Journal of

Manufacturing and
Materials Processing

Article
Experimental Investigation of Laser Surface
Transformation Hardening of 4340 Steel Spur Gears
Al Khader Borki, Abderrazak El Ouafi * and Ahmed Chebak
PARL Research Team, Department of Engineering, University of Quebec at Rimouski,
Rimouski, QC G5L 3A1, Canada
* Correspondence: Abderrazak_elouafi@uqar.ca; Tel.: +1-(418)-723-1986




Received: 29 June 2019; Accepted: 13 August 2019; Published: 15 August 2019


Abstract: This paper presents an experimental investigation of laser surface transformation hardening
(LSTH) of 4340 steel spur gears using regression analysis. The experimental work is focused on
the effects of various LSTH parameters on the hardness profile shape and the hardened depth
variation. The investigations are based on a structured design of experiments and improved statistical
analysis tools. The experimentations are carried out on AISI 4340 steel spur gears using a commercial
3 kW Nd:YAG laser system. Laser power, scanning speed, and rotation speed are used as process
parameters to evaluate the variation of the hardened depth and to identify the possible relationship
between the process parameters and the hardened zone physical and geometrical characteristics.
Based on the experimental data and analysis of variance, the direct and interactive contributions of
the process parameters on the variation of the hardness profile shape and the hardened depth are
analyzed. The main effects and the interaction effects are also evaluated. The results reveal that all the
process parameters are relevant. The cumulative contribution of the three parameters in the hardened
depth variation represents more than 80% with a clear predominance of laser power. The contribution
of the interactions between the parameters represents 12% to 16%. The resulting hardness values are
relatively similar for all the experimental tests with about 60 HRC. The evaluation of the produced
regression models for hardened depth prediction shows limited performance suggesting that the
predictive modeling process can be improved.

Keywords: laser surface transformation hardening; AISI 4340 alloy steel; spur gears; hardness profile;
hardened depth; design of experiments; analysis of variance

1. Introduction
Surface transformation hardening processes are designed to produce wear-, corrosion-, and
fatigue-resistant surfaces with increased lifetime for various industrial applications by hardening the
superficial critical areas using localized heat gains. Among these processes, laser surface transformation
hardening process (LSTH) is considered as an interesting alternative process to the traditional methods
due to several advantages such as high-power density, rapidity, flexibility, full process control,
minimized total energy, reduced distortion, and reduced thermal affected zone.
LSTH of steels is a potentially promising industrial application considering the broad employment
of steels in a variety of industries. Many different types of alloy steel mechanical components such as
bearing, cams, gears, shafts, pistons, and rings, that need to have a very hard surface, can be hardened
using LSTH [1–3]. The ability to deliver a controlled rate of heat energy to the expected surface area of
the part is one of the most significant advantages of using laser as a heating source. The absorbed
energy is dependent on the absorptivity of the part surface, which depends on the wavelength of the
laser beam. For steels, the absorptivity increases when the wavelength is short. The Nd:YAG laser
system with short wavelength at 1064 nm is suggested for surface hardening of steel [4]. In LSTH,

J. Manuf. Mater. Process. 2019, 3, 72; doi:10.3390/jmmp3030072 www.mdpi.com/journal/jmmp

J. Manuf. Mater. Process. 2019, 3, 72 2 of 16

the laser sources provide a controllable heat in the selected area to produce thin hardened surface
layers. When a region of the part surface is exposed to the laser beam, the temperature increases
rapidly, resulting in a thin layer converted into austenite, while the temperature of the rest of part
remains unaffected. When the laser beam moves with respect to the part, the part areas in the beam
track are rapidly heated and cooled because the self-quenching caused conduction into the relatively
cool bulk of the material. In LSTH, the control of rapid heating and cooling is the basic ingredient to
transform the austenite into hard martensite structure at the surface [5–11].
To exploit appropriately the benefits presented by LSTH, it is necessary to develop a comprehensive
strategy to control the heating process in order to produce the appropriate martensite structure, and
then the desired hardened surface characteristics without being forced to use the traditional and
fastidious trial and error procedures. In this process, the thermo-physical properties of the material and
the heating system parameters have significant effects [12]. To address this problem, several researchers
proposed various experimental and modeling approaches as attempts to establish the relationships
between laser processing parameters and hardness profile. In the majority of the proposed approaches,
laser power, scanning speed, beam spot size, focal length, and rotation speed are considered as
process parameters to evaluate hardness profile attributes, such as hardened depth, hardened width,
or hardened area. In the LSTH process, the key objective is to achieve a hardened surface with
minimum hardened depth and maximum hardened width. The traditional procedure used to select
the appropriate combination of parameters to reach these objectives consists of setting the parameters
by conducting a number of experiments by varying one parameter at a time in order to evaluate
the parameter’s effects. As the process parameters are interrelated by nonlinear relationships, this
procedure cannot lead to convincing results despite the prohibitive number of experiments, which
leads to more time and costs. These problems can be avoided if appropriate prediction models are
designed. These models can not only predict the most suitable process parameters for the desired
hardness profiles but also help to a better understanding of their relative influence on the treated
material properties.
Several modeling approaches to LSTH are available [11–16]. In LSTH, two modeling approaches
can be proposed: Numerical modeling and empirical modeling. The first case refers essentially to the
combination of material non-linear properties, laser sport form and motion and heat-flow equation
solved using the finite element method (FEM) in order to define the temperature distribution during
the heating process [15–17]. The hardness is then approximated by the equations of Ashby and
Easterling [18]. The obtained model can be validated using experimental data and then used for
simulation, prediction, and optimization. In the second case, data collected using a structured design
of experiments are processed using analysis techniques such as analysis of variance, regression analysis
and response surface methodology to establish the relationships between the process parameters and
the hardness profile attributes [19,20]. Other modeling approaches use expert system [21] and neural
networks [22,23].
Interesting results are established using different modeling approaches on several types of
materials and different part geometries. However, some laser surface hardening issues are far from
complete and many research works are needed. When LSTH is applied to relatively complex surfaces,
such as gears and spline shafts, particular attention must be focused on the desired hardness profile for
different regions of the gear tooth. Indeed, gears are often surface hardened to increase their resistance
to wear and fatigue. The thickness of the hardened layer must be deep enough to endure the maximum
stress on the gear without affecting the rest of the tooth. However, an excessive hardened depth leaves
the tooth too brittle to withstand tensile stresses.
The type of gear, tooth geometry, and hardness profile, as well as the operating load conditions to
be tolerated, are important factors to consider when looking for the ideal thickness of the hardened
layer [24,25]. Usually, a uniform hardened layer is considered to be the best profile for hardening gear.
This is not necessarily always the case. In many applications, some variety of non-uniform profiles can
provide gears with better performance. Uniform hardness profile is popular for medium-size gears.
J. Manuf. Mater. Process. 2019, 3, 72 3 of 16

As the gear teeth are not completely hardened, a relatively ductile core (35 to 45 HRC) combined with
a hard surface (55 to 65 HRC) offer very remarkable gear properties, such as excellent wear resistance,
toughness, and bending strength, and allows greater gear durability. However, it is essential to harden
all parts of the gear surface, including flanks and roots zones. An uninterrupted hardened pattern
on all contact areas of the tooth indicates good wear properties of the gear and ensures a relatively
continuous distribution of desirable compressive stresses at the gear surface.
An ideal hardness pattern consists of uniform thickness along the flank of the tooth and between
the fillets. This implies that the energy must be applied in a non-uniform manner to generate a uniform
hardened depth, which is not easy to achieve in practice. However, it is possible to approach it by
judiciously controlling the process parameters that have an influence on heat transfer and energy
absorption [25]. Laser power, scanning speed and rotation speed can then play a major role in the case
depth control.
The objective of this work is to conduct an experimental investigation of the process parameter
effects on the hardness profile of 4340 steel spur gears. The used methodology is based on three
steps. First, a preliminary exploration is conducted to determine the appropriate range of the process
parameters and evaluate their main effects on the hardness profile in order to develop a structured
strategy for an efficient design of experiments. The second step consists of developing and performing
a systematic experimental investigation using Taguchi method. Finally, the third step consists of
analyzing the experimental data in order to evaluate direct and interactions effects of the process
parameters on the hardened depth. The experimentations are performed on a commercial 3 kW
Nd:YAG laser system using laser power, scanning speed, and rotation speed as principal factors.

