Applied Sciences: Fficiency of Agricultural Tractors For
Applied Sciences: Fficiency of Agricultural Tractors For
Applied Sciences: Fficiency of Agricultural Tractors For
sciences
Article
Engine Speed Control System for Improving the Fuel
Efficiency of Agricultural Tractors for
Plowing Operations
Jin Woong Lee 1 , Su Chul Kim 2 , Jooseon Oh 3 , Woo-Jin Chung 3 , Hyun-Woo Han 3 , Ji-Tae Kim 3
and Young-Jun Park 3,4, *
1 Convergence Agricultural Machinery R&D Group, Korea Institute of Industrial Technology,
119 Jipyeongseonsandan 3-gil, Baeksan-myeon, Gimje-si, Jeollabuk-do 54325, Korea; [email protected]
2 Department of Smart Machine Technology, Korea Institute of Machinery and Materials, 156 Gajeongbuk-Ro,
Yuseong-gu, Daejeon 34103, Korea; [email protected]
3 Department of Biosystems and Biomaterials Science and Engineering, College of Agriculture and Life
Sciences, Seoul National University, Gwanak-ro 1, Gwanak-gu, Seoul 08862, Korea; [email protected] (J.O.);
[email protected] (W.-J.C.); [email protected] (H.-W.H.); [email protected] (J.-T.K.)
4 Research Institute for Agriculture and Life Science, Seoul National University, Gwanak-ro 1, Gwanak-gu,
Seoul 08862, Korea
* Correspondence: [email protected]; Tel.: +82-2-880-4602
Received: 1 August 2019; Accepted: 6 September 2019; Published: 17 September 2019
Abstract: This study was conducted to develop a load-sensitive engine speed control system to
maximize the fuel efficiency of an agricultural tractor. The engine speed controller was developed
through a model-based design approach using a tractor simulation model. The simulated engine
speed and torque values were measured with an average error range of 1.4–4.9% compared to
results obtained from field experiments. Using the tractor model, the gain parameters of the
proportional–integral (PI) controller were optimized under the step, ramp, and actual load conditions.
The simulation results using the actual load showed that the engine speed could be adjusted to
within 2–3% of the desired value using the proposed engine speed controller. The throttle control
system was constructed using four parts of a tractor engine, a microprocessor with an engine speed
control algorithm, a throttle actuator, and a data acquisition system. Using the developed system,
the operating engine speed values showed an average 1.17 % error compared to the desired engine
speed. Three fuel efficiency parameters were used for evaluating the fuel-saving performance of
the control system: specific volumetric fuel consumption (SVFC), fuel consumption per tilled area
(FCA), and fuel consumption per work hour (FC). The values for SVFC, FCA, and FC obtained
from the engine speed control system during plowing operations were 23.03–57.87%, 4.11–42.06%,
and −7.24–38.48%, respectively, showing an improvement over the same operations without the
control system.
Keywords: agricultural tractor; fuel-saving; control system design; control algorithm; load sensing
1. Introduction
Agricultural tractors are multi-purpose machines used for performing various types of field
operations. Different field operations subject agricultural tractors to varying workloads, which results
in varying levels of fuel consumption under different conditions. These conditions include engine
speed, soil condition, implementation type, and transmission gear step. Even though the same field
operations may be conducted, workload and fuel consumption characteristics change frequently, and
improper adjustments to operating conditions decrease tractor fuel efficiency. Engine speed and
transmission gear selection are the most significant factors that affect the fuel efficiency of agricultural
tractors [1]. Therefore, in order to improve fuel efficiency, each field operation requires proper
adjustment of transmission gear selection and engine speed.
The fuel efficiency of agricultural tractors can be maximized by operating the engine at full load
conditions. To operate under full load conditions, engine speed, transmission gears, and workload
should be adjusted appropriately. Park et al. reported that during field operations in Korea, tractors
were operated in the 24% for plowing and 41% for rotary tillage overpowered condition due to the
usage of inappropriate transmission gear and engine speed [2]. In order to develop an integrated
control system for maximizing fuel consumption efficiency, an engine speed control system should
be designed to maintain the maximum economic speed of the diesel engine according to workload.
Developing an engine speed control system requires the maximization of fuel efficiency and the
incorporation of a method that allows engine speed to be controlled automatically.
Various methods have been proposed to maximize the fuel efficiency of tractors using engine
speed adjustment control. Park et al. designed a passive eco-driving system in order to assist tractor
operators for selecting the optimal operating engine speed and transmission gear depending on the
workload. They reported that fuel consumption can be reduced by an average of 69% and 54% in
plowing and rotavating operations [2]. Chancellor and Thai also proposed an automatic system
to control the engine speed and transmission gear ratio as optimal conditions using a mechanical
instrument for adjusting the governor in five discrete speed ranges. The results showed that the
fuel consumption can be reduced by 5–12%, compared to that of a good operator [3]. Zhang et al.
developed a simple mechanical engine throttle control device for maintaining an optimal engine speed
based on the measured power take-off (PTO) load. The operation of this device was based on a specific
volumetric fuel consumption map. They reported that more than 8% of specific fuel consumption could
be reduced by using the throttle control device [4]. Jiang designed a diesel engine speed controller
using an optimal gain scheduling proportional–integral–derivative (PID) controller to improve engine
performance and reduce fuel consumption. The engine operating region was divided into 15 sections
by engine load and speed, and controller gains were determined through experiments. The fuel-saving
performance of the system was reported to be better than existing mechanical speed governor [5].
The objective of this study is to develop a load-sensitive engine speed control system to maximize
fuel efficiency. In order to measure workload, torque transducer was attached to input shaft of
transmission. This transducer can measure applied load torque to engine in real time not only PTO
load but also draft load. Engine speed controller was developed using model-based controller design
method. Simple mechanical system was designed for controlling mechanical fuel pump governor to
engine speed within optimal engine speed point based on real time input shaft torque. To validate the
performance of the system, an experiment was conducted by comparing plowing operations performed
with and without the use of the control system.
2. Controller Design
In this study, the engine speed controller was developed through a model-based design using
a tractor simulation model. The engine speed controller is composed of the tractor longitudinal
dynamics model with engine performance map, proportional–integral (PI) controller, and desired
engine speed determination algorithm for calculating the optimal engine speed. Matlab, Simulink
R2015a (MathWorks, Inc., Natick, MA, USA) was used for designing the engine speed controller.
