Journal of Materials Processing Technology 231 (2016) 468–478

Contents lists available at ScienceDirect

Journal of Materials Processing Technology

journal homepage: www.elsevier.com/locate/jmatprotec

Effect of real-time cooling rate on microstructure in Laser Additive

Manufacturing
Mohammad H. Farshidianfar ∗ , Amir Khajepour, Adrian P. Gerlich
Mechanical and Mechatronics Engineering Department, University of Waterloo, ON N2L3G1, Canada

a r t i c l e i n f o a b s t r a c t

Article history: An infrared thermal imaging system is developed to compare the effects of cooling rate and melt pool
Received 25 May 2015 temperature on the microstructure of 316L stainless steel produced by Laser Additive Manufacturing
Received in revised form 14 January 2016 (LAM) with injected powder melting. The system captures the real-time cooling rate to provide online
Accepted 19 January 2016
monitoring and control of the microstructure at the onset of solidification. Further study is conducted
Available online 22 January 2016
to evaluate cooling rate effects on grain size, phase transformations and hardness. Cooling rate changes
are also studied with respect to the traveling speed, in order to identify a suitable controlling action for a
Keywords:
microstructural controller. The study provides a strong experimental justification of the current theories
Real-time cooling rate
Microstructure
in solidification and microstructural evolution during the LAM with injected powder melting process.
Laser Additive Manufacturing © 2016 Elsevier B.V. All rights reserved.
Infrared thermal imaging

1. Introduction the fine equiaxed grains in LAM has two main advantages (Kou,
2003). First, the fine grains help reduce susceptibility to crack-
Laser Additive Manufacturing (LAM) with injected powder melt- ing during deposition (Kou, 2003) in LAM as shown by Mitzner
ing implies layer by layer shaping of materials in the form of powder et al. (2012). Second, fine grains can improve ductility and fracture
to arbitrary configurations and geometries, using a comprehensive toughness of the deposit in the case of steels and stainless steels
integration of materials science, mechanical and control engineer- (Kou, 2003). These advantages make it possible to produce locally
ing, and laser technology. The technology is currently being used in tailored microstructures.
rapid prototyping, coating, tooling and parts repair, cladding and Therefore, there has been numerous attempts to understand
design of novel alloys or functionally graded materials (Toyserkani microstructural evolution in LAM to control material properties.
et al., 2005). Parts can be produced with material properties supe- These researches were mainly focused on trying to obtain a set
rior to those of the wrought material with a finer grain size using of optimized process parameters in an offline manner to obtain
LAM (Griffith et al., 2000). In general, the rapid solidification and desired properties. Ultimate tensile strength, yield strength and
high cooling rate values result in detailed finer microstructure. LAM total elongation of 316 stainless steel depositions produced by the
with injected powder melting is the technology that is applied and LAM process were compared to annealed bar stainless steels by
studied in this paper. For brevity, the technology will be abbreviated Ready (2001). In all cases, the strength properties of the LAM fabri-
as LAM throughout the whole paper. cated 316 stainless steel bars significantly exceeded that for the
Despite being a promising manufacturing technology, it is reported value of the annealed material, whereas, their elonga-
essential to understand and control the material behavior during tions did not differ. Pinkerton and Li (2004) studied the connections
the LAM process to be capable of producing functional parts for crit- between input parameters and final material properties and sur-
ical industries such as the aerospace, energy and automotive. The face finish in laser cladding of stainless steel 316L. Delgado et al.
real promise of LAM is the potential to integrate superior material (2011) examined the role of scanning speed, layer thickness, and
properties with a precise fabrication process. LAM provides attrac- building direction as process parameters, and noted these all affect
tive characteristics including minimal dilution, concentrated heat the part quality and mechanical properties of iron-based prod-
source and smaller heat affected zones compared to the more tra- ucts manufactured by Direct Metal Laser Sintering (DMLS) and SLM
ditional subtractive manufacturing techniques. The formation of technologies. Song et al. (Song et al., 2014) obtained different tensile
strengths of the Fe/SiC depositions through variations of the SLM
inputs parameters. The processing conditions, microstructural fea-
∗ Corresponding author. tures, and mechanical properties of DMLS-processed porous metals
E-mail address: mhfarshi@uwaterloo.ca (M.H. Farshidianfar). were assessed and the formation mechanisms of pores were elu-

