemblage. Foam was added exactly
fore achieving the late suggested sequence. With more collected data
before the test to achieve the final target consistency. Once the first test
from tunnel drives, the number of drops might change, expecting an
was performed with a certain foam parameter, the variables were kept
improvement of this testing methodology.
practically the same between all the tested products, to assure a fair
The next step of the current testing routine was to clean the tools,
comparison. All the tests with foam were performed with the samples
which is a stage not included previously by Oliveira et al. (2018a). This
containing 50% of clay (Ksfine50, Bsfine50) and all the variables can be
stage adjusts the final values of the clogging potential, detailed in
accessed in Table 4. The final tests with Kmedium50 were not included
Section 3.1. The cleaning process should also be correlated in the future
in this table and will be described directly in Section 3.4.
with the intervention time in tunnel boring machines. Fig. 4 illustrates
At the end of each test, a sample was collected to verify the moisture
the classification for the removing and cleaning of the MT and bowl,
content and consistency index. However, the plasticity index was not
called the CBFactor. The water used to clean the tools was always luke-
verified again for the mixture with the additive. The water content for
warm tap water.
the soil tested with foam was directly assessed from the tested sample
The cleaning of the beater and the bowl also did not change de-
(with foam). For all the graphs, the presented consistency is the one in
pending on the mass of the soil. The classification current used for the
which the test was conducted.
cleaning was mostly based on the easiness (or difficulty) to remove the
Once the target IC was achieved, the next step was to proceed with
soil from the beater or the bowl, more than on the time to do that. This
the HOBART® mixing. Fig. 2b and c show that whenever possible (i.e.
stage might be affected by the perception of different operators. Yet,

5
D.G.G.d. Oliveira, et al. Tunnelling and Underground Space Technology 94 (2019) 103110

Fig. 4. Cleaning beater and bowl factor (CBFactor), with categories (based on the IC) and sub-categories description on the top, including the respective value range to
be added to Eq. (7). On the bottom, some examples of stiff and soft sub-categories are shown.

Table 5 excavated by shield machines, correlating the test results with the ex-
Clogging potential categories defined by parameter λF. perience of the excavation.
Clogging Potential
After testing around eight different types of natural cohesive soils,
F
in Singapore, it was realized that the CBFactor should be values between
≤0.27 Low 0.5 and 3.0, whereas 0.5 would be the hardest sample to be cleaned,
0.27–0.48 Medium and 3.0, the easiest. The value 0.5 was the one chosen for pure ben-
≥0.48 High
tonite, which rarely occurs in nature. The value 3.0 is only for mixed
soils with low % of low plasticity clay.
With numerous tests using different soil specimens, it was possible
after doing many tests, including latest tests with natural soils, it was
to observe that the cleaning process depended on the soil properties, as
clear that there was a significant difference between soils that were very
well as the beater metal/coating. Specimens with similar properties
easy to clean, and soils that would present a major challenge in the
surely might be not as easy to differentiate between two sub-categories
cleaning process. The only way to further verify the cleaning factor is to
as soils with very different properties (i.e. low versus high plasticity).
conduct many tests with natural soils that were excavated or will be

6
D.G.G.d. Oliveira, et al. Tunnelling and Underground Space Technology 94 (2019) 103110

to assist in future with a further correlation or adjustment of the

CBFactor.
For samples with IC equal to or smaller than 0.8 (not feasible for
stiffer samples), the next step was to conduct the flow table test, as
proposed by Oliveira et al. (2018b) and shown in Fig. 2g. The samples
were the same ones as used in the clogging test procedure. For that, a
cone was greased with oil; the soil was moulded inside the cone,
pressing it in layers and against the cone walls to assure complete
filling. The cone was then lifted, and the diameter of the soil base was
measured (m0). The table was jolted 40 times and an average of three
diameter measurements was taken, obtaining m40. Pictures of the cone
were also taken, so that whenever soil was left stuck in the cone even
with an oil-greased surface, it could indicate some tendencies for
Fig. 5. Main testing target points for the initial ATUR calibration considering clogging. This procedure was quite useful for the samples tested with
clogging potential boundaries as defined in the chart by Hollman and Thewes conditioned soils, as detailed in the Section 3.4.
(2012), with clogging categories presented on the right side. The last stage of the test was to squeeze the sample, ideally re-
cording a video (Fig. 2h and Video 1). This procedure is similar to what
Extreme conditions, such as the easiness to clean the bowl after a test is already done inside EPB machines, when a site worker touches and
with a 30% kaolinite sample, against the extreme difficulty to clean squeezes the material that comes out of the screw conveyor, feeling the
after a 100% bentonite specimen, were extremely obvious and shall not muck. Even though it could be a slightly subjective procedure, it can
be missed by any operator. Besides recording the difficulties to remove still provide a useful information of the muck consistency and its
and clean the soil by describing it in detail, pictures and even videos properties, allowing some adjustments of the conditioning parameters.
were recorded, making it easier later to compare between samples and Still to be correlated in the future with real tunnel drives, combined

Table 6
Calibration table for kaolinite samples for IC above 0.54 compared with clogging potential chart by Hollman and Thewes (2012, 2013), shown as H&T.

