High Performance Ground Prediction Ahead of TBMs - The NeTTUN System "TULIPS"
High Performance Ground Prediction Ahead of TBMs - The NeTTUN System "TULIPS"
High Performance Ground Prediction Ahead of TBMs - The NeTTUN System "TULIPS"
Ahead of TBMs
The NeTTUN System TULIPS
Thomas Camus
NFM Technologies
1
The French Tunnel Association (AFTES) set up a Working Group (nr. 24) to prepare and issue a
recommendation concerning forward probing ahead of TBMs. This group comprised construction
companies, civil engineering firms, tunnel designers, geotechnicians and geophysicists, and TBM
manufacturers. The recommendation (Robert et al. 2014) includes a survey of ground prediction
techniques and of existing systems available on the market.
The changes in the geotechnical conditions appear as contrasts in the ground, which indirect techniques
then have to detect. Three approaches can be used relying on the measurement or analysis of:
The propagation of the stress field generated by a source (mechanical waves): detection is
based on differences in the propagation velocity within the various media through which the
waves travel;
The propagation of an electromagnetic field (electromagnetic waves): the detection process
relies on the differences in the electrical permittivity of the media;
The shape of the electrical and/or magnetic field: this is generally obtained by measuring the
electrical resistivity, assuming that the resistivity varies with the geotechnical conditions.
Without entering into details, each of these methods exhibit limitations that make it insufficient
when a large spectrum of situations need to be detected (boulders, anthropic elements, faults, presence
of water, etc.) over a large variety of ground conditions (clay, sand, gravel, rock, etc.).
2
TULIPS CONCEPTS
One aspect that is common to all the current prediction (and also to recognition) techniques is that such
systems always do provide output information, but the quality of this information is unknown ; even if
the quality is known it is often quite low; these two factors obviously affect the confidence level of
tunnelling contractors. In other domains involving recognition from sensor data (for example in
character recognition and medical diagnosis), this problem has been overcome by implementing several
approaches as complementary as possible in parallel and by merging the corresponding outputs in a
voting algorithm, providing a much more reliable result.
This is the strategy implemented in TULIPS, the ground prediction system developed through the
NeTTUN project. TULIPS stands for Tunnel Lookahead Imaging Prediction System.
a) Theoretical optimal scan b) Circular scan using the TBM c) Seismic scan at different cutter head
cutter head rotation positions
The architecture of TULIPS is designed to be open and flexible, offering built-in scalability vs. TBM
diameter and type, and expandability through the potential addition of other complementary
subsystems. The concept is patented under FR2995627.
3
TULIPS IMPLEMENTATION
Implementation constraints
The most efficient approach consists in sending and receiving the signals (mechanical or electromagnetic
waves) directly from the TBM cutter head into the ground ahead.
This creates the main constraints on the system: the number of active devices that are required to
cover the desired volume ahead of the excavation face and produce good quality results is dictated by
the laws of physics. This number needs to be balanced with the size of these devices and the available
area on the cutter head, on which the number and size of disc cutters, drag bits, scrapers, etc., are
imposed by the ground characteristics. Obviously the TULIPS active devices should be made as small as
possible.
The next constraint is due to the environment: the cutter head is the most severely exposed part of
the TBM in terms of operating conditions (presence of water, vibrations and shocks). The consequence
is that part of the system electronics also have to be integrated in the cutter head in order to reduce the
number of connections to be made through the slip ring.
First implementation
The first implementation of TULIPS is focused on soft ground, as this is the domain in which there are
few available solutions, specifically in terms of seismic systems. It comprises two sensing subsystems:
radar and seismic. The combination of these techniques exhibits the highest potential in terms of
situations covered as clearly demonstrated in the AFTES recommendation.
The focus on soft ground led to the use of seismic shear waves (S-waves) rather than compression
waves (P-waves). Recent advances in very shallow seismic surveying have shown that shear waves are
superior to compression waves in soft soils: shear waves are not sensitive to the fluid contents, and the
relative contrasts in seismic shear-wave properties are much larger than the ones in compression-wave
properties. Under soft soil conditions, shear waves have been used as a soil-indicator for decades in the
Netherlands. It is also known that minute fractions of gas within the ground change the properties of
compression waves dramatically, including the damping, thereby making P-wave imaging very difficult.
Since shear waves are not sensitive to the fluid contents of the ground, shear-wave imaging is not
affected by this situation.
Radar is known to have lower performance in soil than in rock, because soil conducts electricity.