2. Experimental Investigation
The base material used is the AISI 4340 alloy steel. Due to its high tensile strength, AISI 4340 alloy
steel is very frequently used in the aerospace and automotive industries and largely used in power
transmission gears and shafts, aircraft landing gears, and other structural parts. When heat-treated,
this nickel-chromium-molybdenum alloy steel is known to reach great tenacity and high strengths.
Tables 1 and 2 present AISI 4340 alloy steel chemical composition and properties respectively. Relevant
information about the dimensional characteristics of the gear design is presented in Table 3.

Table 1. Chemical composition of AISI 4340 alloy steel [22].

Element Content (wt.%)

Iron, Fe 95.195–96.33
Nickel, Ni 1.65–2.00
Chromium, Cr 0.700–0.900
Manganese, Mn 0.600–0.800
Carbon, C 0.380–0.430
Molybdenum, Mo 0.200–0.300
Silicon, Si 0.150–0.300
Sulfur, S 0.0400
Phosphorous, P 0.0350

Table 2. Material properties of AISI 4340 alloy steel [22].

Property Symbol Unit Value

Reflexion coefficient Rc 0.6
Steel absorptivity Ac m−1 800
Eutectoid temperature Ac1 K 996
Austenitization temperature Ac3 K 1063.15
Activation energy of carbon diffusion in ferrite Qa kJ/mol 10
Pre-exponential for diffusion of carbon D0 m2 /s 80
Gas constant R J/mol·K 6 × 10−5
J. Manuf. Mater. Process. 2019, 3, 72 4 of 16

Table 2. Cont.

Property Symbol Unit Value

Steel carbon content C 0.34%
Austenite carbon content Ce 0.8%
Ferrite carbon content Cf 0.01%
Critical value of carbon content Cc 0.05%
Volume fraction of pearlite colonies fi 0.5375

Table 3. Dimensional characteristics of the gear.

Characteristics Symbols Values (in) Characteristics Symbols Values (in)

Pitch diameter D 4 Circular pitch CP 0.262
Number of teeth N 48 Circular thickness T 0.131
Diametral pitch Dp 12 Working depth WD 0.167
Pressure angle α 20 Internal diameter ID 1.181
Addendum A 0.083 Face width FW 0.256
Dedendum B 0.096 Surface roughness Ra 6.3
Outside diameter OD 4.167 Material Mat SAE 4340
Root diameter RD 3.833 Hardness core HRC 40 HRC

2.1. Process Parameters Identification

The variables that have a significant influence on the hardness profile are clearly identified in the
literature [26]. Several studies on LSTH of various grades of steel identified laser power (P), scanning
speed (SS) and rotational speed (RS) as the factors that most affect the hardness profile and produce
important effects on the hardened depth [25–27]. Therefore, these three parameters are selected for the
present study.

2.2. Experimental Setup

An overview of the laser cell used as an experimental setup for the gear’s laser hardening is
presented in Figure 1. The laser cell consisted of an articulated robot with six degrees of freedom,
an Nd:YAG high-power laser system, a mini lathe, an infrared imaging device, a data acquisition
system, and safety systems. The Fanuc M-710iC robot was used to handle the Highyag Bimo laser
processing head.
The laser heat source was an IPG YLS-3000-ST2 generating a power output up to 3.2 kW at
1064 nm wavelength. The mini lathe was equipped with a variable-speed drive capable of speeds up to
9500 rpm. The mini lathe was fixed to ensure that the spot could translate along the axis of the rotating
part. An accurate thermographic imaging method using a FLIR A325 infrared camera was adopted
for the surface temperature measurements. The FLIR was located close to the part at the adequate
focal distance for better spatial resolution needed to understand the physical phenomena and later to
control the process parameters. To protect the mini-lathe components, it was necessary to avoid the
thermal transfer from the gear assembly to the mini-lathe components (live center, chucks...). To do so,
alumina ceramic rings were inserted between the shaft and the gear as illustrated in Figure 2. The laser
setup configuration is illustrated in Figure 3. The temperatures in selected points of the assembly were
constantly monitored, as shown in Figure 4. Figure 5 presents an illustration of typical temperature
distribution during the heating process. Note that the temperature was not uniform throughout the
gear. However, since the gear was rotating, it was possible to assume that different portions of the
exposed surface would receive the same quality of heat.
 J. Manuf. Mater. Process. 2019, 3, 72 5 of 16
J. Manuf. Mater. Process. 2019, 3, x FOR PEER REVIEW 5 of 16
J. J.Manuf.
Manuf.Mater.
Mater.Process.
Process.2019,
2019,3,3,x xFOR
FORPEER
PEERREVIEW
REVIEW 5 5ofof1616

(b)
(b)
(b)

(a) (c) (d)

(a)
(a) (c)
(c) (d)
(d)
Figure 1. Experimental
Experimentalsetup setup for gear
setup gearlaser
lasersurface
surfacetransformation
transformationhardening
hardening (LSTH): (a) Overview
Figure
Figure
Figure 1.
1.1. Experimental
Experimental setup forgear laser
for lasersurface
surfacetransformation hardening(LSTH):
transformationhardening (LSTH):(a)(a)Overview
(a) Overview
Overview of
of
thethe LSTH
LSTH cell,
cell, (b) major cell components, (c) gear assembly and (d) gear during the heating process.
ofof the
the LSTH
LSTH cell,(b)
cell, (b)major
majorcell
cellcomponents,
components,(c) (c)gear assembly
gearassembly and
assemblyand (d)
and(d) gear
(d)gear during
gearduring the
duringthe heating
theheating process.
heatingprocess.
process.

(a) (b)
(a)
(a) (b)
(b)
Figure 2. Gear
Gear assembly:
assembly: (a)
(a)Overview
Overview of gear assembly and (b) ceramic rings assembly.
Figure
Figure
Figure 2.2.2.Gear
Gear assembly:
assembly: (a)(a) Overview
Overview ofofof gear
gear
gear assembly
assembly
assembly and
and
and (b)
(b)
(b) ceramic
ceramic
ceramic rings
rings
rings assembly.
assembly.
assembly.

