Академический Документы
Профессиональный Документы
Культура Документы
a r t i c l e i n f o a b s t r a c t
Article history: Design of waste rock barriers forming safety berms for haul trucks requires knowledge of complex in-
Received 31 July 2018 teractions which cannot readily be tested by physical means. An advanced numerical model based on
Received in revised form non-smooth multi-domain mechanics is presented together with model calibration using limited full-
21 October 2018
scale experimental data. Waste rock is represented by spherical particles with rolling resistance, and
Accepted 26 November 2018
Available online 31 December 2018
an ultra-class haul truck is represented by a rigid multibody system interconnected with mechanical
joints. The model components are first calibrated and then the calibrated model is used for simulating
various collision scenarios with different approach conditions and safety berm geometries. Numerical
Keywords:
Discrete element method (DEM)
predictions indicate that the width of the berm is most critical for efficiently stopping a runaway truck.
Multibody dynamics (MBD) The model can also predict if a certain berm geometry is capable of stopping a runaway truck. Results are
Non-smooth multi-domain dynamics summarised in a series of diagrams intended for use as design guidelines by practitioners and engineers.
Granular materials Ó 2019 Institute of Rock and Soil Mechanics, Chinese Academy of Sciences. Production and hosting by
Ultra-class haul truck Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/
Safety berm licenses/by-nc-nd/4.0/).
Surface mining
1. Introduction a haulage vehicle upon impact, the height of a safety berm should
be at least equal to the rolling radius of the vehicle’s tyre. In 1999,
Waste rock barriers, such as safety berms or windrows, are a the U.S. Department of Labour’s Mine Safety and Health Adminis-
crucial component of any surface mining environment. They are tration released the Haul road inspection handbook (MSHA, 1999). It
used around dump points and along haul roads to protect heavy states that safety berms must be at least as high as the mid-axle
mining vehicles from running over an edge. However, incidents height of the largest piece of equipment using the haul road. The
happen on a regular basis (see Fig. 1). The behaviour of waste rock material of the berms should be firm and the inside slope of the
barriers is poorly understood, with rational methods for predicting berm must be steep. In addition, the handbook clearly states that
their stability upon vehicle impact being absent. It is common sense equipment operators should know that due to the large size and
that the current design guidelines might not apply to the new weight of mining equipment, the typical axle height berms cannot
generation of ultra-class haul trucks, but it is unclear how safety be relied on, by themselves, to completely stop a vehicle except at
berms need to be designed for efficiently stopping a runaway haul low speeds. Berms much larger than axle height would be required
truck. Hence, the need for a more rigorous design approach is to completely restrain a vehicle for the full range of possible con-
emerging within the mining industry (NSW Mine Safety, 2017; ditions of speed and impact. For this reason, larger than typical
Queensland Government, 2010). berms should be used in areas where it is reasonable to expect
The design of safety berms is currently based on rules of thumb, more adverse conditions, such as where equipment would have
and the height of the berm is considered as the main factor in the more speed or would contact the berm more head-on. However,
design. The rules of thumb were established by Kaufman and Ault the handbook does not indicate how much larger these berms
(1977). They suggested that in order to efficiently stop or redirect should be. Later, other guidelines were developed based on the
handbook and some provided updated dimensions for safety
berms. Tannant and Regensburg (2001) focused on haul road
* Corresponding author.
E-mail addresses: klaus.thoeni@newcastle.edu.au (K. Thoeni), martin.servin@
design for haul truck payload capacities greater than 200 t. They
umu.se (M. Servin), scott.sloan@newcastle.edu.au (S.W. Sloan), anna.giacomini@ recommended increasing the height of safety berms from 2 m for
newcastle.edu.au (A. Giacomini). 240 t trucks to 2.9 m for 360 t trucks. Chapter 10.6 of the SME
1
Deceased. Mining Engineering Handbook (Thompson, 2011a) instead suggests
Peer review under responsibility of Institute of Rock and Soil Mechanics, that for large haul trucks, the berm height should be at least 66% of
Chinese Academy of Sciences.
https://doi.org/10.1016/j.jrmge.2018.11.005
1674-7755 Ó 2019 Institute of Rock and Soil Mechanics, Chinese Academy of Sciences. Production and hosting by Elsevier B.V. This is an open access article under the CC BY-
NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
660 K. Thoeni et al. / Journal of Rock Mechanics and Geotechnical Engineering 11 (2019) 659e675
2. Numerical framework
the truck wheel diameter. More recently, dimensions for high risk
area safety berms have been suggested, where for a 360 t truck (e.g. 2.1. Non-smooth multi-domain dynamics
CAT 797F), the height should be 4.1 m (Reynoldson, 2015). Never-
theless, all the guidelines are based on experience and not on In the classical DEM, or smooth discrete element method
geotechnical design, and thus there is still a need to develop a more (SDEM), the particles are assumed to be locally deformable, i.e. they
rigorous design method. can overlap. Contacts between particles are modelled using penalty
In order to provide more accurate design guidelines, it is crucial forces described by viscoelastic springs obeying the Coulomb fric-
to better understand the behaviour of waste rock barriers upon tion law (e.g. Cundall and Strack, 1979). The forces generally depend
collision with a haul truck under various approach conditions. on the contact overlap, relative velocity and other reaction forces.
