Professional Documents
Culture Documents
Environment
To cite this article: Weiguo Zeng , Ernest Baafi & David Walker (2017): A simulation model to
study bunching effect of a truck-shovel system, International Journal of Mining, Reclamation and
Environment, DOI: 10.1080/17480930.2017.1348284
Article views: 67
in a truck-shovel mining system. This paper presents the development of Accepted 26 June 2017
discrete-event simulation model to estimate the impacts of bunching on
KEYWORDS
the productivity and efficiency of a truck-shovel system. The bunching Discrete-event simulation;
effect on production, BEP, is defined to estimate the production sensitivity truck-shovel system; truck
to the bunching effect on the truck fleet. The simulation results show that bunching
the mixed truck fleets with varying performance can cause significant
bunching effect if the hauling trucks are from multiple loading sites or
dumps. Depending on whether the fleet is over-trucked or under-trucked,
the bunching has significant impact on both the fleet productivity and the
equipment utilisation. Furthermore, the fleet with higher BEP takes a priority
of overtaking the fleet with lower BEP to increase the productivity.
1. Introduction
A truck-shovel mining system generally consists of shovels and the associated trucks. Ore and waste
are loaded into trucks by shovels and the trucks haul between loading sites and dumps or crushers.
The main operational elements for one truck cycle include spotting, loading, hauling loaded, dumping,
hauling empty, queuing and operational delays. Although the maximum system productivity depends
on the output of the shovels, in real terms, this maximum productivity is reduced due to mismatch of
trucks and shovels and the bunching effect [1], which is the effect of a faster truck following behind
a slower truck if overtaking is prohibited or impossible.
Surplus trucks in the fleet cause queues at shovels and increase waiting time of the trucks, whereas
insufficient number of trucks in the system increases idle time of the shovels. The efficiency of the
haulage system is influenced significantly by the size of the truck fleet [2]. Match factor (MF) has
been widely used in the earthmoving industry for selecting the best fleet for truck-shovel system for
homogeneous truck-shovel fleets [3], using Equation (1).
Bunching results in heterogeneous truck cycle times and the non-synchronisation in the truck-shovel
mining system [4]. In Equation (1), the bunching effect is ignored in the calculation of MF. In practice,
reducing factors are used to account for the bunching effect in estimating both the efficiency and pro-
ductivity of truck-shovel system [5,6]. Burt and Caccetta [7] have proposed the need for further mod-
elling work on bunching effect for a better estimation of the performance of the truck-shovel system.
The complex and interactive features of the truck-shovel mining system determine that analytical
and deterministic approach is not feasible for model development [8]. The loading time and amount
vary according to the truck type, shovel type, material characteristics, operator’s performance, etc.
The truck hauling time is influenced by:
• the weight of the truck, the truck performance and retarder curves [9], truck driver’s skill,
• the haul routes design and conditions such as rolling resistance, haul grade and road mainte-
nance, and
• the traffic constraints including bunching effect, intersection passing priority and speed control,
etc.
With discrete-event simulation, it is possible to evaluate the stochastic and dynamic elements of a
truck-shovel mining system and also support management to estimate and compare alternatives for a
Downloaded by [University of Florida] at 23:19 06 August 2017
better decision-making [10]. According to the level of modelling detail, there are currently two kinds
of simulation approaches to the traffic control problem: the macroscopic and microscopic approaches
[11]. The macroscopic approach describes the traffic process via the low level detailed traffic objects
such as the traffic flow and density; while the microscopic approach is known as the high detailed
modelling approach, considering the traffic elements, for instance, individual vehicle units, haul routes,
the interaction between the vehicle units and the influence of the traffic network on the vehicle units.
Previous work on road transport [12–14] has pointed out that the macroscopic approach is not able
to reproduce the individual vehicle movement and to capture the traffic interaction on the haul route
networks. Furthermore, Jaoua et al. [15] proved that with no consideration of the real-time haulage
network constraints or traffic congestion and truck interaction, the macroscopic models could bias
simulation results and thus ‘can significantly increase the gap between simulated versus practical
algorithm performance’ [11]. However, the most of previous truck-allocation models [2,16–20] were
developed using the macroscopic approach. The microscopic traffic influences including the bunching
effect and the truck behaviour at the intersection area are simplified or ignored in those models. Jaoua
et al. [21] proposed a microscopic real-time fleet management model which reflected the haul road
network environment and the traffic congestion level of the haul route. Bastos et al. [22] also consid-
ered bunching as a delay in their time-dependent truck dispatching model with no truck overtaking.
