applied

sciences
Article
A Comparative Study of Regression Model and the
Adaptive Neuro-Fuzzy Conjecture Systems for
Predicting Energy Consumption for Jaw Crusher
Khaled Ali Abuhasel
Mechanical Engineering Department, College of Engineering, University of Bisha, Bisha 61922, Saudi Arabia;
kabuhasel@ub.edu.sa




Received: 5 August 2019; Accepted: 12 September 2019; Published: 18 September 2019


Abstract: Crushing is a vital process for different industrial applications where a significant portion
of power is consumed to properly blast rocks into a predefined size of fragmented rock. An accurate
prediction of the energy needed to control this process rarely exists in the literature, hence there
have been limited efforts to optimize the power consumption at the crushing stage by a jaw crusher;
which is the most widely used type of crusher. The existence of accurate power prediction as well
as optimizing the steps for primary crushing will offer vital tools in selecting a suitable crusher for
a specific application. In this work, the specific power consumption of a jaw crusher is predicted with
the help of the adaptive neuro-fuzzy interference system (ANFIS). The investigation included, aside
from the power required for rock comminution, an optimization of the crushing process to reduce this
estimated power. Results revealed the success of the model to accurately predict comminution power
with an accuracy of more than 96% in comparison with the corresponding real data. The obtained
results introduce good knowledge that may be used in future academic and industrial research.

Keywords: neuro-fuzzy; energy consumption; ANFIS; regression; rock strength

1. Introduction
During the processing of raw materials in different industries such as the mining and cement
sectors, the size reduction of fed rocks is primarily a mechanical process [1]. The rock blasting process
is the most primary, and the first significant stage in these industrial sectors through which massive
rocks are broken and fragmented into suitable smaller sizes before feeding to the processing plant.
This process can be realized in mechanical systems commonly known as crushers. Primary crushers
are capable of handling huge rocks of considerable size (typically around 1.5 m) to provide blasted
rocks with a size reduction ratio varying from 3 to 10 [2]. The reduction ratio is the feed size to the
product size.
The crushing process is a multi-stage dry process where each stage has a small size reduction ratio
within a range of 3 to 6. Rock breakage is accomplished by crushing, impact, and abrasion corresponding
to known modes of rock fracture including compressive, tensile, and shear. The applied mode can be
defined according to rock mechanics and the load type. Rocks meet crushing or compressive failure,
where rocks of two distinct size ranges are obtained. In this mode, the coarse rocks are produced
due to tensile failure, while the small size rocks result from compressive failure occurring at loading
points or due to shear stress between projected rocks [3]. In tensile failure mode (impact crushing),
the rock possessing a higher stress over the stress needed to achieve fracture has a great tendency
to break rapidly, producing smaller rock sizes and shapes. In the final mode, shear failure (attrition
mode), the rocks are broken due to the particle–particle interaction producing a significant part of fine
size rocks. The later mode can occur when too fast feeding of a crusher is applied, which is usually

Appl. Sci. 2019, 9, 3916; doi:10.3390/app9183916 www.mdpi.com/journal/applsci

Appl. Sci. 2019, 9, x FOR PEER REVIEW 2 of 12
Appl. Sci. 2019, 9, 3916 2 of 12
significant part of fine size rocks. The later mode can occur when too fast feeding of a crusher is
applied, which is usually undesirable. Crushing in closed circuit operations produce more unwanted
undesirable. Crushing in closed circuit operations produce more unwanted fine material than open
fine material than open circuit operations. The crushing action comes from stresses applied to rock
circuit operations. The crushing action comes from stresses applied to rock particles by moving parts
particles by moving parts of the machine.
of the machine.
One of the most famous and oldest crushers is the jaw crusher [4,5]. Jaw crushers have been in
One of the most famous and oldest crushers is the jaw crusher [4,5]. Jaw crushers have been in
practical usage for about 175 years. There are different types of jaw crushers; these can be
practical usage for about 175 years. There are different types of jaw crushers; these can be distinguished
distinguished by the presence of two plates where crushed materials are fed between them. One of
by the presence of two plates where crushed materials are fed between them. One of these plates is
these plates is fixed while the other swings. Jaw crushers are classified, according to the location of
fixed while the other swings. Jaw crushers are classified, according to the location of this pivoted
this pivoted swinging plate, into Blake, Dodge, and Universal crushers. The Blake crusher is
swinging plate, into Blake, Dodge, and Universal crushers. The Blake crusher is considered the most
considered the most common, where the swinging plate is pivoted at the top [6]. This crusher can be
common, where the swinging plate is pivoted at the top [6]. This crusher can be a double toggle or
a double toggle or single toggle. Due to its simplicity, lower cost, and its higher efficiency, the single
single toggle. Due to its simplicity, lower cost, and its higher efficiency, the single toggle jaw crusher is
toggle jaw crusher is the most realized form in new applications.
the most realized form in new applications.
Jaw crushers achieve size reduction mainly by compressing particles between relatively slow-
Jaw crushers achieve size reduction mainly by compressing particles between relatively
moving, inclined surfaces. The material being fed into the machine enters from above, where the
slow-moving, inclined surfaces. The material being fed into the machine enters from above, where
crushing surfaces are furthest apart, and is crushed into smaller fragments as it descends into the
the crushing surfaces are furthest apart, and is crushed into smaller fragments as it descends into the
narrowest zone of crushing and is finally discharged by gravity.
narrowest zone of crushing and is finally discharged by gravity.
The crushing surface in a jaw crusher consists of two rectangular plates, one fixed crushing face
The crushing surface in a jaw crusher consists of two rectangular plates, one fixed crushing face
and an inclined mobile face, which moves a small distance back and forth from the fixed face [7]. The
and an inclined mobile face, which moves a small distance back and forth from the fixed face [7].
significant variables in jaw crushing are the angle of the jaws, rate of jaw movement, displacement of
The significant variables in jaw crushing are the angle of the jaws, rate of jaw movement, displacement
the movable plate, and the distance between the jaws at the discharge end, which controls the product
of the movable plate, and the distance between the jaws at the discharge end, which controls the
size as shown in Figure 1[8] .
product size as shown in Figure 1 [8].