Experimental testing can provide valuable information, but it is The equations of motion are solved using an explicit integration
generally limited to specific scenarios such as backward motion and scheme. The use of springs requires a simulation time step smaller
low velocities up to around 10 km/h (Giacomini and Thoeni, 2015). than the elastic response time for numerical stability. Hence, the
Scenarios where the haul truck travels at higher speeds cannot be time steppffiffiffiffiffiffiffiffiffiffiffi
is limited by the contact elasticity timescale,
considered in an experiment due to safety issues and costs (i.e. ffi
hSDEM < p m=kn , with particle mass m and normal spring stiffness
damage to the truck). Hence, numerical modelling is a good kn . This means that an infinitesimal small time step would be
K. Thoeni et al. / Journal of Rock Mechanics and Geotechnical Engineering 11 (2019) 659e675 661
needed in the limit of infinitely stiff springs. This limits the smooth
approach to simulations of material with finite elasticity as smaller g
t lt þ Gt v ¼ 0
ðaÞ ðaÞT ðaÞ (5)
time steps imply more computational time. The name smooth refers lt mt Gn ln
to the motion of the particles. The trajectories of the particles are
twice differentiable with respect to time and velocity jumps are not
g
considered. In the non-smooth discrete element method (NDEM), r lr þ Gr v ¼ 0
ðaÞ ðaÞT ðaÞ (6)
particles are assumed to be perfectly rigid and the elastic contact lr mr r Gn ln
response between particles is generally neglected. The contact
forces are modelled using impact laws (i.e. impulses) and kinematic εj lj þ hj g j þ sj Gj v ¼ 0 (7)
constraints for unilateral contacts and Coulomb friction. These
equations are written as non-differentiable relations involving ve- Eq. (3) is the NewtoneEuler equation of motion for rigid bodies
locity jumps, percussions and thresholds (Jean, 1999; Moreau, with external (smooth) forces fext and constraint forces GT l with
1999; Servin et al., 2014). This means that indefinitely stiff mate- Lagrange multipliers l and a Jacobian G. The constraint forces are
rials can be simulated. The NDEM can be seen as a time-implicit divided into normal (n), tangential (t), rolling (r) and articulated
version of the SDEM. The contact forces are calculated implicitly, and possibly motorised joints (j). M is the generalised mass matrix
which includes solving for unilateral constraint forces. The effort of and v is the generalised velocity vector. Eqs. (4) and (5) are the
solving these contact problems is larger than in the smooth Signorini-Coulomb conditions with constraint regularisation and
approach, but the time step is much larger. The time step in non- stabilisation terms εn , sn and gt . The symbol t denotes compli-
smooth simulations hNDEM is limited by mentary condition (Murty, 1988). Each individual contact is
and
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
hNDEM < ε2dmin =gacc (2)
_ ¼ f
M v_ þ Mv ext þ Gn ln þ Gt lt þ Gr lr þ Gj lj
T T T T
(3)
Fig. 2. Representation of the waste rock material: (a) in situ and numerical PSD, (b)
realistic in situ safety berm, and (c) example of generated safety berm using spherical
0 εn ln þ g n þ sn Gn vtln 0 (4) discrete elements (particles are coloured according to their radius).
662 K. Thoeni et al. / Journal of Rock Mechanics and Geotechnical Engineering 11 (2019) 659e675
indexed with a. With εn ¼ sn ¼ 0, Eq. (4) states that bodies should nonlinear Hertzian contact law, or linear spring and dashpot, from
be separated or have zero overlap, i.e. g n ðxÞ 0 with gn being the conventional SDEM to the constraint based NDEM. The NDEM
gap function, and so the normal force should be non-cohesive, i.e. contact constraint forces can be understood as the time-average of
ln 0. With gt ¼ 0, Eq. (5) states that contacts should have zero SDEM contact forces:
relative slide velocity, i.e. Gt v ¼ 0, provided that the friction forces
3=2 !
remain bounded by the Coulomb friction law with friction coeffi- GTn ln )fn ¼ kn gn þ cgn 1=2 g_ n n (8)
cient mt. Similarly, Eq. (6) constrains the relative rotation of con-
tacting bodies, provided that the constraint torque does not exceed Z
the rolling resistance law with rolling resistance coefficient mr and !
GTt lt )ft ¼ proj ! kt u t dt (9)
radius r. The constraint forces GTj lj arise for articulated rigid bodies mt f n
jointed with kinematic links and motors represented with the
Z
generic constraint of Eq. (7). With εj ; sj ¼ 0 and hj ¼ 1, it becomes !
an ideal holonomic constraint, i.e. gj(x) ¼ 0. For εj, hj ¼ 0 and sj ¼ 1, it GTr lr )sr ¼ proj ! kr w r dt (10)
mr f n
becomes an ideal Pfaffian constraint, Gj x_ ¼ 0. With εj ; hj ; sj s0, it
can represent a generic constraint with compliance and damping. !
where n is the contact normal; gn is the normal overlap (corre-
The Lagrange multipliers l become an auxiliary variable to solve !
sponds to gap function of the individual contact); and u t and w r
!
