Вы находитесь на странице: 1из 10

Discrete Element Method to Predict Soft Ground Cutterhead Performance

Glenn Mongillo
Caterpillar Tunneling Canada Corp., Toronto, ON, Canada
Mustafa Alsaleh
Caterpillar Inc., Mossville, IL, USA
ABSTRACT: Soft ground TBM Cutterhead design has historically been driven by iterative designs based on
empirical data, observations over various projects and industry rules of thumb. The ability to analyze the
performance of a cutterhead has been limited to studies conducted in-situ during TBM operations. With the
use of Caterpillar proprietary Discrete Element Method (DEM) software, Rocks3DTD, critical operational
characteristics can be analyzed much earlier on, during the design phase, and allow for comparison of such
key parameters as required torque, advance rate, material flow paths, and effectiveness of loading
mechanisms. Rocks3DTD is a proprietary discrete element code, which allows for time-elapsed particle
interactions with structures, and considers particle size, shape, soil cohesive properties and density. The
resulting data allows for optimization of cutterhead opening ratio, opening geometry, tool positions, and
muck removal from the plenum. This paper will review simulations performed to date and examine how
discrete element analysis can be applied to cutterhead performance optimization.
INTRODUCTION
Discrete Element Method (DEM), also called a distinct element method, is the family of numerical
methods used to compute the motion of a large number of particles with micro-scale size and above. In recent
years DEM has become more mature and widely accepted as a robust method to treat engineering problems
in granular and discontinuous materials, especially in granular material flows, pharmaceutical applications,
rock and powder mechanics. The various branches of the DEM family are the distinct element method
proposed by Cundall in 1971, and the generalized discrete element method proposed by Hocking, Williams
and Mustoe in 1985. The theoretical basis of the method was established by Sir Isaac Newton in 1697.
Williams, Hocking, and Mustoe in 1985 showed that DEM could be viewed as a generalized finite element
method. Its application to geomechanics problems is described in the book, Numerical Modeling in Rock
Mechanics, by Pande, G., Beer, G. and Williams, J.R..
In geomaterials or rock-mechanics, DEM treats material particles as individual rigid bodies, where their
motion is governed Newtons law of motion. A cohesion model is usually introduced into the formulations to
represent the cohesive forces that exist in fine-grained materials and cementation that exists in rocks. In a
DEM simulation, the model is initially generated using a certain shape of particles (most commonly used are
spherical particles) with a pre-determined particle size distribution. The fact that the actual particles are not
spherical in shape made developers consider clusters of spheres to represent the shape irregularities.
Moreover, single spherical particles have been also used along with virtual shape representation. Developers
usually prefer using spherical particles to reduce the computational cost associated with the contact detection
processes within a DEM model. Once the neighboring search process is done, force calculations will take
place followed by an integration stage, which will be employed to compute the incremental change of the
particle position and velocity for the next time step calculation. These processes will continue for the entire
simulation period.
Caterpillar Inc. researchers have been developing and using a DEM code for the past fifteen years; the
code is called Rocks3DTD. It uses a very computationally efficient contact detection algorithm and can deal
with any particle shape specified by the user. The contact frictional and normal forces are computed using the
well-known Hertzian contact model for cohesion-less materials. Additional algorithms are implemented to
treat cohesive-like bonds when modeling fine-grained materials and rocks. The cohesion is modeled using
cohesive pillars that bond neighboring particles together; this pillar can be strained until a strain threshold is