Figure
Figure 3. Experimental
Experimental setup
setup configuration
configuration for
for LSTH of gears.
Figure
Figure 3.3.Experimental
Experimental setup
setup configuration
configuration for
for LSTHofofgears.
LSTH gears.
J. Manuf. Mater. Process. 2019, 3, 72 6 of 16
J.J. Manuf.
Manuf. Mater.
Mater. Process.
Process. 2019,
2019, 3,
3, xx FOR
FOR PEER
PEER REVIEW
REVIEW 66 of
of 16
16

1200
1200 Live
Live center
center
Shaft
Shaft
1000 Bottom
Bottom
1000

C)
C)
(Deg.C)
800
800
(Deg.
Temperature(Deg.

600
Temperature

600
Temperature

400
400

200
200

00
0 40 80 120 160 200
0 40 80 120 160 200
Time
Time (s)
(s)
Figure
Figure 4.
4. Typical
Typical variation
variation of
of the
the measured temperatures
measured temperatures during
temperatures during the
during the heating
the heating process.
heatingprocess.
process.

5. Typical temperature
Figure 5.
Figure temperature distribution during the heating process.
Figure 5. Typical
Typical temperature distribution
distribution during
during the
the heating
heating process.
process.

Before
Before the laser surface transformation hardening, the parts were machined, hardened, and
Before thethe laser
laser surface
surface transformation
transformation hardening,
hardening, the
the parts
parts were
were machined,
machined, hardened,
hardened, and and
tempered
tempered to reach an initial core hardness of about 45 HRC. Following the heat treatment, the
tempered to reach an initial core hardness of about 45 HRC. Following the heat treatment, the
to reach an initial core hardness of about 45 HRC. Following the heat treatment, the
specimens
specimens were cautiously prepared, polished, and etched using Nital chemical solution (95% of
specimens were were cautiously
cautiously prepared,
prepared, polished,
polished, and
and etched
etched using
using Nital
Nital chemical
chemical solution
solution (95%
(95% of of
ethanol
ethanol and 5%ofofnitricnitric acid). The hardness profiles construction was based on programmed
ethanol andand 5%5% of nitric acid).
acid). The
The hardness
hardness profiles
profiles construction
construction was
was based
based on on programmed
programmed micro-micro-
micro-hardness
hardness measurements ausing
measurements a Clemex machine. The temperature of the preheating was480set at
hardness
◦ measurements using using a Clemex
Clemex machine.
machine. The
The temperature
temperature of of the
the preheating
preheating was
was set
set at
at 480 °C°C
480
in C in the gear bottom. Particular attention was paidmeasuring
when measuring the micro-hardness at the
in the
the gear
gear bottom.
bottom. Particular
Particular attention
attention was
was paid
paid when
when measuring the the micro-hardness
micro-hardness at at the
the bottom
bottom
bottom
of of the spurteeth
gear teeth where there is normally less hardening. As illustrated inFigure
Figure 6,aa visible
visible
of the
the spur
spur gear
gear teeth where
where there
there is
is normally
normally less
less hardening.
hardening. As As illustrated
illustrated in in Figure 6,6, a visible
crack
crack appeared in the hardened gear bottom, suggesting that the bottom is the most fragile location.
crack appeared
appeared in in the
the hardened
hardened gear
gear bottom,
bottom, suggesting
suggesting that
that the
the bottom
bottom is is the
the most
most fragile
fragile location.
location.

Figure 6.
6. Fatigue
Figure 6. Fatigue test
test on
on hardened
hardened gear.
gear.
J. Manuf. Mater. Process. 2019, 3, 72 7 of 16
J. Manuf. Mater. Process. 2019, 3, x FOR PEER REVIEW 7 of 16

2.3. Preliminary
2.3. PreliminaryTests
Tests and
and Results
Results
In order
In order to to determine
determinethe theappropriate
appropriate range
rangeof of
thethe
process
process parameters
parameters and and
to approximate
to approximate their
mainmain
their effects on the
effects onhardness
the hardness profile, a preliminary
profile, a preliminary experimental
experimental exploration
explorationwas wasperformed.
performed.The
experimentation tests were done on spur gears made of AISI 4340 steel
The experimentation tests were done on spur gears made of AISI 4340 steel using laser P, SS, and RS asusing laser P, SS, and RS as
experimental factors. During this stage, these factors were varied in
experimental factors. During this stage, these factors were varied in specific ranges. The hardnessspecific ranges. The hardness
profile evaluation
profile evaluation started
started by by examining
examining visually
visually thethe cross-section
cross-section of of the
the specimen,
specimen, and and then
then by by
measuring the micro-hardness as a function of the depth from the
measuring the micro-hardness as a function of the depth from the surface to the end of the transformed surface to the end of the
transformed
zone. The case zone.
depthThewas case depth by
defined wasthedefined by the depth
depth depicting depicting area.
the hardened the hardened
The hardness area. was
The
hardness was
measured on the measured
hardenedon thecross-section
gear hardened gear cross-section
at 18 at 18 points
points at a separation at a separation
distance of 250 µmdistance
along two of
250 µm along two parallel lines passing through the middle of the tooth
parallel lines passing through the middle of the tooth tip and the middle of the tooth root. When one tip and the middle of the
tooth root.
factor When one
was varied, thefactor
otherswas were varied,
set tothe others
their werevalues.
average set to their
P is average
varied from values.1800P isWvaried
to 2500from W,
1800 W to 2500 W, SS from 0.15 to 0.45 mm/s and RS from 3000 to 4000
SS from 0.15 to 0.45 mm/s and RS from 3000 to 4000 rpm. The hardened depth and the smoothness of rpm. The hardened depth and
the transition
the smoothness of thehardened
between transitionzonebetween hardened
and base material zonewithand base material
negligible with negligible
heat-affected zone wereheat- the
affected
major zone used
criteria were for
thethemajor criteria used
identification for the
of the identification
working ranges. of the working ranges.
Figure 77and
Figures andFigure 8 represent
8 represent the hardness
the hardness curves curves for typical
for typical preliminary
preliminary tests tests representing
representing the
the variation
variation effects
effects of SS,ofRS,SS, and
RS, P,
and P, respectively.
respectively. TheseThese hardness
hardness curves curves
can be can be divided
divided typically
typically into
into three
three regions.
regions. Recognized
Recognized by itsby its hardness
high high hardness
compared compared
to the topartthecore,
partthecore,
firstthe first was
region region was
totally
austenitized and transformed to martensite upon rapid cooling. The second region was identifiedwas
totally austenitized and transformed to martensite upon rapid cooling. The second region by
aidentified by a radical
radical decreasing decreasing
in hardness in hardness
before reaching before reaching
the core hardness.the coreThishardness.
region was This region was
characterized
characterized
by the hardness byloss
the hardness
caused byloss causeddrop
a severe by a in
severe drop in
hardness to hardness
reach initial to reach initialThis
hardness. hardness.
zone wasThis
zone was considered as a mixture of hard and over-tempered martensite
considered as a mixture of hard and over-tempered martensite when the temperature was between Ac1 when the temperature was
between
and Ac3. Ac1 andthe
Finally, Ac3. Finally,
third region the third region to
corresponded corresponded
the zone nottoaffected
the zone bynot
the affected
thermal by theThe
flow. thermal
case
flow. The case depth was then characterized by the first
depth was then characterized by the first zone, called the hard zone [27]. zone, called the hard zone [27].

65
SS = 0.45 mm/s
SS = 0.15 mm/s
RS = 3000 rpm
60 RS = 4000 rpm
P = 1800 W
P = 2500 W
55
Hardness (HRC)

50

45

40

35

30
0 500 1000 1500 2000 2500 3000 3500 4000

Depth (μm)
Figure 7. Typical
Figure 7. Typical hardness
hardness curves
curves for
for gear
gear teeth
teeth top
top versus
versus process
processparameters.
parameters.