http://dx.doi.org/10.1016/j.jmatprotec.2016.01.017
0924-0136/© 2016 Elsevier B.V. All rights reserved.
 M.H. Farshidianfar et al. / Journal of Materials Processing Technology 231 (2016) 468–478 469

cidated by Gu et al. (Shen et al., 2008; Gu and Shen, 2008). Li to control the microstructure of the LAM deposition by means of
et al. (2009) later fabricated 316L stainless steel samples with a real-time monitoring of the cooling rate and melt pool temperature.
graded porous microstructure along the speed gradient direction. The main objective of the present study is to develop a gener-
Zhao et al. (2015) studied the morphology of AISI 420 stainless alized methodology to monitor and control local microstructural
steel deposition during the SLM process and observed a cellular properties of the LAM process with injected powder melting using
microstructure along the direction of the maximum heat flow. In real-time cooling rate measurements. By integrating infrared ther-
addition, Shifeng et al. (Shifeng et al., 2014) observed a coarse mal imaging and processing techniques, the real-time melt pool
grain zone below the Molten Pool Boundaries (MPBs), and a long temperature and cooling rate measurements are obtained. A set
and thin columnar grains with the same orientations on two sides of combined energy and powder deposition density parameters
of the MPBs. Though, all these reports offer a good insight into are used to evaluate the pivotal role of the cooling rate and melt
microstructural variations during LAM, each only cover microstruc- pool temperature on defining the LAM microstructure. The effect
tural analysis of a specific material inside a specific window of of process parameters on the real-time cooling rate and melt pool
process parameters. In other words, none of the above researches temperature is evaluated. Moreover, the microstructural evolution
developed a general methodology which may be applied to control is studied based on the cooling rate and melt pool temperature
material properties during LAM in various materials and process variations to produce a general methodology for understanding
conditions. microstructural formation during the LAM with injected powder
Since LAM is a solidification process, the initial step towards melting process. This methodology is further expanded to under-
achieving a fully controlled material property is to understand the stand the influence of the cooling rate on the grain size and
thermal behavior (Flemings 1974; Kobryn and Semiatin, 2003). By geometrical dilution. The real-time cooling rate is found to be an
understanding the correlation between the thermal history and effective parameter for local microstructure control purposes, such
microstructure, precise locally tailored microstructures can be fab- as local grain refinement in ultra-precision parts.
ricated through graded deposition of single or multi-materials and
optimized process parameters. Elmer et al. (1989a,b) were the
first to characterize the effects of different processing conditions 2. Thermal characteristics of the LAM process
on the microstructure using the cooling rate in an electron beam
deposition process. They indicated the cooling rate and chemical Any solidification microstructure is defined by two main solid-
composition primarily determine the mode of solidification, solid ification characteristics: (1) solidification mode (planar, cellular,
distribution and nucleation. Their cooling rates were estimated columnar dendritic and equiaxed dendritic), and (2) size of the
indirectly (Elmer et al., 1989a) based on the dendritic arm spac- solidification structure. Additionally, the two most important
solidification parameters are the growth rateR mm/s , and tem-
ing measurements on 300-series stainless steel alloys, and were  
found to range from 7 ◦ C/s to 7.5×106 ◦ C/s . In 1999, Griffith et al. perature gradient G ◦ C/mm . The travel speed of the solid/liquid
(1999) used a thermocouple for measuring the melt pool temper- interface is called the growth rate, R. The temperature gradient, G, in
ature at different passes of a hollow block during LAM. Although the liquid metal is defined as the tangent of the temperature profile
in-situ thermal measurements were acquired, no correlation was with respect to distance. The ratio G/R determines  solidifica-