reached and then the bond is broken. Recently, Cosserat rotation has been added to the code kinematics along
with particle shape indices described by Pande, Beer and Willams. This additional degree of freedom enabled
the code to capture more of the micro-structural properties for the material being modeled (angularity, size,
sphericity etc.). While Rocks3DTD is used to model particulate force responses and material flows, it is
capable of linking to full machine models using in-house built codes for modeling machine dynamics, tireground interaction, machine hydraulics, etc.
Rocks3DTD is also capable of interacting with tracked-type tractors to pass proper forces to the machine
through the track shoes. The machine tools can be treated as either rigid and/or flexible bodies. The code had
been parallelized to take advantage of mutli-threaded processors. It has been benchmarked against other
commercial codes; to-date, Rocks3DTD usability, simulation speeds and accuracy have been found more
encouraging. As acknowledged by many researchers, it is very challenging to obtain DEM model parameters
that best represent real materials, especially when dealing with fine-grained materials. Rocks3DTD developers
have been successful in defining an engineered procedure to map these micro quantities to some material
physical and macro quantities. Both small scale laboratory testing and full machine testing are being utilized
to develop micro-macro parameter mapping functions. The particle size for instance, a very important DEM
parameter, must be chosen carefully. Choosing the particle size for a given model will always have a great
deal of trade-off between simulation accuracy and computational cost. Special attention had been given to
this matter; the particle size distribution is established for a given model in a way to ensure highest simulation
accuracy at the lowest computational cost. This way, the model parameters (micro-mechanical properties
such as friction, stiffness, etc.) can be linked to macro properties to achieve better physical representation.
Rocks3DTD can predict the dynamic forces and flows of different discrete systems geometries under
dynamic loading. As mentioned earlier, the contact parameters are micromechanical parameters that are very
difficult to physically measure, and very challenging to evaluate due to the fact that it is almost impossible to
represent the actual shape and size of real materials. A real material is very complex to mimic in terms of
shape, size, and size distributions.
DESIGN VARIABLES WHICH INFLUENCE SOFT GROUND CUTTERHEAD PERFORMANCE
The design of soft ground TBM cutterheads typically includes the control of several parameters that
drive the overall look and performance of a TBM cutterhead. Over the years, these structures have seen a
great deal of evolution in the attempt to improve overall TBM performance.
The intention of using DEM software was to develop a tool to analyze the effects of changes to some or
all of the major parameters which are known to affect TBM cutterhead performance. By isolating specific
parameters (such as percentage opening or face opening geometry) and keeping other parameters constant
(such as tool position, and rpm) we can determine which factors may have the most significant impact or if
indeed there is an impact.
Cutterhead percentage opening
Within the Earth Pressure Balance TBM industry, a cutterhead face opening of 30% has been the typical
standard requirement for many project specifications. The value of 30% has been based primarily on base
empirical performance data and experience. This parameter is comparable to the flow coefficient of a fluid
through a restriction, however the complex interactions within the soil itself and between the soil and
cutterhead make this assumption an over simplification. While care must be taken to ensure that sufficient
material can enter the cutterhead chamber at a fast enough rate to meet the TBM designed advance rate, too
great of a percentage opening can be detrimental as well.
In some soil types such as clay, larger openings can in fact lead to clogging and increase the requirement
for torque and yield less efficient muck flow into the chamber. In flowing silts it has also been observed that
a reduction of cutterhead percentage opening has allowed for greater TBM thrust offsets while controlling
TBM advance rate, with an observed improvement in TBM steerability in difficult silt ground conditions.
This is especially important when the primary steering method is based on the creation of a net resultant
moment with the TBM propulsion system. Generally in soft ground and EPM projects a suitable cutterhead
structure is still required to provide adequate face support, which will limit the maximum allowable
percentage face opening.