Radar wave attenuation increases with ground conductivity; wet clay is a worst case in which the radar
performance is severely limited. In order to get the best of the technique, focus was given on increasing
the penetration length as much as possible: this can be achieved with the cost of a reduced resolution.
In order to cover the whole range of detection capability, the radar subsystem therefore uses two
complementary types of antennas: one with the maximum range (depth of penetration), one with high
resolution (short range).
4
Tx Rx
Vibrator(s) Geophones
Antenna(s) Antenna(s)
Electromagnetic Seismic
Radar Subsystem
Long and Short range S-waves
(WP3) (WP4)
Storage
Future expansion
The mid-term plan is to incorporate extensions to the above system. Two approaches are considered:
Add a P-wave seismic subsystem and use both S-waves and P-waves, which will provide the
mechanical characteristics of the media;
Add an electric resistivity measurement subsystem, which will offer improved performance, for
instance in situations involving the presence of water.
Operation mode
The TULIPS measurements are performed while the TBM is not excavating, when erecting the segmental
lining. When measuring, the seismic devices (and also the electric devices) have to be pushed into the
ground, which means they protrude from the cutter head front face. However, they need to be
protected while the cutter head rotates, namely during excavation and while moving from one image
capture position to the next. Therefore these devices are actuated and moved from passive to active
position and vice versa.
The radar antennas need to remain in contact with the ground during scanning; as they are
mounted flush on the cutter head front face, they can be left in place during excavation.
DETAILED DESCRIPTION
Radar
The radar subsystem is composed of several modules as shown in Figure 3:
Antennas
Passive devices that convert electric signals into electromagnetic waves and vice-versa
5
Front-end electronics
Electronic devices connected directly to the antenna. The transmitter converts a trigger signal
into a high-energy pulse of controlled shape and amplitude. The receiver can be seen as a very
low noise preamplifier.
Digital antenna drivers
Generates the triggering signal for the transmitter (and also for the receiver)
Converts into digital the analog signal received from the antenna preamplifier
Transmits the data to the radar control computer
Radar Control computer
Triggers the radar transmission and reception
Performs digital signal processing of the radar data
Ensures communication with the other TULIPS components
Monitors the system operation and diagnoses failures
Encoder
Pulse Tx Antenna
Power supply generator Long range
DC/DC
Slip ring
Rx Antenna
Digital Preamplifier
Long range
Antenna
Driver
Ethernet
Timing
Data circuit Pulse Tx Antenna
Switch
Seismic
The seismic subsystem comprises the following components:
Source(s)
Active devices that convert electrical signals into mechanical motion (e.g. as a loudspeaker)
Sensors
Active devices that convert mechanical motion into electrical signals (e.g. as a microphone).
6
Seismograph
Receives the signals from all sensors and preprocesses them, converts them into digital data,
stores them, and transfers the digital data to the processing computer
Seismic controller
Manages the sequence of operations
Seismic data processor
Applies a sequence of processing steps to the digital seismic data including the Full Wave
Inversion algorithm to deliver the seismic image
Seismic
Encoder
controller
Power supply
Slip ring
DC/DC
Motion Seismic
Controller source(s)
Data Seismic
Ethernet Seismograph
processing sensor
computer
Hydraulics
Manifold
Rotary Seismic
joint sensor
Data fusion
The Data Fusion module is currently being designed. Work is based on:
the results of numerical simulations: data was produced for both the radar and seismic
subsystems, based on models of seven representative geotechnical scenarios;
the results of the field tests conducted in the Netherlands, where both radar and seismic
subsystems were used to image the same five scenarios (see Tests and Results).
The Data Fusion module takes input from the radar and seismics subsystems, together with any
background data available (such as the geological profile) and produces a most probable interpretation
of these data. Fusion also uses data from the previous rings.
The various sets of TULIPS data then need to be displayed. The visualisation system is also in the
design phase at the date this paper is written. It is based on the following concepts:
7
several visualisation approaches are needed in order to perform a full analysis;
visualisation should be dynamic, for instance allow changing the data displayed (on/off), the
time window, the spatial scope, and the visualisation method;
the default visualisation should be as informative as possible and the mechanisms for switching
visualisations should be kept simple.
Several features will be implemented such as superimposition of 2D images to produce a 3D
approximation, volume rendering, panoramic imaging, animation (controlling the timepoint of the
visualised data and images) and a combination of these using selective time based transparency.
Examples are shown in Figure 5.