for, in addition to position and velocity. The regularisation and are the tangential relative velocity and relative angular velocity at
stabilisation terms, ε and g, introduce compliance and dissipation the contact point, respectively; and kt and kr are the corresponding
in motion orthogonal to the constraint manifold. In the absence of contact stiffnesses. Assuming Hertzian contact and spherical par-
the inequality and complementarity conditions, the regularised ticles, the normal spring stiffness kn and damping coefficient c are
constraints may be viewed as Legendre transforms of a potential
and Rayleigh dissipation function of the form U ε ðxÞ ¼ 2ε 1 g T g and pffiffiffiffiffiffi.h i
kn ¼ E 2d 3 1 m2 (11)
Rg ðx; vÞ ¼ 21gðGvÞT ðGvÞ (Bornemann and Schütte, 1997), where g
represents any of the constraints in Eqs. (4)e(7). This enables .
modelling of arbitrarily stiff elastic and viscous interactions in c ¼ 4 1 m2 ð1 2mÞn 15Em2 (12)
terms of constraint forces with direct mapping between the regu-
larisation and stabilisation terms to physical material parameters.
where E is the Young’s modulus, m is the Poisson’s ratio, n is the
This is applied to map the stiffness and damping terms from the 1 1
material viscosity constant, and d ¼ ðd1 ½a þ d½b Þ is the effective
diameter from the interaction between spheres with diameter d1 ½a
and d1½b of particle a and b, respectively. The contacts are divided
into impacts and continuous contacts, depending on the magnitude
of the incoming relative normal velocities Gn v . The impulse
transfer through the system should satisfy the Newton impact law
ðnÞ ðnÞ
Gn vþ ¼ eGn v , with coefficient of restitution e for the im-
pacts (n), as well as preserve all remaining constraints (m) on ve-
ðmÞ
locity level, Gn vþ ¼ 0.
The implicit numerical time integration scheme is based on the
SPOOK stepper (Lacoursière, 2007) derived from discrete varia-
tional principle for the augmented system (x;v; l; l_ ) and applying a
semi-implicit discretisation. The stepper is linearly stable and
O ðh2 Þ accurate for constraint violations. Stepping the system
positions and velocities, ðxi ; vi Þ/ðxiþ1 ; viþ1 Þ, from time ti to
tiþ1 ¼ ti þ h involves solving a mixed complementarity problem
(Murty, 1988). This is solved using a hybrid solver where a projected
Gauss-Seidel solver is applied for the NDEM subsystem and a direct
solver is used for articulated rigid multibodies (Lindmark and
Servin, 2018).
The waste rock material with which the safety berm is con-
structed of is modelled using spherical particles. Each particle is
represented by a dynamic discrete element obeying the equations
of motion introduced in Section 2.1. The particle shape effect (in
reality the particles are not spherical) is taken into account by
applying a rolling resistance model. Generally, more realistic sha-
ped particles could also be taken into account. However, the
simulation time would increase considerably.
The particle size distribution (PSD) of the waste rock material
can be approximated and only particles in a certain range are
considered in order to keep the computational costs at a manage-
able level. The particle size of real waste rock material generally
Fig. 3. CAT 797F: (a) picture of the real haul truck, and (b) the geometric model seen varies considerably and ranges from sub-millimetre particles (silt
from front, side, back and bottom. and sand) to large boulders in excess of 2 m (Simmons and
K. Thoeni et al. / Journal of Rock Mechanics and Geotechnical Engineering 11 (2019) 659e675 663
Table 1 Table 5
Key dimensions of the CAT 797F haul truck model (CAT, 2018). Validation of weight distribution.
Wheelbase Overall Overall rear Tyre Tyre Truck Force (kN) Distribution front:rear (%)
length (m) length (m) tyre width (m) diameter (m) width (m)
Front Rear Total Model CAT (2018)
7.195 15.08 9.529 4.02 1.47
Unloaded truck 1188 1461 2649 45:55 47.2:52.8
Loaded truck 2123 4056 6180 34:66 33.3:66.7
Table 2
The bodies of the vehicle model.
elements. It should be noted that in reality the shape (i.e. triangle or
Name Mass (inertial) Number DoFs trapezoid) of the in situ safety berm is quite irregular whereas its
Chassis 180 t 1 6 numerical representation is rather regular. This is because in the
Empty or loaded body 52.9 t or 412.9 t 1 6 numerical model, the safety berm was generated within a smooth
Rear bogie 1t 1 6 mould with a specific geometry.