The experimental results show that the hardness remains relatively constant within the hardened
The experimental results show that the hardness remains relatively constant within the
depth before dropping off to reach the initial hardness value despite the variation of the process
hardened depth before dropping off to reach the initial hardness value despite the variation of the
parameters. From these preliminary results, several observations can be made. The surface hardness
process parameters. From these preliminary results, several observations can be made. The surface
value is similar in all cases through the preheating stage. As expected, decreasing the SS or the
hardness value is similar in all cases through the preheating stage. As expected, decreasing the SS or
revolution speed results in increasing the hardened depth. However, increasing the power results
the revolution speed results in increasing the hardened depth. However, increasing the power results
in increasing the hardened depth. Using these preliminary experimental observations, the working
in increasing the hardened depth. Using these preliminary experimental observations, the working
limits of the process parameters to achieve the desired hardening conditions are selected. The selected
limits of the process parameters to achieve the desired hardening conditions are selected. The selected
(lower–upper) ranges for P, SS, and RS are (1900–2300 W), (0.25–0.45 mm/s), and (3000–4000 rpm),
(lower–upper) ranges for P, SS, and RS are (1900–2300 W), (0.25–0.45 mm/s), and (3000–4000 rpm),
respectively. Other than the limits, the middle level is selected in order to evaluate the effect of the
J. Manuf. Mater. Process. 2019, 3, 72 8 of 16

J. Manuf. Mater. Process. 2019, 3, x FOR PEER REVIEW 8 of 16

respectively. Other than the limits, the middle level is selected in order to evaluate the effect of the
process
process parameters
parameters in
in the
the middle
middle region.
region. Consequently, oneone more
more level
level in
in the
the middle
middle of
of each
each factor
factor
range
range is
is selected.
selected. The increasing of the quantity
quantity ofof levels
levels will
will dramatically
dramatically increase
increase the
the number
number ofof
experiments,
experiments, which
which will
will be
be more
morematerial-,
material-,time-,
time-,and
andcost-consuming.
cost-consuming.

65
SS = 0.45 mm/s
SS = 0.15 mm/s
RS = 3000 rpm
60 RS = 4000 rpm
P = 1800 W
P = 2500 W
55
Hardness (HRC)

50

45

40

35

30
0 500 1000 1500 2000 2500 3000

Depth (μm)

Figure 8. Typical
Figure 8. Typical hardness
hardness curves
curves for
for gear
gear teeth
teeth bottom
bottom versus
versus process
processparameters.
parameters.

2.4. Design of Experiments

2.4. Design of Experiments
In any experimental investigation, the data collection method used is the basic ingredient to ensure
In any experimental investigation, the data collection method used is the basic ingredient to
the quality and the significance of the acquired data. Orthogonal arrays-based design of experiments
ensure the quality and the significance of the acquired data. Orthogonal arrays-based design of
is one of the most appropriate strategies that can lead to an efficient and robust design that can collect
experiments is one of the most appropriate strategies that can lead to an efficient and robust design
statistically significant data with the minimum number of repetitions [28]. In the case of heat treatment,
that can collect statistically significant data with the minimum number of repetitions [28]. In the case
it is important to avoid the non-transformation to martensite as well as melting of the superficial
of heat treatment, it is important to avoid the non-transformation to martensite as well as melting of
layer. For this reason, the experimental investigation is started by defining the working limits of the
the superficial layer. For this reason, the experimental investigation is started by defining the working
process parameters using preliminary tests. The evaluation of the effect of the process parameters on
limits of the process parameters using preliminary tests. The evaluation of the effect of the process
the middle region and the possible nonlinearity in the process parameter effects justify the choice of
parameters on the middle region and the possible nonlinearity in the process parameter effects justify
three levels for each factor. As a result, three factors with three levels are considered. Table 4 shows
the choice of three levels for each factor. As a result, three factors with three levels are considered.
the factors and levels used in the experimental design. These factors are the laser power, the beam
Table 4 shows the factors and levels used in the experimental design. These factors are the laser power,
scanning speed, and the part revolution speed. The central values of the variation ranges of these
the beam scanning speed, and the part revolution speed. The central values of the variation ranges of
factors represent the nominal values used to reach the nominal case depth of the hardness. The design
these factors represent the nominal values used to reach the nominal case depth of the hardness. The
that fits this problem corresponds to an L9 matrix. The experimental design is shown in Table 5. As the
design that fits this problem corresponds to an L9 matrix. The experimental design is shown in Table
experiments are planned using an orthogonal design the estimate of the average effects of a given
5. As the experiments are planned using an orthogonal design the estimate of the average effects of a
factor on various responses will not be biased.
given factor on various responses will not be biased.
Table 4. Factors and levels for experimentation.
Table 4. Factors and levels for experimentation.
Levels
Factors Levels
Factors 1 1 22 3 3
Power: P Power:
(W) P (W) 1900 1900 2100
2100 2300 2300
ScanningScanning
speed: SS speed:
(mm/s) SS (mm/s)
0.25 0.25 0.35
0.35 0.45 0.45
Revolution speed: RS (rpm) 3000 3500 4000
Revolution speed: RS (rpm) 3000 3500 4000

Table 5. Experimental design using L9.

Tests P (W) SS (mm/s) RS (rpm)

T1 1900 0.25 3000
T2 1900 0.35 3500
J. Manuf. Mater. Process. 2019, 3, 72 9 of 16

Table 5. Experimental design using L9.