Distance away from the cutter head
Detected features
8
Installation of anthropogenic structures
Figure 6. Construction and use of test site Left: installation of anthropogenic structures (concrete
and metal pipes); right: radar acquisition
Each scenario was used in sequence by moving the equipment to each corresponding central
position, from where scanning was done by rotating the equipment around this centre. For the seismic
subsystem, the 8-metre line of source and sensors was positioned at three angles (0, 60, 120), the
mid-point being the centre of rotation: this actually also generates data for 180, 240, and 300. Once
data was captured from the five scenarios, TBM advance was simulated by removing a 1 m layer of
ground and the measurements were repeated: this was done six times, representing six strokes (rings)
and seven complete sets of data.
The data are being processed but establishing the results is not complete at the date this paper is
written. Figure 7 gives the outputs for both the radar and seismic subsystems, for the karst simulated by
a block of polystyrene of dimensions 4 x 1 x 0.5m (L x W x H). The karst is clearly detected by both
subsystems.
+6.0m
TULIPS Top of
the hill
Top of
the block
-7.5m
0m
Bottom of
-2.0m
the pit
0 line
Figure 7. Preliminary test results Left: test configuration; centre: seismogram corresponding to the
simulated karst; right: 2D migrated radar image, the detected object being shown in orange colour
The first results are promising. As expected, radar penetration is shorter than that of seismic
S-waves in the sandy clay that composes the test site. In such ground, radar penetration is in the order
of 2-3 m while seismic waves penetrate up to 8 m as witnessed by the reflections given by the bottom of
the pit. The preliminary test on the limestone slope in Switzerland shows that the seismic subsystem
actually works well up to 40 m. The detailed TULIPS test results will be presented during the congress.
9
Future tests
The TULIPS radar will be installed on the 11.2 m diameter TBM that will excavate the Lyon-Turin (LTF)
exploratory tunnel, starting early 2016. This 9 km tunnel will be bored starting from the Saint-Martin-la-
Porte adit and will later on be included in the complete 57 km twin-tube base tunnel.
Due to the geological conditions (rock) only the radar system will be implemented. This will serve as
a test on both the radar performance (mainly, penetration and resolution) and its resistance to the
harsh environment of the TBM. The cutter head has been designed to house two long range antennas
and four high resolution antennas (short range), together with the driving electronics as shown in
Figure 3. The radiation pattern of the long range antenna allows covering the whole tunnel section at
two metres away from the cutter head except for a blind central zone of one metre in diameter.
Integration tests, including radar, seismic, and data fusion will be performed, both replaying the
already acquired data and from new scans to be organized.
PERSPECTIVES
The NeTTUN TULIPS team is working on parallel fronts as follows:
finalizing the seismic processing software, including the automatic Full Wave Inversion
(Bharadwaj et al. 2015), which is a challenge in itself;
developing the data fusion algorithms and several prototypes of data visualization. These will be
presented to a panel of end-users in order to choose the most appropriate solutions;
redesigning the control electronics of the seismic source so as to reduce its size and make it an
industrial product, adapted to the TBM environment.
The next step and final challenge consists in integrating a full-blown TULIPS on a TBM.
ACKNOWLEDMENTS
The NeTTUN project is funded by the European Commission Seventh Framework Programme for
Research, Technological Development and Demonstration under Grant Agreement 280712.
I am grateful to the NeTTUN partners participating in the TULIPS development: Technical University
of Delft and MI-Partners (Netherlands); IDS and Sial.Tec (Italy); University of Limoges, CISTEME, Systra,
and my colleagues at NFM (France); University of Leeds (United Kingdom); Geo2X (Switzerland).
REFERENCES
Bharadwaj, P., Mulder, W.A., Drijkoningen, G., and Reijnen, R. 2015. Looking ahead of a tunnel boring
machine with 2-D SH full waveform inversion. In Proceedings of the 77th EAGE Conference &
Exhibition, Madrid, June 1-4, 2015.
Camus, T. 2015. TULIPS An innovative multi-sensor ground prediction system for TBMs. In Proceedings
of the International Conference on Tunnel Boring Machines in Difficult Grounds, Singapore,
November 18-20, 2015.
Goel, R. K. 2014. Tunnel Boring Machines in the Himalyan tunnels. In Indorock 2014 Volume I
Proceedings of the 5th Indian Rock Conference, New Delhi, November 12-14.
Manacorda, G. et al. 2015. The NeTTUN Project: development of a ground prediction sensor. In
Proceedings of the 8th IWAGPR, Florence, July 7-10, 2015.
Robert, A. et al. 2014. Forward probing ahead of tunnel boring machines. TES. 242: 132169.
10