7
D.G.G.d. Oliveira, et al. Tunnelling and Underground Space Technology 94 (2019) 103110

Table 7
Calibration table for kaolinite samples for IC below 0.55 compared with clogging potential chart by Hollman and Thewes (2012, 2013), shown as H&T.

with successful excavation results, the ideal muck most probably is si- Table 5 presents the main clogging potential categories. These cate-
milar to a toothpaste being squeezed out of its tube, for the sample that gories were also defined with the calibration procedure, as detailed in
is collected directly from the conveyor belt. If this paste is extremely Section 3.1. Table 5 shall also be adjusted with more acquired data,
liquid, it will fall aside, or if it is extremely stiff, squeezing and flowing especially with tests done with tunnel drives.
are not possible (Video 1).
Clogging tendencies can also be observed when squeezing the F = x
CBFactor (7)
sample, or even excess in air due to high FER values (detailed in Section
2.3). A compilation of several recorded videos from several tests can be + + + +
0 1 2 3 7
accessed in Video 1. =
x
5 (8)

2.3. Calculations of parameters Additionally, for all the samples classified as stiff (St in Fig. 4;
IC > 0.75), a low clogging value would result in lumps, as the clogging
Parameters regarding tendencies for clogging and flow behaviour category.
can be drawn from the above-mentioned methodology. The calculation For the flow behaviour, as explained also by Oliveira et al. (2018b),
for clogging potential obtained from weighing the mass of the soil stuck the flow can be defined by Eq. (9).
in the MT, as proposed by Zumsteg and Puzrin (2012), is calculated by m40 m 0
Flow40 = 100
Equation (5), obtaining λ, which is equivalent to λ0 (no drop of the m0 (9)
beater).
where
GMT 0
0 =
GTOT (5) m0 = average of three measurements of initial specimen diameter
where GMT0 is the total mass of soil stuck in the MT (B-flat beater) and m40 = average of three measurements of specimen diameter after 40
GTOT the total mass of soil. times jolting
As the MT is dropped, the other parameters can be calculated si-
milarly, with the same general equation (Eq. (6)), as initially proposed 3. Results and discussions
by Oliveira et al. (2018a).
This section presents all the results obtained from the testing cam-
GMTx
x = paign with the combined methodology as described in Section 2.2. It
GTOT (6)
also considers some of the results gained from previous testing phases,
where GMTx is the total mass of soil stuck at the MT after x numbers of which can be accessed in Oliveira et al. (2018a,b). Initially, it is de-
drops and GTOT the total mass of soil. monstrated how a final calibration of the method was performed by
The final clogging parameter is λF, which was adjusted after several comparing the results with the universal clogging graph from Hollman
tests with pure kaolinite samples, comparing results with the universal and Thewes (2012). Furthermore, aiming to investigate particularities
clogging potential diagram of Hollmann and Thewes (2012), as ex- of the cohesive mixed soil, tests with different proportions of clay and
plained in detail in Section 3.1 and by Oliveira et al. (2018a). The final minerals, in diverse consistency ranges, were conducted. Tests were
clogging parameter, λF is defined by Eq. (7), where x is the average of conducted as well with four different beaters, detailed in Section 3.3.
all the obtained λx values, as indicated in Eq. (8). The CBFactor is cal- Finally, the influence of additives when added to these mixed soils was
culated according the categories/sub-categories in Fig. 4. Finally, also investigated.

8
D.G.G.d. Oliveira, et al. Tunnelling and Underground Space Technology 94 (2019) 103110

Table 8
Calibration table for bentonite samples compared with clogging potential chart by Hollman and Thewes (2012, 2013), shown as H&T.

3.1. Calibration with pure clay The target IC values, for the tests conducted with water, were ap-
proximately 0.5 points above and below the clogging limits defined in
This calibration was done mainly to define how many drops of MT the graph by Hollman and Thewes (2012), as illustrated in Fig. 5, and
had to be performed to provide reasonable results of the clogging also presented in the study by Oliveira et al. (2018a,c). Numerous tests
parameter (λF), which directly depends on the height of the drop of the for each IC were conducted and the results for kaolinite can be found in
beater and stiffness of the device base (ATUR), besides the properties of Tables 6 and 7. To facilitate visualisation, the cells with the resulting λ
the soil. An initial calibration trial was presented by Oliveira et al. and clogging category, as defined in the graph by Hollman and Thewes
(2018a); however, these tests were conducted starting with dried (2012), have been coloured as grey for lumps, yellow for little clogging,
powdered mixtures, where water was added as the test went along orange for medium clogging, and red for strong clogging.
without curing time, which seemed to increase the standard deviation After realising that the cleaning and removing of the stuck soil from
of the results. By modifying the soil assemblage, the standard deviation the beater and the bowl were also providing valuable information, the
values decreased, obtaining results that were more reliable. CBFactor was incorporated. The values of CBFactor and the resulting
This calibration was performed by comparing the results with the equations that finally lead into the resulting clogging parameter λF
universal clogging chart of Hollman and Thewes (2012). Considering have been defined by adjusting the results presented in Tables 6 and 7.
that this chart was elaborated based on real tunnel drives, indirectly, To differentiate little clogging from lumps, a rule was applied that
this calibration would have also a correlation with tunnel drives. The whenever the clogging was little, considering the values given in
data for this universal chart, however, were obtained with excavations Table 5, and the CBFactor category was St (stiff), then the final resulting
along sedimentary soils, mainly with pure clay layers of low to medium clogging category was, logically, lumps (grey colour), and not little
plasticity index (average IP of approximately 40%, with the highest clogging.
IP = 72%) (Hollman and Thewes, 2012, 2013). The database did not As it can be observed in these tables, the results represented by the
include residual soils or mixed clay-sand soils, neither very highly two last columns, ‘Clogging Category’ and ‘Clogging Potential H&T’, are
plastic clays, such as the bentonite tested in this research (IP = 433%; consistent, except for only one point out of 49, in ID 6, which is rela-
Table 2). Therefore, it was concluded that to guarantee a good com- tively close to the limit of little/medium clogging category.
parison and, consequently, calibration, tests with pure kaolinite should Considering this effective correlation with pure kaolinite samples,
be conducted and compared first. tests with bentonite were then conducted. The results are presented in

9
D.G.G.d. Oliveira, et al. Tunnelling and Underground Space Technology 94 (2019) 103110

Fig. 6. Clogging parameters for kaolinite (purple triangle) and bentonite (red diamond). (For interpretation of the references to colour in this figure legend, the
reader is referred to the web version of this article.)