Gap
Eccentric shaft

Throw
CSS
OSS
Figure 1. Kinematic of a single jaw crusher [3].
Figure 1. Kinematic of a single jaw crusher [3].
Several studies have dealt with size reduction, focusing on developing a theory or criterion
Severalbestudies
that would usefulhave
duringdealt
thewith size reduction,
selection focusing
and evaluation on developing
of crushing a theory
equipment. or criterion
However, nonethat
of
would be useful during the selection and evaluation of crushing equipment. However,
these studies has satisfactorily successfully predicted the power consumption, a major source none of these
of
studies has
running costssatisfactorily successfullyDonovan
in crushing equipment. predicted[9]the power consumption,
provided a majorreview
an in-depth historical sourceon
ofthe
running
most
costs in crushing equipment. Donovan [9]provided an in-depth historical review on
proposed physical basis regarding the criteria of crusher selection, prediction of crusher performance, the most
proposed physical basis regarding the criteria of crusher selection, prediction of crusher
laws of comminution, mechanisms of rock fracturing, and the corresponding application. As stated by performance,
laws of comminution,
Donovan [9], among the mechanisms
common laws of rock fracturing, andthe
of comminution, thetheories
corresponding
proposed application. As stated
by Von Rittinger in
Appl. Sci. 2019, 9, 3916 3 of 12

1867, Kick in 1883, and Bond in 1952 were found to correctly demonstrate the relationship between the
product size reduction and the corresponding required input energy throughout the main three laws of
comminution. The essential problem within these theories is their limited range of applicability as they
are based on empirical equations fitted from experimental data applied in some instances. Eloranta [10]
used Bond’s theory to estimate the crusher power consumption and recorded a 240% higher predicted
power than the actual value. Thus, it is vital to consider data provided by the crusher manufacturers
and designers who may rely on alternative methods of sizing and select crushing equipment for
specific blasting operations. Bearman et al. [11] stated that these methods were subjective and relied on
individual judgment, which led to the conservation over-design of crushers, so additional improvement
efforts should be made to include rock fracture toughness in addition to all of the factors in a real
crushing plant to be able to predict the corrected input power.
During the last decades, most of the cited work regarding rock blasting was focused on the
quantity of matter and the related energy of particle fracture in addition to material properties relevant
to fragmentation during the crushing process [12]. Single-particle breakage was used, which has the
objective of relating the breakage pattern and nature of broken material to the resultant fragmented
size distribution. Studies using single-particle breakage induce mathematical models that describe the
size reduction of different breakage materials. These efforts can be extended to relate to the fracture
consumption of energy and the produced broken size distribution to the physical property of the
broken material.
While attempting to link the energy consumption and performance to major rock properties of
the crushing system, Bearman et al. [13] performed extensive tests. The work provided an empirical
relationship between the rock strength properties and crusher power intake as well as the produced
broken size for a cone crusher. In this work, the fracture properties of the rock material were categorized
in terms of rock particle strength, breakage energy, and the fragment size distribution of broken particles.
Any inefficiency in crusher power consumption within the energy-intensive equipment leads to the loss
of gigawatt-hours of electricity per year [14] (. Therefore, the most valuable step to reduce this power
consumption is to properly improve comminution regardless of the applied technology to realize
the crushing process [15]. The improvements in feed size operation lead to beneficial optimization
in the performance due to the lowering of the system capital costs, reducing unit operating costs,
and increasing of plant productivity [15]. The use of inefficient crushers may lead to many difficulties
as the process quality may mainly depend on the quality of the crushers to feed the downstream process
with product in an acceptable reduced size [16]. Thus, the necessity of optimizing the performance
and power consumption of the primary crusher to reduce the operating costs of quarrying tools is
urgent [17].
To optimize crushing energy efficiency, proper modeling relating to the stone strength and jaw
crusher parameters is required to successfully estimate the power consumption [18]. Modeling based
on energy consumption data can be accomplished by soft computing techniques [13,19–21]. The soft
computing techniques are useful in providing accurate mathematical relations rather than computing
techniques when exact relations are not available [22,23]. Two famous forms of artificial intelligence
including neural networks and belief networks have been used to enhance the developed models
of onsite aggregation system. The adaptive neuro-fuzzy interference system, ANFIS, is an example
of soft computing techniques playing a great role in modeling an accurate input–output matrix
relationship [24]. As such, ANFIS is a suitable model to predict specific energy consumption based on
the input independent variables in the process of jaw crushing.
As the reducing size process depends on different performance characteristics of the crusher
as well as various properties of the feeding rocks, the objective of the current work was to properly
combine these parameters to achieve a low power consumption while maintaining a high product
quality for a sustainable production process. In this work, the power consumption of the jaw crusher
was predicted to provide a specific reduced rock size with the help of ANFIS modeling as one of the
computing techniques playing a great importance in modeling the relationship of the input–output
Appl. Sci. 2019, 9, 3916 4 of 12
Appl. Sci. 2019, 9, x FOR PEER REVIEW 4 of 12