J. Manuf. Mater. Process. 2019, 3, x FOR PEER REVIEW 9 of 16
Tests P (W) SS (mm/s) RS (rpm)
T1 T3 1900
1900 0.45
0.25 4000 3000
T2 T4 1900
2100 0.35
0.25 3500 3500
T3 T5 1900
2100 0.45
0.35 4000 4000
T4 2100 0.25 3500
T6 2100 0.45 3000
T5 2100 0.35 4000
T6 T7 2300
2100 0.25
0.45 4000 3000
T7 T8 2300
2300 0.35
0.25 3000 4000
T8 T9 2300
2300 0.35
0.45 3500 3000
T9 2300 0.45 3500
3. Experimental Data Analysis
3. Experimental Data Analysis
Three major statistical tools are used to analyze the produced experimental data: The graph of
Three major
the average effectsstatistical
of each tools
factorare used
level ontotheanalyze
mean the produced
hardened experimental
depth (bottom and data: The
top), thegraph of
percent
the average effects
contribution of each
of factors in thefactor
meanlevel on thedepth
hardened meanvariation
hardenedextracted
depth (bottom
from the and top), the
analysis percent
of variance
contribution
(ANOVA), and of factors in the mean
the correlation hardened
between depthhardened
the mean variation depth
extracted
andfrom the analysis
various of variance
factors. The percent
(ANOVA),
contribution andofthean correlation
experimental between
factorthe mean hardened
represents depthofand
the portion thevarious factors.
variation The percent
observed in the
contribution
experimental of dataanthatexperimental
is attributedfactor
to thatrepresents the portion
specific factor. Typically,of the
the total
variation observed
percent contributionin the
of
experimental data that is attributed to that specific factor. Typically, the total
all considered experimental factors must add up to 100. The difference from 100 is credited to the percent contribution
of all considered
contribution experimental
of other uncontrolled factors mustasadd
factors welluptotoexperimental
100. The difference from 100
errors. Using theseis results,
crediteditto is
the contribution
possible to extractof other
some uncontrolled
accurate andfactorsuseful as well to experimental
indicators errors. Usingofthese
on the best combination results,
the process
itparameters
is possibletotoconsider
extract some accurate
for eventual and useful
predictive indicators on the best combination of the process
modeling.
parameters
As theto consider
first for eventual
observation of datapredictive
indicates modeling.
that the minimum values of the hardened depth (top
and As the first
bottom) observation
are obtained of in data
test indicates
3 using that
1900theW,minimum
0.45 mm/s values
and of4000
the hardened
rpm for P, depth
SS, (top
and and
RS,
bottom) are obtained
respectively. The maximal in testvalues
3 using 1900hardened
of the W, 0.45 mm/s
depthand(top4000 rpm for P,
and bottom) areSS, and RS,inrespectively.
obtained test 8 using
The
2300maximal values
W, 0.35 mm/s andof 3000
the hardened
rpm for P,depth
SS, and(top
RS,and bottom) are
respectively. Theobtained in test 8 using
second observation 2300 W,
indicates, as
0.35 mm/s and 3000 rpm for P, SS, and RS, respectively. The second observation
illustrated in Figure 9, that T1, T4, and T9 produce the same result in terms of hardness profile and indicates, as illustrated
in Figure 9,depth
hardened that T1demonstrating
, T4 , and T9 produce the sameprocess
that different result inparameter
terms of hardness profile can
combinations and lead
hardened depth
to identical
demonstrating
gear teeth top thatand different process parameter
bottom hardened combinations
depths, suggesting can lead
multiple to identical
options for an gear teethcost
eventual top and
and
bottom hardened
production depths, suggesting
optimization multiple
process. Finally, optionscorrelation
a strong for an eventual cost andbetween
is observed production topoptimization
and bottom
process.
hardened Finally,
depths. a strong correlation is observed between top and bottom hardened depths.

65
T1_Top
T4_Top
T9_Top
60 T1_Bottom
T4_Bottom
T9_Bottom
55
Hardness (HRC)

50

45

40

35

30
0 500 1000 1500 2000 2500 3000 3500 4000

Depth (μm)

Figure9.9.Typical
Figure Typicalhardness
hardnesscurves
curvesatatgear
gearteeth
teethtop
topand
andbottom
bottomvs.
vs.depth
depthfor
fortests
testsTT1 ,1,TT44,, and
and T
T99..

As indicated, the objective of ANOVA is to investigate whether the parameters and the
interactions present significant effects on the hardened depth in order to identify the variables to be
included in the prediction model. Extracted from ANOVA, Figure 10 shows that the top hardened
depth is affected at different degrees by the process parameters. The direction and the slope of each
graph determine the effect of each of the three parameters. The factor most affecting the hardened
J. Manuf. Mater. Process. 2019, 3, 72 10 of 16

As indicated, the objective of ANOVA is to investigate whether the parameters and the interactions
present significant effects on the hardened depth in order to identify the variables to be included
in the prediction model. Extracted from ANOVA, Figure 10 shows that the top hardened depth is
affected at different degrees by the process parameters. The direction and the slope of each graph
determine the effect of each of the three parameters. The factor most affecting the hardened depth
is P. SS and RS have a lower impact. The results are in concordance with the observations made in
the preliminary experimental results. It can be observed that the top hardened depth increases as
P increases. However, the hardened depth decreases when the SS and RS increase. These results
are anticipated, since increasing SS and RS reduces the heat transmitted from the laser beam to the
part surface.

Figure 10. Main effects of the process parameters on the tooth top hardened depth.

These results are confirmed by the average effect of each parameter in terms of percent contribution
in the top hardened depth variation. Table 6 reveals that the cumulative contributions of P, SS, and
SR in the hardened depth variations represent more than 84%. P is the most significant parameter
having the greatest effect (52.92%). The SS is the second most significant parameter with 20.04%, and
RS is the least significant parameter with the smallest effect (11.27%). These results are confirmed
by the correlation analysis between process parameters and hardened depths presented in Table 7.
The contribution of the interactions between the three parameters is less important, representing about
12.5%. However, it is surprising that P and SS interaction have a less significant effect with about 0.2%,
which is less than the error contribution. Given the impact of the increasing SS on the power effect,
SS interaction is assumed to be at the same level as RS.

Table 6. ANOVA analysis for the tooth top hardened depth.

Characteristic Sum of Squares Contribution F-Value p-Value

Regression 12.89 96.85% 10.26 0.091
P 7.04 52.92% 2.90 0.231
SS 2.67 20.04% 0.54 0.539
RS 1.50 11.27% 0.11 0.767
P·SS 0.02 0.19% 0.71 0.488
P·RS 0.93 7.01% 0.11 0.768
SS·RS 0.72 5.41% 3.44 0.205
Error 0.42 3.15%
Total 13.3 100%
J. Manuf. Mater. Process. 2019, 3, 72 11 of 16

Table 7. Correlations between process parameters and hardened depth.

P SS RS
Tooth top hardened depth 72.17% 43.30% 31.75%
Tooth bottom hardened depth 72.75% 44.77% 33.58%
J. Manuf. Mater. Process. 2019, 3, x FOR PEER REVIEW 11 of 16
Globally, the total contributions of the parameters and the interactions exceed 96.5%, suggesting
the most
good important
experimental process The
planning. parameters influencing
error remains theacceptable
within an hardenedleveldepth are all
(3.15%), includedthat
signifying in the
the
experiments.
most importantThese results
process suggestinfluencing
parameters that there the
are hardened
many options
depthtoare
consider in building
all included an efficient
in the experiments.
hardened depth (HD) estimation model. As an example, all the process parameters
These results suggest that there are many options to consider in building an efficient hardened and the
depth
interactions are included in a regression model derived from ANOVA and proposed in
(HD) estimation model. As an example, all the process parameters and the interactions are included in Equation (1).
aThe relationship
regression modelbetween
derivedpredicted and measured
from ANOVA tooth top
and proposed hardened (1).
in Equation depthTheis relationship
represented in Figure
between
11.
predicted and measured tooth top hardened depth is represented in Figure 11.
HDTop = −3262 + 8.30·P − 11088·SS − 1.72·RS − 7.96·P·SS − 0.00064·P·RS + 7.00·SS·RS (1)
HDTop = −3262 + 8.30·P − 11,088·SS − 1.72·RS − 7.96·P·SS − 0.00064·P·RS + 7.00·SS·RS (1)
Given the strong correlation between top and bottom hardened depths, ANOVA led to the same
typesGiven
of results. Figure
the strong 12 showsbetween
correlation the average effects
top and of the
bottom process depths,
hardened parameters
ANOVAon theledbottom
to the
hardened
same typesdepth. Again,
of results. P is the
Figure 12 most
showsdominant factor.
the average Thisofobservation
effects the processisparameters
confirmed on by the bottom
percent
contribution.
hardened Table
depth. 8 presents
Again, P is theamost
detailed ANOVA.
dominant P has
factor. This the greatest effect
observation (52.08%).by
is confirmed The
theSS is the
percent
second most significant
contribution. parameter
Table 8 presents with 18.75%
a detailed ANOVA. and RS isthe
P has the least significant
greatest parameter
effect (52.08%). The SSwith the
is the
smallestmost
second effectsignificant
(10.08%). These results
parameter are18.75%
with confirmed
andbyRSthe correlation
is the analysis between
least significant parameterthewith
process
the
parameters
smallest and
effect the hardened
(10.08%). depthsarepresented
These results confirmedinbyTable 7. The contribution
the correlation of the the
analysis between interaction
process
between the
parameters andthree parameters
the hardened is less
depths important,
presented representing
in Table about 16%.
7. The contribution Here
of the again, Pbetween
interaction and SS
interaction
the has a lessissignificant
three parameters contribution.
less important, The about
representing error 16%.
remains
Here within
again,acceptable levels (2.90%)
P and SS interaction has
asuggesting a good
less significant experimental
contribution. The design and planning.
error remains withinAccordingly, all the
acceptable levels process
(2.90%) parameters
suggesting and
a good
the interactions
experimental are and
design included in a Accordingly,
planning. regression model
all theproposed in Equation
process parameters and(2).
theThe relationship
interactions are
between predicted
included and measured
in a regression tooth bottom
model proposed hardened
in Equation (2).depth is represented
The relationship in Figure
between 13.
predicted and
measured tooth bottom hardened depth is represented in Figure 13.
HDBottom = −4982 + 9.89·P − 10515·SS − 2.76·RS − 11.90·P·SS − 0.00055·P·RS +
(2)
9.14·SS·RS
HDBottom = −4982 + 9.89·P − 10,515·SS − 2.76·RS − 11.90·P·SS − 0.00055·P·RS + 9.14·SS·RS (2)