the
obtained between microstructure and thermal behavior. In a later tion mode (morphology) while the product GR ◦ C/s governs the
study, Hofmeister and Griffith (2001) used CCD thermal imaging to size of the solidification structure (Kou, 2003). Thus, to control the
study the microstructural variation of LENS processed 316 stainless whole solidification microstructure, one has to monitor and control
and H13 tool steels. Spatial temperature gradients were measured the ratio and product of the growth rate and the temperature gra-
online and later used to calculate the cooling rate since it was diffi- dient. Going from one solidification mode to another requires great
cult to measure online cooling rate values. Hofmeister and Griffith variations inside the LAM process, whereas, the size and morphol-
(2001) reported cooling rate values ranging from 200–6000 ◦ C/s at ogy is very sensitive to process parameters. Since LAM is a very fast
the solid-liquid interface. Doubenskaia et al. (Doubenskaia et al., solidification process compared to casting and welding, it is diffi-
2004; Pavlov et al., 2011; Doubenskaia et al., 2006) used pyrome- cult to capture real-time values of the growth rate and temperature
ters and infrared cameras to measure melt pool temperatures for gradient. Moreover, since in-situ real-time cooling rate measure-
monitoring laser cladding, however, no microstructural analysis ments are difficult to achieve (Hofmeister and Griffith, 2001), no
was carried out. Hua et al. (2008) found a relationship between the researcher has obtained a generalized logic on how to obtain
molten pool temperature and the cladding thickness. Zhang et al. optimized process parameters in order to obtain desired mate-
(2012) indicated that the thermal history of a deposition has an rial properties. However, it could be possible to measure real-time
important effect on the microstructure, and consequently on the cooling rate values with the current thermal monitoring apparatus.
final properties of a multi-layer stainless steel 410 (SS 410) thin Therefore, there is a great possibility that by measuring the cooling
wall. Despite having studied the microstructures in LAM based on rate and correlating its variations with microstructural evolution,
well-established metallurgy of solidification versus thermal behav- one could easily understand and control the local microstructure
ior, none of the researchers were successful to develop an online formation in real-time.
monitoring and control methodology for the microstructure. There are four main mechanisms for grains to nucleate during
There are widely used passive microstructure controlling tech- LAM (Kou, 2003): (1) dendrite fragmentation, (2) grain detach-
niques in the welding and casting industries (e.g. grain refinement ment, (3) heterogeneous nucleation, and (4) surface nucleation. The
and inoculation), however, they are only ideal for bulk material of clad-substrate interface provides nucleation sites for the colum-
low dimensional tolerances, and not suitable for local microstruc- nar growth. The axes of the columnar crystals are normally in the
ture control in graded structures. Furthermore, all grain structure direction of heat flow (Davies, 1973). Two phenomena are respon-
control techniques in welding are open loop, meaning that they sible for production of equiaxed grains, which are desirable in
cannot be monitored online and adapted accordingly. Yet again, the process (Flemings, 1974): grain multiplication and convection.
the need for a generalized microstructure control method that Modeling or predicting convection is a difficult process, however,
can locally control the microstructure and grain size in real-time convection can be monitored and predicted more easily through
is required in LAM. However, in a recent work by the authors cooling rate monitoring.
(Farshidianfar et al., 2015), which is also the basis for the current The amount of actual heat input during the LAM process defines
study, a real-time thermal monitoring apparatus was developed the values of the melt pool temperature and cooling rate. A large
470 M.H. Farshidianfar et al. / Journal of Materials Processing Technology 231 (2016) 468–478