Table 8 and illustrated by Fig. 6. For bentonite, the chosen CBFactor was feasible to verify the reality of clogging occurrence in tunnel drives, as
between 0.5 and 1.0, especially after the experience of testing several it is most unlikely such a tunnel project to happen along pure bentonitic
natural soils from Singapore and realising that the pure bentonite was clays.
indeed the worst case encountered in terms of cleaning the beater and
the bowl. For the highest consistencies (IC > 0.9), the samples would 3.2. Tests performed with water
get totally stuck, almost cooked, on the bottom of the bowl, being ex-
tremely hard to remove, even with the use of a spoon to scrape it. For Figs. 7–9 show the results from tests for different clay fractions and
the case of the pure bentonite, it might happen that it will never be IC values, without any additive. The term water refers to the moisture

10
D.G.G.d. Oliveira, et al. Tunnelling and Underground Space Technology 94 (2019) 103110

Fig. 7. Tests with different contents of kaolinite and fine sand.

content of the samples. All the graphs show the original parameter sample without dropping might lead into very different results for
proposed by Zumsteg and Puzrin (2012), labelled here as λ0, and the certain soils that are not really stuck at the beater (e.g. Ksfine 30). If a
new clogging category, λF. On all graphs, labels represent the average soil is really stuck at the beater (e.g. alluvium clay from Singapore), the
values and it includes the standard deviation. In some points it is not beater can be dropped many times and the result is the same, nothing is
possible to visualise any standard deviation because it was too low. dropped, leading into a low standard deviation. In the opposite direc-
Tables 9 and 10 resume the used values for those graphs, including the tion, a soil that is not really stuck, might lose mass at the first, or at the
CBFactor, only for the mixed samples, as the pure one is already pre- second or third drop, this varies, but after seven drops it has lost all it
sented in Tables 6–8. had to lose, therefore, what was not really stuck to it.
There is an overestimation of the clogging values λ0 for mixed soils, For kaolinite samples mixed with sand (Fig. 7), only from 50% of
when compared to the final drop λ7 and the final clogging category λF, clay content that medium clogging starts to occur for a consistency of
as there was soil stuck in the beater that was not really stuck, any vi- approximately 0.6, being practically absent for 30% of kaolinite. Even
bration would easily detach the soil. As the clay content increases, the for λ7 there was little clogging happening with 30% of kaolinite.
differences between λ0, λ7 and λF decrease, especially for the con- Clogging becomes more evident with 70% of clay content, with medium
sistencies between 0.5 and 0.7 for kaolinite, and 0.7 and 1.0 for ben- clogging for IC between 0.5 and 0.8. The highest clogging values seem
tonite, suggesting that when clogging is prominently present, the drops to be slightly towards the right of the x-axis for mixed soils, when
of the beater do not significantly decrease the resulting λ values. In compared to pure clay (in Fig. 6).
general, λ0 presented higher standard deviation values than λ7 and λF, For bentonite (Fig. 8), medium clogging already occurs with 20% of
which also indicates that only conducting the original weight of the clay, intensifying with increasing clay content. For most of the four

11
D.G.G.d. Oliveira, et al. Tunnelling and Underground Space Technology 94 (2019) 103110

Fig. 8. Tests with different contents of bentonite and fine sand.

graphs, except for 50% of clay content, the highest clogging values are procedure was also slightly harder for samples mixed with larger grains.
at IC between approximately 0.7 and 0.95. For 50% of clay content, The clogging is also higher (λ7 and λF) towards the right of the x-axis
there is a peak clogging value for a higher consistency around the for the medium sand.
plastic limit. High clogging values in this consistency also occur for 70% Flow table tests were conducted for all the tested samples and
of clay content. If only λ7 is considered, then the resulting clogging Fig. 10 presents a resume of the results for the tests without additives
potential would be lower than the samples with kaolinite, which might for fine and medium sand, with only water. The flow table, as detailed
not be the case in tunnel drives, so this needs to be further considered in the study by Oliveira et al. (2018b), was applied to evaluate the flow
when comparing material excavated from different tunnel projects. of cohesive soils, with and without conditioning. Several conclusions
During the testing procedure, the samples with bentonite would were already drawn from the flow behaviour of mixed clay–sand
clearly provide a bigger challenge to clean the beater and the bowl, no samples, such as the significant difference in the flow between samples
matter if there was sand mixed or not. The pure bentonite, as well as the containing kaolinite and bentonite, where the kaolinite samples had a
70%, especially for consistencies above 0.7, would get stuck in the higher flow than the ones with bentonite. The samples with bentonite
bottom of the bowl, implying in great difficulties to remove the re- also showed the lowest flow values for soils containing 50% and 70% of
maining soil. This is why the CBFactor would make a difference in the clay, and not 100%. The main concern was the relation between flow
ending result. and clogging tendencies, if for the same consistency, soils that flow less
For the case of larger grains (Kmedium50, Fig. 9), there was a slight would have a higher tendency for clogging.
increase of the clogging values for consistencies above 0.8. The cleaning As observed in Fig. 10 all the samples with kaolinite had larger flow

12
D.G.G.d. Oliveira, et al. Tunnelling and Underground Space Technology 94 (2019) 103110

Fig. 9. Comparison between tests with kaolinite and two different sizes of sand.

Table 9
Resume of the data for the mixed samples with kaolinite, comparing with the clogging potential from Hollman and Thewes (2012, 2013), shown as H&T.

13
D.G.G.d. Oliveira, et al. Tunnelling and Underground Space Technology 94 (2019) 103110

Table 10
Resume of the data for the mixed samples with bentonite, comparing with the clogging potential from Hollman and Thewes (2012, 2013), shown as H&T.