the calculations
parameters. arenext
In the presented, discussed,
subsection, detailsand compared
of the AFNFISwith modelrealaredata from an applied
introduced, then thecrusher
results toof
determine the level
the calculations are of accuracy discussed,
presented, to predict the andrequired
compared power
withconsumed
real data from by theancrusher.
appliedItcrusher
is known to
that the issue
determine theof the of
level prediction
accuracy of to the power
predict theconsumption
required power of aconsumed
jaw crusher by using the ANFIS
the crusher. model
It is known
has
thatnot
the been
issue addressed until now.
of the prediction of theHopefully, the given of
power consumption approach and the
a jaw crusher findings
using from this
the ANFIS model study
has
will be used
not been for further
addressed untilresearch in the area
now. Hopefully, theofgiven
crushing performance.
approach and the findings from this study will be
used for further research in the area of crushing performance.
2. Materials and Methods
2. Materials and Methods
2.1. ANFIS Modeling
2.1. ANFIS Modeling
ANFIS is a neural-fuzzy predictive computing system based on an adaptive neural network [24]
(. WithANFIS is a neural-fuzzy
the help of a hybrid predictive computing
learning sequence, system
ANFIS is based
used to ongenerate
an adaptive neural network
input–output [24]
relations
(. With the help
considering fuzzyofif–then
a hybrid learning
rules to providesequence, ANFIS
different is used to
membership generateThe
functions. input–output
parametersrelations
of each
considering fuzzy if–then rules to provide different membership
function are determined by the ANFIS technique to follow already given empirical functions. The parameters of each
input–output
function
data. are determined by the ANFIS technique to follow already given empirical input–output data.
ANFIS uses
ANFIS uses five
five network
networklayers
layersto toachieve
achievethe thefuzzy
fuzzyinterpretation
interpretation steps
stepsshown
shown below
below (Figure
(Figure2),
where layer one is the parameters entering the system, layer two is the set database
2), where layer one is the parameters entering the system, layer two is the set database layer of the layer of the fuzzy,
layer three
fuzzy, layeristhree
the fuzzy
is therule base
fuzzy rulestructure layer, layer
base structure four
layer, is the
layer decision
four is the making
decisionlayer,
making andlayer,
layer and
five
is the output defuzzification layer; more information is available in the literature
layer five is the output defuzzification layer; more information is available in the literature [25-27]. [25–27].

Figure2.2. Adaptive
Figure Neuro-FuzzyInterference
Adaptive Neuro-Fuzzy InterferenceSystem(ANFIS)
System(ANFIS) network
networkfor
fora atwo-input
two-input ‘Sugeno’
‘Sugeno’
fuzzy model. fuzzy model.

This system
This system can
can be
be explained
explained in in terms
terms of
of two
two suggested
suggested laws laws and
and literal
literal values
values in
in each
each input,
input,
considering the following five layers:
considering the following five layers:
Layer 1:1:where
Layer wherethethe
output is theisstep
output thetostep
maketoa specified
make a input
specified satisfy the verbal
input label
satisfy thecorresponding
verbal label
corresponding to the current node. In this layer, Gaussian membership functions verbal
to the current node. In this layer, Gaussian membership functions are used to represent values
are used to
as a connection of aggregate production limits (see Figure 3).
represent verbal values as a connection of aggregate production limits (see Figure 3).
First parameter function
 1 u − ai1 2 
 ! 
Ai (u) = exp−  (1)
2 bi1
Second parameter function:
 1 v − ai2 2 
 ! 
Bi (v) = exp−  (2)
2 bi2 

where {ai , . . . , bi2 }4 denote set of parameters.

Appl. Sci. 2019, 9, 3916 5 of 12
Appl. Sci. 2019, 9, x FOR PEER REVIEW 5 of 12

(a)

(b)
Figure 3. Initial
Figure 3. Initial and
and final
final membership
membership functions
functions of stone strength
of stone strength (s);
(s); (a)
(a) initial,
initial, (b)
(b) final.
final.

First
As parameter
the function
limit values of modification, the shapes of the model vary considerably as seen below,
displaying several forms of the functions on linguistic − 𝑎 Ai , Bi . The parameters in this layer are
1 𝑢tags
described as attitude limits. 𝐴 (𝑢) = 𝑒𝑥𝑝 − (1)
2 𝑏
In Layer 2, each node calculates the firing strength of the related law. Here, the nodes are called
Second
the ‘rule parameter
nodes’. function:
The outputs of the top and the bottommost neurons are as follow:
1 𝑣−𝑎
Upper (𝑣) = 𝑒𝑥𝑝 −
𝐵 neuron α12 = A𝑏1 (x) × B1 (y) (2)
(3)

where {ai, …………., bi2}4 denote Bottom parameters. α2 = A2 (x) × B2 (y)