Figure 1 shows a simple block diagram of the engine speed controller using Matlab Simulink [6].
Appl. Sci. 2019, 9, 3898 3 of 19
Appl. Sci. 2019, 9, x FOR PEER REVIEW 3 of 20
Figure1.1. Block
Figure Block diagram
diagramof
of engine
enginespeed
speedcontroller.
controller.
2.1.
2.1. Tractor
TractorDynamic
DynamicSimulation
SimulationModel
Model
Figure 1. Block diagram of engine speed controller.
AAschematic
schematicrepresentation
representationof ofthe
thedynamic
dynamicsimulation
simulationmodel
modelof ofthe
thetractor
tractorused
usedfor
forcalculating
calculating
2.1.performance
its Tractor Dynamic Simulation
parameters is Model in Figure 2. The simulation model solves the mathematical
shown
its performance parameters is shown in Figure 2. The simulation model solves the mathematical
equations
equations using
A schematic
using therepresentation
previously
the calculated
previously engine
ofcalculated
the dynamic speed,
engine throttle
simulation
speed, command,
model of the
throttle and workload
tractor
command, used
andforto determine
calculating
workload to
the
its newly calculated
performance
determine engine
parameters
the newly speed. The
is shown
calculated engine
engineinspeed. model
FigureThe was
2. The designed
simulation
engine model wasusing
model the look-up table
solves using
designed module
the mathematical
the look-up in
Matlab/Simulink,
equations
table module using and
the engine torquecalculated
previously
in Matlab/Simulink, wasengine
and defined as a function
engine
torque speed,
was ofthrottle
definedrotational speed of
as a command,
function of engine andspeed
and workload
rotational throttle
to
of
opening,
determine given
engine andthe in Equation
newlyopening,
throttle (1).
calculated The tractor
engine
given simulation
speed. The
in Equation model
engine
(1). The was
model
tractor based on
was designed
simulation a 48 kW
model using MFWD diesel
the look-up
was based on a 48
tractor
table
kW MFWD (Tongdiesel
module Yang Moolsan,
in Matlab/Simulink,Seoul,
tractor (Tong South
and
Yang Korea).
engine
Moolsan, Specifications
torque
Seoul,was defined
South ofas
Korea).the tractor areofgiven
aSpecifications
function in Table
rotational
of the 1 [7].
speed
tractor of
are
engine
given inand throttle
Table 1 [7].opening, given in Equation (1). The tractor simulation model was based on a 48
Te = f (Ne , θ) (1)
kW MFWD diesel tractor (Tong Yang Moolsan, Seoul, South Korea). Specifications of the tractor are
given in Table 1 [7].
where Te is engine torque in N·m; Ne is engine rotational speed in RPM; θ is throttle opening in %.
Figure
Figure 2.2. Block
Table 1. Specifications diagram
of MFWD
Block of
of tractor
tractor
diagram dynamic
used for
tractor model.
modelling
dynamic in this study [7].
model.
Table 1. Specifications of MFWD tractor used for modelling in this study [7].
Item
Table 1. Specifications of MFWD tractor used for modellingSpecification
in this study [7].
Item kN
Unballasted weight, Specification
24.4
Item Specification
Unballasted
Weight distribution weight,
(Front: kN%
Rear), 24.4
47.6: 52.4
Unballasted weight,(Front:
Weight distribution kN Rear), % 47.6: 52.424.4
Transmission (main/sub/creep)
Transmission (main/sub/creep) Forward
Forward 24, Backward
24, Backward 24 (4/3/2)
24 (4/3/2)
Weight distribution (Front: Rear), % 47.6: 52.4
Manufacturer and Model
Manufacturer and Model Caterpillar
Caterpillar C3.3TC3.3T 2232/2200
2232/2200
Engine
EngineTransmission (main/sub/creep) Forward
Rated power, kW 48 kW 24, Backward
@ 2200 rpm 24 (4/3/2)
Rated power, kW 48 kW @ 2200 rpm
Manufacturer
Ratedand Model
power, kW Caterpillar
47.61C3.3T 2232/2200
Engine PTO Rated power, kW 47.61
PTO Rated speed,
Rated power, kW rpm 540, 733 @ 48 kWengine
rated @ 2200speed
rpm
Rated speed, rpm
Front 540, 733
11.2-20 8PR@(drive
rated tires)
engine speed
Tire Rated power, kW 47.61
PTO FrontRear 14.9-30 8PR (drive
11.2-20 tires) tires)
8PR (drive
Tire Rated speed, rpm 540, 733 @ rated engine speed
Rear 14.9-30 8PR (drive tires)
The engine torque for theFront
simulation model was based 11.2-20
on the PTO8PRperformance
(drive tires) data of the
Tire
T623 tractor tested by the OECD Reartest code considering a power 14.9-30 8PR (drive efficiency
transmission tires) of 90%.
𝑇𝑒 = 𝑓 (𝑁𝑒 , 𝜃)
The full-load torque with full throttle was expressed as a second order polynomial of engine speed, (1)
where 𝑇𝑒 is engine torque in N·m; 𝑁𝑒 is engine rotational speed in RPM; 𝜃 is throttle opening in %.
𝑇𝑒 = 𝑓 (𝑁𝑒 , 𝜃) (1)
where 𝑇𝑒 is engine torque in N·m; 𝑁𝑒 is engine rotational speed in RPM; 𝜃 is throttle opening in %.
Appl. Sci. 2019, 9, 3898 4 of 19
and the partial-load torque with both full and reduced throttles (θ) by a linear function of engine
speed (Ne ) [8–13]. The engine torque from Equation (1) was used to calculate the tractive force while
considering the applied engine load torque, the selected transmission gear ratio, overall transmission
efficiency, and draft load using Equations (2) and (3) [9–13].
(Te × ηt × Gr )
TF = − FL (2)
rd
FL × rd
TL = (3)
Gr × ηt
where TF is tractive force in N; TL is applied engine load torque; ηt is transmission efficiency; Gr . is
transmission gear ratio; rd is dynamic radius of driving wheel in m; FL is draft load in N.
Finally, Equation (2) can be modified to the equation shown below using Equations (2) and (3).