variety of operating parameters and physical phenomena deter- The cooling rate estimation therefore, is not very sensitive to the
mine the amount of heat input. The main input parameters of the emissivity changes for the current application. Moreover, because
current LAM system can be categorized as the travel speed, laser a comparative study is used, no knowledge of the exact tempera-
power, laser beam diameter and powder feed rate. Since these tures are required. Hence, an average emissivity of 0.6 is calculated
parameters affect the heat input simultaneously, it is essential to for the material (stainless steel) being deposited and this emissiv-
use a combined parameter to address their combined influence. The ity is kept constant throughout the whole process and real-time
effective energy per unit area E, and the powder deposition density temperature calculations.
PDD (Toyserkani et al., 2005) are two widely used parameters in lit- The deposition powder used in this work is stainless steel (SS)
erature that can express combined influence of all the main input 316L powder, –325 mesh, from Praxair Surface Technologies. The
parameters. SS 316L powder had a composition of 0.003 wt% C, 17 wt% Cr,
The effective energy measures the amount of energy delivered 65.5 wt% Fe, 1 wt% MN, 3 wt% Mo, 13 wt% Ni, less than 0.005 wt%,
to the process by the laser. This energy is principally responsible 0.01 wt% S and 0.45 wt% Si. All experiments conducted used AISI
for melting the substrate surface and powder, and is defined by 1030 medium carbon steel substrates with a composition of:
(Toyserkani et al., 2005): 0.28–0.34 wt% C, 0.6–0.9 wt% Mn, less than 0.04 wt% P, and less than
0.05 wt% S. Specimens were sectioned in the longitudinal direction
P
E= (1) for microstructural examination. Samples were prepared using SiC
VD
grit paper with grit mesh sizes from 240 to 1200, polished with 1m
in which P is the laser power (W ), V is the traveling speed of the sub- alumina powder and auto polished with 1to 0.5m diamond paste.
strate (mm/s), and D is the laser beam diameter (mm). The effective After each polishing step, the samples were placed in an ultrasonic
energy E is measured in units of J 2 . machine to remove contamination and later air dried. Stainless
mm
The powder deposition density is also a good indicator of the steel samples were etched in Marble’s reagent (10g CuSO4 in 50mL
amount of powder fed to a unit area of the substrate during deposi- HCl and 50mL H2 O), to expose grain structure and morphology.
tion. The powder deposition density is defined by (Toyserkani et al., The microstructures were analyzed using optical microscopy with
2005): images obtained from an Olympus AH microscope with objective
R magnifications of up to50×. All of the height and dilution measure-
PDD = (1) ments were also conducted by the Olympus AH microscope.
VD
Where R is the powder feed rate. The powder deposition density
PDD is measured in units of g 2 . 3.1. Thermography analysis and processing techniques
mm
The effective energy and powder deposition density are two
parameters that describe the combined effects of the laser power, Thermal images taken using an IR camera only provide temper-
travel speed, laser beam diameter and powder feed rate on the atures of each pixel and thus, require processing and analysis to be
amount of heat input during the LAM process. The amount of heat functional for exact microstructural analyses and control applica-
input defines values for the melt pool temperature and cooling rate. tions.
Thus, E and PDD are directly related to the melt pool temperature The intensity of each pixel is directly related to its temperature,
and its cooling rate, and indirectly related to the microstructure the greater the RGB value (the brighter the pixel) the higher the
formation. temperature. The camera was located at a distance of 20 − 25cm
mm
from the substrate that provided a resolution of 0.5 to0.65 pixel .
The image array has a size of X × Y in which X is the total num-
3. Experimental setup
ber of pixel rows and Y is the total number of pixel columns of
the image. Each element of the array has an assigned row r, and
A schematic view of the state of the art Laser Additive Manufac-
columnc, with a position vector (c, r). The intensity (RGB value) of
turing (LAM) apparatus with injected powder melting used in this
each pixel indicates the magnitude or size of each element.
research is illustrated in Fig. 1. This system is composed of several
As stated earlier, to be capable of monitoring changes in the
integrated instruments such as a high power laser (1.1kW IPG fiber
solidification mode and structure size, it is required to have in-
laser YLR-1000-IC with 300m feeding fiber and a Gaussian power
process knowledge of the temperature gradient and growth rate. It
distribution), powder feeder (Sulzer Metco TWIN 10-C), multi-axes
was also realized that the cooling rate, which is the product of the
CNC machine (5-axis Fadal VMC 3016), nozzle and an intelligent
temperature gradient and growth rate, has direct influence on the
controller with a data acquisition system and sensors.
solidification mode and solidification structure size. Therefore, con-
A Jenoptik IR-TCM 384 camera module with a resolution of
trolling microstructure during the LAM process requires a complete
384 × 288 pixles and a measuring accuracy of ±2% is used to moni-
knowledge of the real-time cooling rate values. A general descrip-
tor the thermal behavior of the LAM process. The IR imaging is used
tion of the thermal algorithm that extracts the real-time cooling
to directly measure the cooling rate and temperature of the melt
rate is discussed below.
pool in real time.
An IR camera defines the real temperature of an object using
a constant emissivity valueε. The emissivity of metallic powders 3.1.1. Measuring the melt pool temperature and cooling rate
deposited during LAM (such as stainless steel and titanium) are When a thermal image is taken, the pixel with the maximum
highly dependent on temperature, especially at the elevated tem- RGB value Imax in each frame, represents the melt pool. Thus, the
peratures of the LAM process 800–2000 ◦ C. The change of emissivity center of the melt pool and temperature is found by identifying
with temperature in steel is attributed to the formation of an oxide the maximum element and averaging its’ value with the surround-
layer on the surface of the steel as its temperature increases (Sadiq ing elements inside the image array. The cooling rate is calculated
et al., 2011). This change of emissivity may lead to errors in captur- by deducting the melt pool temperature reduction with time at
ing the real body temperature as high as 100 ◦ C (Felice, 2008). the onset of solidification. In other words, the real-time melt pool
Since in the current application two temperatures at different temperature is stored and later subtracted from the reduced tem-
instances of time are subtracted to find cooling rate values, and also perature of that point in time. Therefore, temperature reduction in
cooling rates are compared and correlated with microstructural the melt pool is compared with a measured line of points, having
evolutions, a comparative study of the cooling rate values are used. the highest temperature progressively further from the previous
 M.H. Farshidianfar et al. / Journal of Materials Processing Technology 231 (2016) 468–478 471