Opening distribution with respect to radius from the center


Perhaps more important than the total percentage opening of the cutterhead face, is the distribution of
the openings. Well distributed openings on the cutterhead face can make a considerable difference in muck
loading. Openings positioned where greater volume of excavated material is expected can support greater
TBM advance rates, less muck clogging and more consistent loading of cutting tools. Openings at the center
of the cutterhead are important where low material velocities exist and less mixing occurs. These points tend
to support triangular opening as a better choice to promote a more balanced opening ratio with respect to the
radius on the face. However, in certain bolder conditions the opening shape may allow for more irregular
shaped boulders to enter the chamber and ultimately slow TBM operations. In such cases rectangular
openings provide better protection from this occurring and additional rectangular openings may be used to
increase the total percentage opening.
Spacing, quantity and position of tools
Each soft ground tool represents a source of drag, whether performing a cutting action or simply being
engaged with the ground during cutterhead rotation. The combination of all the drag forces during cutterhead
rotation results in the required operational torque of the cutterhead. Too few cutting tools can result in more
revolutions required to excavate sufficient material for the TBM to advance. Too many cutting tools can
increase the required cutterhead torque by increasing the overall drag and promoting blockages of material
that do not enter the chamber. Optimizing the quantity of soft ground cutting tools is a key factor to achieving
the required TBM advance rates and keeping the required cutterhead torque to an achievable level.
An equally important parameter is the spacing of tools on the cutterhead face. Typical four spoke
cutterhead designs within the 4m to 6m range allow for staggered tool spacing on each spoke where
successive tools along the increasing cutting radii are positioned on alternating spokes. This allows for a
staggered cutting action and space for material to follow between tools on the same spoke. If the spacing
between successive tools on the same spoke does not permit the flow of excavated material between them,
material may build up and eventually render the tools ineffective on subsequent revolutions of the cutterhead,
creating more drag, and thus increasing cutterhead torque required and reducing efficiency.
In some cases, specific tool locations on the outer rim or gauge area of the cutterhead may include
additional cutting tools on the same path. This generally is expected to reduce the wear on any one cutting
tool by allowing the cutting action to be shared equally by all the tools on the same path. Excessive tools on
the gauge area can lead to a significant increase in required cutterhead torque since drag forces at the furthest
radius from the center have a greater effect due to the larger moment arm. As well this increase in tools may
lead to material blockages and inefficient loading of material into the chamber. This would also have a
contributing factor in the overall performance of the cutterhead.
Obstructions to flow of material into and within the chamber
Scraper collecting tools are designed to gather excavated material from the tunnel face and direct it into
the cutterhead chamber. This action is every bit as important as the cutting action itself. A TBMs
performance would be short lived if the material from the tunnel face was broken off but not collected and
transported away to allow for continuous cutting action. The scraper tools are critical to this action. They
must be positioned such that once the material is cut by the cutting tools the scraper tools must direct the
material into an adjacent opening. Cutting tools positioned too far from scraper collection tools may result in
excessive travel time for muck, the muck adhering to the cutterhead face, reconsolidation with the tunnel
face, or regrinding or muck resulting in excessive wear of the structure.
In some cases grizzly bars or opening restrictions are used in soft ground cutterhead designs to limit the
size and/or shape of material entering the chamber. This can be especially useful in geology containing
boulders. Boulders pose a danger of damage to the cutterhead chamber, and blockage of screw conveyor
augers in Earth Pressure Balance TBMs. While beneficial in certain geologies, they may actually hinder
material ingress into the chamber in other geologies. Where clays and cohesive soil conditions exists it is
often recommended to remove such opening restrictions. Where geological conditions are varying and
unpredictable, this may not always be possible, and therefore could lead to unexpected performance of the
cutterhead when such changing geology is encountered.

SUMMARY OF SIMULATIONS PERFORMED


Methodology
For the purposes of this study, it was decided to use two past cutterhead design types of a common size
and vary only the shape of the opening (and consequently the overall percentage opening). (see Figure 1)
This was done to limit the number of variable that would affect the results of the simulation. By applying a
constant cutterhead rotational speed and axial thrust to both models onto the same discrete particle pile, the
resulting outputs, torque and advance rate, would then be directly compared as to the effectiveness of one
cutterhead design over the other. Certain conditions needed to be defined to provide a realistic simulation and
much of the initial work was focused on determining the properties of the discrete element pile as far as
density, particle size, particle quantity and cohesion.