(Te − TL ) × ηt × Gr
TF = (4)
rd
The actual velocity of the tractor was calculated from the following equations using the draft force
and the equivalent mass of the tractor. meq was calculated using Equation (6). It is the summation of
the mass of the tractor chassis and moment of inertia of the rotating part of transmission components,
which is divided by the dynamic radius of driving wheel square. The tractor mass (mt ) and dynamic
radius of rear wheel (rd ) for the simulation model was 2485 kg and 0.665 m based on the T623 tractor
OECD test report. In order to determine the moment of inertia of the rotating part of transmission (Ir ),
engine speed rising time from idle to a certain speed was measured with the gear engaged and engine
torque was also measured during engine speed acceleration with full throttle open. Measured engine
speed rising time and engine torque were used to calculate engine speed angular acceleration (α).
For excluding the effect of load torque and tractor mass, the rear wheel of the tractor was suspended
in air. Finally, Equation (7) was used for calculating the moment of inertia of the rotating part of
transmission (Ir ).
.
meq × v = TF (5)
Ir
meq = mt + (6)
rd 2
Ir × α = (Te − TL ) × ηt × Gr (7)
where meq is the summation of equivalent mass of tractor and rotating part of transmission components
.
in kg; v is the derivative of actual velocity in m s−2 ; α is angular acceleration of wheel in rad s−2 ; mt is
tractor mass in kg; Ir is moment of inertia of the rotating part of transmission components in kg m2 .
.
The actual velocity of the tractor was calculated by integrating the calculated v from Equation (5).
The theoretical velocity of the tractor was calculated using Equations (8)–(9) using measured driving
wheel slip conditions for 0.1 based on field experiment results. The engine speed for Equation (1) was
obtained by multiplying the theoretical velocity and the transmission gear ratio [9–13].
va = (1 − s) × vt (8)
Z
.
vt = v0 + v dt (9)
where va is the actual velocity of the tractor in m s−1 ; vt is the theoretical velocity of the tractor in m s−1 ;
s is slip of driving wheels (decimal); v0 is initial velocity of the tractor in m s−1 .
In order to characterize the behavior of the diesel engine and the throttle control motor at nominal
operating speed, time delays are required in the model [5]. To determine time delays for the diesel
engine, the step input was applied to the diesel engine using the throttle command from 0 to 100%,
Appl. Sci. 2019, 9, x FOR PEER REVIEW 5 of 20
100%,
Appl. Sci.and the engine speed response
9, 3898 time was measured finally. For the throttle control motor, ofthe
Appl. 2019,
Sci. 2019, 9, x FOR PEER REVIEW 5 of5 20 19
step input was also applied, and the response time measured. Thus, the time delay of the diesel
engine
100%,(0.4
ands)the
andengine
the throttle
speed control
responsemotor
time (0.15 s) are represented
was measured in the
finally. For the throttle
feedback gain of
control the engine
motor, the
and
speed the engine speed response time was measured finally. For the throttle control motor, the step input
stepcontroller.
input was also applied, and the response time measured. Thus, the time delay of the diesel
was also applied, and the response time measured. Thus, the time delay of the diesel engine (0.4 s) and
engine (0.4 s) and the throttle control motor (0.15 s) are represented in the feedback gain of the engine
the throttle
2.2.speed
Desired control motorDetermination
Engine Speed
controller. (0.15 s) are represented
Algorithm in the feedback gain of the engine speed controller.
The desired
2.2. Desired Engineengine speed requiredAlgorithm
Speed Determination to maximize fuel efficiency was determined by an optimum
2.2. Desired Engine Speed Determination Algorithm
fuel consumption curve as a function of the applied engine load torque (y-axis), which was measured
at theTheThedesired
desired
transmission engine
inputspeed
engine shaft.required
speed required
A method to
to maximize
maximize
proposedfuel fuel efficiency
efficiency
by Kim was
et al.was determined
wasdetermined byby
used to construct anan optimum
optimum
a specific
fuel consumption
fuel consumption
volumetric curve
curveasasaafunction
fuel consumption (SVFC) of
function ofthe
map applied
theunder
applied engine
theengine loadtorque
given load
load torque(y-axis),
conditions(y-axis), which
which
[13,14]. waswas
Contours measured
measured
of equal
at the
SVFC transmission
at the
of transmission input
the model tractorinputshaft.
shaft.
are shownAA method
method
in Figureproposed
proposed
3 [13]. To byconstruct
by Kim et
Kim etal.
al.the
was
was usedtotoconstruct
used
optimum construct
fuel a specific
a specific
efficiency curve,
volumetric
volumetric fuel consumption
fuel consumption (SVFC)
(SVFC) map
map under
under the
the given
given load
load conditions
conditions
the minimum SVFC point was taken from each applied engine load torque and connected to a single [13,14].
[13,14]. Contours
Contours of of equal
equal
SVFC
SVFC of the
of themodel
model tractor
tractorare
are shown
shown in
in Figure
Figure 3 [13].
[13]. To
To construct
construct the
the optimum
optimum
regression curve point, such as the exponential functions shown in Figure 3. The low-torque region fuel
fuel efficiency
efficiency curve,
curve,
the minimum
of the
the minimum
optimum SVFC
SVFC point
curve was
point
was was taken
takenfrom
slightly from each
shifted toapplied
each the rated engine
engine load
load
speed torque
totorque andconnected
and
guarantee connected
a stable to to a single
a single
response to
regression
regression curve point,
curve point,such
such
unexpected workload fluctuations. as the
as exponential
the exponential functions
functions shown
shown in
inFigure
Figure 3.
3. The
The low-torque
low-torque region
region of
theofoptimum
the optimum
curve curve was slightly
was slightly shiftedshifted to thespeed
to the rated ratedtospeed to guarantee
guarantee a stable to
a stable response response to
unexpected
unexpected
workload workload fluctuations.
fluctuations.
2.3. PI
2.3. Controller
PIPI
Controller
2.3. Controller
A graphical
A graphicalrepresentation
representation of of
thethecontroller
PI controller used instudy
this isstudy is in
shown in Figure 4. This
A graphical representation of PI usedused
the PI controller in thisin this studyshown
is shownFigure
in 4. This controller
Figure 4. This
controller
produced produced
a throttle
controller produced a throttle
command command
a throttlefor for
engine speed
command engine speed
control
for engine using
speed control using
the difference
control the difference
using the between
differencethe between
desired
between the
and
the
desired
real
desiredand
engine real
realengine
speeds
and and the
engine speeds and
and the
PI control
speeds the PI
PI control
algorithm. algorithm.