Fig. 1. Experimental setup in Automated Laser Fabrication (ALFa) lab.

melt pool point. More details on the algorithm are provided in prior
work (Farshidianfar et al., 2015).

3.2. Experiments

To explore the influence of cooling rate or melt pool temper-

ature on the microstructure, a set of samples denoted by series
A with identical effective energy and powder deposition density,
but dissimilar process parameters were prepared. Laser processing
parameters for samples A1 − A5 are listed in Table 1.
The parameters of the B series samples are shown in Table 2,
in which the travelling speed of the laser increases gradually from
Fig. 2. Clad cross section in LAM (Toyserkani et al., 2005).
sample B1 to B4. These samples were designed to study the effect
of the travelling speed on the cooling rate and consequent effect
of the cooling on microstructural properties such as hardness and
length of the test line, PL , is obtained. The mean lineal intercept
grain size.
length,l, is obtained as follows:

3.3. Methods for dilution and grain size measurements 1

l= (3)
PL
Clad height and dilution are two very important parameters
that also define final characteristics of a deposited part. Therefore, which has a unit of mm. Finally, the ASTM grain size number, G, is
to understand cooling rate and melt pool temperature effects on derived by the following:
these parameters, clad height and dilution values were measured.
In Fig. 2, h is the clad height, w is the clad width,  is the angle of wet- G = (−6.643856 log l) − 3.288 (4)
ting, and b is the clad depth representing the thickness of melted
substrate during the cladding and added to the clad region. The 4. Results and discussion
geometrical definition of dilution is defined as follows (Toyserkani
et al., 2005): In this section, steps are taken towards understanding the direct
h effect of the real-time cooling rate on the solidification microstruc-
dilution = (2) ture characteristics. By understanding these effects, the real-time
h+b
cooling rate can act as an online monitoring and control signal to
The Heyn (Heyn, 1903) lineal intercept procedure was used to tailor the local microstructure in real-time at the onset of solid-
obtain the ASTM grain size of the current samples. In this proce- ification. To develop a general methodology for controlling the
dure three concentric circles were sketched on a random location of local microstructure during the LAM process, it is necessary to
each sample and the number of grain boundary intersections with determine a physical parameter, which directly affects the clad
the circles, P, were counted. By dividing P with the actual length of microstructure. Thus, by controlling such a parameter, a locally
the circles, L, the number of grain boundary intersections per unit tailored microstructure can be achieved.
472 M.H. Farshidianfar et al. / Journal of Materials Processing Technology 231 (2016) 468–478

Table 1
J g
Laser processing parameters of A samples with similar E = 110 and PDD = 0.008 parameters (beam diameter is2.5mm).
mm2 mm2

Sample No. (W )Laser power Travel Speed (mm/min) Powder Feed Rate (g/min)

A1 1031 225 4.5

A2 917 200 4.0
A3 802 175 3.5
A4 688 150 3.0
A5 573 125 2.5

Table 2
Laser processing parameters of B samples for studying the effects of cooling rate on microstructure, and effects of travelling speed on cooling rate.