Steering kingpin 0.28 t 2 6
Wheel hub 0.5 t 6 6
Tyre 5.3 t 6 6
Engine shaft 10 kg m2 1 1
2.3. Representation of the haul truck
Main shaft 10 kg m2 1 1
Rear shaft 10 kg m2 1 1 The ultra-class haul truck considered in this study is a CAT 797F
Total 269.3 t or 629.3 t 20 105 (Fig. 3a). It was also one of the models considered in the experi-
mental study presented in Giacomini and Thoeni (2015). A CAD
model (CAT, 2018) was adapted to represent the haul truck in the
Table 3 numerical simulation. The truck model was parameterised based
The constraints of the vehicle model.
on the technical specifications from the manufacturer. The key di-
Name Constraint type Number DoFs removed mensions used in the numerical model are listed in Table 1. The
Chassis-body lock Lock6D 1 6 total mass of the truck without load is assumed to be 269.3 t and
Front suspension Prism þ lock1D 2 4þ1 the weight distribution according to the technical specifications is
Rear suspension Prism þ lock1D 2 4þ1 45% on the front axle and 55% on the rear axle. A fully-loaded truck
Wheel axle Hinge 6 5 is assumed to weigh 629.3 t. The load, i.e. the material inside the
Tyre-hub lock Locktyre 6 6
body, is not modelled explicitly. Instead, the load is considered by
Torque converter 1D Pfaffian 1 1
Gear box 1D Pfaffian 1 1 giving the truck body an additional mass of 360 t such that the total
Differential 1D Pfaffian 1 1 mass is 629.3 t, and by displacing its centre of mass to achieve an
Total 22 95 approximate weight distribution with 34% on the front axle and
66% on the rear axle. The maximum speed according to the speci-
fications is 67.6 km/h for forward driving and 11.9 km/h for
Table 4 reversing.
Parameters of the suspension and the tyre model. The geometric model of the truck is depicted in Fig. 3b. It con-
Name Elasticity Damping sists of 20 rigid bodies. Some minor parts such as handrail and
Front suspension 6
26 10 N/m 0.2 106 N s/m
ladder are not included for simplicity. The main bodies and their
Rear suspension 52 106 N/m 0.4 106 N s/m masses are listed in Table 2. The three drivetrain shafts have only a
Radial tyre deflection mode 27 106 N/m 0.6 106 N s/m single (rotational) degree of freedom (DoF). The other 17 bodies
Lateral tyre deflection mode 135 106 N/m 2 106 N s/m have 6 DoFs. This gives a total of 105 DoFs for the whole truck
Bending tyre deflection mode 135 106 N m/rad 2 106 N m s/rad
model.
Torsion tyre deflection mode 108 106 N m/rad 2 106 N m s/rad
The rigid bodies are interconnected by kinematic constraints
which restrict their relative movement. Constraint compliance and
damping are introduced and parameterised to physical viscoelas-
McManus, 2004; Fityus et al., 2008). In the following, a typical fresh ticity. The constraints and the number of DoFs they remove are
waste rock material is considered, e.g. the Fresh Spoil Site A in listed in Table 3. The constraints include standard joints like hinges,
Giacomini and Thoeni (2015). A comparison of the PSD of the locks and prisms. The suspension joint is a combination of a pris-
original material with the PSD used in the numerical model is matic joint and a one-dimensional (1D) lock constraint which re-
shown in Fig. 2a. The in situ PSD ranges from 0.08 mm to 1200 mm stricts the motion along the prismatic sliding axis. The drivetrain is
whereas the PSD in the numerical model ranges from 60 to modelled using Pfaffian constraints on the rotational velocities u,
700 mm. Fig. 2b and c shows, respectively, a realistic in situ safety i.e. Gu ¼ 0. The resulting total number of DoFs for the truck model
berm and its numerical representation using spherical discrete after applying all constraints is 10.
Fig. 4. Illustration of the two-body tyre model (wheel hub and tyre are treated as two distinct rigid bodies): (a) reference configuration, and (b) general deformation. The general
deformation is a combination of four modes of deformation: (c) torsional, (d) radial, (e) lateral and (f) bending deflection of the tyre relative to the hub.
664 K. Thoeni et al. / Journal of Rock Mechanics and Geotechnical Engineering 11 (2019) 659e675
Table 7
Contact parameters for the interactions of the spherical particles with other bodies.
Fig. 7. Comparison of the experimental results with the numerical predictions of the final calibrated model.
K. Thoeni et al. / Journal of Rock Mechanics and Geotechnical Engineering 11 (2019) 659e675 667
5. Data analysis
Berm In total, more than 250 simulations were carried out and ana-
lysed. Each configuration (i.e. combination of berm geometry and
scenario) was run several times by gradually increasing the
approach velocity of the truck. The aim was to evaluate the effec-
S4 tiveness of the specific berm geometry in stopping the truck and to
identify a critical velocity at which the truck would pass the berm
Approach direction (i.e. the berm would fail). Two passing criteria, C1 and C2 (see
Fig. 10), are used to identify two corresponding critical velocities.
Skidding direction Both criteria are based on one of the wheels of the truck passing a
0.5·v
specific line (i.e. exceeding a specific horizontal wheel position
v which is measured from the centre of the wheels). For the trian-
gular berm, C1 and C2 indicate the critical velocity before passing
the centreline and the end of the berm, respectively (Fig. 10a). For
the trapezoidal berm, C1 and C2 indicate the critical velocity before
(d) passing the top width and the end of the berm, respectively
Fig. 9. Scenarios with corresponding approach direction: (a) scenario S1 reversing, (b)
(Fig. 10b). Criterion C1 is more conservative than criterion C2. When
scenario S2 forward, (c) scenario S3 forward at shallow approach angle, and (d) sce- using criterion C1, the truck is definitely stopped by the safety
nario S4 skidding. berm. When using criterion C2, it is very likely that the truck is
stopped by the safety berm, but the likelihood generally depends
668 K. Thoeni et al. / Journal of Rock Mechanics and Geotechnical Engineering 11 (2019) 659e675
Table 9
Configurations used in the considered scenarios.