4.2 m [167] Dia. 30% Opening


No Door Design

4.2 [167] Dia. 26.5% Opening


With Door Design

Figure 1. Cutterhead models used for simulation


Study #1 - Initial Feasibility
A first pass feasibility simulation was performed with the lowest level of complexity to allow for
efficient computational time while still demonstrating the capabilities of the code for this TBM cutterhead
application. (see Figure 2) Spherical elements with a common size of 150 mm in diameter were chosen, while
TBM input parameters were selected to ensure some visual element motion was noticeable and the TBM
model was moved as expected. The results of this simulation were by no means realistic as far as normal
TBM operations are concerned, but they do indicate that the dynamic motion of the elements and the
interaction with the rotating cutterhead structure can be further refined to better simulate actual TBM
conditions. This simulation was useful in debugging the initial TBM model and particle parameters.
One key observation is that the control of the element size is important in this application. Elements
should be comparable to the average size of the cutterhead tools performing the cutting action. In this case,
static ripper tools have been used with a typical cross section of 75mm x 150mm. Using elements larger than
the minimum size of the cutting tools presents a concern with the scale of interactions between the elements
and the structure. This may affect the applied force on cutting tools, the observed material flow paths and the
overall TBM performance expected from the simulation. For subsequent simulations it was determined that
element size should be set less than 75mm in diameter.

Figure 2. Initial DEM Simulation with simplified TBM model


Study #2 - Particle and Cohesion Refinement
The second simulation focused on providing a more realistic representation of the TBM cutterhead
application with an analysis of critical performance measures of required torque and advance rate. To
facilitate a comparison between the two cutterhead designs, constant values for cutterhead rotational speed
and applied thrust were used. As well, in an effort to validate the soil pile, varying degrees of particle
cohesion were used to account for high and low soil stiffness. The particle size was set to 70 mm within a pile
consisting of 400,000 particles. (see Figure 3)
The simulations performed as part of this second study were as follows:
1.
2.
3.
4.

High stiffness pile & 30% opening No Door cutterhead


High stiffness pile & 26.5% opening With Door cutterhead
Low stiffness pile & 30% opening No Door cutterhead
Low stiffness pile & 26.5% opening With Door cutterhead

The resulting animations provided a more realistic representation of soil loading into the chamber
compared to the initial feasibility study. The reduced element size allowed for particles to collect and travel
with the rotation of the cutterhead, similar to real world conditions.
Some validation of the soil pile was possible with the results of the simulation. The output data
confirmed that with higher soil stiffness (i.e., greater element cohesion) the required torque to rotate the
cutterhead was greater and the resulting TBM advance rate was lower. This is consistent with operational
expectations of TBMs. Between the two soil stiffness models an average of 16% to 20% greater torque was
required for the respective cutterhead design. (see Figure 4)
Additionally, there was a demonstrated difference in the required torque between the two cutterhead
designs used when the soil stiffness remained constant. The 30% opening No Door cutterhead design
required on average 7% to 11% more torque within either stiffness of soil compared to the 26.5% opening
With Door cutterhead design. This correlated well with the increase in payload or excavated material

within the same timeframe. The larger percentage opening design excavated more material, had a greater
advance rate, and hence required more torque.

Figure 3. Refined DEM simulation with 70mm particle size


2000

1750
Avg. Torque - No Door Model
(High Stiffness)

Avg. Torque - With Door Model


(High Stiffness)

Avg. Torque - No Door


Model

Avg. Torque - With Door Model


(Low Stiffness)

Torque (T-m)

1500

1250

1000

750

500
0

10

15

20

25

30

Time (s)

Figure 4. Study #2 torque comparison of both cutterhead models with high and low particle stiffness