However, the researchers [21,22] ignored the impact of road conditions and truck performance on
truck’s speed and did not explore the impact of the bunching on the productivity of the truck-shovel
system. Soofastaei et al. [23] estimated the relationship between the variability of the truck payload
and the increased cycle time caused by bunching. But the haul road network associated with the
interaction between trucks in the network system was not considered in their model.
This paper presents a microscopic discrete-event simulation model that is able to estimate the key
performance indicators (KPIs) of a truck-shovel system. The model considers not only the stochastic
and dynamic operational elements of the system but also the microscopic traffic behaviours, including
the interaction between the individual trucks and the impact of the haul road network. The bunch-
ing effect and the influence of the traffic at the intersection area on the system performance can be
evaluated and a more efficient strategic management support can be offered by the simulation model.
standard Java programming language. The framework of the developed model is composed of the
operational elements of a truck-shovel mining system, including material, loader, dump, queue, truck,
hauling route and the operator. The following simulation model objects were developed for modelling
a truck-shovel mining system in a microscopic approach:
(1) OreEntity: material that flows through the system.
(2) OreGenerator: generates the OreEntity.
(3) OreSink: destroys the OreEntity.
(4) Loader: receives the OreEntity from the OreGenerator object and sends it to the Truck object
for transportation.
(5) LoaderOperator: interacts with the Loader object to represent the influence of the shovel
operator on the shovel performance.
(6) Dump: receives the OreEntity from the Truck object. After process, it sends the OreEntity
object to the Sink object and releases the Truck object.
(7) Truck: interacts with the Route object and transports the OreEntity objects between the
Downloaded by [University of Florida] at 23:19 06 August 2017
of the haul section. In addition to the truck performance and retarder charts, the average hauling speed
is also affected by the speed limit on certain route imposed by management.
In the model, the Route object is divided into various segments according to the route conditions
such as grade, rolling resistance and traffic infrastructure. Trucks travel on divided segments of hauling
route with varied mean travelling speeds. Depending on whether the truck ahead (slower truck) and the
following truck (faster truck) are travelling in the same segment, there are three bunching possibilities:
three-stage bunching possibility, two-stage bunching possibility and safe correction distance possibility.
• Three-stage bunching possibility
When bunching occurs (refer to Figure 2), if the truck ahead is still within the route segment,
the speed change for the following truck to pass this route segment contains three stages or options:
hauling with its own speed in the current segment, bunching with the speed of the truck ahead in the
current segment and bunching with the speed of the truck ahead in the next segment. As shown in
Figure 2, in Stage 1, after reaching point A and before bunching occurs, Truck 2 travels with v2B, the
speed for Truck 2 to haul in segment B; after the bunching occurs and before Truck 1 arrives at point
B, in Stage 2, the bunching speed of Truck 2 is equal to the speed of Truck 1, v1B, the speed for Truck
1 to haul in segment B. When Truck 1 arrives at point B and Truck 2 is still in segment B, in Stage 3,
the bunching speed of Truck 2 is v1C, the speed for Truck 1 to haul in segment C. Therefore the sequent
hauling speeds for Truck 2 to pass segment B include v2B, v1B and v1C.
• Two-stage bunching possibility
When bunching occurs (refer to Figure 3), if the truck ahead is in segment C whereas the following
truck is still in segment B, the speed change for the following truck to pass the segment B contains
two stages: hauling with its own speed in the current segment and bunching with the speed of the
truck ahead in the next segment. As shown in Figure 3, in Stage 1, before the bunching occurs, Truck
2 travels with v2B, the speed for Truck 2 to haul in segment B; in Stage 2, after the bunching occurs,
although Truck 2 is still in segment B, as Truck 1 is already in segment C, the speed of Truck 2 changes
to v1C, the speed for Truck 1 to haul in segment C. Thus the sequent speeds for Truck 2 to pass segment
B include v2B and v1C.
• Safe correction distance possibility.