Fig. 10. Flow results for all the clogging tests conducted without any additive, for kaolinite and bentonite samples.

values than the samples with bentonite. Frequently, the samples with 3.3. Tests with different beaters
the smallest clay fraction presented the highest flow (Ksfine30 and
Bsfine20), except for kaolinite and close to the liquid limit. The samples Tests with mixed soils were also conducted with different beaters to
with higher amount of bentonite (Bsfine50, 70 and 100) flow similar in investigate the influence of the metal or addition of a coating. Both
consistencies above 0.4, but close to their liquid limit samples with clogging parameters (λ0 and λF) were plotted. Again, λ0 presented
100% of bentonite flow more than samples with 50 and 70%. For the higher standard deviation values, as well as higher clogging (Figs. 11
kaolinite samples, samples with 50 and 100% of kaolinite had the and 12). Tests were conducted with kaolinite and fine sand (50% each;
lowest flow, except close to the liquid limit where only 50% of kaolinite Fig. 11) and with 20% of bentonite and 80% of fine sand (Fig. 12).
had the lowest flow. The results obtained with a stainless-steel beater, used also for all
There is not a clear a correlation between the flow pattern and the the other tests, were compared with three aluminium beaters, where
clogging results. For kaolinite samples, for example, the larger the clay two of them were covered with a sandblasted coating and a special red
fraction, the higher the clogging potential, which does not correlate coating3. Several machine cutterheads are sandblasted, justifying this
with the sample with 50% of clay having the lowest flow. For bentonite choice. The sandblasted beater was the one that produced the highest
it is hard to say because it should also be expected stronger clogging
potential with pure clay, whereas 50 and 70% of bentonite showed the
lowest flows. 3
The coating material cannot be revealed as it is currently under industrial
research. The supplier of the coated beater did not provide the material details.

14
D.G.G.d. Oliveira, et al. Tunnelling and Underground Space Technology 94 (2019) 103110

Fig. 11. Comparison between tests with Ksfine50 and different beaters.

clogging values. It was also more difficult to remove and clean the soil the soil to an ideal paste, such as decreasing temperature in the cut-
stuck at the MT for the sandblasted case, which justified the decrease of terhead or the torque values. Therefore, tests were conducted with foam
the CBFactor added value to 0.5. Additionally, the special coating did not and foam with two different polymers, looking to compare the differ-
produce a very significant reduction in clogging when compared to the ences between the chosen additives and its absence.
aluminium or stainless steel beater. These preliminary tests can be helpful as a guidance to choose the
most effective additive for a certain soil, before trying directly on-site,
3.4. Tests with additives which requires a larger quantity of the chemical product, especially for
large shields. Preliminary tests could also decrease the costs of ex-
For the case of clayey soils, EPB foam is not much required as it is cavation by avoiding the usage of a certain product that might not make
for sandy soils, as the soil is already around the ideal composition for an such a difference, or even the opposite, using a product that it does. So,
EPB operation (minimum of fines around 30%). The FER values are instead of the typical trial-and-error approach directly done inside the
lower for clayey soils than for sandy ones, with typical values of FER machine, simpler laboratory tests with a mixing machine and the device
around 5 to 8 for clayey soils, against 15 to 20, for sandy soils and the ATUR could provide insights about the behavior of different soils, when
amount of added water seems to be more efficient as conditioner for conditioned with a certain product. This combined testing methodology
cohesive soils than foam alone. However, adding polymers with anti- could certainly provide a small-scale and easy-to-perform experiment,
clay properties might be efficient to reduce the occurrence of clogging, avoiding trying different chemical products or even unorthodox solu-
and adding foam would have other advantages than merely bringing tions directly during a tunnel operation.

15
D.G.G.d. Oliveira, et al. Tunnelling and Underground Space Technology 94 (2019) 103110

Fig. 12. Comparison between tests with Bsfine20 and different beaters.

For the samples tested with additives, as it can be seen in Fig. 13 table cone where some material got stuck on it. Fig. 15 presents the
(kaolinite samples) and 14 (bentonite samples), the foam products had flow results, as well as three examples of the soil sample after the 40
not much influence on the clogging values. However, the foams with drops of the flow table. Product A was the one leaving a higher amount
anti-clay polymers produced slightly better results than those with of soil specimen stuck on the mortar cone for kaolinite and bentonite. B
none, except for kaolinite, at IC = 0.68. There, the Product A, only seemed to be more effective for kaolinite, and C, even though slightly
foam, resulted in more clogging than the test without additives. For the better than A, did not work so well for bentonite.
same case, the anti-clay products B and C, resulted in slightly fewer The manual squeezing stage provided valuable information, even
clogging effects. For bentonite as well, both products with anti-clay though on a more descriptive basis, on the effectiveness of a certain soil
polymers had the best results than only foam or without, though, with conditioning. As it can be observed in the Video 1, all the conditioned
very small differences. For low plastic soils, such as the kaolinite samples had spongy aspect with excessive air, which could imply as
samples, the addition of foam had a higher influence, in terms of flow having too much air inside the excavation chamber.
increase, than samples with high plasticity, such as the bentonite The tests conducted with foam and polymers were the most labor-
samples. ious ones, if compared to the tests with only water, due to several
The tests with the flow table contributed to some limited informa- reasons, such as achieving the target FER, the mixing procedure of the
tion regarding the influence of certain products on the clogging po- wetted soil with foam, and the choice of parameters for possible com-
tential. Figs. 13 and 14 in the bottom shows some pictures of the flow parison between soils, especially without using unrealistic soil

16
D.G.G.d. Oliveira, et al. Tunnelling and Underground Space Technology 94 (2019) 103110

Fig. 13. Results for Ksfine50 conditioned with water, products A, B, and C.

conditioning parameters. Testing with foam might be the case that most soil samples. However, the required amount of foam between kaolinite
requires comparison with the sample directly acquired on-site for later and bentonite was extremely different and bringing down the con-
comparison with the laboratory preliminary results. sistency to the final target IC resulted in too little foam for kaolinite, on
The difficulties in producing a foam with the target FER were one side, and extremely high and unrealistic FIR values for bentonite.
greater whenever the surfactant was changed from one product to an- The solution was to focus on applying realistic values, which are
other. It took several trials to achieve the desired FER with the used already used in tunnel projects for a cohesive mixed soil (Maidl et al.,
foam generator. There was also a limitation in producing FER of small 2015; Peixoto da Silva et al., 2017). The FIR was then always ap-
values, which is usually applied for cohesive soils (FER of approxi- proximately 80% and 120%, as indicated in Table 4.
mately 5–7). The lowest feasible values were approximately 9.5–10, The other difficulty was in mixing the wet soil homogeneously with
which allowed the testing procedure to be conducted within the time the foam, which for many cases, was practically impossible. This fact
before the foam was separated. was already pointed out by Peila et al. (2016), showing that only the
The preliminary idea was to have an initial consistency of the wet surface of lumps was really conditioned in the case of clayey soils. Thus,
soil and lower it with a foam (or foam plus anti-clay polymer), to force a complete mixing was also far from the reality in tunnel drives.
achieving a final consistency, which is 0.25 points lower than the initial Several options of mixing were tried: from a higher speed of machine
one, thereby always comparing the same consistency index among all mixing to the use of a spoon. This is one of the limitations of the testing