The comparison of the output values was done at the steady state once the initial peak torque was
overcome. In this case all simulations during this phase of the study reached a steady state close to the 10
second point of the 30 second simulation. The peak torque observed would be analogous to real world inertial
effects and the initial penetration of the cutterhead into the soil face.
However, the absolute values of the output torque and TBM advance rate were well beyond real world
expectations for this size of TBM studied. Simulated required torque values were on average 10X greater
than typically expected. Similarly, the TBM advance rates were on average 8X to 10X greater than expected.
With respect to torque it was clear that with a decrease in particle cohesion a decreased amount of torque was
calculated. Therefore, to bring subsequent simulations closer to reality a significant reduction in element
cohesion was required. This, however, did not address the concern of the TBM advance rate, where it was
observed to increase with decreasing particle cohesion. Both of these TBM performance parameters would
seem to be at odds in this simulation when compared to real world applications. In typical soft ground TBM
applications it is reasonable to expect that with a decrease in soil stiffness there would exist less resistance to
the cutterhead motion in both rotational and axial velocity and as such would require less cutterhead torque
and the resulting TBM advance rate would be greater. In this particular simulation the TBM advance rate is
already far too high. This indicates that expected torque values may be achieved with a reduction in particle
cohesion, but an additional modification must be made to correct the already high advance rate. Some
improvement in advance rate may be possible even with a reduction in overall particle cohesion by the
control of some or all of the parameters which govern the Rocks3D cohesion model. Since particle cohesion
is controlled via cohesion pillars that bond elements, variations of the pillars modulus of elasticity, strain
limit, damage threshold, or pillar size may help to improve the high observed cutter torque and TBM advance
rate. This alone may not be enough to reduce the advance rate and further improvement to the TBM model
may be required.
Study #3 Cohesion Refinement and Comparison to Actual Data
The third study was focused on refining the element properties to better simulate expected soil
conditions similar to a silty clayey fine grain material. This was required to achieve a more realistic value for
TBM cutterhead torque and hopefully TBM advance rate, thus validating the element pile and determining a
good set of parameters to simulate a typical soil pile for a TBM application.
The DEM cohesion model has five parameters that control the mechanical properties of the cohesive
pillar between particles to represent either a fine grained cohesive material or a rock mass. These parameters
were refined in the final simulation to better represent a cohesive material, such as silty clay. The two main
parameters refined in this simulation were the cohesion pillar modulus of elasticity and the cohesion pillar
damage parameter. These basically allow for a reduction in the cohesive pillar strength to decrease the
required cutterhead torque and an increase in the particle to particle stiffness to try to reduce the TBM
advance rate.
The adjustments made to the particle cohesion did allow for the overall cutterhead torque to be reduced
to reasonable values within the range of expectations in both the No Door and With Door models. (see
Figure 5) There was however an increase in the TBM advance rate with the lower values of particle cohesion.
(See table 1) This would suggest that the soil parameters may be closer to actual conditions for this
application, but other TBM based conditions exist that would otherwise lower the overall advance rate, which
are not included in any of the simulations to date. These external factors may include; external friction
between the TBM structure and the soil, pressure at the TBM face in the case of Earth Pressure Balance
TBMs, interactions of soil in the cutterhead chamber, and controlled extraction of material from the
cutterhead chamber. With the current simplified TBM model these secondary factors are not possible to
analysis; nonetheless the particle cohesion parameters determined thus far have produced torque values of a
more realistic nature, which may allow for further comparison of the two different cutterhead designs.
This final simulation still suggested that the No Door model required more torque compared to the With
Door model. The resultant torque curve over time did, however, differ compared to the second study
performed. There was a general increase in torque over the simulation time with both design models
compared to the previous results where the torque values leveled off towards the end of the simulation. A
logarithmic trend line was fitted to both torque curves for comparison. At the end of the simulation the No

Door model still displayed approximately a 6.5% higher torque value. This is still comparable to the results
of study #2.

Table 1. Advance rate summary


Avg. Advance Rate
(mm/min)
1

26.5% With Door - Higher Stiffness

795

26.5% With Door - Lower Stiffness


26.5% With Door - Study #3

1279
1972 2

30% No Door - Higher Stiffness

999

30% No Door - Lower Stiffness

1524
2133 2

1
1

30% No Door - Study #3


1.

Avg. Advance Rate taken in Study #2 between 10s and 30s.