set ofneuron (4)
AsLayer
In the limit values
3, every of in
node modification, the shapes by
this layer is considered of N,
thewhich
modelindicates
vary considerably as seen
the regulation of thebelow,
firing
displaying several forms of the functions on linguistic tags A i, Bi. The parameters in this layer are
powers. The output of the upper and bottom neuron is made normal as follow:
described as attitude limits.
In Layer 2, each node calculates Topthe firing strength
neuron β1 =of theα1related law. Here, the nodes are called (5)
the ‘rule nodes’. The outputs of the top and the bottommost α1 neurons
+ α2 are as follow:
Upper neuron
Bottom neuron αβ12==A1(x)α×2 B1(y) (3)
(6)
α1 + α2
Bottom neuron α2 = A2(x) × B2(y) (4)
Layer 4 provides the upper and lower neuron outputs as the result of normal firing intensity and
In particle
inbuilt Layer 3,energy
every for
node
thein this layer is considered by N, which indicates the regulation of the
process.
firing powers. The output of the upper and bottom neuron is made normal as follow:
Upper neuron β1 z1 = β1 (a1 x + b1 y) (7)
Top neuron 𝛽 = (5)
Bottom neuron β2 z2 = β2 (a2 x + b2 y) (8)
Bottom neuron 𝛽 = (6)
Appl. Sci. 2019, 9, 3916 6 of 12

Layer 5 where the system overall output is determined by each node summing all the incoming
signals. i.e.,
z= β1 z1 + β2 z2 (9)

Here, the combination set of neutral parameters are learned after providing the system training
set (xk , yk ), k = 1, . . . , K . The corresponding sensed function for iterated k is determined by


Ek = ( yk − ok )2 (10)

where yk is the expected output and ok is the calculated output by the improved neural net.
The ANFIS model was built in MATLAB using a set of 32 readings (provided in Table 1). Various
related functions were used to learn ANFIS; among them, two were closed side sets ((CSS), Gape G),
and the reduction ratio (RR) and four functions of the rock strength (S) were selected to generate the
ANFIS model.

Table 1. Measured power consumption at different crushing conditions.

No. Reduction Ratios Gape (mm) CSS (mm) Strength Power Consumption kWh/t
1 5.697 0.103
2 7.798 0.161
3 18.576 0.02
4 9.899 0.094
5 22.493 0.018
6 12.931 0.001
7 9.994 0.008
8 1.5 284 31.75 26.662 0.16
9 9.211 0.139
10 7.067 0.045
11 8.893 0.141
12 12.96 0.198
13 11.293 0.208
14 11.461 0.149
15 10.008 0.11
16 8.71 0.079
17 6.097 0.106
18 7.205 0.213
19 9.098 0.23
20 11.99 0.359
21 12.598 0.307
22 6.567 0.321
23 6.696 0.138
24 2.97 16 12.129 0.091
224
25 10.558 0.178
26 18.164 0.169
27 13.233 0.169
28 13.902 0.313
29 12.874 0.454
30 12.269 0.212
31 9.72 0.282
32 4.863 0.148

This study presents a jaw crusher, which is a key tool in the dry process of making a cement plant.
The procedure involves inputting the raw material for producing the cement into the jaw crusher.
The size of the used rocks for the procedure on average amounted to a similar quantity for all of the
experimentations conducted. The settings of the jaw crusher were set as shown in Table 1 so that all
values in the experiment could be based purely on strength. Two different sets of experiments were
 25 10.558 0.178
28
26 13.902
18.164 0.313
0.169
29
27 12.874
13.233 0.454
0.169
30
28 12.269
13.902 0.212
0.313
31
29 9.72
12.874 0.282
0.454
Appl. 32
Sci. 2019, 9, 3916 4.863 0.148 7 of 12
30 12.269 0.212
31 9.72 0.282
From the Gaussian membership function, the lowermost error of power consumption was
32
conducted, and the results are recorded in Table 1. In each set 4.863 0.148data were fixed
of tests, all of the input
determined to be implemented for ANFIS training. The input–output system construction of ANFIS
apart from the strength.
whenFromGaussian membership function is used and entails 32 fuzzy rules produced from the related
From the Gaussian membership
membership function,
function, the
the lowermost
lowermost error
error of power consumption
consumption was
system of the
determined toinput–output dataset
be implemented corresponding
for ANFIS training.to the Sugeno fuzzysystem
The model, as shown in Figure 4.
The input–output
input–output system construction of ANFIS
whenDuring
Gaussian the training stage,
Gaussian membership
32 efficiencies
membership function
function is
is used
of the performance
used and
and entails
entails 32
32 fuzzy
(training dataset)
fuzzy rules
rules produced
were used to
produced from the related
conductof500
system theintervals of learning,
input–output datasetaccommodating
corresponding an estimated error of 0.0755, as shown in Figure
Figure 4.5.
to the Sugeno fuzzy model, as shown in Figure 4.
During the training stage, 32 efficiencies of the performance (training dataset) were used to
conduct 500 intervals of learning, accommodating an estimated error of 0.0755, as shown in Figure 5.

Figure 4. Fuzzy rule architecture of the Gaussian membership function

Figure 4. Fuzzy rule architecture of the Gaussian membership function
During the training stage, 32 efficiencies of the performance (training dataset) were used to
conduct 500 intervals of learning,
Figure accommodating
4. Fuzzy rule an estimated
architecture of the error of 0.0755,
Gaussian membership as shown in Figure 5.
function

Figure 5. Training error.