17
D.G.G.d. Oliveira, et al. Tunnelling and Underground Space Technology 94 (2019) 103110

Fig. 14. Results for Bsfine50 conditioned with only water, and with Products A, B, and C.

methodology when conducting tests with foam that still requires im- not very significant difference in the clogging (Fig. 16c, d, e, and f). For
provements. Obtaining a homogenous soil–foam mixture was most the second case, the beater was turned several times inside the bowl
difficult in the tests done with bentonite, which produced a variability with the liquid, waiting several minutes after each turn, which resulted
of the results especially for λ0, not so much for λF. Surely, tests with in a significant lower clogging effect (Fig. 16i, j, k, and l). This could
foam and cohesive soil shall include further trials and search a feasible suggest, for example, that in cases of major stops of the machine due to
solution, comparing the results with on-site samples. clogging occurrence, by filling the chamber with a highly concentrated
An experimental trial was performed to provide other alternative anti-clay polymer mixed with water, the removal of the clogged soil
testing possibilities for the effectiveness of anti-clay polymers, which could be facilitated. A third trial was conducted by lubricating the MT
can be visualised in Fig. 16. After a test with only water and with with an anti-clay polymer, which did not result in any clogging de-
clogging occurrence for the sample Kmedium50 (Figs. 9 and 16a, b, g, crease. It was even tested to add sugar to the sample and check if that
and h), an anti-clay polymer (Product B) was mixed with water and could affect clogging, as an additional experiment. Sugar increased
added to the MT with soil, without foam production. In the first test, clogging and deteriorated the entire cleaning process.
10% of water content was added, and in the second, 20%. The surfac- Moreover, tests conducted with foam in laboratory in atmospheric
tant concentration was 12% and 20%, respectively, instead of 3%. pressure are not representative of the real tunnelling conditions in an
For the first test, the liquid was spilled, and after waiting for 20 min, EPB excavation. In future, tests should be conducted on-site, with the
the liquid was mixed again, and the test was performed, resulting in a already conditioned muck, directly obtained from the conveyor belt,

18
D.G.G.d. Oliveira, et al. Tunnelling and Underground Space Technology 94 (2019) 103110

Fig. 15. Flow table results for the tests conducted with foam (Products A, B and C), for Ksfine50 (Graph 1) and Bsfine 50 (Graph 2).

which is also not entirely representative, as it also in atmospheric with the excavation parameters. A testing set can be installed inside an
pressure, however, the mixing conditions should be more realistic. EPB machine, close to the conveyor belt, where samples can be directly
acquired and tested. Another option is to place the samples from the
4. Conclusions conveyor belt inside closed recipients and bring them to be tested just
outside the tunnel entrance, still at the jobsite. Both options will surely
It was demonstrated here that by a simple yet complete laboratory present limitations. Parameters such as torque of the cutterhead and
routine, it is possible to characterise and classify cohesive samples, with screw conveyor can be analysed, as well as clogging issues that might
or without sand content, making assumptions and forecasting the soil occur during the tunnel drive. Even spending some time speaking with
behaviour regarding EPB excavation. It is also feasible to include and the machine operator and jobsite personal can be supportive of ac-
compare changes in behaviour of soils after conditioning, which will quiring valuable insights.
not only avoid future issues along the excavation, but also reduce the The conditioning of soil in an EPB operation, and consequently, its
costs of tunnel operation by avoiding the use of unnecessary products or flow behaviour, depends on many factors, and not solely on the soil and
adjusting the initial suggested quantities to the minimum required. conditioning characteristics, but also on the project logistics, such as
The methodology routine here proposed is a product of an extensive details of the available water, or the price and environmental regula-
testing campaign and has been modified to achieve better results with tions of the muck disposal, for example. Decreasing the cost of additives
an effective procedure. Undoubtedly, there is space for improvements, while avoiding clogging by increasing the water content of the muck
especially when tests are conducted with natural soils for tunnel pro- could create a major issue with the muck transportation and disposal,
jects, by crosschecking the results of preliminary laboratory investiga- possibly justifying the initial costs with conditioning polymers, for ex-
tion with the results obtained from tunnel drives. This procedure might ample. Or, in the other direction, the environmental regulations might
be time demanding due to restrictions related to tunnel projects, be the bigger limitation of which product could be used for con-
especially concerning the confidentiality aspects and involved claims. ditioning the soil.
Therefore, the main intention of sharing the work at this stage is to Additionally, the wide range of soil/rock types, together with the
invite others to try similar methodology, especially on jobsites. This can influence of the local groundwater, makes it a challenge to define any
increase the data obtained directly from the tunnel excavation, al- ideal state of the excavated material in a broader scenario. Including, as
lowing the improvement of this methodology. well, the details of any project, it is unrealistic to talk about ‘ideal
The intention of the testing routine is to provide a qualitative as- mucks’, or ‘ideal consistencies’, as this will vary from point to point and
sessment of the clogging and flow of soils to be excavated by an EPB from a tunnel project to another. The ‘ideal’ must be found for each
machine. It does not intend to recreate exactly the EPB excavation project and for that, a good characterisation methodology is essential.
process. Hopefully, with more data, it would be feasible to be more The greater the understanding about the natural soil and conditioned
precise in the variables used for this testing routine. Once this proce- muck before the excavation occurs, the easier it will be to adjust after-
dure is conducted on-site, then it could be feasible to provide a more wards any operation parameters, including making the best choices of
reliable number of drops and a cleaning factor, for example. soil conditioning. The most important aspect in a geological-geotechnical
Regarding clogging specifically, there are too many variables in- characterisation, and that should include the characterisation of the
fluencing its occurrence, such as the cutterhead design (opening ratio), muck, is the description. A proper and extensive description should al-
the machine operation, and even the geological characteristics of a ways be the focus of any geological–geotechnical characterisation, not
certain soil deposit (e.g. over consolidation, residual ground, marine only by a detailed summary of the soil and muck characteristics, but also
environment). All these variables will need to be accounted when by complete photographic and video documentation. The classification
correlating the test results with the excavation parameters. may be altered or adjusted in the future, as it is still under development,
There are some ideas of how to correlate the occurrence of clogging requiring an extensive validation with real tunnel drives. However, with