Scenario Berm geometry Approach direction Approach angle, b ( ) Approach velocity, v (km/h) Skidding ratio
S1 Tri Reversing 90 5 to 30 e
S2 Tri, Tra1, Tra2 Forward 90 10 to 70 e
S3 Tri, Tra1, Tra2 Forward 10, 30 10 to 70 e
S4 Tri, Tra1, Tra2 Skidding 0 10 to 70 0.5
Fig. 11. Example of typical data obtained from the simulations: (a) velocities and positions of the wheels of the truck in one simulation (the orange star indicates that the front left
wheel has passed the line of criterion C1), and (b) summary of stopping positions (condensed into one abscissa) for various approach velocities and different safety berm geometries
(corresponding point to (a) is marked with the orange star). Orange and red circles in (b) indicate determined interpolated critical velocities according to criteria C1 and C2,
respectively.
K. Thoeni et al. / Journal of Rock Mechanics and Geotechnical Engineering 11 (2019) 659e675 669
Fig. 12. Berm height vs. critical reversing velocity for scenario S1. vmax indicates the
maximum modelled velocity.
Fig. 14. Berm height vs. critical forward velocity for scenario S2.
haul truck is initialised with a velocity of 30 km/h and approaching
the berm in forward motion at an angle of 30 . The truck stops just
after the front left wheel passes the line for criterion C1. The
stopping positions are then summarised in one graph for the
Fig. 13. Typical screenshots of scenario S1 Tri, H ¼ 2 m, v ¼ 10 km/h: (a) just before Fig. 15. Typical screenshots of scenario S Tra1, H ¼ 2 m, B ¼ 1 m, v ¼ 40 km/h: (a) just
contact with the safety berm, (b) during the impact, and (c) configuration at the before contact with the safety berm, (b) during the impact, and (c) final configuration
maximum horizontal wheel position (v ¼ 0). See Appendix online version for video. after the truck has stopped (v ¼ 0). Appendix online version for video.
670 K. Thoeni et al. / Journal of Rock Mechanics and Geotechnical Engineering 11 (2019) 659e675
different velocities (Fig. 11b). Only velocity increments of 5 km/h specific scenario, only C1 should be applied as the truck is
are considered in the simulations and hence, the values for critical approaching the berm with its rear wheels at more or less 90 . If
velocities are linearly interpolated in-between if the stopping point loaded, the rear axle carries most of the load and any passing of the
is not close enough to the line of criteria C1 or C2 (orange and red berm increases the risk of a slope failure. In addition, in a tip-head
circles in Fig. 11b). The calculated critical velocities are then re- dumping scenario, a truck is extremely unlikely to be in a runaway
ported in a separate graph for all safety berm geometries for each condition unless it is reversing downhill or deliberately reverses at
scenario. These graphs summarise the main results and can be used the maximum speed.
to develop design guidelines. They are discussed in Section 6 below. Fig. 12 shows that a typical safety berm with 2 m height is
effective in stopping a CAT 797F travelling at its maximum
6. Results and discussion reversing speed of 11.9 km/h. The predicted critical approach
velocity for criterion C1 is 13.9 km/h. It is worth noting that
6.1. Scenario S1 Giacomini and Thoeni (2014) observed a critical velocity between
12 km/h and 15 km/h in small-scale experimental tests. The nu-
Scenario S1 represents a typical tip-head dumping scenario merical prediction is within this range and gives some confidence
where the haul truck is reversing into the safety berm. Approach that the complex numerical model provides reasonable pre-
velocities from 5 km/h to 30 km/h and triangular safety berms dictions. The predicted critical velocity is very close to the
with 2 m, 3 m and 4 m height are considered. Only triangular maximum reversing speed. Hence, increasing the berm height
berms are considered since trapezoidal berms prevent free dump would be beneficial. Simulations with berm heights of 3 m and
tray movement and are difficult to construct for scenario S1. The 4 m predict critical velocities for C1 of 19 km/h and 25.9 km/h,
main results are summarised in Fig. 12. Although results for criteria respectively. It can also be noted that the increase in velocity is
C1 and C2 are reported, it should be pointed out that for this almost linear. Fig. 13 shows typical screenshots of simulations of
scenario S1. In this particular case, the truck is reversing into a
triangular berm of 2 m height at a speed of 10 km/h. It can be
seen that the truck is stopped before the rear wheels pass the
centre of the berm.