Sample no. Laser power (W ) Travel speed (mm/min) Powder feed rate (g/min) Effective energy (J/mm2 ) Powder deposition density (g/min)

B1 700 25 4.0 672 0.064

B2 50 336 0.032
B3 100 168 0.016
B4 200 84 0.008

Fig. 3. Low magnification micrograph of sampleA3.

4.1. Combined effect of real-time cooling rate and melt pool

temperature on clad microstructure

To examine the microstructure and morphology of each A-

sample in Table 1, optical micrographs were captured. Different
magnifications were used to analyze both the general clad cross
section geometry and grain size in the deposit.
The macroscopic structure of a deposition is of great importance
since many mechanical and material properties depend on grain
size and grain shape. The microstructures of each A-sample differ-
entiate in different regions of each cross section across the deposit.
Thus, there was no constant morphology for any of the samples, in
contrary to what the theory suggested in terms of a similar heat
and mass input to the system. A low magnification micrograph of Fig. 4. Real time cooling rate of A samples with similar E = 110 J
and PDD =
mm2
sample A3 is shown in Fig. 3 for better understanding of the deposit 0.008 g
parameters.
mm2
microstructure. Generally, columnar zones dominated the bottom
of the claddings at the clad substrate interface (Fig. 3). All other
regions of the clad had a cellular structure (Fig. 3). The macrostruc- process. Thus, real-time cooling rate and melt pool temperature
tures of the A samples were greatly different in terms of grain size of each sample were captured to understand microstructural evo-
and volume percentage of each cellular and columnar zone. lutions during the process. In Fig. 4 and Fig. 5, the real-time cooling
Since the effective energy and powder deposition density of all rates and melt pool temperatures of each sample are illustrated.
samples are similar, none of the microstructural changes can be The average values of the cooling rate and melt pool temperature
interpreted theoretically considering the combined values or laser for samples A1 − A5 are also listed in Table 3.
processing parameters individually. As discussed above, cooling The effect of laser power and travelling speed on the amount of
rate and melt pool temperature directly influence the microstruc- heat input can be interpreted by comparing the process parame-
ture and solidification modes in the melt pool during the LAM ters of the samples, in Table 1, with the cooling rate and melt pool
 M.H. Farshidianfar et al. / Journal of Materials Processing Technology 231 (2016) 468–478 473

Table 4
Geometrical dilution and clad height of the A samples with similar E = 110 J/mm2
and PDD = 0.008 g/mm2 parameters.