3500
Predicted top hardened depth (μm)

3000

2500

2000

1500
1500 2000 2500 3000 3500
Measured top hardened depth (μm)

Figure 11. Measured and predicted

Figure 11. predicted tooth
tooth top
top hardened
hardened depth.
depth.
J. Manuf. Mater. Process. 2019, 3, x FOR PEER REVIEW 12 of 16

J. Manuf. Mater. Process. 2019, 3, 72

P SS RS 12 of 16

Bottom hardened depth (microns)

1500

1250

1000

750

500

1900 2100 2300 0,25 0,35 0,45 3000 3500 4000

Figure 12. Main effects of the process parameters on the tooth bottom hardened depth.

Figure 12. Main

Tableeffects of the analysis
8. ANOVA process parameters onbottom
for the tooth the tooth bottomdepth.
hardened hardened depth.

Table 8. ANOVA
Characteristic analysis
Sum of for theContribution
Squares tooth bottom hardened
F-Value depth.
p-Value
Regression Sum of12.14
Characteristic Squares 97.10%
Contribution 11.14
F-Value 0.085
p-Value
P 6.51 52.08% 3.59 0.999
Regression 12.14 97.10% 11.14 0.085
P SS 2.34
6.51 18.75%
52.08% 0.42
3.59 0.582
0.999
SS RS 1.26
2.34 10.08%
18.75% 0.26
0.42 0.663
0.582
RS P·SS 1.26
0.06 10.08%
0.45% 0.26
1.39 0.663
0.360
P·SSP·RS 0.06
1.03 0.45%
8.29% 1.39
0.07 0.360
0.811
P·RS 1.03 8.29% 0.07 0.811
SS·RS 0.93 7.44% 5.12 0.152
SS·RS 0.93 7.44% 5.12 0.152
ErrorError 0.36
0.36 2.90%
2.90%
TotalTotal 12.5
12.5 100%
100%

2500
Predicted bottom hardened depth (μm)

2000

1500

1000

500

0
0 500 1000 1500 2000 2500
Measured bottom hardened depth (μm)

Figure 13. Measured and

Figure 13. and predicted
predicted tooth
tooth bottom
bottom hardened
hardened depth.
depth.

ANOVA
ANOVAisisused to investigate
used whether
to investigate the process
whether parameters
the process and the interactions
parameters have significant
and the interactions have
effects on theeffects
significant hardness
on profile and to identify
the hardness profile the
andvariables to include
to identify in the model.
the variables ANOVA
to include in is also
the used
model.
to estimate
ANOVA is the
alsosignificance of thethe
used to estimate regression model
significance at 5%
of the significance
regression modellevel. Thesignificance
at 5% model is considered
level. The
significant if the significance probability value (p-value) is less than 0.05.
model is considered significant if the significance probability value (p-value) is less than Despite the0.05.
F-value of
Despite
10.26 and 11.14
the F-value for top
of 10.26 andhardened
11.14 fordepth and bottom
top hardened hardened
depth depth,hardened
and bottom respectively, suggesting
depth, that
respectively,
the models are significant, the p-values for the model are bigger than the significance
suggesting that the models are significant, the p-values for the model are bigger than the significance level of 5%,
and
levelthe
of model
5%, andterms are statistically
the model terms are insignificant. The coefficients
statistically insignificant. The of determination
coefficients used to measure
of determination used
the performance of the models do not exceed 87% and 88% for top and bottom
to measure the performance of the models do not exceed 87% and 88% for top and bottom hardened hardened depths,
respectively. These observations suggest that the regression approach is inappropriate for this problem.
J. Manuf. Mater. Process. 2019, 3, x FOR PEER REVIEW 13 of 16

depths, respectively. These observations suggest that the regression approach is inappropriate for
this problem.
J. Manuf. Mater. Process. 2019, 3, 72 13 of 16
In order to produce accurate and efficient models for hardened depth prediction, an alternative
modeling approach able to handle variables exhibiting complex and nonlinear relationships is
In order
needed. to produce
Although severalaccurate
modeling and efficient models
techniques for hardened
can be used to produce depth
suchprediction, an alternative
a model, artificial neural
modeling approach able to handle variables exhibiting complex and nonlinear
networks (ANN) is an effective tool for this type of applications [29,30], thus ANN is proposed as an relationships is needed.
Although
illustrative several
examplemodeling
for thetechniques
hardened depthcan be prediction
used to producemodel. such a model,
Although artificial
many ANN neural
models networks
can be
(ANN)
used inisthisan application,
effective toolthe formultilayer
this type offeedforward
applicationsnetwork
[29,30], thusseems ANN is one
to be proposed
of the as an convenient
most illustrative
example for the hardened
for its suppleness depth Before
and simplicity. prediction model.
starting Although
to train the ANN many ANN
model, it ismodels
importantcan to
beestablish
used in
this application,
the number and thethe multilayer
size of the feedforward
hidden layersnetwork
and adjust seems thetotraining
be one performance.
of the most convenient
The idea isfor to
its suppleness and simplicity. Before starting to train the ANN model,
identify the relationship between the number of inputs, the hidden layer size, and the parameters to it is important to establish
the number
estimate. Forandthisthe size of
reason, the hidden
several layers and adjust
ANN architectures the training
are studied. The bestperformance. The ideausing
results are achieved is to
identify the relationship
an architecture combining between the number
three inputs, of inputs,
one hidden layer the hidden
with layer size,
10 neurons, andand theoutputs.
three parameters to
estimate. For this reason,
The experimental dataseveral
from ANNthe L9 architectures
design is usedare tostudied.
train theThe best results
ANN-based are achieved
prediction model. using
The
an architecture combining
experimentation three inputs,
results suggest onethree
that the hidden layer with
process 10 neurons,
parameters have and three outputs.
significant effects on the
The experimental
hardened data from
depth. Consequently, P,the
SS, L 9 design
and RS areisused
used asto traintothe
input theANN-based
ANN model.prediction
Moreover,model.
to test
The experimentation results suggest that the three process parameters
the ANN-based prediction model, additional experimental data are produced for the model have significant effects on the
hardened
validationdepth.phase.Consequently,
The experimentationP, SS, andisRS are used
built usingasaninput to the
L8 and theANN
middle model. Moreover,
values to test
of the factors
the ANN-based prediction model, additional experimental data are produced
considered in the L9. Table 9 shows the factors and levels used in the experimental design for the for the model validation
phase.
model The experimentation is built using an L8 and the middle values of the factors considered in the
validation.
L9 . Table
The 9measured
shows theand factors and levels
predicted used indepths
hardened the experimental
are shown design in Figurefor 14
theandmodel validation.
Figure 15 for the
The measured and predicted hardened depths are shown
regression and the ANN-based prediction models. Figure 16 and Figure 17 present in Figures 14 and 15 for the regression
measured and and
the
the ANN-based
predicted curves prediction
fittingmodels.
for toothFigures 16 and
top and 17 present
bottom hardenedthe measured and the predicted
depth, respectively. These curves
results
fitting for tooth
demonstrate topthe
that andANN bottom hardened
presents moredepth, respectively.
enhanced modelingThese results demonstrate
performance that the ANN
than the regression. With
presents moreofenhanced
a coefficient determinationmodeling performance
at more than 98%, than
thethe regression.
modeling With a can
approach coefficient
be veryofeffective
determination
in this
at more
case. than 98%, the modeling
Experimentation coveringapproach
more factors can and
be verymoreeffective in this
levels for morecase. Experimentation
training covering
and validation data,
more
as well factors
as an and more levels for
improvement of more training and
the modeling validation
conditions, candata,
leadastowell
moreas an improvement
accurate of the
and efficient
modeling
predictionconditions,
models. can lead to more accurate and efficient prediction models.