6.2. Scenario S2
berm geometries are considered in this scenario. Fig. 14 sum- stopping the truck increases. As an example, for criterion C2, the
marises the main results. Similar to scenario S1, an almost linear critical velocity increases from 30.8 km/h (Tri) to 34.8 km/h and
increase of velocity is observed with increasing height for all 44.2 km/h for the trapezoidal berms with a top width B of 1 m
geometries and both criteria, C1 and C2. When comparing the (Tra1) and 2 m (Tra2), respectively. The effect of an increase of
results of the triangular berm to those of scenario S1 in more the top berm width from 1 m to 2 m can also be seen in Figs. 15
detail, it can clearly be seen that the values for the critical ve- and 16. Clearly, the truck in Fig. 15c stops about one tyre
locity are decreased by about 30%. This can be related to the diameter after the one in Fig. 16c. Finally, Fig. 14 also indicates
ploughing effect of the front body of the truck which is also that a trapezoidal berm would perform very similarly to a
evident in Figs. 15b and 16b. This effect was also observed in triangular berm that is 1 m higher. This is related to the fact
small-scale experimental tests (Giacomini and Thoeni, 2014) and that the base width of the two geometries is very similar. As an
can also be seen in Fig. 1a. This suggests that berms along haul example, a 3 m high triangular berm has the same critical ve-
roads should be designed differently to berms at dump points. locity as that of a 2 m high trapezoidal berm with top width
Fig. 14 also shows that by increasing the width of the berm, i.e. 2 m. The base width of the two berms is 7.2 m and 6.8 m,
using trapezoidal berms, the performance of the berm in respectively (see Table 8).
Fig. 18. Typical screenshots of scenario S3 Tri, H ¼ 3 m, v ¼ 40 km/h, b ¼ 10 : (a) just Fig. 19. Typical screenshots of scenario S3 Tri, H ¼ 3 m, v ¼ 40 km/h, b ¼ 30 : (a) just
before contact with the safety berm, (b) during the impact, and (c) final configuration before contact with the safety berm, (b) during the impact, and (c) final configuration
after the truck has stopped (v ¼ 0). See Appendix online version for video. after the truck has stopped (v ¼ 0). See Appendix online version for video.
672 K. Thoeni et al. / Journal of Rock Mechanics and Geotechnical Engineering 11 (2019) 659e675
Fig. 20. Berm height vs. critical forward velocity for scenario S4. vmax indicates the
maximum modelled velocity.
6.3. Scenario S3
In scenario S3, the truck runs into the safety berm in a forward
motion but at a shallow approach angle. Two approach angles b of
10 and 30 are considered, and the corresponding results are
summarised in Fig. 17a and b. It can immediately be seen that the
critical velocity is different for the two approach angles. Whereas a
2 m high berm can be effective in stopping a truck approaching at
10 (the truck is running almost parallel to the berm), this is not the
case for a truck approaching at 30 . As an example, according to
criterion C2, the critical velocity for a 2 m high trapezoidal berm
with top width 2 m (Tra2) is reduced by more than 50% from
65.5 km/h to 30.9 km/h for b equal to 10 and 30 , respectively. This
is related to the fact that the ploughing effect is increased with
increasing approach angle (see also Figs. 18b and 19b). Another
interesting observation from Fig. 18c is that, at a very shallow
approach angle of 10 , the rear of the truck is turning outwards (i.e.
counter clockwise). This observation was also made during real
incidents (Thoeni and Giacomini, 2017). Finally, similar to the
previous scenarios S1 and S2, an almost linear trend is observed
with increasing height for b equal to 30 but this seems not to be
the case for b equal to 10 . However, the latter is partly affected by
the fact that 70 km/h is the maximum modelled velocity. Hence,
for the approach angle of 10 , the critical velocities indicated as
70 km/h could, in theory, be higher.
6.4. Scenario S4
particles and various particle size distributions for different mate- Hugo D, Heyns P, Thompson R, Visser A. Haul road defect identification using
measured truck response. Journal of Terramechanics 2008;45(3):79e88.
rial types.
Jean M. The non-smooth contact dynamics method. Computer Methods in Applied
Mechanics and Engineering 1999;177(3e4):235e57.
Conflict of interest Kaufman W, Ault J. Design of surface mine haulage roads - a manual. Technical Report.
U.S. Department of the Interior, Bureau of Mines, Information Circular 8758; 1977.
Lacoursière C. Regularized, stabilized, variational methods for multibodies. In:
The authors wish to confirm that there are no known conflicts of Bunus PDF, Führer C, editors. The 48th scandinavian conference on simulation
interest associated with this publication and there has been no and modeling (SIMS 2007), 30-31 october 2007. Göteborg (Särö), Sweden:
Linköping University Electronic Press; 2007. p. 40e8.
significant financial support for this work that could have influ- Layton R, Fabien B. Systematic modelling using Lagrangian DAEs. Mathematical and
enced its outcome. Computer Modelling of Dynamical Systems 2001;7(3):273e304.
Lindmark D, Servin M. Computational exploration of robotic rock loading. Robotics
and Autonomous Systems 2018;106:117e29.
Acknowledgements Lommen S, Lodewijks G, Schott D. Co-simulation framework of discrete element
method and multibody dynamics models. Engineering Computations
2018;35(3):1481e99.
This study was financially supported by the Australian Coal Moreau J. Numerical aspects of the sweeping process. Computer Methods in
Association Research Program (ACARP C21032 e Stage 2). The au- Applied Mechanics and Engineering 1999;177(3e4):329e49.
thors would like to acknowledge Neil Reynoldson from the MSHA. Haul road inspection handbook. Mine Safety and Health Administration, U.S.
Department of Labor; 1999.