Analgorithm.
control anti-wind-up Ananti-wind-up
An anti-wind-up
controller was controller
used forwas
controller was used
preventing
used
for preventing
thefor
saturation
preventing the
of thesaturation
the saturation of
integratorof the
[15]. integrator [15].
the integrator [15].
Figure 4. Feedback controller for engine speed control based on proportional–integral (PI) algorithm.
Appl. Sci. 2019, 9, x FOR PEER REVIEW 6 of 20
Figure 4. Feedback controller for engine speed control based on proportional–integral (PI)
algorithm.
Appl. Sci. 2019, 9, 3898 6 of 19
3. Computer Simulation
3. Computer
Computer Simulation
simulations were conducted to validate the previously developed dynamic tractor
simulation model, and to adjust
Computer simulations theconducted
were gains of thetoPIvalidate
controller.
theAdditional
previouslysimulations
developed were
dynamicconducted
tractor
to observe variations
simulation model, and in to
fuel consumption
adjust the gains and thePIdynamic
of the behavior
controller. of thesimulations
Additional engine. In the
weresimulations,
conducted
the same conditions
to observe variationsasinanfuel
actual plowing operation
consumption were applied
and the dynamic to the
behavior of developed
the engine.tractor
In the model with
simulations,
the
the PI controller.
same conditions as an actual plowing operation were applied to the developed tractor model with
the PITocontroller.
verify the tractor simulation model, plowing operations were conducted [13]. The throttle
opening and workload
To verify the tractorinput data model,
simulation for the plowing
simulation model were
operations were conducted
collected during
[13]. The thethrottle
field
operations,
opening andasworkload
shown ininputFigure 5.for
data Tractor performance
the simulation data,
model were including
collectedthe inputthe
during shaft
fieldtorque and
operations,
engine
as shown speed, were5. monitored
in Figure during the
Tractor performance operations
data, includingand the results
the input were and
shaft torque compared with those
engine speed, were
simulated
monitoredby the tractor
during model under
the operations and the
the same
resultsthrottle opening and
were compared withworkload conditions
those simulated by from field
the tractor
plowing
model underoperation data.throttle opening and workload conditions from field plowing operation data.
the same
Figure 5.
Figure Throttleopening
5. Throttle openingand
andworkload
workloaddata
datafor
for the
the simulation
simulation model.
model.
To optimize
To optimize thethe PI
PI controller,
controller, the
the engine
engine speed
speed responses
responses to to step
step and
and ramp
ramp inputs
inputs of
of workload
workload
under various PI controller gain parameters were simulated, and the optimal
under various PI controller gain parameters were simulated, and the optimal controller parameters controller parameters
were determined
were determined withwith the
the aim
aim to
to minimize
minimize thethe root
root mean
mean square
square error
error (RMSE)
(RMSE) between
between the
the desired
desired
and actual engine speeds.
and actual engine speeds.
The initial
The initial values
values of of the
the gain
gain parameters
parameters (K (Kpp,,K
Ki)i )of
ofthe
thePI
PIcontroller
controllerwere
were determined
determined using
using the
the
Ziggler and Nichols gain tuning methods, based on the engine speed response
Ziggler and Nichols gain tuning methods, based on the engine speed response from the step input of from the step input
of engine
engine throttle
throttle [15].[15]. In research,
In this this research, the main
the main objective
objective of gainoftuning
gain tuning is minimizing
is minimizing the engine
the engine speed
speed overshoot by commanding a desired engine speed. If engine speed
overshoot by commanding a desired engine speed. If engine speed overshoot occurs during field overshoot occurs during
field operation,
operation, the engine
the engine couldcouldbecomebecome unstable.
unstable.
Finally, the controllability and response performance of
Finally, the controllability and response performance of the
the controller
controller was
was simulated
simulated using
using the
the
dynamic tractor
dynamic tractor simulation
simulation model,
model, and
and the
thecollected
collected workload
workload from from plowing
plowing operations
operations was
was used
usedasas
input data for the simulation
input data for the simulation model. model.
Figure 7.
Figure 7. Comparison of simulation and
and measured
measured input
input shaft
shaft torque.
torque.
Figure 7. Comparison of simulation and measured input shaft torque.
3.2.
3.2. Optimization
Optimization Results
Results of
of Controller
Controller Parameter
Parameter
3.2. Optimization Results of Controller Parameter
In
In order
order to
to optimize
optimize thethe engine
engine speed
speed controller
controller parameter,
parameter, the
the load
load torque
torque acting
acting on
on the
the
In order input
transmission to optimize
shaft theassumed
was engine speed
to be controller
a step and parameter,
ramp the
input, as load
showntorque
in acting8on
Figures andthe9.
transmission input shaft was assumed to be a step and ramp input, as shown in Figures 8 and 9. The
transmission
The input
engine speed shaft was assumed
determination to be
algorithm a step andthe
determined ramp input,
desired as shown
engine speedinrequired
Figures to
8 and 9. The
maximize
fuel efficiency by using the load torque of the input shaft, as shown in Figure 3.
Appl. Sci. 2019, 9, x FOR PEER REVIEW 8 of 20
Appl. Sci. 2019, 9, x FOR PEER REVIEW 8 of 20
engine speed determination algorithm determined the desired engine speed required to maximize
engine speed determination
fuel efficiency by using algorithm
the load determined
torque of the desired
the input shaft, engine
as shown speed3.required to maximize
in Figure
Appl. Sci. 2019, 9, 3898 8 of 19
fuel efficiency by using the load torque of the input shaft, as shown in Figure 3.
Figure 9.
Figure Ramp input
9. -Ramp input for
for engine
engine speed
speed controller.
controller.
Figure 9. -Ramp input for engine speed controller.
