See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/330159805

Application of Advanced Data Analytics to Improve Haul

Trucks Energy Efficiency in Surface Mines

Chapter · December 2017

DOI: 10.1201/b22353-12

CITATIONS READS

0 246

3 authors:

Ali Soofastaei P. Knights

VALE The University of Queensland
68 PUBLICATIONS   104 CITATIONS    88 PUBLICATIONS   546 CITATIONS   

SEE PROFILE SEE PROFILE

M. S. Kizil
The University of Queensland
108 PUBLICATIONS   652 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Coal Mining; Solid backfilling coal mining View project

ACARP project View project

All content following this page was uploaded by Ali Soofastaei on 05 January 2019.

The user has requested enhancement of the downloaded file.

12
Application of Advanced Data
Analytics to Improve Haul Trucks
Energy Efficiency in Surface Mines

Ali Soofastaei, Peter Knights, and Mehmet Kizil

CONTENTS
12.1 Introduction............................................................................................... 163
12.2 Context: Reducing Energy Consumption via Data Analytics............ 164
12.3 Modeling Haul Trucks’ Fuel Consumption.......................................... 166
12.3.1 Artificial Neural Network......................................................... 167
12.3.2 Optimization of Effective Parameters on Haul Truck
Fuel Consumption...................................................................... 168
12.3.3 The Developed Model................................................................ 170
12.4 Results........................................................................................................ 172
12.5 Conclusions................................................................................................ 177
References.............................................................................................................. 178

12.1 Introduction
Truck haulage is responsible for a majority of cost in a surface mining
operation. Diesel fuel, which is costly and has a significant environmental
footprint, is used as a source of energy for haul trucks in surface mines.
Accordingly, improving truck energy efficiency would lead to a reduction in
fuel consumption and therefore greenhouse gas emissions.
The determination of haul trucks fuel consumption is complex and
requires multiple parameters including the mine, fleet, truck, speed, pay-
load, operator inputs, fuel, climate, tire, and road conditions as inputs. Data
analytics can be used to simulate the complex relationships between the
input parameters affecting haul trucks fuel consumption. The aim of this
chapter is to introduce an advanced data analytics model to improve the
energy efficiency of haul trucks in surface mines.
The most important controllable parameters affecting fuel consump-
tion are payload, truck speed, and total rolling resistance. From these

163
164 Extracting Innovations

parameters a comprehensive analytical framework can be developed to

determine the opportunities for minimizing truck fuel consumption.
The first stage of the analytical framework includes the development of
the artificial neural network (ANN) model to determine the relationship
between truck fuel consumption and payload, truck speed, and total
resistance.
This model can be trained and tested using real data collected from some
large surface mines in Australia, the United States, and Canada. A fitness
function for the haul truck fuel consumption can be successfully generated.
This fitness function is then used in the second stage of the analytical frame-
work to develop a digital learning algorithm based on a novel multiobjective
genetic algorithm (GA). The aim of this algorithm is to establish the optimum
set points of the three controllable parameters to reduce the diesel fuel con-
sumption, with these set points being specific to individual mines and fleet
operations.

12.2 Context: Reducing Energy Consumption

via Data Analytics
Energy efficiency has gradually become a more important consideration
worldwide since the rise of the cost of fuel in the 1970s. The mining indus-
try annually consumes trillions of British thermal units (BTUs) of energy
in operations such as exploration, extraction, transportation, and process-
ing. Mining operations use energy in a variety of ways, including excava-
tion, materials handling, mineral processing, ventilation, and dewatering.
It also uses significant quantities of power. The Mining industry consumed
520  petajoules (PJ) of energy in 2014–2015 or 9% of the national energy
end use in Australia (Allison et  al. 2016). Energy consumption in min-
ing is rising at around 6% annually in Australia due to lower grade ores
located deeper underground (EEO 2012), a trend seen in other developed
countries (DOE 2012). As well as improving margins through efficiency
savings, energy streamlining in the sector can also result in the reduction
of millions of tons of gas emissions because the primary energy sources
used in the mining industry are petroleum products: electricity, coal, and
natural gas.
The potential for energy (and financial) savings has motivated the min-
ing industry and governments to conduct research into the reduction of
energy consumption. Consequently, a large number of research studies and
industrial projects have been carried out in an attempt to do this in min-
ing operations across the world (Soofastaei 2016). Current investments in the
improvement of mining technologies and energy management systems have
Application of Advanced Data Analytics 165