Sample A1 A2 A3 A4 A5

Clad height 81m 76m 66m 70m 7m

Geometrical dilution 48% 51% 52% 43% 63%

Consequently, convection has a vital role in forming the colum-

nar and cellular zones. On the other hand, increasing convection
results in increased cooling rate that is monitored by the IR camera.
The melt pool temperature has a great effect on the formation of
the equiaxed grains. When the melt pool temperature is high, the
Fig. 5. Real time melt pool temperature of A samples with similar E = 110 J
and separated dendrite arms melt and disappear. Hence, as the melt
mm2
PDD = 0.008 g
mm2
parameters. pool temperature increases, the tendency to form equiaxed grains
should decrease. Generally, the equiaxed grains that form in high
melt pool temperatures have an increased grain size (Chalmers,
Table 3 1963).
Average cooling rate and melt pool temperatures A samples with similar
It is essential to understand microstructural changes and their
E = 110 J/mm2 and PDD = 0.008 g/mm2 parameters.
correlation with cooling rate and melt pool temperature variations.
Sample no. Average cooling rate (◦ C/s) Average melt pool temperature (◦ C) High magnification micrographs are required to compare grain size
A1 505 1440 and grain shapes in more detail, where micrographs of each sample
A2 518 1510 are shown in Fig. 6. These micrographs have all been taken at similar
A3 511 1576 locations at the center of the claddings.
A4 426 1621
Samples A1 − A3 have very similar macrostructural and
A5 712 1683
microstructural (Fig. 6) properties. Their macrostructures consist
of an initial columnar zone at the clad substrate interface with a
region of cellular zone in the center. Moreover, according to Fig. 6,
temperatures in Table 3. According to Table 1, the power, travel- generally the size of these grains for samples A1 − A3 are also iden-
ling speed and powder feed rate decrease gradually from A1 to A5. tical. As observed in Table 3, these samples were solidified with
Decreasing power results in less heat input, whereas decreasing similar cooling rate values but different melt pool temperatures.
traveling speed results in more concentrated energy and heat. Thus, The macrostructural consistency with the cooling rate indicates
the power and travelling speed have an opposing influence on the a strong relation between these two parameters. Compared to
amount of heat during the LAM process. A1 − A3, sample A4 has a greater columnar region and less frac-
In Table 3, the melt pool temperature increases from sample tion of cellular grains. This change is due to the decreasing value of
A1 to A5, which indicates an increase in the absorbed heat. As it the cooling rate for A4 compared to A1 − A3. The lower cooling rate
is stated in Table 1, the A-samples have similar effective energy in sample A4 provides more time for the columnar grains to grow,
and powder deposition density. Therefore, the amount of available making it difficult for the cellular grains to further nucleate and
powder per unit area is the same for all of the A-samples. Since precipitate. In addition, the higher melt pool temperature in sam-
the effective energy is also the same, there is equal powder and ple A4 further reduces the chances of the cellular crystals to survive
energy density for all samples in theory. Thus, it cannot be stated in the melt pool. On the other hand, sample A5 has the largest cel-
that there is less powder to melt while the PDD is constant. Neither lular region amongst all the samples. This greater cellular region
can the effect of less powder be neglected. It can be concluded that is caused by the high cooling rate value measured in sample A5.
the changing process parameters (which also includes the feed rate) The increased cooling rate is a burden for dendrites to grow, thus,
result in different absorption of the laser and change in the absorbed resulting in early coarsening and separation of their tips that form
heat, which cannot be predicted by the theoretical parameters. The the equiaxed grains. In spite of the increased cellular region due to
increasing melt pool temperature and absorption can be attributed higher cooling rate, the equiaxed grains in A5 are much bigger in
to the decreasing travelling speed, which increases the heat absorp- size compared to the ones in sample A4. Examining Table 3, it can be
tion. Although, according to Eq. (1), the laser power and travelling noted that sample A5 has a higher melt pool temperature than sam-
speed have similar effects on the theoretical value of the effective ple A4. As stated earlier, it is harder for the cellular grains to grow in
energy, the traveling speed is found to be more influential on the high melt pool temperatures since the smaller grains re-melt due
actual heat input than the laser power. On the other hand, the pres- to the high temperature of the melt pool. Thus, only the bigger cel-
ence of less powder at lower feed rates also contributes to the melt lular grains survived and grown in sample A5 due to a higher melt
pool temperature increase. pool temperature. Moreover, since there are fewer columnar grains
On the other hand, the cooling rate behaves in a nonlinear fash- to limit the growth of the cellular region, the cellular grains grow
ion. Samples A1 − A3 have an equal value of cooling rate, whereas, rapidly and yield a larger size in sample A5.
samples A4 and A5 have different values. Since the effective energy
and powder deposition density of the samples are similar in theory 4.1.1. Effect of real-time cooling rate and melt pool temperature
(Table 3) however, a similar cooling rate and melt pool tempera- on geometrical dilution and clad height
ture were predicted. This inconsistent behavior of the cooling rate In Table 4, clad height and geometrical dilution are calculated
makes it difficult to correlate it with the combined influence of for the samples listed in Table 1. The clad height decreases from A1
travel speed or laser power. It is noteworthy to point out that due to A5. This decrease is in good agreement with the increasing melt
the inconsistent behavior of the cooling rate and melt pool tem- pool temperature for these samples as listed in Table 3, whereas, it
peratures in Table 3, and also to study the repeatability of the does not have any relation with the cooling rates of these samples.
measurements, each sample was deposited two times and similar Thus, as the melt pool temperature increases, the clad height of the
results were observed. samples decrease.
474 M.H. Farshidianfar et al. / Journal of Materials Processing Technology 231 (2016) 468–478