Therefore,
Therefore, thethe desired
desired engine
engine speed
speed fromfrom thethe algorithm
algorithm (Figure
(Figure 3)3) could
could also
also bebe determined
determined in in
terms Therefore,
terms ofof the
the step the
step and desired
and ramp engine
ramp shaped speed
shaped input
input loadfrom the
load torque.
torque. algorithm (Figure 3) could also be determined in
terms of the10
Figure
Figure 10step andthe
shows
shows ramp
the shaped input
simulation
simulation resultsload
results for torque.
for the
the step
step response
response of of the
the engine
engine speed
speed controller
controller using
using
the Figure
tuned gain 10 shows the
parameter. simulation
The maximum results for
overshoot the step
for response
the step of
input
the tuned gain parameter. The maximum overshoot for the step input was approximately the
was engine speed
approximately controller
53 rpm using
53 from
rpm
the
the
from tuned
desired gain
engine
the desired parameter.
speed,
engine The maximum
whichwhich
speed, overshoot
is approximately
is approximately for
3.6%3.6% the step
of theoftargetinput was
value.
the target approximately
The The
value. settling 53
of rpm
timetime
settling the
of
from
engine the desired
speed engine
controller speed,
was set towhich
2% of is
theapproximately
target value. 3.6%
The of the
settling target
time of value.
the
the engine speed controller was set to 2% of the target value. The settling time of the step input The
step settling
input time
simulation of
the engine
results speed
required controller
approximately was
2.2 set
s. to
These 2% of the
responses target
were value.
due to The
the settling
constant time
time
simulation results required approximately 2.2 s. These responses were due to the constant time delay of
delay the
of step
the input
diesel
simulation
engine
of (0.4 s)results
the diesel engine required
and the motor
(0.4 approximately
(0.15
s) and thes).
motor (0.15 2.2 s).
s. These responses were due to the constant time delay
of the diesel engine (0.4 s) and the motor (0.15 s).
Appl. Sci. 2019, 9, 3898 9 of 19
Appl. Sci. 2019, 9, x FOR PEER REVIEW 9 of 20
Figure
Figure 11
11 shows
shows the
the simulation
simulation results
results of
of the
the ramp
ramp response
response of
of the
the engine
engine speed
speed controller
controller using
using
the
the tuned gain parameter. The engine speed controller was found to show high traceability within
tuned gain parameter. The engine speed controller was found to show high traceability within aa
2.1%
2.1% averaged
averagederror from
error the desired
from engine
the desired speed.speed.
engine This was concluded
This to be acceptable
was concluded to be for application
acceptable for
to the actual load using the determined gain parameters.
application to the actual load using the determined gain parameters.
Figure 12.
12. Simulation results of engine speed controller using actual load data.
Figure
Figure 12. Simulation results of engine speed controller using actual load data.
Figure 14.
Figure 14. Schematic
Schematic diagram
diagram of
of throttle control system.
throttle control system.
Figure 14. Schematic diagram of throttle control system.
4.1. Tractor
4.1. Tractor
4.1. Tractor
T623,
T623, thethe4848kWkWMFWDMFWD tractor used
tractor for the
used for simulation
the simulation model, was also
model, wasused alsotoused
develop the throttle
to develop the
T623,
control
throttle the 48
system.
control kW has
T623
system. MFWD a C has
T623 tractor
3.3T used
a 2232/2200
C 3.3T for the simulation
(Caterpillar,
2232/2200 Peoria,
(Caterpillar, model, was
IL, USA)
Peoria, also used
IL,turbocharged,
USA) to develop
four-stroke,
turbocharged, the
four-
throttle 3 . The
stroke, control
three-in-line system.
diesel
three-in-line T623
engine
diesel has
with a C 3.3T
direct
engine fuel
with 2232/2200
injection
direct fuel (Caterpillar,
andinjection
a capacityPeoria,a IL,
of 3300
and USA)
cm
capacity turbocharged,
3300 cm 3.four-
of specifications Theof
stroke,
T623 are three-in-line
listed in Table diesel
1. engine
specifications of T623 are listed in Table 1. with direct fuel injection and a capacity of 3300 cm 3. The
Figure 15. Strain gages and pick-up sensor installed on input shaft.
Figure 15. Strain gages and pick-up sensor installed on input shaft.
To
To determine
determine the
the fuel
fuel consumption
consumption of the engine, inlet flow and drain flow of the fuel supply
To determine the fuel consumption
system were measured using two flow of sensors
flow rate
rate the engine,
[13].inlet
[13]. flow and drain
A pulse-output flowflow ofwas
meter the used
fuel supply
for the
system were measured
measurements. using two
Specifications flow
of the rate
flow sensors
meter are [13].
listedAin
listed pulse-output
in Table
Table 2.
2. flow meter was used for the
measurements. Specifications of the flow meter are listed in Table 2.
Appl. Sci. 2019, 9, 3898 12 of 19
Appl. Sci. 2019, 9, x FOR PEER REVIEW 12 of 20
Table2.2.Specifications
Table Specificationsofofthe
theflow
flowmeter
meterused
usedininthis
thisstudy.
study.
Item
Item Specification
Specification
Model OF10-ZAT-AR
Model OF10-ZAT-AR
Manufacturer
Manufacturer AichiTokei
Aichi Tokei Denki,
Denki, Japan
Japan
Pulse out (mL/Pulse)
Pulse out (mL/Pulse) 2.5
2.5
Measuring range(L/h)
Measuring range (L/h) 5–300
5–300
Max. Pressure
Max. Pressure (MPa)
(MPa) 0.5
0.5atat
2020
°C℃
DEWE-3010(DEWETRON
DEWE-3010 (DEWETRON GmbH,
GmbH, Austria)
Austria) was as
was used used as the
the data data acquisition
acquisition system forsystem
receivingfor
receiving
signals fromsignals from
a sensor a sensor
attached toattached to the
the tractor. tractor. DEWE-3010
DEWE-3010 haschannels
has 15 digital 15 digitaland
channels and 17
17 analogue
analoguewith
channels channels
16-bitwith 16-bit resolution
resolution and 100 kHz andmaximum
100 kHz maximum sampling
sampling rate. rate.
In this In this
study, thestudy, the
throttle
throttle potentiometer
potentiometer and inputand input
shaft shaft
torque torque
signal signal
were were through
collected collectedathrough a 10 Hz filter
10 Hz low-pass low-pass filter
within a
within a 100 Hz sampling
100 Hz sampling rate. rate.
4.3.
4.3.Throttle
ThrottleActuating
ActuatingSystem
System
The
Thethrottle
throttleactuating
actuatingsystem
systemwaswasdeveloped
developedtotoadjust
adjustthethethrottle
throttleopening
openingshown
shownin inFigure
Figure16.
16.
