Queueing Theory For Optimizing The Productivity In Haulage System

Introduction

Surface mining is the most common mining method worldwide, and open pit
mining accounts for more than 60% of all surface output (Hartman & Mutmansky,
2002). For most surface mines, truck haulage represent as much as 60% of their total
operating cost, so it is desirable to maintain an efficient haulage system (Ercelebi &
Bascetin, 2009). As the size of the haulage fleet being used increases, shovel
productivity and truck productivity decreases, so an effective fleet size must be chosen
that will efficiently utilize all pieces of equipment (Najor & Hagan, 2004). Having more
trucks in service than necessary wastes fuel, as trucks must spend time idling while
waiting for service, and the company must pay the vehicle operator to drive a truck that
is not actually needed. Alternately, having too few trucks causes idle time for the
loaders, which causes a drop in production.
By applying queueing theory to mining haulge system, the inherent stochastic
nature of haul truck and loader behavior can be accounted for and the model created can
be used to adjust fleet sizes to better serve loading needs. In this project a queueing
model was generated that can be used to model truck and loader behavior in an open pit
mine. The model is then applied to actual haulage data from an active mining operation.
Queuing Theory Background
Queuing theory was developed to provide models capable of predicting the
behavior of systems that provide service for randomly arising demands. A queuing
system is defined as one in which customers arrive for service, wait for service if it is
not immediately available, and move on to the next server or exit the system once
service is complete. Queuing theory was originally developed to model telephone
traffic. Randomly arising calls would arrive and need to be handled by the switchboard,
which had a finite maximum capacity. There are six basic characteristics that are used to
describe a queuing system: arrival distribution of customers, service distribution of

1
servers, queue discipline, system capacity, number of service channels, and number of
service stages (Gross & Harris, 1998).
a.) Customer Arrivals
In most queuing situations the arrival process of new customers to the system is
stochastic. In these cases it is necessary to know the distribution of the times between
successive customer arrivals, or the interarrival times. It is also important to understand
the behavior of customers upon entering the system. Some customers may wait for
service no matter how long the queue is, while others may see that a queue is too long
and decide not to enter the system. When this happens the customer is described as
having balked. Other customers may enter the system, but lose patience after waiting in
the queue and decide to leave the system. These customers are said to have reneged. In
situations with two or more parallel waiting lines a customer who switches from one line to
the other is said to have jockeyed for position. Any or all of these behaviors may be present
when a queuing system has what are classified as impatient customers. Impatient customers
cause state-dependent arrival distributions, since the arrival pattern of new customers
depends on the amount of congestion in the system at the time of their entry.
b.) Service Distributions
A probability distribution is also necessary to describe customer service times,
since it will not always take the same amount of time for each customer to receive
service. Single service, where one customer is serviced at a time, or batch service,
where multiple customers receive simultaneous service from a single server are both
service options. A common example of a queuing system utilizing batch service
involves waiting in line for a roller coaster. In this scenario, the people waiting in line
are the customers and the roller coaster car is the server. A single line is formed to wait,
and when the roller coaster car arrives the first four people in line who get into the car
receive simultaneous batch service.
In some cases the service process may be dependent upon the number of
customers waiting in the queue. The server may work more quickly due to the
lengthening queue, or alternately the server may become flustered by the large number
of customers waiting and the service rate may slow as a result. Situations in which the