19
D.G.G.d. Oliveira, et al. Tunnelling and Underground Space Technology 94 (2019) 103110

Fig. 16. Alternative experimental trials with anti-clay polymer and Kmedium50 soil.

a precise and abundant description, the initial classifications can be ad- drives, doing a back analysis of the consistency of the excavated muck
justed afterwards, whenever necessary. and the excavation parameters.
The video compilation here provided is essential for further ob- One of the significant limitations of this methodology is conducting
servations about clogging and flow behaviour. It seemed that an ideal tests with conditioned samples. First, because it is performed in atmo-
muck for bentonite was not in the same IC range as the kaolinite sam- spheric pressure in an open space and ignores the effect of the pressure.
ples. The bentonitic samples would have an ideal muck range of ap- Second, because it is not realistic the mixing process in the lab as it
proximately 0.4 to 0.75, as defined by Maidl (1995); however, the occurs in the EPB machine. Also, mixing the samples in the lab was not
kaolinite samples, when mixed with sand, would have a lower IC ideal easy and many times would lead either into the destruction of the foam,
range, around the liquid limit. This is essential to be verified in tunnel or to a non-homogenous mixture.

20
D.G.G.d. Oliveira, et al. Tunnelling and Underground Space Technology 94 (2019) 103110

The lack of pressure ignores one important aspect for soil con- References
ditioning: for some foam parameters, the entrained air could separate
from the muck and would be restrained on top of the excavation Avunduk, E., Copur, H., 2019. Effect of clogging on EPB TBM Performance: a case study in
chamber, which could represent a challenge for certain scenarios of face Akfirat waste water tunnel, Turkey. Geotech. Geol. Eng. https://doi.org/10.1007/
s10706-019-00938-6.
support (i.e. loose soil layers on top). In laboratory tests, this char- Basmenj, A.K., Mirjavan, A., Ghafoori, M., Cheshomi, A., 2016. Assessment of the adhe-
acteristic is lost and what could seem as a successful conditioning in the sion potential of kaolinite and montmorillonite using a pull-out test device. Bull. Eng.
laboratory could result in an ineffective tunnel operation. The same Geol. Environ. 1–13. https://doi.org/10.1007/s10064-016-0921-3.
Borio, L., 2010. Soil conditioning for cohesionless soils. PhD Thesis. Politecnico di Torino,
goes for all the differences between tests conducted in atmospheric Department of Environment, Land and Infrastructure Engineering.
pressure, while tunnels are being driven with higher pressures, espe- BTS, 2005. Closed-face tunnelling machines and ground stability. British Tunnelling
cially in terms of foam behaviour, as already pointed out by Mooney Society, Institution of Civil Engineers, Thomas Telford, London.
Budach, C., 2011. Untersuchungen zum erweiterten Einsatz von Erddruckschilden in
et al. (2016). Similar can be said regarding the swelling of the clays, as grobkörnigem Lockergestein. Doctoral Thesis. Faculty of Civil and Environmental
the swelling in the laboratory test has happened in an unconfined space, Engineering of the Ruhr-University Bochum.
which is not the reality of natural soils, changing its characteristics. Comulada, M., Maidl, U., Peixoto da Silva, M.A., Aguiar, G., Ferreira, A., 2016.
Experiences gained in heterogeneous ground conditions at the twin tube EPB shield
That could justify why the results with the bentonite samples showed
tunnels in São Paulo metro line 5. Proceedings of the ITA-WTC 2016, San Francisco.
many times less clogging potential than the ones with kaolinite, because Della Valle, N., 2001. Boring through a rock-soil interface in Singapore. Proceedings of
the samples were tested unconfined. RETC 2001, pp. 633–645.
In the case of different geological scenarios, such as soils originating EFNARC, 2005. Specification and guidelines for the use of specialist products for me-
chanised tunnelling (TBM) in soft ground and hard rock. EFNARC, UK.
from a sedimentary deposit versus a residual ground, for example, it Feinendegen, M., Ziegler, M., Weh, M., Spagnoli, G., 2011. Clogging during EPB-tun-
will be necessary to identify the main differences in the properties of nelling: occurrence, classification and new manipulation methods. In: Proceedings of
the ground that would affect the particularities of the excavation with the ITA-WTC 2011, Helsinki, pp. 767–776.
Feinendegen, M., Ziegler, M., 2016. Verklebungen beim maschninellen Tunnelvortrieb:
an EPB machine, always questioning if a certain classification scheme Normative Regelungen, Klassifikation und Prognose. Bauingenieur 91, 350–359.
would entirely fit a specific scenario. Galli, M., Thewes, M., 2014. Investigations for the application of EPB shields in difficult.
Most of all, this methodology should be applied in real tunnel Geomech. Tunnel. 7 (1), 31–44. https://doi.org/10.1002/geot.201310030.
Galli, M., 2016. Rheological characterisation of Earth-Pressure-Balance (EPB) support
projects, before excavation, and, afterwards, as a back analysis of the medium composed of non-cohesive soils and foam. Doctoral Thesis. Faculty of Civil
assumptions made here, being then feasible to improve the proposed and Environmental Engineering of the Ruhr-University Bochum.
methodology. A flow table, for example, could even be installed at the Galli, M., Thewes, M., 2018. Rheological characterisation of foam-conditioned sands in
EPB tunneling. Int. J. Civil Eng. https://doi.org/10.1007/s40999-018-0316-x.
back of an EPB machine, where the muck can be directly checked and Gong, Q., Yin, L.M.H., Zhao, J., 2016. TBM tunnelling under adverse geological condi-