The
Thesystem
systemconsists
consistsofofa aDCDCservo
servomotor
motor(which
(whichserves
servesasasthetheactuator),
actuator),aamicroprocessor,
microprocessor,and andanan
accelerator
acceleratorpedal
pedalwith
witha apotentiometer.
potentiometer.The Thewire
wirebetween
betweenthe thethrottle
throttlelever
leverofofthe
thefuel
fuelpump
pumpandandthe
the
accelerator
acceleratorpedal
pedalwaswasseparated
separated from the accelerator
acceleratorpedal
pedalandandattached
attachedtotothethe designed
designed pulley
pulley of of
the
the
DCDC servo
servo motor.The
motor. Thepulley
pulleypulled
pulledthe
theconnecting
connectingwire
wire ofof the fuel pump
pump governor,
governor,allowing
allowingthis
this
mechanism
mechanismtotobebeavailable
availablefor
forengine
enginespeed
speedcontrol.
control.AADC DCservomotor
servomotorwith with44.044.0kg·cm
kg·cmmaximum
maximum
torque, ◦ of revolution angle were used for adjusting the fuel
torque,an anindividual
individualservo
servosystem,
system,and
and120120° of revolution angle were used for adjusting the fuel
injection
injectionquantity
quantityofoffuel
fuelpump.
pump.TheThesize
sizeofofthe
therotating
rotatingpulley
pulleywaswasdetermined
determinedby bythe
themaximum
maximum
rotating
rotatingangle
angleofofthe
theDCDCservomotor,
servomotor,andandit itwas
wasused
usedfor
forimproving
improving control
controlprecision.
precision.
Flowchart of
Figure 17. Flowchart of the
the control
control algorithm.
algorithm.
Figure 18. Measured engine speed and desired engine speed during plowing operation without the
control system.
Appl. Sci. 2019, 9, x FOR PEER REVIEW 15 of 20
Appl. Sci. 2019, 9, x FOR PEER REVIEW 15 of 20
Figure 18. Measured engine speed and desired engine speed during plowing operation without the
Figure 18. Measured engine speed and desired engine speed during plowing operation without the
Appl. Sci. 9, 3898
2019,system.
control 15 of 19
control system.
Figure 19.
Figure 19. Measured
Measuredandand
filtered input
filtered shaftshaft
input torque during
torque plowing
during operation
plowing withoutwithout
operation the control
the
Figure 19. Measured and filtered input shaft torque during plowing operation without the control
system.
control
system.system.
Figure 20
Figure 20 shows thethe engine operating
operating speed and the desired
desired engine speed
speed from the
the plowing
Figure 20 shows
shows the engine
engine operating speed
speed and
and the
the desired engine
engine speed from
from the plowing
plowing
operation
operation using
using the
the control
control system.
system. With
With the
the control
control system,
system, the
the engine
engine operating
operating speed
speed was
was closer
closer to
operation using the control system. With the control system, the engine operating speed was closer
to the
the desired
desired engine
engine speed speed
than than without
without the the control
control system.system.
The The engine
engine operating
operating speed speedshowed
values values
to the desired engine speed than without the control system. The engine operating speed values
ashowed
showed
a 1.17%
1.17% error
a 1.17%
error on
on average
error on
average to
compared compared
the desired
average compared
to engine
the desired
speed.engine
to the desired
speed.
Therefore,
engine theTherefore,
speed.
the throttle
throttle control
Therefore, system
the throttle
control
can be system cansuitable
considered be considered
for the suitable for
purpose of the purposeand
maintaining of maintaining
adjusting theand adjusting
engine the engine
operating speed
control system can be considered suitable for the purpose of maintaining and adjusting the engine
operating
close to speed
the close
desired to thespeed.
engine desired engine speed.
operating speed close to the desired engine speed.
Figure
Figure 20. Measured engine
20. Measured engine speed
speed and
and desired
desired engine
engine speed
speed during
during plowing
plowing operation
operation using
using the
the
Figure system.
control
20. Measured engine speed and desired engine speed during plowing operation using the
control system.
control system.
Figure
Figure 21 shows the
21 shows the measured
measured input
input shaft
shaft transmission
transmission torque
torque using
using the
the throttle
throttle control
control system.
system.
The Figure 21 shows
characteristics of the
the measured
input input
shaft shaft
torque transmission
using the control torque
systemusing
were the throttle
very control
similar to thesystem.
input
The characteristics of the input shaft torque using the control system were very similar to the input
The
shaft characteristics of the input shaft torque using the control system were very similar to the input
shaft torque
torquewithout
withoutthethesystem. However,
system. the engine
However, operating
the engine speed showed
operating speedsignificant
showed differences.
significant
shaft torque
Moreover, without the system. However, the engine operating speed showed significant
differences.adjusting
Moreover, theadjusting
throttle only affectedonly
the throttle the affected
engine operating
the enginespeed and not
operating speedthe and
inputnotshaft
the
differences.
torque from Moreover,
the draft loadadjusting
during theplowing
the throttle operation.
only affected
In the engine
other words, operating
this system speed
could and
be not the
used for
input shaft torque from the draft load during the plowing operation. In other words, this system
input shaft
adjusting torque from the draft load during the plowing operation. In other words, this system
could be theusedengine speed without
for adjusting affecting
the engine the results
speed withoutof the operation.
affecting Consequently,
the results of the the engine
operation.
could be speed
operating used can
for be
adjusting
controlledthebyengine speed
the system forwithout affecting
the desired enginethe
speedresults of the operation.
that maximizes the fuel
efficiency by using real-time input shaft torque.
Consequently, the engine operating speed can be controlled by the system for the desired engine
Appl. Sci. 2019, 9, x FOR PEER REVIEW 16 of 20
speed that maximizes the fuel efficiency by using real-time input shaft torque.
Consequently, the engine operating speed can be controlled by the system for the desired engine
Appl. Sci. 2019, 9, 3898 16 of 19
speed that maximizes the fuel efficiency by using real-time input shaft torque.
Figure 21. Measured and filtered input shaft torque during plowing operation using the control
system.
21. Measured
Figure 21. Measuredand
andfiltered input
filtered shaft
input torque
shaft during
torque plowing
during operation
plowing using the
operation control
using the system.
control
Figure 22 shows the throttle command for engine speed control during field operation. The
system.
throttle opening
Figure wasthe
22 shows maintained within 34.6–43.5%,
throttle command and itcontrol
for engine speed showedduring
a similar
fieldtendency
operation.toThe
the throttle
engine
speed.
opening was22
maintained
Figure shows thewithin 34.6–43.5%,
throttle commandandforitengine
showed a similar
speed tendency
control duringtofield
the engine speed.
operation. The
throttle opening was maintained within 34.6–43.5%, and it showed a similar tendency to the engine
speed.