2.2. Multiple Regression Model

Figure
Figure 5.
5. Training
Training error.
error.
Several engineering problems include identifying the association between two or more
2.2.
2.2. Multiple
Multiple
variables, Regression
inRegression Model as correlation [28]. Finding how the variables are correlated is called
Model
this case, known
regression
Severalanalysis;
Several theseproblems
engineering
engineering are powerful
problems includestatistical
include methods
identifying
identifying that association
the association
the have between
been employed
two or more
between by
tworesearchers
variables,
or more
in many
this areas[29].
case, known Regression
as correlationmodeling
[28]. is the
Finding process
how the by which
variables correlation
are correlated
variables, in this case, known as correlation [28]. Finding how the variables are correlated is called are fitted
is called to data. In a
regression
nonlinear
analysis; model
these of
are regression,
powerful considered
statistical as
methods a special
that case,
have the
been nonlinear
employed
regression analysis; these are powerful statistical methods that have been employed by researchers independent
by researchersfunctions
in are
many
used
areas
in manyto
[29].work on
Regression
areas[29]. the data, and
modeling
Regression the output
is the process
modeling values
is theby are
whichby
process compared
correlation from the values.
are fitted are
which correlation to data. In this
fitted toa data. study,
nonlinear
In a
multiple-input
model of variables
regression, were
considered used;
as a therefore,
special we
case, theused the
nonlinear “multiple nonlinear
independent
nonlinear model of regression, considered as a special case, the nonlinear independent functions regression”
functions are used function.
to work
are
In this
on
used event,
thetodata,
work andthe
on multiple
the
the data,nonlinear
output values
and theare regression
compared evaluates
from
output values the the relationship
are values.
compared In this between
study,
from the four
multiple-input
values. Invariables
thisvariables
study,by
finding a
multiple-input nonlinear
were used; therefore, fit
variables equation
we used
were the to the
“multiple
used; calculated
therefore,nonlinear data.
we usedregression” Multiple
the “multiple non-linear
function. In this
nonlinear regression includes
event, thefunction.
regression” multiple
refining
nonlinear the data
regressionand examining
evaluates the the correlation
relationship between
between four all variables[30].
variables by The
finding
In this event, the multiple nonlinear regression evaluates the relationship between four variables by ageneralized
nonlinear fitnature
equation of
a multiple
finding non-linear
to the calculated
a nonlinear regression
data.fitMultiple model
equationnon-linear is shown in
regression
to the calculated Equation
includes
data. (11):
refining
Multiple the dataregression
non-linear and examining includesthe
correlation between all variables [30]. The generalized nature of a multiple
refining the data and examining the correlation between all variables[30]. The generalized nature ofnon-linear regression model
aismultiple
shown innon-linear
Equation regression
(11): model is shown in Equation (11):
=n,
iX
b j+1
Y = b0 + b j Xi j = 2i − 1 (11)
i=1

where Y is the output; Xi is the independent variables of the system; and b0 , b1 , . . . , bn are the
correlation parameters. The coefficients were then trained to coincide with the model’s output and the
training output.
3. Results and Discussion

3.1. ANFIS Model

Figure 3a,b illustrate the original and the last membership functions of the stone strength. It was
noted that
Appl. Sci. tuning
2019, 9, 3916the final membership function led to remarkable changes in the low and high areas, 8 of 12
but in the small and medium regions, there were minor changes. The significant changes in the very
low and high areas indicated that all ranges of stone strength had a different effect on energy
3. Results and(E).
consumption Discussion
Additionally, Figure 3 shows that stone strength had the most significant impact
on energy consumption.
3.1. ANFIS Model
Figures 6 and 7 show the effects of the crushing parameters and stone material properties on
energy consumption.
Figure According
3a,b illustrate to Figures
the original and the 6 and
last 7, the reduction
membership ratio (RR),
functions of thegap (G),strength.
stone and stone It
strength
was noted (S)that
hadtuning
a considerable effect on energy
the final membership consumption,
function while thechanges
led to remarkable closed side setlow
in the (CSS)
andhad
high a
minor impact
areas, but in theonsmall
energyandconsumption.
medium regions, there were minor changes. The significant changes in the
very Inlow Figure
and high6 at areas
a lowindicated
strengththat
level,
all it can be
ranges of seen
stonethat the closed
strength side set effect
had a different did not on have
energy a
considerable
consumptioneffect such as at the
(E). Additionally, high 3level
Figure showsof strength,
that stoneand the consumption
strength had the mostenergy increased
significant with
impact on
the increase
energy in the closed side set. Moreover, the energy consumption increased with the decrease in
consumption.
stoneFigures
strength. 6 and 7 show the effects of the crushing parameters and stone material properties on
energyFigure 7 shows the
consumption. variationtoofFigures
According energy 6consumption with theratio
and 7, the reduction variation
(RR), in
gapgape
(G),width and
and stone
reduction
strength (S)ratio.
had aNote the gap effect
considerable rangeonofenergy
230–250 mm; the energy
consumption, consumption
while the closed side setwas(CSS)
directly
had
proportional
a minor impact to the reduction
on energy ratio across the whole range of the reduction ratio.
consumption.