resulted in a significant reduction of energy consumption. Based on com-

pleted industrial projects, significant further opportunities exist within the
mining industry to reduce energy consumption. The case study presented
here—haulage equipment—is one of these potential areas.
Service trucks, front-end loaders, bulldozers, hydraulic excavators, rear
dump trucks, and ancillary equipment, such as pick-up trucks and mobile
maintenance equipment, are prominent examples of the diesel equipment
and associated energy footprint of mining operations. In surface mines,
the most commonly used means of mining and hauling of materials is by
a truck and shovel operation. The trucking of overburden constitutes a
major portion of energy consumption. However, as will be discussed, the
rate of this energy consumption is the result of many different parameters
(EEO 2010) which can be analyzed and altered to obtain optimal levels of
performance.
Data analytics represents a very appropriate approach to pulling together
these disparate data sources since it is the science of examining raw data
to draw conclusions about that information. The main advantages of data
analytics can be presented by cost reduction, faster and better decision-
making, and finally new products and services (Soofastaei and Davis
2016). The uses of data analytics are many and can apply to areas that
many might not have thought of before. One area that sees much potential
in data analytics is the mining industry. For an industry that does trillions
of dollars in business every year, data analytics should be considered a
necessity not a luxury. Indeed, there are many phases of the mining pro-
cess where data analytics can be put to use. The four main phases are the
(1) extraction of ore, (2) materials handling, (3) ore comminution and sepa-
ration, and (4) mineral processing. Of particular focus for some compa-
nies is efficiency improvements in the second phase, materials handling.
Without data analytics at the heart of this phase, operators are more than
likely to be subject to suboptimal functioning of their equipment, includ-
ing in haulage vehicles and infrastructure.
As Figure 12.1 illustrates, use of data analytics in organizations cover
two dimensions: time frame (the past, present, or future) and competitive
advantage (value of insight generated). At the lowest level, analytics are rou-
tinely used to produce reports and alerts. These are simple, retrospective
processing and reporting tools, such as pie graphs, top-ten histograms and
trending plots, typically addressing variations of the basic question of “what
happened and why?” Increasingly, sophisticated analytical tools, capable
of working at or near real-time and providing rapid insights for process
improvement, can show the user “what just happened” and assist them in
understanding “why” as well as the next best action to take. However, at the
top of the pictured comparative advantage scale, there are predictive models
and optimization tools, aimed at evaluating “what will happen” and identi-
fying the best available responses.
166 Extracting Innovations

What is the best that could happen?

Reactive enterprise
What will happen?

Why did it happen?

Completive advantage

Business optimization

Predictive analytics

What happened?

Advanced analytics

Proactive enterprise
Generic analytics

Online analytic processing

Standard data

Cleaned data
Raw data

Analytics maturity

FIGURE 12.1
Competitive advantages of data analytics in organizations. (From Davenport, T. H. et  al.,
Analytics at Work: Smarter Decisions, Better Results, Harvard Business Press, New York, 2010.)

12.3  Modeling Haul Trucks’ Fuel Consumption

In this chapter, the effects of the three most important and effective param-
eters on fuel consumption of haul trucks are examined. These param-
eters are payload (P), truck speed (S), and total rolling resistance (TR). On
a real mine site, the correlation between fuel consumption and the three
parameters is complex. We use two artificial intelligence methods to create
an advanced data analytic model to estimate and reduce haul truck fuel
consumption in surface mines. The model can estimate the energy con-
sumption of haul trucks in surface mines using an artificial neural net-
work (ANN) and can also find the optimum values of P, S, and TR using a
GA. We analyze each of these in turn and then present the results of our
modeling.
Application of Advanced Data Analytics 167

12.3.1  Artificial Neural Network

ANNs, also known as neural networks (NNs), simulated neural networks
(SNNs), or parallel distributed processing (PDP), are the representation of
methods that the brain uses for learning (Hammond 2012). ANNs are a
series of mathematical models that imitate a few of the known characteris-
tics of natural nerve systems and sketch on the analogies of adaptive natural
learning (Rodriguez et al. 2013). The key component of a ANN paradigm is
the unusual structure of the data processing system. ANNs are utilized in
various computer applications to solve complex problems. They are fault-
tolerant and straightforward models that do not require information to iden-
tify the related factors and do not need the mathematical description of the
phenomena involved in the process (Beigmoradi et al. 2014).
The main part of a NN structure is a “node.” Biological nodes sum the sig-
nals received from numerous sources in different ways and then carry out
a nonlinear action on the results to create the outputs. NNs typically have
an input layer, one or more hidden layers, and an output layer (Figure 12.2).
Each input is multiplied by its connected weight, and in the simplest state,
these quantities and biases are combined. They then pass through the activa-
tion functions to create the output (Equations 12.1 through 12.3).