It may happen that the initial distance between the two trucks is shorter than the required safety
bunching distance. In this case, the following truck has to slow down to increase the distance to the
truck ahead to the required safety bunching distance. The reduced speed for the following truck (Truck
2), which ensures that the safety bunching distance can be obtained before the truck ahead (Truck 1)
leaves the segment (segment B), can be calculated by
Δd + v1B × Δt1 − v2 × Δt1 ≥ Δs (2)
INTERNATIONAL JOURNAL OF MINING, RECLAMATION AND ENVIRONMENT 5
Δd + v1B × Δt1 + Δs
0 ≤ v2 ≤ (3)
Δt1
where Δd = initial distance, m; Δs = safety bunching distance, m; v1B = hauling speed for Truck 1 in
segment B, m/s; v2 = modified hauling speed for Truck 2 increasing the bunching distance, m/s; Δt1 =
time for Truck 1 to reach point B, s.
where t = time length for the faster truck to reach the slower truck, s; d = distance between the two
trucks, m; Δv = difference between the speed of the faster truck and that of the slower truck, m/s.
The chasing time for three-stage following is different from that for two-stage following, as the Δv
is different. Therefore the condition for applying the three-stage following process is
entry timecurrent + t3s < arrivetimefirst (5)
The condition for applying the two-stage following process is
entry timecurrent + t2s < arrivetimefirst (6)
where entrytimecurrent = time for the current truck to enter the segment, s; arrivetimefirst = time for the
first truck in the bunch to arrive at the next point on the route, s; t3s = duration for the current truck
to reach the truck ahead in three-stage following, s; t2s = duration for the current truck to reach the
truck ahead in two-stage following, s.
If the conditions for three-stage following and two-stage following are not satisfied, indicating that
the current truck is not fast enough to form a bunch before the truck ahead exits from the segment,
then the truck will travel in a normal situation.
an intersection over the trucks hauling on other routes. When a main route truck is travelling within
the safe distance from an intersection, as shown in Figure 5, a non-main route truck at the intersec-
tion has to wait until the main route truck has passed through the intersection. The behaviours of the
non-main route truck, waiting, moving forward and entering the intersection, are highly influenced
by the dynamic traffic conditions in the intersection area.
The state of the non-main route truck (Truck 2 and Truck 3 in Figure 5) is specified by two variables:
waiting time and moving time. The waiting time is the expected waiting time for Truck 2 to queue
and the moving time is the expected hauling time either when the queue is moving or when Truck
2 is passing through the intersection. The two variables keep updated as the traffic condition in the
intersection area varies. Figure 6 shows the bunching algorithm at intersection area in a flow chart.
Initially when a truck enters a segment near intersection, the programme firstly checks if the truck
ahead is waiting or not. If it is waiting in a queue, then the waiting time of the truck ahead, tAw, will
be used to see if the following truck can reach it before it moves. If not, then the following truck will
firstly travel for tAw, and then this bunching algorithm will be repeated. If the truck ahead is moving
in a queue, then the hauling time of the truck ahead, tAh, will be used to see if the following truck can
reach the truck ahead before it stops. If not, then the following truck will travel for tAh, and then the
algorithm will be repeated to see if the truck ahead is waiting now.
8 W. ZENG ET AL.
In Figure 6, T = the following truck, TA = the truck ahead, tAw = the time for TA to wait in the queue,
Downloaded by [University of Florida] at 23:19 06 August 2017
productivity may be low. As more trucks are added into the system, the queue forms at the shovel
and there is a potential for bunching; both the efficiency and productivity for each truck are lowered.
Downloaded by [University of Florida] at 23:19 06 August 2017
Without considering dispatching and other operation delays, it is clear that the main loss of truck
efficiency and productivity is due to truck queuing and bunching.
Consider two trucks hauling the same distance on the same haul route, Truck 1 (in front) with
average velocity v1 and Truck 2 (behind Truck 1) with average velocity v2, v1 < v2; the travelling time
for Truck 1 is t1 and that for Truck 2 is t2, obviously, t1 > t2. If bunching happens, Truck 2 continues
to follow Truck 1 with speed of v1, the lost time caused by bunching (or bunching time) Δt = t1 − t2.
Since v1 × t1 = v2 × t2, the lost time due to bunching (bunching time) can be written as:
t1 ⋅ (v2 − v1 )
Δt = (7)
v2
The queuing time and bunching time result in no production at all. For a truck-shovel system with
no bunching effect considered, the lost production is only due to truck queuing. However, if the
bunching effect is considered, the lost production can be the combined result of truck bunching and
queuing. To estimate the bunching effect, the bunching time and the increased queuing time due to
bunching should be considered. A factor, bunching effect on production (BEP, t∕s or min), is defined
to represent the changing rate of production caused by bunching. It can be expressed by Equation
(8). This factor shows the production change every lost time unit caused by bunching (including the
bunching time and increased queuing time due to bunching). For instance, if BEP of a truck is x t/
min, it indicates that shovel production change is x t every minute the truck spend on bunching and
queuing due to bunching.