Appl. Sci. 2019, 9, x FOR PEER REVIEW 9 of 12

Figure 6.
Figure Energy (E)
6. Energy (E) in
in relation
relation to
to change
change of
of closed
closed side
side set
set (CSS)
(CSS) and
and strength
strength (S).
(S).

Figure 7. Energy (E) in relation to change of gape (G) and the reduction ratio (RR).
Figure 7. Energy (E) in relation to change of gape (G) and the reduction ratio (RR).
In Figure 6 at a low strength level, it can be seen that the closed side set did not have a considerable
3.2. ANFIS Model Verification
effect such as at the high level of strength, and the consumption energy increased with the increase in
the closed side set. power
The predicted Moreover, the energy(Ep)
consumption consumption
versus the increased
measuredwith
powerthe(Em)
decrease in stone strength.
consumption from the
Figure
real case 7 shows
study the variation
are compared of energy
in Table 2 andconsumption
Figure 8. with the variation in gape width and reduction
ratio.ANote
grouptheofgap
real dataofcomprised
range 230–250 mm;of eight cases was
the energy used to run
consumption wasthe ANFIS,
directly then the ANFIS
proportional to the
provided
reduction the predicted
ratio power
across the wholeconsumption.
range of the Next, the Ep
reduction obtained by the ANFIS was compared with
ratio.
the measured Em. It was noticed that the maximum deviation was less than 6.5%, thus the ANFIS
model provides a comparable value of power consumption that was very close to the actual values.
The error per cent Ei for any sample of data i (i varied from 1 to m, here m = 8) between the
predicted values of energy by the ANFIS model (Epi) and the measured values (Emi) was estimated
from the following equation:
⌊𝐸𝑚 − 𝐸𝑝 ⌋
𝐸 = 100 (12)
𝐸𝑚
While the corresponding average error percent Eav was computed using the following relation:
 Figure 7. Energy (E) in relation to change of gape (G) and the reduction ratio (RR).

Appl. Sci. 2019, 9, 3916 9 of 12

3.2. ANFIS Model Verification
The predicted power consumption (Ep) versus the measured power (Em) consumption from the
3.2. ANFIS Model Verification
real case study are compared in Table 2 and Figure 8.
A group
The of real
predicted dataconsumption
power comprised of eight
(Ep) cases
versus thewas used topower
measured run the
(Em)ANFIS, then thefrom
consumption ANFIS
the
provided
real the predicted
case study powerinconsumption.
are compared Next, the
Table 2 and Figure 8. Ep obtained by the ANFIS was compared with
the measured Em. It was noticed that the maximum deviation was less than 6.5%, thus the ANFIS
modelTable 2. The
provides a ANFIS predicted
comparable powers
value versusconsumption
of power the measured that
energy
wasconsumed by the
very close jawactual
to the crusher.values.
The error per cent Ei for any sample of data i (i varied from 1 to m, here m = 8) between the
Parameters Power Consumption kWh/t
predicted
Test No.values of energy by the ANFIS model (Epi) and the measured values (Emi) was estimated
Error (%)
RR Gape (mm) CSS (mm) S (MPa) Measured E Predicted E
from the following equation:
1 21.666 0.197 0.185 6.09
𝐸𝑝 ⌋
⌊𝐸𝑚 −8.33
2
1.5 284 𝐸 =
31.75 100 0.114 0.109 4.39 (12)
3 𝐸𝑚
6.286 0.06 0.061 1.67
4 8.069 0.094 0.1 6.38
While the corresponding average error percent Eav was computed using the following relation:
5 4.897 0.033 0.034 3.03
6 ∑ 𝐸
8.071 0.152 0.153 0.66 (13)
2.97 224 16 𝐸 =
7 𝑚
8.635 0.336 0.351 4.46
8 11.021 0.379 0.366 3.43
Based on the average error per cent provided in Table 2, the ANFIS model successfully predicted
the power consumption with a 3.67% deviation from the measuredAverage Errorthe ANFIS model
data. Thus, 3.76 with
gaussmf had an accuracy of more than 96% to predict the energy consumedR2 by the jaw crusher.
0.9947

Figure 8. Error
Figure 8. Error bars
bars for
for the
the predicted
predicted energy
energy with
with the
the help
help of
of the
the ANFIS
ANFIS model
model versus
versus the
the measured
measured
energy values for different test data.
energy values for different test data.