Hidden

Input Weights Summation Activation

X1 Wi1

Output Xj Wij Ek
Σ E
⋅ ⋅ ⋅

Xq Wiq
bi

Weights Summation Activation

F1 Wo1
Fk Wok Out
Σ out
⋅ ⋅ ⋅

Fm Wom
bo

E1
Fk
E2 f
⋅ ⋅ ⋅

Ek

FIGURE 12.2
Artificial neural network structure. (From Soofastaei, A., Development of an Advanced Data
Analytics Model to Improve the Energy Efficiency of Haul Trucks in Surface Mines, PhD thesis,
The University of Queensland, School of Mechanical and Mining Engineering, Brisbane,
Australia, 2016.)
168 Extracting Innovations

Ek = ∑ (w
j =1
i , j ,k x j + bi , k ) k = 1, 2,..., m (12.1)

where:
x is the normalized input variable
w is the weight of that variable
i is the input
b is the bias
q is the number of input variables
k and m are the counter and number of NN nodes, respectively, in the
­hidden layer

In general, the activation functions consist of linear and nonlinear equa-

tions. The coefficients associated with the hidden layer are grouped into
matrices wi,j,k and bi,k. Equation 12.2 can be used as the activation function
between the hidden and the output layers (in this equation, f is the transfer
function).

Fk = f (Ek ) (12.2)

The output layer computes the weighted sum of the signals provided by
the hidden layer, and the associated coefficients are grouped into matrices
Wo,k and bo. Using the matrix notation, the network output can be given by
Equation 12.3.

 m

Out = 


∑w
k =1
F  + bo (12.3)
o ,k k



Network training is the most important part of NN modeling and is car-

ried out using two methods: controllable and uncontrollable training. The
most common training algorithm is that of back-propagation. A training
algorithm is defined as a procedure that consists of adjusting the coefficients
(weights and biases) of a network to minimize the error function between
the estimated network outputs and the real outputs.

12.3.2 Optimization of Effective Parameters on Haul Truck

Fuel Consumption
Optimization is a part of computational science that represents a very effec-
tive way to find the best measurable solution for problems. To solve a given
problem, it is important to consider two components: (1) search area and
(2)  objective function. In the search area, all the possibilities of the solu-
tion are considered. The objective function is a mathematical function that
Application of Advanced Data Analytics 169

associates each point in the search area to a real value, applicable to evaluate
all the members of the search area.
Traditional optimization methods are characterized by the stiffness of
their mathematical models, making their application limited in represent-
ing “real-life” dynamic and complex situations (Selvakumar et  al. 2013).
Introducing optimization techniques based on artificial intelligence, under-
pinned by heuristic rulings, have reduced the problem of stiffness. Heuristic
rules can be defined as reasonable rules derived from experience and obser-
vations of behavior tendencies within a system of analysis.
Using analogies with nature, some heuristic algorithms were proposed
during the 1950s by trying to simulate biological phenomena in engineering.
Accordingly, these algorithms were termed natural optimization methods. One
of the best advantages of using the mentioned algorithms is their random char-
acteristic. Due to their innate flexibility, they have been found to be appropriate
to solve all types of problems in engineering (Singh and Rossi 2013; Soleimani
et  al. 2013; Soofastaei et  al. 2016). Rapid advances in computing during the
1980s made the use of these complex algorithms for optimization of functions
and processes more practicable when traditional methods were not success-
ful in this area. During the 1990s some new heuristic methods were created
by the previously completed algorithms, such as swarm algorithms, simulated
annealing, ant colony optimization, and the method used in this study, GAs.
GAs were proposed by Holland (1975) as an abstraction of biological evolu-
tion, drawing on ideas from natural evolution and genetics for the design and
implementation of robust adaptive systems (Sivanandam and Deepa 2008). Use
of the new generation of GAs is comparatively novel in optimization methods.
They do not use any derivative information and, therefore, have good chances of
escape from local minimums. As a result, their application in practical engineer-
ing problems can bring more optimal, or at least more satisfactory, solutions than
those obtained by other traditional mathematical methods (Whitley et al. 1990).
GAs are analogous with the evolutionary aspects of natural genetics. From
randomly selected “individuals” in any search area, the fitness of the solu-
tions, which is the result of the variable that is to be optimized, is determined
subsequently from the “fitness function.” The individual that generates the
best fitness within the population has the highest chance to return in the
next generation with the opportunity to reproduce by crossover with another
individual, thus producing decedents with both characteristics. If a GA is
developed correctly, the population (a group of possible solutions) will con-
verge to an optimal solution for the proposed problem (Xing and Qu 2013).
The processes that have more contribution to the evolution are the crossover,
based on the selection and reproduction and the mutation.
GAs have been applied to a diverse range of scientific, engineering, and eco-
nomic problems due to their potential as optimization techniques for c­ omplex
functions (Singh and Rossi 2013; Stanković et al. 2013; Tian et al. 2013). There
are four significant advantages when using GAs to optimize problems (Yousefi
et al. 2013). First GAs do not have many mathematical requirements regarding
170 Extracting Innovations