decisions about conditioning will not rely merely on the experience of tions: an overview. Tunnel. Undergr. Space Technol. 57, 4–17.
an individual feeling the muck. Hernandez, Y.A., Kang, C., Yi, Y., Bayat, A., 2018. Mechanical properties of clayey soil
relevant for clogging potential. Int. J. Geotech. Eng. 12 (6), 316–323. https://doi.
Finally, the greatest value of this laboratory routine is the easiness
org/10.1080/19386362.2017.1311086.
of reproduction in any other laboratory facility, making future com- Herrenknecht, M., Thewes, M., Budach, C., 2011. The development of earth pressure
parisons possible, and, perhaps, allowing a standardisation that can be shields: from the beginning to the present. Geomech. Tunnel. 4 (1), 11–35.
applied in tunnel projects worldwide, establishing the choices of soil Hollmann, F., Thewes, M., 2012. Evaluation of the tendency of clogging and separation of
fines on shield drives. Geomech. Tunnel. 5, 574–580.
conditioning and tunnel operation in more scientific-based conclusions. Hollmann, F., Thewes, M., 2013. Assessment method for clay clogging and disintegration
It is essential to mention that this research to achieve success will of fines in mechanised tunnelling. Tunnel. Undergr. Space Technol. 37, 96–106.
require a cooperation between different parties. Machine suppliers, https://doi.org/10.1016/j.tust.2013.03.010.
Hollmann, F., 2014. Bewertung von Boden und Fels auf Verklebungen und
contractors, designers, chemical suppliers and research institutes can Feinkornfreisetzung beim maschinellen Tunnelvortrieb. Doctoral Thesis. Faculty of
gain by achieving a methodology to characterise and evaluate soils to Civil and Environmental Engineering of the Ruhr-University Bochum.
be excavated with shield machines. The current state-of-the-art relies Huat, B.B.K., Toll, D.G., Prasad, A., 2012. Handbook of Tropical Residual Soils
Engineering. CRC Press, Taylor & Francis Group.
very much on the trial-and-error approach, with little being done to- Khabbazi, A., Mohammad, G., Azali, S.T., Cheshomi, A., 2017. Experimental and la-
wards a preliminary investigation followed by a back analysis, com- boratory assessment of clogging potential based on adhesion. Bull. Eng. Geol.
paring testing data with excavation parameters. This testing routine Environ. 1–13. https://doi.org/10.1007/s10064-017-1044-1.
Leinala, T., Grabinsky, M., Delmar, R., Collins, J.R., 2000. Effects of foam soil con-
needs to be improved by doing this exercise of preliminary assessment
ditioning on EPBM performance. North American Tunneling Boston, pp. 6-11.
and later back analysis, and indeed learning from the best testing de- Maidl, U., 1995. Erweiterung des Einsatzbereiches von Erddruckschilden durch
vice: the tunnel boring machine itself. Konditionierung mit Schaum. Dissertation Ruhr-Universität Bochum. Technisch-
Wissenschaftliche-Mitteilungen des Instituts für konstruktiven Ingenieurbau, p. 192.
Maidl, B., Herrenknecht, M., Maidl, U., Wehrmeyer, G., 2012. Mechanised Shield
Tunneling, second ed. Ernst & Sohn Verlag, Berlin.
Acknowledgements Maidl, U., Turolla Maia, C.H., Comulada, M., Mahfuz, A., Coutinho, A.A., 2015. First
experiences gained with the hybrid EPB technology in the Rio de Janeiro sands.
Proceedings of the ITA-WTC 2015, Croatia.
Support for this research was received from the Natural Science and Milatz, M., Grabe, J., Delle, M.M., 2019. Bewertung von Werklebungen und
Engineering Research Council (NSERC), Canada and funding from Werkzeugverschleiß im Lockergestein anhand von Modellversuchen. Geotecknik 2
Queen’s University. The authors would like to acknowledge the support doi: gete.201800017.
Mooney, M.A., Wu, Y., Mori, L., Bearce, R., Cha, M., 2016. Earth Pressure Balance TBM
of the Collaborative Research Centre SFB 837, ‘Interaction Modelling in soil conditioning: it’s about the pressure. In: Proceedings of the ITA-WTC 2016, San
Mechanized Tunnelling’, funded by the German Research Foundation Francisco, pp. 22–28.
(DFG). We are also grateful to the laboratory team at TLB, at Ruhr Oliveira, D.G.G., Diederichs, M.S., 2016. TBM interaction with soil-rock transitional
ground. Proceedings of the TAC 2016 Annual Conference, Ottawa.
University Bochum. Special thanks to Yvonne Ueberholz from Ruhr Oliveira, D.G.G., Diederichs, M.S., Rasmussen, L.L., Cecílio Jr, M.O., 2017a. Tropical
University Bochum, for the device assemblage, to Gustavo Aguiar, from residual soil data compilation as guidance for laboratory tests and EPB excavation
BabEng GmbH, for the discussions regarding soil conditioning, and process simulation. In: Proceedings of the 9th International Symposium on
Geotechnical Aspects of Underground Construction in Soft Ground IS – São Paulo, pp.
Herrenknecht AG, for the modified mixing tools and the tests with
185–194.
natural soils conducted in Singapore. Oliveira, D.G.G., Diederichs, M.S., Thewes, M., Freimann, S., Aguiar, G., 2017b. EPB
conditioning of mixed transitional ground: investigating preliminary aspects.
Proceedings of the ITA-WTC 2017, Bergen.
Oliveira, D.G.G., Thewes, M., Diederichs, M.S., Langmaack, L., 2018a. EPB tunnelling
Appendix A. Supplementary material through clay-sand mixed soils: proposed methodology for clogging evaluation.
Geomech. Tunnel. 4 11. https://doi.org/10.1002/geot.201800009.
Supplementary data to this article can be found online at https:// Oliveira, D.G.G., Thewes, M., Diederichs, M.S., Langmaack, L., 2018b. The consistency
index and its correlation with EPB excavation of mixed clay-sand soils. Geotech. Geol.
doi.org/10.1016/j.tust.2019.103110.