A group of real data comprised of eight cases was used to run the ANFIS, then the ANFIS provided
the predicted power consumption. Next, the Ep obtained by the ANFIS was compared with the
measured Em. It was noticed that the maximum deviation was less than 6.5%, thus the ANFIS model
provides a comparable value of power consumption that was very close to the actual values.
The error per cent Ei for any sample of data i (i varied from 1 to m, here m = 8) between the
predicted values of energy by the ANFIS model (Epi) and the measured values (Emi) was estimated
from the following equation:
bEmi − Epi c
Ei = × 100 (12)
Emi
While the corresponding average error percent Eav was computed using the following relation:
Pm
i=1 Ei
Eav = (13)
m
 2 8.33 0.114 0.109 4.39
1.5 284 31.75
3 6.286 0.06 0.061 1.67
4 8.069 0.094 0.1 6.38
5 4.897 0.033 0.034 3.03
6 8.071 0.152 0.153 0.66
2.97
Appl. Sci. 2019, 9, 3916 224 16 10 of 12
7 8.635 0.336 0.351 4.46
8 11.021 0.379 0.366 3.43
Based on the average error per cent provided in Table 2, the Average
ANFIS Error
model successfully3.76 predicted
the power consumption with a 3.67% deviation from the measured data. R2 Thus, the ANFIS 0.9947
model with
gaussmf had an accuracy of more than 96% to predict the energy consumed by the jaw crusher.
3.3. Working with the Regression Model
3.3. Working with the Regression Model
The regression system was designed, and the regression statistics of the Energy (E) were also
The regression
calculated using MATLAB systemandwas
thedesigned,
results areand the regression
as shown in Figurestatistics of the
9 and Table 3. ItEnergy (E) were also
can be summarized
calculated
that the value using MATLAB
of the andcorrelation,
coefficient the results are
R, ofas the
shown in Figure
energy 9 andwas
prediction Table 3. It0.41.
about can be summarized
This indicates
that the value of the coefficient correlation, R, of the energy prediction was about
that the regression system of best fit accounted for 42% of the variability of energy. The inferences0.41. This indicates
thatthe
and thecalculated
regression system
values of best fit
of varying accounted
energy valuesfor are42%
shownof the variability
below, of energy.
which further The that
indicate inferences
these
and the calculated values of varying energy values are shown below, which
values varied with each other as opposed to the ANFIS model. This may be the result of the varying further indicate that
these values varied with each other as opposed to the ANFIS model. This may
coefficients of correlation between the independent variables. This can well be shown by the ANFIS be the result of the
varying coefficients of correlation between the independent variables. This can well be shown by the
systems.
ANFIS systems.

0.3

0.25
Predicted Energy

0.2

R² = 0.4128
0.15

0.1

0.05

0
0 0.1 0.2 0.3 0.4 0.5
Measured Energy

Figure 9. Comparing calculated and inference energy consumption by the regression model.

Figure
Table 9. Comparing
3. Regression calculated
statistics and inference
and coefficient energy
values consumption
for the by the regression
regression equation of energymodel.
prediction.

Regression
Table 3. Regression Statistics
statistics Coefficient
and coefficient values for the regression equation of energy prediction.
Root Mean Squared Error 0.086 Intercept b0 = −79.4983
Regression Statistics Coefficient
2
R Squared Error 0.413
Root Mean 0.086 reduction
Interceptratios bb1 0== −14.417,
−79.4983b2 = 0.30816
Adjusted R2 0.413
0.35 reduction
Gaperatios b3 = 0.7962,
b1 = −14.417, b4 = 0.9599
b2 = 0.30816
Adjusted R2
p-value 0.35
0.00169 Gape
Css b3 = 0.7962, b4 = 0.9599
b5 = −3.0399, b6 = 0.961
p-value 0.00169 Css b5 = −3.0399, b6 = 0.961
Strength
Strength b7 = 70.200,
b7 = 70.200, b8 = −4.117
b8 = −4.117

4. Conclusions
In this work, the ANFIS model with gaussmf was used to obtain an accuracy relation to estimate
the jaw crusher power consumption. The predictor input data included the closed side set, gap, stone
strength, and targeted reduction ratio. A set of 32 specific energy consumption values were measured
at various crushing conditions to generate the corresponding ANFIS model. Then, another set of eight
measured data were used to evaluate the accuracy of the generated model to provide the specific
energy consumption. Based on the average error percent, it was found that the ANFIS model with
gaussmf had a high level of accuracy (more than 96%) to predict the specific energy consumption of jaw
crusher. It was observed that the ANFIS model was more efficient than the regression model in terms
of R2 . Alternatively, the results showed that in some cases, the regression analysis, although a standard
Appl. Sci. 2019, 9, 3916 11 of 12

method in the modeling of such systems, failed to be reliable (R2 = 0.41), and hence, high-end models
like the ANFIS model are preferred. The current study results proved the effectiveness of ANFIS as
an accurate means of predicting the amount of energy applicable in the process of crushing. Therefore,
the management team could further use the new model to assess the consumption of power early
enough to enable them to take any necessary measures to avoid mistakes. The usefulness of the
developed model could be improved further by a support system of decisions to assist the professionals
working in the cement production plant.

Conflicts of Interest: The authors declare no conflicts of interest.