optimization problems. Second, they can handle many types of objective func-
tions and constraints (i.e., linear or nonlinear) defined in discrete, continuous,
or mixed search spaces. Third, the periodicity of evolution ­operators makes
them very efficient at performing global searches (in probability). And finally,
they provide us with great flexibility to hybridize with domain-dependent heu-
ristics to allow an efficient implementation for a problem.
It is also important to analyze the influence of certain parameters on the
behavior and the performance of the GA, to establish their relationship with
the problem necessities and the available resources. The influence of each
parameter on algorithm performance depends on the context of the chal-
lenge being treated. Thus, determining an optimized group of values to
these parameters will depend on a good deal of experimentation and testing.
There are a few main parameters in the GA method. Details of these five core
parameters are illustrated in Figure 12.3 and tabulated in Table 12.1.
The primary genetic parameters are the size of the population that affects
the global performance and the efficiency of the GA, the mutation rate that
avoids that a given position remains stationary in value, or that the search
becomes essentially random.

12.3.3  The Developed Model

An innovative combined model was introduced to improve the three key effec-
tive parameters on the energy consumption of haul trucks. Taking the facets of

Genetic operations

Start Initialisation Encoding

Control Crossover
End stopping Replacement
Population

Selection

Mutation

Evaluation Decoding

New generation

FIGURE 12.3
A simple structure of the genetic algorithm. (From Soofastaei, A., Development of an Advanced Data
Analytics Model to Improve the Energy Efficiency of Haul Trucks in Surface Mines, PhD ­thesis, The
University of Queensland, School of Mechanical and Mining Engineering, Brisbane, Australia, 2016.)
Application of Advanced Data Analytics 171

TABLE 12.1
Genetic Algorithm Parameters
GA Parameter Details

Fitness function The primary function for optimization.

Individuals An individual is any parameter to apply the fitness function. The value of
the fitness function for an individual is its score.
Populations and A population is an array of individuals. At each iteration, the GA
generations performs a series of computations on the current population to produce
a new population. Each successive population is called a new generation.
Fitness value The fitness value of an individual is the value of the fitness function
calculated for that individual.
Parents and To create the next generation, the GA selects certain individuals in the
children current population, called parents, and uses them to create individuals in
the next generation, called children.

the GA approach, in this model P, S, and TR are the individuals and the main
function for optimization of the fitness function is fuel c­ onsumption. The fit-
ness function was created by an ANN model. This function is a c­ orrelation
between the fuel consumption of the haul truck, P, S, and TR. After the first
step, the completed function goes to the GA phase of the computer code as
an input. The developed code starts all GA processes under stopping criteria
defined by the model (MSE and R2). Finally, the improved P, S, and TR will be
presented by the model. These optimized parameters can be used to minimize
the fuel consumption of haul trucks (Figures 12.4 and 12.5).