2.

Avg. Advance Rate taken in Study #3 between 5s and 50s.

300
275
No Door Model Torque
(Operational Trendline)

Torque (T-m)

250
225

With Door Model Torque


(Operational Trendline)

200
175
150
Historical Calculated Torque Range
= (1.5 to 2.5)*Diameter^3

125
100
0

10

15

20

25

30

35

40

45

50

Time (s)

Figure 5. Study #3 torque comparison of both cutterhead models with lower particle stiffness
The increasing torque over time suggested that the mass of particle accumulation in the chamber was
more significant in this simulation compared to the previous. Study #2 used much greater particle cohesion

pillar strength and as such the cutting action was the predominant source of the required torque to keep the
cutterhead rotating. However in study #3 the particle cohesion was decreased to bring the simulated torque
closer to a value that would typically be observed in actual TBM operation. As such the cutting action
between the cutterhead and the particle pile was less significant, where as the accumulation of particles in the
chamber accounted for a steady increase in the required torque. This value would typically level off over
time, once the chamber was full of excavated material.
CONCLUSIONS
In all the simulations performed to date the No Door model with 30% percentage opening required
greater torque, had a higher advance rate and subsequently allowed for greater flow of material into the
cutterhead chamber when compared to the With Door model having only 26.5% percentage opening. This
would suggest that the DEM software did demonstrate that the parameter of cutterhead percentage opening is
in fact critical to the performance of soft ground TBM cutterheads. Furthermore, if the true benefit of greater
TBM advance rates is to be realized, the higher percentage opening parameter is not without its compromise
with the requirement for greater torque.
In addition, the DEM software did allow for some validation of the soil pile by refinement of the particle
cohesion parameters to better predict the required torque of a soft ground cutterhead. It was observed that
greater particle cohesion required greater torque, which is consistent with TBM operational conditions and
expectations. This suggests that with an accurate representation of given geological conditions for a tunnel
project, the DEM software would be able to predict the required cutterhead torque with a good degree of
accuracy.
Finally, to better predict other TBM operational conditions such as advance rate, more refinement of the
model is required; along with the addition of external factors that were not considered in the studies
performed. These may include:
A more detailed TBM model with sealed cutterhead chamber
A controlled method to extract muck from the chamber, such as a screw conveyor
Addition of frictional forces between the TBM and surrounding geology
Face pressure to simulate Earth Pressure Balance conditions
An upper threshold limit on advance rate based on physical TBM specifications
It is the authors contention, given the studies performed thus far, that the Discrete Element Method is a
viable tool to predict the performance of soft ground TBM cutterheads and that with on-going improvements
better simulations will advance the predictive nature of DEM to handle tunnel boring operations in the virtual
world.
ACKNOWLEDGEMENTS
The authors would like to acknowledge Ben Conners and Paul Yassa of Caterpillar Inc. for their
assistance in the preparation of 3D models and the collection of past operational TBM data.
REFERENCES
Cundall, P.A. 1971. A computer model for simulating progressive large scale movements in blocky rock systems.
Proc. Symp. Int. Soc. Rock Mech.,Nancy, pap. II-8.
Williams, J.R., Hocking, G., and Mustoe, G.G.W. 1985. The Theoretical Basis of the Discrete Element Method.
NUMETA 1985, Numerical Methods of Engineering, Theory and Applications, A.A. Balkema, Rotterdam,
January 1985
Williams, J.R., Hocking G., et al. 1985. The Theoretical Basis of the Discrete Element Method. NUMETA '85
Conference, Swansea

Pande, G., Beer, G. and Williams, J.R. 1990. Numerical Methods in Rock Mechanics, John Wiley and Sons,
Chichester, England.
SELECTED ADDITIONAL READINGS
Burger, W., 2007. Design Principles for Soft Ground Cutterheads. In Rapid Excavation and Tunneling
Conference Proceedings, Littleton CO.:SME

Вам также может понравиться