Table9.9. Factors
Table Factors and
and levels
levels for
forvalidation.
validation.

FactorsFactors Level
Level 1 1 Level 2 2
Level
PowerPower
(W) (W) 2000
2000 22002200
Scanning
Scanning speed mm/s) 0.30.3
speed (mm/s) 0.4 0.4
Revolution speed (rpm)
Revolution speed (rpm) 3250
3250 37503750

3500
Mesured
Predicted_Reg
Predicted top hardened depth (μm)

Predicted_ANN
3000

2500

2000

1500
1500 2000 2500 3000 3500
Measured top hardened depth (μm)

Figure 14. Measured and

Figure 14. and predicted
predicted tooth
tooth top
top hardened
hardened depth.
depth.
J.J. Manuf.
Manuf. Mater.
Mater. Process.
Process.2019,
2019,3,
3,x72FOR PEER REVIEW 1414of
of 16
16
J. Manuf. Mater. Process. 2019, 3, x FOR PEER REVIEW 14 of 16
J. Manuf. Mater. Process. 2019, 3, x FOR PEER REVIEW 14 of 16
2500
2500 Mesured
2500 Mesured
Predicted_Reg
Mesured
Predicted_Reg
Predicted_ANN

(μm)
2000 Predicted_Reg
Predicted_ANN

(μm)
2000 Predicted_ANN

(μm)
2000

depth
depth
depth
1500

hardened
1500
1500
hardened
hardened 1000
1000
bottom

1000
bottom
bottom

500
500
Predicted

500
Predicted
Predicted

0
0 0 500 1000 1500 2000 2500
00 500 1000 1500 2000 2500
0 Measured bottom
500 1000 hardened
1500depth (μm)
2000 2500
Measured bottom hardened depth (μm)
Measured bottom hardened depth (μm)
Figure 15. Measured and predicted tooth bottom hardened depth.
Figure 15. Measured and predicted tooth bottom hardened depth.
Figure 15. Measured
Figure15. Measured and
and predicted
predictedtooth
toothbottom
bottomhardened
hardeneddepth.
depth.
4000
4000 Measured
4000 Measured
Predicted_Reg
Measured
Predicted_Reg
Predicted_ANN
Predicted_Reg
(μm)

3500 Predicted_ANN
Predicted_ANN
(μm)

3500
(μm)

3500
depth
depth

3000
depth

3000
hardened

3000
hardened
hardened

2500
2500
2500
Top
Top

2000
Top

2000
2000

1500
1500 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
1500 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
1 2 3 4 5 6 7Test
8 number
9 10 11 12 13 14 15 16 17
Test number
Test number
Figure
Figure 16.
16. Curve
Curve fitting
fitting of
of measured
measured and
and predicted
predicted tooth
tooth top
top hardened
hardened depth.
depth.
Figure 16. Curve fitting of measured and predicted tooth top hardened depth.
Figure 16. Curve fitting of measured and predicted tooth top hardened depth.
2500
2500 Measured
2500 Measured
Predicted_Reg
Measured
Predicted_Reg
(μm)

Predicted_ANN
2000 Predicted_Reg
(μm)

Predicted_ANN
(μm)

2000 Predicted_ANN
depth

2000
depth
depth

1500
hardened

1500
hardened

1500
hardened

1000
1000
Bottom

1000
Bottom
Bottom

500
500
500

0
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
1 2 3 4 5 6 7Test
8 number
9 10 11 12 13 14 15 16 17
Test number
Figure 17.
Figure Curvefitting
17. Curve fittingof
ofmeasured andTest
measured and numbertooth bottom hardened depth.
predicted
predicted tooth bottom hardened depth.
Figure 17. Curve fitting of measured and predicted tooth bottom hardened depth.
Figure 17. Curve fitting of measured and predicted tooth bottom hardened depth.
4. Conclusions
4. Conclusions
4. Conclusions
This paper presents an experimental investigation on laser surface hardening of AISI 4340 steel
4. Conclusions
This paper presents an experimental investigation on laser surface hardening of AISI 4340 steel
spurThis
gears. The presents
paper investigation was conducted
an experimental using a structured
investigation on laser design
surfaceofhardening
experiments and 4340
of AISI improved
steel
spur gears.
This The investigation
paper presents an was conducted
experimental using a structured
investigation on design
laser surfaceof hardening
experimentsof and
AISIimproved
4340 steel
statistical analysis tools. The effect of process parameters, such as laser power, scanning
spur gears. The investigation was conducted using a structured design of experiments and improved speed, and
statistical
spur gears.analysis tools. The effect
The investigation of process using
was conducted parameters, such as
a structured laser of
design power, scanning
experiments speed,
and and
improved
statistical analysis tools. The effect of process parameters, such as laser power, scanning speed, and
statistical analysis tools. The effect of process parameters, such as laser power, scanning speed, and
J. Manuf. Mater. Process. 2019, 3, 72 15 of 16

revolution speed, and their interactions on tooth top hardened depth and tooth bottom hardened
depth were evaluated and their variations were analyzed. The experimentations were performed using
a 3 kW Nd:YAG laser system. Based on the experimental results, regression models for predicting the
hardened depth for the top and the bottom of the tooth were developed and evaluated. The direct and
the interactive effects of the process parameters on the hardened depth were investigated. The results
confirm a significant similarity between the top and bottom hardness profiles, as anticipated. According
to the results of ANOVA, laser power is the factor with greater influence on the hardness profile followed
by scanning speed and revolution speed. The cumulative contribution of the three parameters in the
hardened depth variation represents more than 80% with a predominance of laser power with more than
50%. The effects of scanning speed and rotation speed are limited to less than 20%. The contribution of
the interactions between the parameters represents 12% to 16%. The resulting hardness values are
relatively similar for all the experimental tests with about 60 HRC. The evaluation of the performance
of the regression prediction models reveals a significant lack of precision, suggesting that the regression
approach is inappropriate for the studied problem. An alternative modeling approach handling
variables exhibiting complex and nonlinear relationships is proposed to produce accurate and efficient
models for hardened depth prediction. A simplified artificial neural network-based perdition model is
presented as an explanatory example.