Data analytics Data analytics

Mine site
classical methods advanced methods

Data analytics
Prediction
Artificial neural network (ANN)

Data cleaning

Data collection Optimization Genetic algorithm (GA)

FIGURE 12.4
A schematic of the developed idea to create a combined artificial intelligence model. (From
Soofastaei, A., Development of an Advanced Data Analytics Model to Improve the Energy Efficiency of
Haul Trucks in Surface Mines, PhD thesis, The University of Queensland, School of Mechanical
and Mining Engineering, Brisbane, Australia, 2016.)
172 Extracting Innovations

Hidden

Inputs Transfer function

P
Payload (ton)

S
Output
Truck speed (km/h)
FCIndex: Fuel consumption index
(L/(ton.h))
TR
Total resistance (%)

Bias Bias

Artificial neural network (prediction) model

Evaluation Selection
POP

Optimized payload (ton)

SOP

Optimized truck speed (km/h)

Mutation Crossover
TROP
Genetic algorithm (optimization) model
Optimized total resistance (%)

FIGURE 12.5
Details of developed model. (From Soofastaei, A., Development of an Advanced Data Analytics
Model to Improve the Energy Efficiency of Haul Trucks in Surface Mines, PhD thesis, The University
of Queensland, School of Mechanical and Mining Engineering, Brisbane, Australia, 2016.)

12.4 Results
The indicated artificial intelligence model that was developed was then
tested against real data taken from some types of popular trucks in four big
surface mines in the United States, Canada, and Australia. Some informa-
tion about these mines and trucks is presented in Table 12.2 (Figures 12.6
through 12.9).
To test the developed networks and validate the developed model,
1,000,000 independent samples collected from four mines were used. As our
figures illustrate, the results show good agreement between the actual and
estimated values of fuel consumption. Figure 12.10 presents sample values
Application of Advanced Data Analytics 173

TABLE 12.2
Case Studies
Investigated
Case Study Location Mine Type Mine Details Truck

Mine 1 Queensland, Surface coal The mine has coal CAT 793D
Australia mine reserves amounting to
877 million tons of
coking coal, one of the
largest coal reserves in
Asia and the world. It
has an annual
production capacity of
13 million tons of coal.
Mine 2 Arizona, Surface The mine represents one CAT 777D
United States copper of the largest copper
mine reserves in the United
States and the world,
having estimated
reserves of 3.2 billion
tons of ore grading
0.16% copper.
Mine 3 Arizona, Surface The deposit had CAT 775G
United States copper estimated reserves (in
mine 2017) of 907 million
tons of ore grading
0.26% copper and 0.03%
molybdenum.
Mine 4 Ontario, Surface gold This mine produced CAT 785D
Canada mine 235,000 ounces of gold in
2016, at the cost of sales
of $795 per ounce, and
all-in sustaining costs of
$839 per ounce. The
mine’s proven mineral
reserves as of December
2016, were 1.6 million
ounces of gold.

for the estimated (using the ANN) and the independent (tested) fuel con-
sumption to highlight the insignificance of the values of the absolute errors
in the analysis for the four mines that were studied.
All processes in the developed model, then, certainly work based on the
present dataset collected from four large surface mines. The results of using
developed model for the selected real-life mines are given in Tables 12.3
through 12.6. They, therefore, could presumably be replicated using the same
method for other surface mines.
174 Extracting Innovations

2.20
S = 20 km/h
TR = 12%
S = 15 km/h
2.16
FCIndex L/(ton.h)

S = 10 km/h

2.12

2.08
TR = 8% S = 20 km/h
S = 15 km/h
2.04 S = 10 km/h

2.00
320 340 360 380 400 420 440
Gross vehicle weight (ton)

FIGURE 12.6
Correlation between Gross Vehicle Weight, S, TR, and FCIndex based on the developed ANN
model for CAT 793D. All data were collected from a surface coal mine located in Central
Queensland, Australia (Mine 1).

TR = 15%
1.35

1.25
TR = 10%
FCIndex L/(ton.h)

S = 20 km/h
1.15
S = 15 km/h
S = 10 km/h
1.05
S = 20 km/h
S = 15 km/h
0.95 S = 10 km/h

0.85
60 80 100 120 140 160 180 200
Gross vehicle weight (ton)

FIGURE 12.7
Correlation between GVW, S, TR, and FCIndex based on the developed ANN model for CAT
777D. All data were collected from a surface copper mine located in Arizona, United States
(Mine 2).
Application of Advanced Data Analytics 175

4.0

TR = 16%
3.5

S = 15 km/h
3.0
FCIndex L/(ton.h)

S = 10 km/h

2.5

TR = 13%
2.0

S = 15 km/h
1.5
S = 10 km/h

1.0
20 40 60 80 100 120 140
Gross vehicle weight (ton)

FIGURE 12.8
Correlation between GVW, S, TR, and FCIndex based on the developed ANN model for CAT
775G. All data were collected from a surface copper mine located in Arizona, United States
(Mine 3).