2
service rate depends on the number of customers in the queue for service are referred to
as state-dependent services.
c.) Queue Discipline
The manner in which customers in a queue are selected for service is referred to
as the queue discipline. The most common queue discipline is first come, first served, or
FCFS, where customers receive service in the order in which they arrived. This
discipline is also commonly referred to as FIFO, or first in, first out. Another common
queue discipline is LCFS, or last come, first served. This is commonly used in inventory
situations where the most recently placed items waiting to be used are the most easily
reached to be selected. RSS is a service discipline in which customers are selected for
service in random order, independent of their order arriving to the queue.
There are a variety of different priority queue disciplines where different classes of
customers are given higher priorities than other classes. In these disciplines the customer
with the highest priority will be selected for service ahead of lower priority customers,
regardless of how long each customer has been in the queue. If the queue discipline is
preemptive, a customer with the highest priority is allowed to receive service immediately
upon arrival at the server, even if a lower priority customer is already in service. The lower
priority customer whose service is preempted resumes service after the higher priority
customer has left. In nonpreemptive cases the highest priority customer that arrives at the
server moves to the head of the queue, but must wait until the customer currently being
serviced has left (Cooper, 1972).
d.) System Capacity
If a queue has a physical limitation to the number of customers that can be
waiting in the system at one time, the maximum number of customers who can be
receiving service and waiting is referred to as the system capacity. These are called
finite queues since there is a finite limit to the maximum system size. If capacity is
reached, no additional customers are allowed to enter the system.
e.) Number of Service Stations
The number of service stations in a queuing system refers to the number of
servers operating in parallel that can service customers simultaneously. In a single
channel service station, there is only one path that customers can take through the

3
system. Figure 1 below shows the path customers, represented by circles, take through a
single service channel queuing network. The customers arrive at the server, represented
by the rectangle, and form a queue to wait for service if it is not immediately available,
and then proceed through the system once service has been completed.

Figure 1. Single Channel Queuing System

When there are multiple servers available operating in parallel, incoming customers
can either wait for service by forming multiple queues at each server, as shown in (a) of
Figure 2, or they can form a single queue where the first customer in line goes to the next
available server, depicted in (b). Both of these types of queues are commonly found in day-
to-day life. At the grocery store individual lines are formed at each cashier, but a single line
is generally formed when customers are waiting in line at the bank. The first customer in
line then proceeds to the next available teller. A single queue waiting for multiple servers is
generally the preferred method, as it is more efficient at providing service to the incoming
customers.

Figure 2. Multichannel Queuing Systems

Queuing Systems in Mining

In mining operations, queues frequently form during the haulage process as
trucks arrive at loaders, crushers, and dump locations and have to wait their turn in line.
This process can be represented using queuing networks where the haul trucks represent
the customers in the system and the servers are the loaders or crushers that the trucks are
waiting for. When representing loading operations with queuing systems, the time a
truck spends positioning and spotting at the loader can be included either as part of the

4
loading cycle time or as part of the time the truck was waiting in the queue for service.
Figure 3 below depicts a basic mining queuing system composed of haul trucks and
excavators.

Figure 3: Truck and Loader Queuing System

Most basic haul routes have four main components: loading, loaded travel time,
dumping of material, and unloaded travel time. These stages are repeated in sequence
throughout the haulage system, and are easily represented by a cyclic queue, as shown
below in Figure 4. In some cases the haul routes can be classified as servers in addition
to the loader and the crusher, since the haul routes are necessary steps in the production
cycle, and the amount of time it will take individual trucks to complete the trip is not
constant throughout the production shift, so it is possible to assign a service distribution
to the haul routes and treat them as servers, even though no queues will form since
multiple trucks can be on the haul roads at the same time.

Figure 4: Cyclic Queuing System

The above cyclic queuing model can be adjusted to include multiple loaders,
operating in parallel. Figure 5 below shows a possible configuration with three loaders

5
with a single queue formed for trucks to wait to be loaded, but any number of loaders
could be used.

Figure 5: Cyclic Queuing System with Parallel Loaders

The cyclic queues represented above model the haulage systems for basic mine
layouts. As the complexity of mining operations increases more intricate queuing systems
must be used to represent operations. A network queue, such as the one depicted below in
Figure 6 can be used when there are multiple paths available to the haul trucks. For this type
of queuing model to work, metrics are necessary to determine the likelihood of each path
being taken throughout the haul cycle. This could depend on the congestion of part of the
system, the characteristics of each individual server, the contents of the truck’s load, or a
myriad of other factors.

Figure 6: Network Queuing System

Mines that are simultaneously operating from more than one pit can treat each
pit as separate, independent queuing networks provided they do not share any resources.
6
If they do share resources, for example two separate pits sharing a single crusher, the
operation must be treated as one queuing network with subsystems for each pit. An
example of this type of configuration is shown on the following page in Figure 7.

Figure 7: Queuing Schematic with Multiple Pits