Figure 22.
Figure Throttle command
22. Throttle command of
of the
the engine
engine speed
speed controller.
controller.
It is concluded that the operating engine speed of an agricultural tractor can be controlled within
It is concluded that the operating engine speed of an agricultural tractor can be controlled within
a stable range by usingFigure 22. Throttle
the throttle command
control system.of the engine speed controller.
a stable range by using the throttle control system.
5.2. Fuel-Saving Performance
It is concluded that the operating engine speed of an agricultural tractor can be controlled within
5.2 Fuel-Saving Performance
a stable range by using the throttle control system.
In this study, three fuel efficiency parameters were used to evaluate the fuel conservation
performance of the system.
In this study, FCA
three fuel and SVFC,
efficiency were calculated
parameters fromtoFCevaluate
were used using Equations
the fuel (11) and (12),
conservation
5.2 Fuel-Saving Performance
respectively
performance[13].
of the system. FCA and SVFC, were calculated from FC using Equations (11) and (12),
respectively [13]. 10 × FC
In this study, three fuel efficiencyFCA =
parameters were used to evaluate the fuel conservation (11)
3.6 × va × wi
performance of the system. FCA and SVFC, were calculated from FC using Equations (11) and (12),
respectively [13]. 60000 × FC
SVFC = (12)
2π × Te × Ne
where FCA is fuel consumption per tilled area in L ha−1 ; FC is fuel consumption per work hour in
L h−1 ; SVFC is specific volumetric fuel consumption in L(kWh)−1 ; wi is working width in m.
Appl. Sci. 2019, 9, 3898 17 of 19
For evaluating fuel-saving performance, the fuel efficiency parameters were analyzed through
operation with and without the control system. The fuel conservation performance evaluation results
are presented as percentage differences between scenarios with and without the system.
The results of the fuel efficiency parameters from field experiments with and without the control
system are shown in Table 5. A gear step was used from L-4 to M-3 to adjust the operating speed
from 2.70 to 6.94 km/h at the rated engine speed. With the control system, FC was reduced from 23.03
to 57.87%, and FCA was reduced from 4.11 to 42.06%. SVFC was reduced from 15.73 to 38.48% in
almost all gear steps, but two cases showed increases of 0.66 to 7.24% when the control system was not
used. As a result, this system was considered to be more effective at reducing FC. This was attributable
to the tendency of the engine speed to be more easily adjusted by the FC than FCA and SVFC [13].
Almost every desired engine speed for the measured input shaft torque corresponded to the maximum
torque output speed. The maximum torque output speed was approximately half of the rated engine
speed, which was 1300–1400 RPM in this study. Accordingly, a reduction in the operation speed cannot
be avoided. In addition, reduction in the work speed also leads to a reduction in tractive resistance
and tractive power. Therefore, reduced work speed and tractive power led to lower FCA and SVFC
reduction rates than FC, and even increases of SVFC were observed in two cases. As each set of fuel
efficiency performance data were obtained under different driving habits, the range of results were
significantly widened.
The engine operating region during the plowing operations is presented as a set of measured
input shaft torques in Figure 23.
With the control system, the engine operating region was distributed around the optimum
operating curve from the desired engine speed determination algorithm. Without the control system,
the operating region was more widely distributed. As a result, the engine operating region was
maintained under 70% of the rated speed. Consequently, the reduced engine speed attributable to the
control system led to improved fuel efficiency.
Overall, the developed engine speed control system can be considered an effective system for
improving fuel efficiency. Although some experiment results showed less efficiency for reducing SVFC,
it could be concluded that the control system may be effective approach towards reducing FC, FCA,
and SVFC for common drivers.
This system could be useful for developing an integrated control system for engines, automatic
transmissions, and instruments for agricultural tractors. Nevertheless, additional research into the
development of integrated control systems for tractors with automatic transmission are still required.
Appl. Sci. 2019, 9, 3898 18 of 19
Appl. Sci. 2019, 9, x FOR PEER REVIEW 18 of 20
6. Conclusions
With the control system, the engine operating region was distributed around the optimum
operating curve from
This study the desiredto
was conducted engine
develop speed andetermination
engine speedalgorithm. Without
control system the controlthe
to maximize system,
fuel
the operating
efficiency regioninwas
of a tractor moreoperations
plowing widely distributed.
using simple Asmechanical
a result, the engine
devices. Theoperating regionspeed
desired engine was
maintained under
determination 70% ofbased
algorithm the rated speed. Consequently,
on measured the reduced
input shaft torque and PIengine speed
controller forattributable
adjusting the to
the control
engine speedsystem led to improved
was designed fuel efficiency.
using a developed computer simulation model. This model was verified
Overall, the
by comparing developed
measured engineperformance
operating speed control system
data can be considered
and simulated an effective
data. Finally, an enginesystem
throttlefor
improving
control fuelwas
system efficiency.
constructed Although
using the some experiment
desired resultscontrol
engine speed showed less efficiency
algorithm, for reducing
a PI controller, servo
SVFC, torque
motor, it couldandbe concluded that the control
speed measurement systemand
instruments, mayparameters
be effectiveofapproach towards reducing
fuel conservation performance. FC,
FCA,
A and SVFC
summary of theforresults
common drivers.
of this study are listed below.
This
A system
1-DOF could betractor
dynamic usefulsimulation
for developing model an integrated
and a fuelcontrol system for
consumption engines,map
prediction automatic
were
transmissions,
developed and and instruments
verified for agricultural
using measured tractors.
field operation Nevertheless,
data. additional model,
Using the simulation researchthe into
meanthe
development
errors of integrated
of the simulated control
engine speedsystemsand for tractors
input shaftwith automatic
torque transmission
were 2.03% and 4.87%, are still required.
respectively,
compared with those obtained from the field experiment. It means that 1-DOF dynamic tractor
6. Conclusions
simulation model can predict fuel consumption and work performance and also can be used for
model-based
This study controller design methodology.
was conducted to develop an engine speed control system to maximize the fuel
Desired
efficiency of engine
a tractorspeed determination
in plowing operations algorithm was developed
using simple mechanical based on workload
devices. The desiredapplied
engine to
engine using look-up table method. Fuel consumption prediction map
speed determination algorithm based on measured input shaft torque and PI controller for adjustingwas used to determining the
most efficientspeed
the engine point was
of engine speedusing
designed at measured workload
a developed torque. Simple
computer continuous
simulation model.optimal operating
This model was
line for maximizing fuel efficiency was developed and applied to engine
verified by comparing measured operating performance data and simulated data. Finally, an engine speed control algorithm.