1.8
TR = 13%

1.7
TR = 10%

1.6
TR = 8%
FCIndex L/(ton.h)

1.5 S = 15 km/h
S = 10 km/h
1.4
S = 15 km/h
S = 10 km/h
1.3
S = 15 km/h
S = 10 km/h
1.2

1.1

1.0
125 175 225 275
Gross vehicle weight (ton)

FIGURE 12.9
Correlation between GVW, S, TR, and FCIndex based on the developed ANN model for CAT
785D. All data were collected from a surface coal mine located in Ontario, Canada (Mine 4).
176 Extracting Innovations

5
Estimated value of FCIndex (ANN) (L/(ton.h))

4.5

3.5

2.5

1.5
CAT 793D, Surface coal mine, Central Queensland, Australia (Mine 1)
1 CAT 777D, Surface copper mine, Arizona, United States (Mine 2)
CAT 775G, Surface copper mine, Arizona, United States (Mine 3)
0.5 CAT 785D, Surface gold mine, Ontario, Canada (Mine 4)

0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Independent value of FCIndex (tests) (L/(ton.h))

FIGURE 12.10
Sample values for the estimated (using the ANN) and the independent (tested) fuel
­consumption index.

TABLE 12.3
The Range of Normal Values and Optimized Range of Variables by GA Model to
Minimize Fuel Consumption by Haul Trucks (Caterpillar 793D in Mine 1)
Normal Values Optimized Values
Variables Minimum Maximum Minimum Maximum

Gross vehicle weight (ton) 340 430 380 400

Total resistance (%) 8 12 8 9
Truck speed (km/hr) 10 20 10 15

TABLE 12.4
The Range of Normal Values and Optimized Range of Variables by GA Model to
Minimize Fuel Consumption by Haul Trucks (Caterpillar 777D in Mine 2)
Normal Values Optimized Values
Variables Minimum Maximum Minimum Maximum

Gross vehicle weight (ton) 80 180 140 160

Total resistance (%) 10 15 10 11
Truck speed (km/hr) 10 20 10 12
Application of Advanced Data Analytics 177

TABLE 12.5
The Range of Normal Values and Optimized Range of Variables by GA Model to
Minimize Fuel Consumption by Haul Trucks (Caterpillar 775G in Mine 3)
Normal Values Optimized Values
Variables Minimum Maximum Minimum Maximum

Gross vehicle weight (ton) 50 120 70 90

Total resistance (%) 13 26 13 14
Truck speed (km/hr) 10 15 10 13

TABLE 12.6
The Range of Normal Values and Optimized Range of Variables by GA Model to
Minimize Fuel Consumption by Haul Trucks (Caterpillar 785D in Mine 4)
Normal Values Optimized Values
Variables Minimum Maximum Minimum Maximum

Gross vehicle weight (ton) 150 275 200 225

Total resistance (%) 8 13 8 9
Truck speed (km/hr) 10 15 10 12

12.5 Conclusions
The aim of this chapter was to formulate an advanced data analytics model
capable of improving haul truck fuel consumption based on the relationship
between P, S, and TR. From the available “real-life” datasets obtained from
surface mining operations, this relationship is extremely complex to disect
using traditional analysis. Therefore, an artificial intelligence method was
adopted to create a reliable model to analyze the problem.
The first element of this method was to utilize an ANN model to establish
a correlation between P, S, and TR with fuel consumption. The results of
this correlation showed that fuel consumption has a nonlinear relationship
with the investigated parameters. The ANN was then trained and tested
using the collected real mine site dataset, with there being good agreement
between the actual and estimated values of fuel consumption. Building upon
this material, a GA model was developed for considering the optimization of
effective parameters on fuel consumption in haulage trucks, which in turn
could maximize the energy efficiency in haulage operations.
From these amalgamated models, the range of all studied effective
parameters on fuel consumption of haul trucks were optimized, and the
178 Extracting Innovations

best values of P, S, and TR to minimize fuel consumption index (FCIndex)

were highlighted. The developed model was applied to analyze data for
four big coal, copper, and gold surface mines in the United States, Canada,
and Australia.