References
1. Ashby, M.F.; Shercliff, H.; Cebon, D. Materials: Engineering, Science, Processing and Design;
Butterworth-Heinemann: Oxford, UK, 2018.
2. King, R.P. Modeling and Simulation of Mineral Processing Systems; Elsevier: Amsterdam, The Netherlands, 2012.
3. Indraratna, B.; Sun, Y.; Nimbalkar, S. Laboratory assessment of the role of particle size distribution on the
deformation and degradation of ballast under cyclic loading. J. Geotech. Geoenviron. Eng. 2016, 142, 04016016.
[CrossRef]
4. Mwangi, P.N.u.; Muvengei, O.M.; Mbuya, T.O. Review of Discrete Element Modelling in Optimisation
of Energy Consumption of a Single-Toggle Jaw Crusher. In Proceedings of the Sustainable Research and
Innovation Conference, Juja, Kenya, 2–4 May 2018; pp. 51–259. Available online: http://sri.jkuat.ac.ke/ojs/
index.php/proceedings/article/view/684 (accessed on 18 September 2019).
5. Ulsen, C.; Tseng, E.; Angulo, S.C.; Landmann, M.; Contessotto, R.; Balbo, J.T.; Kahn, H. Concrete aggregates
properties crushed by jaw and impact secondary crushing. J. Mater. Res. Technol. 2019, 8, 494–502. [CrossRef]
6. Weiss, N. SME Mining Engineering Handbook, 3rd ed.; American Institute of Mining, Metallurgical and
Petroleum Engineering Inc.: Littleton, CO, USA, 2002.
7. Revuelta, M.B. Mineral Processing. In Mineral Resources; Springer: Berlin, Germany, 2018; pp. 423–530.
8. Sarkar, S.; Kumar, R.; Murthy, V. Experimental and Numerical Simulation of Crack Propagation in Sandstone
by Semi Circular Bend Test. Geotech. Geol. Eng. 2019, 37, 1–13. [CrossRef]
9. Donovan, J.G. Fracture Toughness Based Models for the Prediction of Power Consumption, Product Size, and Capacity
of Jaw Crushers; Virginia Tech: Blacksburg, VA, USA, 2003.
10. Eloranta, J. The efficiency of blasting versus crushing and grinding. In Proceedings of the 23rd Conference of
Explosives and Blasting Technique, Las Vegas, NV, USA, 2–6 February 1997.
11. Bearman, R.; Barley, R.; Hitchcock, A. Prediction of power consumption and product size in cone crushing.
Miner. Eng. 1991, 4, 1243–1256. [CrossRef]
12. Zhang, S.; Huang, Z.; Wang, H.; Li, G.; Sepehrnoori, K.; Wu, X.; Hong, C. Experimental study on the
rock-breaking characteristics of abrasive liquid nitrogen jet for hot dry rock. J. Pet. Sci. Eng. 2019. [CrossRef]
13. Bearman, R.T. Jaw and Impact Crushers. In SME Mineral Processing and Extractive Metallurgy Handbook;
Society for Mining, Metallurgy, and Exploration, Inc.: Englewood, CO, USA, 2019; p. 367.
14. Hofler, A.A. Fundamental Breakage Studies of Mineral Particles with an Ultrafast Load Cell Device. Ph.D.
Thesis, University of Utah, Salt Lake City, UT, USA, 1992.
15. Napier-Munn, T.J. Mineral Comminution Circuits: Their Operation and Optimization. Julius Kruttschnitt
Mineral Research Centre: Indooroopilly, Australia, 1996.
16. Svensson, A.; Steer, J. New cone crusher technology and developments in comminution circuits. Miner. Eng.
1990, 3, 83–103. [CrossRef]
17. DeDiemar, R. New concepts in Jaw Crusher technology. Miner. Eng. 1990, 3, 67–74. [CrossRef]
18. Legendre, D. Numerical and Experimental Optimization Analysis of a Jaw Crusher and a Bubble Column Reactor;
Åbo Akademi University: Turku, Finland, 2019.
19. Larison, B.W. Aggregate Production Modeling Using Neural Networks and Belief Networks. Master’s
Thesis, University of Alberta, Edmonton, AB, Canada, 1999.
20. Shi, J.J. Mathematical Models for Maximizing Aggregate Plant Production. J. Constr. Eng. Manag. 1999, 125,
53–60. [CrossRef]
Appl. Sci. 2019, 9, 3916 12 of 12

21. Halabian, A.M.; Shamsabadi, P.J. Numerical modeling of the RAP construction process and its effects on
RAP behavior. Int. J. Geomech. 2014, 15, 04014085. [CrossRef]
22. Sivacoumar, R.; Raj, S.M.; Chinnadurai, S.J.; Jayabalou, R. Modeling of fugitive dust emission and control
measures in stone crushing industry. J. Environ. Monit. 2009, 11, 987–997. [CrossRef] [PubMed]
23. Wang, B.; Martin, U.; Rapp, S. Discrete element modeling of the single-particle crushing test for ballast stones.
Comput. Geotech. 2017, 88, 61–73. [CrossRef]
24. Karaboga, D.; Kaya, E. Adaptive network based fuzzy inference system (ANFIS) training approaches:
A comprehensive survey. Artif. Intell. Rev. 2018, 1–31. [CrossRef]
25. Kasabov, N.K. Foundations of Neural Networks, Fuzzy Systems, and Knowledge Engineering; Auckland University
of Technology: Auckland, New Zealand, 1996.
26. Fullér, R. Fuzzy systems. In Introduction to Neuro-Fuzzy Systems; Springer: Berlin, Germany, 2000; pp. 1–131.
27. Maher, I.; Eltaib, M.; Sarhan, A.A.; El-Zahry, R. Cutting force-based adaptive neuro-fuzzy approach for
accurate surface roughness prediction in end milling operation for intelligent machining. Int. J. Adv. Manuf.
Technol. 2015, 76, 1459–1467. [CrossRef]
28. Belegundu, A.D.; Chandrupatla, T.R. Optimization Concepts and Applications in Engineering; Cambridge
University Press: Cambridge, UK, 2019.
29. Weigend, A.S. Time Series Prediction: Forecasting the Future and Understanding the Past; Routledge:
London, UK, 2018.
30. Cox, D.R. Analysis of Binary Data; Routledge: London, UK, 2018.

© 2019 by the author. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).