21
D.G.G.d. Oliveira, et al. Tunnelling and Underground Space Technology 94 (2019) 103110

Eng. J. 2018, 1–19. 2010. Investigation of adherence behaviour and related effects on different scales in
Oliveira, D.G.G., Diederichs, M.S., Thewes, M., Langmaack, L., 2018c. Proposed metho- mechanical tunnel driving. In: Proceedings of the Underground Construction Prague
dology for clogging evaluation in EPB machines. Proceedings of the ITA-WTC 2018, 2010, Transport and City Tunnels, pp. 692–699.
Dubai. Spagnoli, G., Feinendegen, M., Ziegler, M., 2019. Assessing the clogging potential of clay/
Oliveira, D.G.G., Diederichs, M.S., Thewes, M., 2019. EPB machine excavation of mixed additive mixtures by cone pull-out tests. Geomech. Tunnel. J. 12 (4), 362–371.
soils – laboratory characterisation. Geomech. Tunnel. J. 12 (4), 373–385. https://doi. https://doi.org/10.1002/geot.201900008.
org/10.1002/geot.201900014. Thewes, M., 1999. Adhäsion von Tonböden beim Tunnelvortrieb mit Flüssigkeitsschilden.
Peila, D., Oggeri, C., Borio, L., 2008. Influence of granulometry, time and temperature on Doctoral Thesis. Bericht aus Bodenmechanik und Grundbau, Bergische Universität
soil conditioning for EPBS applications. In: Proceedings of the ITA-WTC 2008, Agra, Wuppertal, Fachbereich Bauingenieurwesen.
pp. 881–891. Thewes, M., 2004. Schildvortrieb mit Flüssigkeits- oder Erddruckstützung in Bereichen
Peila, D., Picchio, A., Chieregato, A., 2013. Earth pressure balance tunnelling in rock mit gemischter Ortsbrust aus Fels und Lockergestein. Geotechnik 27, 214–219.
masses: laboratory feasibility study of the conditioning process. Tunnel. Undergr. Thewes, M., Budach, C., Bezuijen, A., 2012. Foam conditioning in EPB tunnelling.
Space Technol. 35, 55–66. Geotechnical Aspects of Underground Construction in Soft Ground – Viggiani (Ed.),
Peila, D., Picchio, A., Martinelli, D. Dal, Negro, E., 2016. Laboratory tests on soil con- pp. 127–135.
ditioning of clayey soil. Acta Geotech. 11, 1061–1074. https://doi.org/10.1007/ Vargas, M., 1980. Geotecnia do Terciário de São Paulo – Histórico. Aspectos geológicos e
s11440-015-0406-8. geotécnicos da bacia sedimentar de São Paulo, ABGE/SBG, pp. 37–46.
Peixoto da Silva, M.A.A., Katayama, L.T., Leyser, F.G., Aguiar, G., Ferreira, A.A., 2017. Viana da Fonseca, A., Topa Gomes, A., 2010. Project and construction of underground
Twin tunnels excavated in mixed face conditions. Proceedings of the ITA-WTC 2017, stations and tunnels (TBM and NATM) in heterogeneous masses for Metro do Porto.
Bergen. Excavations and tunnels in granite, pp. 133–172.
Quebaud, S., Sibai, M., Henry, J.P., 1998. Use of chemical foam for improvements in Zhao, J., Gong, Q.M., Eisensten, Z., 2007. Tunnelling through a frequently changing and
drilling by earth-pressure balanced shields in granular soils. Tunnel. Undergr. Space mixed ground: a case history in Singapore. Tunnel. Undergr. Space Technol. 22,
Technol. 13 (2), 173–180. https://doi.org/10.1016/S0886-7798(98)00045-5. 388–400.
Rahardjo, H., Aung, K.K., Leong, E.C., Rezaur, R.B., 2004. Characteristics of residual soils Zumsteg, R., Puzrin, A.M., 2012. Stickiness and adhesion of conditioned clay pastes.
in Singapore as formed by weathering. Eng. Geol. 73, 157–169. Tunnel. Undergr. Space Technol. 31, 86–96. https://doi.org/10.1680/geot.SIP13.P.
Sass, I., Burbaum, U., 2009. A method for assessing adhesion of clays to tunneling ma- 005.
chines. Bull. Eng. Geol. Environ. 68, 27–34. Zumsteg, R., Plötze, M., Puzrin, A., 2013. Reduction of the clogging potential of clays:
Shirlaw, N., 2015. Pressurized TBM tunnelling in mixed face conditions resulting from new chemical applications and novel quantification approaches. Géotechnique 6 (4),
tropical weathering of igneous rock. In: Proceedings of the International Conference 276–286. https://doi.org/10.1680/geot.SIP13.P.005.
on Tunnel Boring Machines in Difficult Grounds (TBM DIGs), Singapore, pp. 1–28. Zumsteg, R., Puzrin, A.M., Anagnostou, G., 2016. Effects of slurry on stickiness of ex-
Shirlaw, N., 2016. Pressurized TBM tunnelling in mixed-face conditions resulting from cavated clays and clogging of equipment in fluid supported excavations. Tunnell.
tropical weathering of igneous rock. Tunnell. Undergr. Space Technol. 57, 225–240. Undergr. Space Technol. 58, 197–208. https://doi.org/10.1016/j.tust.2016.05.006.
Spagnoli, G., Fernández-Steeger, T., Azzam, R., Feinendegen, M., Neher, H.P., Stanjek, H.,

22