References
Allison, B., A. Shamim, M. Caitlin, and P. Pha. 2016. Australian Energy Update 2016.
Canberra, Australia: Office of the chief economist, Department of Industry
Innovation and Science.
Beigmoradi, S., H. Hajabdollahi, and A. Ramezani. 2014. Multi-objective aeroacous-
tic optimisation of rear end in a simplified car model by using hybrid robust
parameter design, artificial neural networks and genetic algorithm methods.
Computers and Fluids 90:123–132.
Davenport, T. H., J. G. Harris, and R. Morison. 2010. Analytics at Work: Smarter
Decisions, Better Results. Vol. 3. New York: Harvard Business Press, p. 93.
DOE. 2012. Mining Industry Energy Bandwidth Study. Washington, DC: Department of
Energy, USA Government.
EEO. 2010. Energy-Mass Balance: Mining. Canberra, Australia: Australian Government,
Department of Resources Energy and Tourism.
EEO. 2012. First Opportunities in Depth: The Mining Industry. Canberra, Australia:
Australian Government, Department of Resources Energy and Tourism.
Hammond, A. 2012. Development artificial neural network model to study the
influence of oxidation process and zinc electroplating on fatigue life of grey
cast iron. International Journal of Mechanical and Mechatronics Engineering
12(5):128–136.
Holland, J. H. 1975. Adaptation in Natural and Artificial Systems: An Introductory
Analysis with Applications to Biology, Control, and Artificial Intelligence. Vol. 2.
Washington, DC: University of Michigan Press, pp. 183–195.
Rodriguez, J. A., Y. El Hamzaoui, J. A. Hernandez, J. C. García, J. E. Flores, and A. L.
Tejeda. 2013. The use of the artificial neural network (ANN) for modelling the
useful life of the failure assessment in blades of steam turbines. Engineering
Failure Analysis 35:562–575.
Selvakumar, S., K. P. Arulshri, and K. P. Padmanaban. 2013. Machining fixture
layout optimisation using a genetic algorithm and artificial neural network.
International Journal of Manufacturing Research 8(2):171–195.
Singh, A., and A. Rossi. 2013. A genetic algorithm based exact approach for lifetime
maximisation of directional sensor networks. Ad Hoc Networks 11(3):1006–1021.
Sivanandam, S. N., and S. N. Deepa. 2008. Introduction to Genetic Algorithms. Vol. 3.
New Delhi, India: Springer, p. 453.
Soleimani, H., M. Seyyed-Esfahani, and M. A. Shirazi. 2013. Designing and plan-
ning a multi-echelon multi-period multi-product closed-loop supply chain
utilising genetic algorithm. The International Journal of Advanced Manufacturing
Technology 68(1–4):917–931.
 Application of Advanced Data Analytics 179

Soofastaei, A., S. M. Aminossadati, M. M. Arefi, and M. S. Kizil. 2016. Development

of a multi-layer perceptron artificial neural network model to determine
haul trucks energy consumption. International Journal of Mining Science and
Technology 26(2):285–293.
Soofastaei, A., S. M. Aminossadati, M. S. Kizil, and P. Knights. 2016. Reducing fuel
consumption of haul trucks in surface mines using genetic algorithm. Applied
Soft Computing.
Soofastaei, A., and J. Davis. 2016. Advanced data analytic: A new competitive advan-
tage to increase energy efficiency in surface mines. Australian Resources and
Investment 1(1):68–69.
Stanković, L. J., V. Popović-Bugarin, and F. Radenović. 2013. Genetic algorithm for
rigid body reconstruction after micro-Doppler removal in the radar imaging
analysis. Signal Processing 93(7):1921–1932.
Tian, J., M. Gao, and Y. Li. 2013. Urban logistics demand forecasting based on regres-
sion support vector machine optimised by chaos genetic algorithm. Advances in
Information Sciences and Service Sciences 5(7):471–478.
Whitley, D., T. Starkweather, and C. Bogart. 1990. Genetic algorithms and neu-
ral networks: Optimising connections and connectivity. Parallel Computing
14(3):347–361.
Xing, H., and R. Qu. 2013. A nondominated sorting genetic algorithm for bi-­objective
network coding based multicast routeing problems. Information Sciences
233(3):36–53.
Yousefi, T., A. Karami, E. Rezaei, and E. Ghanbari. 2013. Optimisation of the free
convection from a vertical array of isothermal horizontal elliptic cylinders via a
genetic algorithm. Journal of Engineering Physics and Thermophysics 86(2):424–430.

View publication stats