spatial information modelled in Vulcan, including the length and elevation of each segment of the
routes as well as the locations of intersections. The curves of the routes were not considered in the
model. In the validation model, the Liebherr 9250 works at P3WC, two Hitachi 1900BEs work at S4C
and P3EC, the CAT 993 works at P4. Four trucks (the CAT 785C) are assigned to route P3WC – ROM
Dump; four trucks (the CAT 789C) to route S4C – P1ED; four trucks (the CAT 785C) to route P3EC –
P3WD; one truck (the CAT 785C) to route P4 – ROM Dump; one truck (the CAT 785C) to P4 – P4WD.
The assumptions for the model implementation are:
• Each truck is assigned to a fixed route (non-dispatching mode).
• The simulation model performs for 11 h per shift during one simulation run, without consider-
ation of shift change and operational delays in the network system.
• The experiment is implemented with 100 simulation replications.
and if the P-value is less than 0.05, then the null hypothesis is rejected.
Table 4 shows the comparison of the actual cycle times and the expected values after 100 replications.
The maximum prediction error for all the output is 0.5% and the P-values for all the cycle times are
significantly greater than 0.05. It is sufficient to assume the model is valid.
785C trucks hauling between P3WC loading site and ROM dump (MF = 1.37). Fleet 2 consists
of 10 Komatsu 860E trucks hauling between S4C loading site and ROM dump (MF=1.48).
The trucks are hauling in the system in a non-dispatching mode, hence the effect of dispatching
is not considered.
Parameter Value
CAT 785C empty weight (kg) 102,150
CAT 785C capacity (kg) 147,330
Komatsu 860E empty weight (kg) 200,351
Komatsu 860E capacity (kg) 254,363
Shovel service time for CAT 785C (s) Normal(122, 18)
Shovel service time for Komatsu 860E (s) Normal(208, 21)
Safe bunching distance (m) 25
Dumping time for CAT 785C (s) Normal(35, 11)
Dumping time for Komatsu 860E (s) Normal(46, 12)
No. of 785Cs No. of 860Es Bunching time (min) Queuing time (min) Shift production (t)
6 1 0 256 56,812.14
5 2 0 259 62,009.79
4 3 0 262 67,179.95
3 4 0 264 72,460.11
2 5 0 261 77,756.31
1 6 0 265 83,050.22
INTERNATIONAL JOURNAL OF MINING, RECLAMATION AND ENVIRONMENT 13
• Case 2: an under trucked system with two loaders and two fleets
In an under trucked system with two separate loading sites, when no overtaking rule was applied,
both fleets were influenced by the bunching effect. More specifically, the bunching only occurred
when the fleets were hauling from P3WC and S4C to ROM dump (from two separate loading sites
to the shared route), no bunching occurred when the fleets were hauling back from the ROM dump
(from a single dump site to the shared route), which confirms that the mixed truck fleets with varying
performance can cause significant bunching effect on the haul route if the hauling trucks are from
multiple loading sites or dumps.
Since the average velocity for fleet 1 to travel from P3WC to ROM dump (17.8 km/h) was greater
than the average velocity for fleet 2 to travel from S4C to ROM dump (16.2 km/h), when the bunching
effect was considered, the bunching time of fleet 1 (20.5 min), as shown in Figure 10, was significantly
greater than the bunching time of fleet 2 (1.2 min). The queuing time of fleet 1 increased from 40.8
min to 89.1 min, yet the queuing time of fleet 2 decreased from 107.3 min to 94.1 min, resulting in
the production decrease of fleet 1 (272.8 t) and the production increase of fleet 2 (662.4 t), as shown
in Figure 11. The faster trucks (CAT 785C) from fleet 1 were often blocked by the slower trucks
Downloaded by [University of Florida] at 23:19 06 August 2017
(Komatsu 860E) from fleet 2, which increased the queuing time and reduced the production of fleet 1
while decreased the queuing time and increased the production of fleet 2. Hence it can be summarised
that for fleet 1, the total lost time (sum of the bunching time and increased queuing time) is 68.8 min
and the decrease of shift production is 272.8 t; for fleet 2, the lost time is −12.1 min and the increase
of shift production is 662.4 t. The BEP of fleet 1 is 3.96 t/min and the BEP of fleet 2 is −s54.96 t/min.