Queensland Department of Natural Resources and Mines and Chris
Murty K. Linear complementarity, linear and nonlinear programming. Heidelberg:
Stubley from the Western Australia Department of Mines and Pe- Helderman-Verlag; 1988.
troleum for providing information on recent incidents involving NSW Mine Safety. Safety bulletin No: SB17-01: industry reports more truck rollover
ultra-class haul trucks. The support of the IT facilities of the Uni- incidents. NSW Australia: Department of Industry; 2017.
Queensland Government. Safety alert No. 247: trucks tipping over the edge.
versity of Newcastle is also gratefully acknowledged. Queensland Australia: Department of Employment, Economic Development
and Innovation; 2010.
Reynoldson N. Specification for design and construction of mine roads. OCE seminar
Appendix A. Supplementary data presentation. Queensland Australia: Department of Natural Resources and
Mines; 2015.
Supplementary videos to this article can be found online at Servin M, Wang D, Lacoursière C, Bodin K. Examining the smooth and nonsmooth
discrete element approach to granular matter. International Journal for Numerical
https://doi.org/10.1016/j.jrmge.2018.11.005. Methods in Engineering 2014;97(12):878e902. https://doi.org/10.1002/nme.4612.
Servin M, Wang D. Adaptive model reduction for nonsmooth discrete element
simulation. Computational Particle Mechanics 2016;3(1):107e21.
References Simmons J, McManus D. Shear strength framework for the design of dumped spoil
slopes for open pit coal mines. In: Jardine RJ, Potts DM, Higgins KG, editors.
Acary V, Brogliato B. Numerical methods for nonsmooth dynamical systems: ap- Proceedings of advances in geotechnical engineering: the skempton confer-
plications in mechanics and electronics. Springer; 2008. ence; 2004. p. 981e91.
Algoryx Simulation. AGX dynamics. 2018. http://algoryx.se [Date accessed: 22 July Tannant D, Regensburg B. Guidelines for mine haul road design. University of
2018]. Alberta: School of Mining and Petroleum Engineering; 2001.
Alsaleh M. Soil e machine interaction: simulation and testing. In: Bonelli S, Thoeni K, Fityus S, Giacomini A, Vaha J. Discrete modelling of a tilt box test for
Dascalu C, Nicot F, editors. Advances in Bifurcation and Degradation in Geo- granular materials. In: Oka F, Murakami A, Uzuoka R, Kimoto S, editors. Com-
materials. Springer Series in Geomechanics and Geoengineering, vol. 11. puter methods and recent advances in geomechanics. The 14th international
Springer; 2011. p. 165e76. conference of the IACMAG. Taylor and Francis; 2015. p. 1557e62.
Barton N, Kjaernsli B. Shear strength of rockfill. Journal of Geotechnical Engineering Thoeni K, Giacomini A. Energy absorption capacity of muck piles and their status as
Division ASCE 1981;107(GT7):873e91. engineered hard barriers e Stage two. Technical Report ACARP C21032.
Bornemann F, Schütte C. Homogenization of Hamiltonian systems with a strong Australian Coal Association Research Program; 2017.
constraining potential. Physica D Nonlinear Phenomena 1997;102(1e2):57e77. Thoeni K, Servin M, Giacomini A. Using non-smooth multi-domain dynamics to
Burger M, Dreler K, Ekevid T, Steidel S, Weber D. Coupling a DEM material model to improve the safety on haul roads in surface mining. In: The 5th international
multibody construction equipment. In: Conference proceedings of 8th ECCO- conference on particle-based methods. CIMNE International Center for Nu-
MAS thematic conference on multibody dynamics. Nakladatelství CVUT (CTN); merical Methods in Engineering; 2017. p. 600e11.
2017. p. 417e24. Thompson R. SME mining engineering handbook: charpter 10.6 design, construc-
CAT. Caterpillar 797F mining truck specification. 2018. www.cat.com/en_GB/ tion, and maintenance of haul roads. 3rd ed. Lyttleton, Colorado, USA: Society
products/new/equipment/off-highway-trucks/mining-trucks/18093014.html for Mining, Metallurgy, and Exploration; 2011a. p. 957e77.
[Date accessed: 22 July 2018]. Thompson R. Mining roads: mine haul road design, construction & maintenance
Coetzee C, Els D, Dymond G. Discrete element parameter calibration and the modelling management. 2011. http://aspasa.co.za/wp-content/uploads/2017/07/road-
of dragline bucket filling. Journal of Terramechanics 2010;47(1):33e44. design-and-maintanance-Molefe.compressed.pdf [Date accessed: 22 July 2018].
Cundall P, Strack O. A discrete numerical model for granular assemblies. Gèo-
technique 1979;29(1):47e65.