Thecontrol
throttle parameter gainswas
system of the PI controller
constructed werethe
using tuned usingengine
desired the simulation modelalgorithm,
speed control by Ziggler aand PI
Nichols gain tuning methods, and the applicability of the controller
controller, servo motor, torque and speed measurement instruments, and parameters was verified with the tractorof model
fuel
and controllerperformance.
conservation using the measuredA summary field experiment
of the results data. Main
of this objective
study of gain
are listed tuning is minimizing
below.
the engine speed overshooting. After simulation, the average
A 1-DOF dynamic tractor simulation model and a fuel consumption prediction percentage errors and standardmap deviation
were
between
developed simulated and measured
and verified using measuredengine field
speed values were
operation data.2.03%
Using and
the1.37%. For input
simulation shaft
model, thetorque,
mean
the average
errors of thepercentage
simulatederror andspeed
engine standard anddeviation
input shaft were 4.87%were
torque and 2.03%
3.95%, andrespectively.
4.87%, respectively,
A measuring system was constructed and used
compared with those obtained from the field experiment. It means that 1-DOF for collecting the input shaft torque,
dynamic fuel
tractor
consumption,
simulation model and can
engine speed
predict fueldata from plowing
consumption andoperations.
work performanceHollow and typealso
slip can
ringbe
and usedstrain
for
gage were installed to input shaft
model-based controller design methodology. of transmission for measuring input shaft torque and engine speed.
In order to measure
Desired enginefuel speedconsumption
determination two flow rate sensors
algorithm were used based
was developed for inletonflow and drain
workload flow of
applied to
the fuel supply system.
engine using look-up table method. Fuel consumption prediction map was used to determining the
most efficient point of engine speed at measured workload torque. Simple continuous optimal
Appl. Sci. 2019, 9, 3898 19 of 19
The mechanical governor in the experimental tractor was converted into an electrical governor
using simple servo motor and throttle potentiometer in order to control the throttle and engine speed
control system. This was conducted to maximize the fuel efficiency.
By plowing field test data, with the engine speed control system in plowing operations can
improve FC, FCA, and SVFC about 23.03–57.87%, 4.11–42.06%, and -7.24–38.48%, respectively, over
those without using the control system.
The results of this study can be applied to the development of an optimal engine control algorithm
to maximize the fuel efficiency of tractors depending on the workload of different tractor field operations.
Author Contributions: Conceptualization, J.W.L.; methodology, J.W.L. and Y.-J.P.; software, J.W.L.; validation,
S.C.K. and J.O.; data curation, W.-J.C., H.-W.H. and J.-T.K.; supervision, Y.-J.P.; writing-original draft preparation,
J.W.L.; writing-review & editing, Y.-J.P.
Funding: The author(s) disclose receipt of the following financial support for the research, authorship, and/or
publication of this article: This work was supported by the National Research Foundation of Korea (NRF) Grant
funded by the Republic of Korea Government (MOE) (2010-0023058).
Conflicts of Interest: The authors declare no conflict of interest.
References
1. Grisso, R.; Pitman, R.; Perumpral, J.V.; Roberson, G.T. Gear up and Throttle down to Save Fuel; Virginia
Cooperative Extension: Blacksburg, VA, USA, 2010; pp. 442–450.
2. Park, S.H.; Kim, Y.J.; Im, D.H.; Kim, C.K.; Jung, S.C.; Kim, H.J.; Jang, Y.; Kim, S.S. Development of eco driving
for agricultural tractor. J. Biosyst. Eng. 2010, 35, 77–84, (In Korean, with English abstract). [CrossRef]
3. Chancellor, W.J.; Thai, N.C. Automatic control of tractor transmission ratio an engine speed. Trans. ASABE
1984, 27, 642–646. [CrossRef]
4. Zhang, N.; Perumpral, J.V.; Byler, R.K. Automatic control system for optimizing diesel engine performance.
Comput. Electron. Agric. 1987, 2, 31–46. [CrossRef]
5. Jiang, J. Optimal gain scheduling controller for a diesel engine. IEEE Control Syst. Mag. 1994, 14, 42–48.
6. Mathworks Inc. Matlab 2015b User Help Manual; Mathworks Inc.: Natick, MA, USA, 2015.
7. Tong Yang Moolsan. T623 User Manual; Tong Yang Moolsan: Seoul, Korea, 2010.
8. Organisation for Economic Co-Operation and Development. OECD Code 2: Standard Code for the Official Testing
of Agricultural and Forestry Tractor Performance; Organisation for Economic Co-Operation and Development:
Paris, France, 2010.
9. Qingwen, S.; Grigoriadis, K.M. Diesel Engine Speed Regulation Using Linear Parameter Varying Control. In
Proceedings of the American Control Conference, Denver, CO, USA, 4–6 June 2003; Volume 1, pp. 779–784.
10. American Society of Agricultural and Biological Engineers. ASABE Standard ASABE D497.5 Agricultural
Machinery Management Data; American Society of Agricultural and Biological Engineers: St. Joseph, MI,
USA, 2006.
11. Kim, K.U.; Kim, K.D.; Koo, Y.M.; Park, K.J. Agricultural Machinery Design 1; Moon Woon Dang: Seoul, Korea,
2009. (In Korean)
12. Kolator, B.; Białobrzewski, I. A simulation model of 2WD tractor performance. Comput. Electron. Agric. 2011,
76, 231–239. [CrossRef]
13. Lee, J.W.; Kim, J.S.; Kim, K.U. Computer simulations to maximise fuel efficiency and work performance of
agricultural tractors in rotovating and ploughing operations. Biosyst. Eng. 2016, 142, 1–11. [CrossRef]
14. Kim, S.C.; Kim, K.U.; Kim, D.C. Modeling of fuel consumption rate for agricultural tractors. J. Biosyst. Eng.
2010, 35, 1–9, (In Korean, with English abstract). [CrossRef]
15. Ogata, K. Modern Control Engineering-International Edition, 5th ed.; Prentice Hall: Upper Saddle River, NJ,
USA, 2009.
© 2019 by the authors. 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/).