This implies if a CAT 785C in fleet 1 is blocked by a Komatsu 860E in fleet 2, every minute lost due to
Downloaded by [University of Florida] at 23:19 06 August 2017
bunching for fleet 1 would cause 3.96 t production loss, while every minute saved from no overtaking
policy for fleet 2 would result in 54.96 t production increase. It is clear that fleet 2 should avoid being
delayed by bunching, and CAT 785C from fleet 1 should not overtake Komatsu 860E from fleet 2,
even if overtaking is allowed.
• Case 3: an over trucked system with two loaders and two fleets
In Case 3, the bunching effect was also significant when the loaders were over trucked. As shown
in Figure 12, the bunching time of fleet 1 was 30.8 min and the bunching time of fleet 2 was 1.8 min.
Different from Case 2, when overtaking is not allowed, both the queuing times of fleet 1 and fleet 2
increased. The increase in the queuing time of fleet 1 was 34.8 min and that of fleet 2 was 30.0 min.
However, both the productions of fleet 1 and 2 were not changed (50,468.6 t for fleet 1 and 89,560.4
t for fleet 2), and the BEPs for the two fleets are 0. Thus if the fleet is over trucked, the bunching has
no impact on the productivity, but efficiency of the fleet drops due to the significant increase in the
queuing time.
8. Conclusions
A truck-shovel simulation model has been developed for estimating the operational elements and
evaluating the performance of the entire truck-shovel mining system. The microscopic traffic simu-
lation approach allows the interaction between the mining equipment and the traffic environment to
be modelled realistically and dynamically. The bunching impact, including the bunching time and the
increased queuing time, on the productivity and efficiency of a truck-shovel mining system has been
estimated. According to the simulation results, there are three conclusions as follows:
(1) The mixed truck fleets with varying performance can cause significant bunching effect on the
haul route if the hauling trucks are from multiple loading sites or dumps.
(2) When bunching occurs, in the case of the under trucked fleets, both the bunching time and
queuing time of the blocked trucks increase, resulting in production decrease; yet for the
unblocked trucks, the bunching time is insignificant and the queuing time decreases, caus-
ing production increase. In the case of the over trucked fleets, both the bunching time and
queuing time of the fleets increase with no impact on productivity.
(3) The truck fleet with higher BEP has more significant bunching effect on the fleet production,
thus the delay due to bunching should be prevented for the truck fleet with higher BEP.
Therefore, it can be concluded that the truck bunching effect should be considered for mining
strategic planning particularly where it is envisaged that a mixed truck fleet will be employed or a
INTERNATIONAL JOURNAL OF MINING, RECLAMATION AND ENVIRONMENT 15
homogeneous truck is employed with varying performance. The developed model provides the capabil-
ities to evaluate the bunching influence on the performance of a truck-shovel mining network system.
Disclosure statement
No potential conflict of interest was reported by the authors.
ORCID
Weiguo Zeng http://orcid.org/0000-0002-4445-7657
References
[1] S.D. Smith, Earthmoving productivity estimation using linear regression techniques, J. Constr. Eng. Manag. 125
(1999), pp. 133–141.
[2] S. Alarie and M. Gamache, Overview of solution strategies used in truck dispatching systems for open pit mines, Int.
Downloaded by [University of Florida] at 23:19 06 August 2017
[22] G.S. Bastos, L.E. Souza, F.T. Ramos, and C.H.C. Ribeiro, A Single-Dependent Agent Approach for Stochastic Time-
Dependent Truck Dispatching in Open-Pit Mining, Proceedings of 2011 14th International IEEE Conference on
Intelligent Transportation Systems (ITSC), 2011, pp. 1057–1062.
[23] A. Soofastaei, S.M. Aminossadati, M.S. Kizil, and P. Knights, A discrete-event model to simulate the effect of truck
bunching due to payload variance on cycle time, hauled mine materials and fuel consumption, Int. J. Min. Sci.
Technol. 26 (2016), pp. 745–752.
[24] D.H. King and H.S. Harrison, Open-Source Simulation Software JAAMSIM, Proceedings of the 2013 Winter
Simulation Conference, IEEE, Piscataway, NJ, 2013, pp. 2163–2171.
[25] G. Shaw, Proactive production estimation and fleet management of a surface mining operation, Unpublished B. E.
thesis, University of Wollongong, Wollongong, 2012, pp. 33–58.
[26] R. Carver, Practical data analysis with JMP®, SAS Institute Inc., Cary, 2010.
Downloaded by [University of Florida] at 23:19 06 August 2017