Curry D, Deng Y. Optimizing heavy equipment for handling bulk materials with
Adams-EDEM co-simulation. In: Proceedings of the 7th international confer-
ence on discrete element methods, DEM 2016. Springer; 2017. p. 1219e24. Dr. Klaus Thoeni is Senior Lecturer at the University of
https://link.springer.com/content/pdf/10.1007/978-981-10-1926-5_126.pdf. Newcastle, Australia. He has more than 10 years’ experi-
Effeindzourou A, Giacomini A, Thoeni K, Sloan S. Numerical investigation of rockfall ence in the development of cutting-edge numerical tools
impacts on muckpiles for underground portals. Rock Mechanics and Rock En- for geotechnical engineering and rock mechanics applica-
gineering 2017;50(6):1569e83. tions. He obtained his PhD and Master’s degrees in civil
Fityus S, Hancock G, Wells T. Geotechnical characteristics of coal mine spoil. engineering from Graz University of Technology (Austria).
Australian Geomechanics 2008;43(3):13e22. After moving to Australia, he expanded his initial research
Giacomini A, Thoeni K. Energy adsorption capacity of muck piles and their status as experience on continuum-based numerical modelling by
engineered hard barriers. Technical Report ACARP C21032. Australian Coal As- taki ng on the Discrete Element Method (DEM), a
sociation Research Program; 2014. discontinuum-based method. He is an active developer
Giacomini A, Thoeni K. Full-scale experimental testing of dump-point safety berms of the open-source DEM framework YADE, an efficient nu-
in surface mining. Canadian Geotechnical Journal 2015;52(11):1791e810. merical tool for the dynamic simulation of geomaterials.
Gruening T, Kunze G, Katterfeld A. Simulating the working process of construction Lately he has been concentrating on the development of
machines. In: Bonelli S, Dascalu C, Nicot F, editors. Proceedings of bulk solids highly innovative numerical methods to solve practical
Europe 2010. Glasgow, UK: Vogel; 2010. p. 1e10. problems in the Australian open cut mining environment. Besides his expertise in nu-
Hess G, Richter C, Katterfeld A. Simulation of the dynamic interaction between bulk merical modelling, Klaus has extensive knowledge in photogrammetry. He is reviewer
material and heavy equipment: calibration and validation. In: The 12th inter- for leading international journals in computational mechanics, rock mechanics, soil
national conference on bulk materials storage, handling and transportation mechanics, structural engineering and remote sensing. Klaus is also member of the
(ICBMH). ICBMH; 2016. p. 427e36. Editorial Board of the Canadian Geotechnical Journal and Computers and Geotechnics.
K. Thoeni et al. / Journal of Rock Mechanics and Geotechnical Engineering 11 (2019) 659e675 675
Dr. Martin Servin is Associate Professor at Umeå Uni- Significantly, this original work has enabled engineers to model complex geo-
versity, Sweden, where he received his PhD in theoretical material behaviour in a robust manner and is now used in academia, research in-
physics in 2004. His research interests are in computa- stitutions, and commercial computer programs worldwide. In the past decade, he
tional multibody dynamics with particular focus on has conducted full-scale field tests on embankments resting on soft soils, and also
granular materials, complex mechanical systems and been active in the field of georemediation, specifically the clean-up of waste from
computational software for design, analysis and control aluminium smelters and the (PFAS) contamination generated by the use of fire-
as well as real-time interactive physics. Martin is co- fighting foams at airports.
founder of the UMIT Research Laboratory, a leading
centre in northern Europe for research in computational
science and engineering. He is also co-founder of the Associate Professor Giacomini has pioneered research
spin-off Algoryx Simulation, a leading provider of soft- in rock mechanics and rockfall analysis as applied to civil
ware and services for visual and interactive physics and mining engineering. She received her PhD in 2003
based simulations. from the University of Parma (Italy), and joined the
University of Newcastle in 2005. She is currently Asso-
ciate Professor in the Priority Research Centre of
Laureate Professor Scott Sloan has pioneered new nu- Geotechnical Science in Engineering and Higher Degree
merical methods which enable engineers to predict the Research Director for the School of Engineering of the
maximum load capacity of general types of geostructures University of Newcastle (Australia). Associate Professor
(such as roads, railways, tunnels, dams, and port facil- Giacomini has worked in the area of Rock Mechanics for
ities). These methods are based on the limit theorems of more than 15 years. She is committed to experimental
plasticity, finite elements, advanced optimisation algo- and computational innovation in modelling rockfall
rithms and adaptive remeshing, and have been extended hazards, which has markedly improved the safety of
to model cyclic loading. Being able to estimate the limit mining environments and major transport corridors in
load accurately is crucial in many forms of infrastructure Australia and abroad. Her internationally renowned work has had a major impact in
design, but is complicated by the fact that geomaterials the Australian mining industry, where rockfalls threaten human lives, machinery
are frequently heterogeneous, anisotropic, and discon- and the portal structures for underground entry. She is leading six major research
tinuous. Sloan’s methods have delivered exciting and projects through the Australian Coal Association Research Program (ACARP) on
useful new tools for engineers to tackle these complex rockfall hazard and mitigation and an Australian Research Council Linkage Project
problems, and have led to cheaper and safer designs for for the development of new innovating monitoring methodologies of rock slopes. A/
civil infrastructure. Sloan has also derived many new algorithms and software for Prof Giacomini is Editorial Board member of Computer and Geotechnics and Rock
implementing realistic soil models, solving nonlinear finite element equations, Mechanics and Rock Engineering journals and serves as a reviewer for over 15 in-
generating finite element grids, and solving large systems of sparse equations. ternational journals in the field.