Author Contributions: A.K.B., A.E.O. and A.C. developed the experimental design and the experimentation
strategy; A.K.B. performed the experiments, the characterization tests, and the statistical analysis of the results;
A.K.B. wrote the manuscript; A.E.O. and A.C. supervised the experiments and the analysis; A.E.O. and A.C.
revised the manuscript.
Funding: This research has been supported by PARL team funds.
Acknowledgments: The authors would like to thanks the PARL team members who made this research possible.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Mazumder, J. Laser heat treatment: The state of the art. J. Mater. 1983, 35, 18–26. [CrossRef]
2. Kennedy, E.; Byrne, G.; Collins, D.N. A review of the use of high power diode lasers in surface hardening.
J. Mater. Process. Technol. 2004, 155, 1855–1860. [CrossRef]
3. Mackwood, A.P.; Crafer, R.C. Thermal modelling of laser welding and related processes: A literature review.
Opt. Laser Technol. 2005, 37, 99–115. [CrossRef]
4. Grum, J.; Sturm, R. Laser surface melt-hardening of gray and nodular irons. Appl. Surf. Sci. 1997, 109,
128–132. [CrossRef]
5. Sandven, O.A. Laser surface transformation hardening. In Metals Handbook; ASM Press: Washington, DC,
USA, 1997; Volume 4, pp. 507–517.
6. Colombo, V.; Mentrelli, A.; Trombetti, T. Time-dependent 3-D modelling of laser surface heating for the
hardening of metallic materials. Eur. Phys. J. D Atomic Mol. Opt. Plasma Phys. 2003, 27, 239–246. [CrossRef]
7. Kannatey-Asibu, E., Jr. Principles of Laser Materials Processing; John Wiley & Sons: New York, NY, USA, 2009.
8. Steen, W.; Watkins, K.G.; Mazumder, J. Laser Material Processing; Springer Science & Business Media:
New York, NY, USA, 2010.
9. Chattopadhyay, R. Laser assisted surface engineering processes. In Advanced Thermally Assisted Surface
Engineering Processes; Springer: Boston, MA, USA, 2004; pp. 157–192.
10. Goia, F.; De Lima, M. Surface hardening of an AISI D6 cold work steel using a fiber laser. J. ASTM Int. 2011,
8, 315–318. [CrossRef]
11. Tobar, M.J.; Álvarez, C.; Amado, J.M.; Ramil, A.; Saavedra, E.; Yáñez, A. Laser transformation hardening of
a tool steel: Simulation-based parameter optimization and experimental results. Surf. Coat. Technol. 2006,
200, 6362–6367. [CrossRef]
12. Shercliff, A.M.F. The prediction of case depth in laser transformation hardening. Metal. Trans. A 1991, 22,
2459–2466. [CrossRef]
J. Manuf. Mater. Process. 2019, 3, 72 16 of 16

13. Cheung, N.; Pinto, M.A.; Ierardi, M.C.F.; Garcia, A. Mathematical modeling and experimental analysis of the
hardened zone in laser treatment of a 1045 AISI steel. Mater. Res. 2004, 7, 349–354. [CrossRef]
14. Komanduri, R.; Hou, Z.B. Thermal analysis of laser surface transformation hardening-optimization of process
parameters. Int. J. Mach. Tools Manuf. 2004, 44, 991–1008. [CrossRef]
15. Patwa, R.; Shin, Y.C. Predictive modeling of laser hardening of AISI5150H steels. Int. J. Mach. Tools Manuf.
2007, 47, 307–320. [CrossRef]
16. Yilbas, B.S.; Shuja, S.Z.; Khan, S.M.A.; Aleem, A. Laser melting of carbide tool surface: Model and
experimental studies. Appl. Surf. Sci. 2009, 255, 9396–9403. [CrossRef]
17. Mioković, T.; Schulze, V.; Vöhringer, O.; Löhe, D. Prediction of phase transformations during laser surface
hardening of AISI 4140 including the effects of inhomogeneous austenite formation. Mater. Sci. Eng. A 2006,
435, 547–555. [CrossRef]
18. Ashby, M.F.; Easterling, K.E. The transformation hardening of steel surfaces by laser beams—I. Hypo-
Eutectoid steels. Acta Metal. 1984, 32, 935–948. [CrossRef]
19. Benyounis, K.Y.; Olabi, A.G.; Hashmi, M.S.J. Multi response optimization of CO2 laser welding process of
austenitic stainless steel. Opt. Laser Technol. 2008, 40, 76–87. [CrossRef]
20. An, T.J.S. DOE based statistical approaches in modeling of laser processing—Review & suggestion. Int. J.
Eng. Technol. 2010, 10, 1–7.
21. Lambiase, F.; Di Ilio, A.M.; Paoletti, A. Prediction of laser hardening by means of neural network. Procedia
CIRP 2013, 12, 181–186. [CrossRef]
22. Hadhri, M.; El Ouafi, A.; Barka, N. Prediction of the hardness profile of an AISI 4340 steel cylinder heat-treated
by laser-3D and artificial neural networks modelling and experimental validation. J. Mech. Sci. Technol. 2017,
31, 615–623. [CrossRef]
23. Jerniti, A.G.; El Ouafi, A.; Barka, N. A predictive modeling based on regression and artificial neural network
analysis of laser transformation hardening for cylindrical steel workpieces. J. Surf. Eng. Mater. Adv. Technol.
2016, 6, 149–163. [CrossRef]
24. Ion, J.C. Laser transformation hardening. Surf. Eng. 2002, 18, 14–31. [CrossRef]
25. Zhang, H.; Shi, Y.; Xu, C.Y.; Kutsuna, M. Surface hardening of gears by laser beam processing. Surf. Eng.
2003, 19, 134–136. [CrossRef]
26. Dinesh Babu, P.; Balasubramanian, K.R.; Buvanashekaran, G. Laser surface hardening: A review. Int. J. Surf.
Sci. Eng. 2011, 5, 131–151. [CrossRef]
27. Barka, N.; El Ouafi, A. Effects of laser hardening process parameters on case depth of 4340 steel cylindrical
specimen—A statistical analysis. J. Surf. Eng. Mater. Adv. Technol. 2015, 5, 124–135. [CrossRef]
28. Ross, P.J. Taguchi Techniques for Quality Engineering: Loss Function, Orthogonal Experiments, Parameter and
Tolerance Design, 2nd ed.; McGraw-Hill: New York, NY, USA, 1996; pp. 23–40.
29. Huang, S.H.; Zhang, H.C. Artificial neural networks in manufacturing: Concepts, applications, and
perspectives. IEEE Trans. Comp. Pack. Manuf. Technol. 1994, 17, 212–228. [CrossRef]
30. Meireles, M.R.; Almeida, P.E.; Simoes, M.G. A comprehensive review for industrial applicability of artificial
neural networks. IEEE Trans. Ind. Electron. 2003, 50, 585–601. [CrossRef]

© 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/).