J g
Fig. 6. High magnification 50× micrographs of samples A1 through A5 with similar values of E = 110 and PDD = 0.008 parameters.
mm2 mm2

On the contrary, the geometrical dilution does not have any

specific trend on samples A1 to A5. However, by comparing the real-
time cooling rates (Table 3) with the geometrical dilution of each
sample, it is observed that there is a positive correlation between
the cooling rate and dilution. Sample A5, which has the highest geo-
metrical dilution also has the highest cooling rate. In addition, A4
has the lowest dilution because of its low cooling rate. On the other
hand, samples A1, A2 and A3, which have identical cooling rates, all
have similar geometrical dilutions.
In order to better illustrate the correlation between the cooling
rate and geometrical dilution of the A series samples, geometri-
cal dilution is plotted as a function of cooling rate in Fig. 7. A
pseudo-linear relation exists between the real-time cooling rate
and geometrical dilution. The increase in relative powder feed from Fig. 7. Clad geometrical dilution as a function of cooling rate.
A5 to A1 is commonly known to cavern the dilution and cooling rate,
since there is less interaction between cold powder and laser beam,
microstructural evolutions inside LAM are understood and con-
which with low powder feeding rate can enter more freely to solid
trolled by online monitoring of the cooling rate. Moreover, dilution,
substrate heating it further.
which is of great importance during the LAM process, was found to
Clearly, it is the cooling rate rather than the melt pool tem-
have a close correlation with the real-time cooling rate as well.
perature that is governing the microstructure at the onset of the
solidification process. Such a conclusion is widely accepted in the-
4.2. Effect of real-time cooling rate on microstructure
ory and literature, however, no online measurements of the cooling
rate were available to further prove the pivotal influence of this
Having defined the cooling rate as the most influential parame-
physical parameter during the LAM process. Therefore, real-time
ter on the microstructural evolution, it is necessary to study the sole
 M.H. Farshidianfar et al. / Journal of Materials Processing Technology 231 (2016) 468–478 475

Table 5
Average cooling rate and melt pool temperatures B samples for studying cooling
rate effects on microstructural evolutions.

Sample no. Average cooling rate (◦ C/s) Average melt pool temperature (◦ C)

B1 22 2053
B2 39 1630
B3 248 1434
B4 764 1153

Table 6
Micro hardness results of B samples.

Sample no. Hardness (HV1000) Average cooling rate ((◦ C/s))

 mm
  mm
  mm
 B1 150.5 22
Fig. 8. Cooling rate of B samples: B1 v = 25 min , B2 v = 50 min , B3 v = 100 min
  B2 150.5 39
andB4 v = 200 mm
min
. B3 153.3 248
B4 368.0 764

is located on the left side of the three-phase equilibrium triangle

it shall remain austenitic upon cooling to room temperature.
Microstructures of the SS 316L depositions in Fig. 10 consist of
two regions: (1) light regions, which represent the -austenite
phase, and (2) dark regions, which represent the ı −ferrite phase.
According to Table 5 and the micrographs shown in Fig. 10, as
the cooling rate decreases from B1 to B3 the ferrite content of the
deposition decreases likewise. The increased cooling rate in B3 and

to some extent  reduces
in B2,  the  chances of ı-ferrite formation.
Since the  + ı + L and  + ı inter-phases are formed at very
narrow temperature ranges, high cooling rates will reduce the
total time of transitions in which ı-ferrite is formed. Thus, at high
Fig. 
9. Melt  pool  
temperature  of  B  samples: cooling rates less time is allocated for the formation of the ı-ferrite
B1 v = 25 mm , B2 v = 50 mm , B3 v = 100 mm and B4 v = 200 mm .
min min min min phase, resulting in a primary uniform austenite phase with cellular
and dendritic austenitic substructures.
effect of real-time cooling rate on the clad microstructure. On the Surprisingly, B4 has a higher content of ferrite compared to B2
other hand, it was also observed that the travel speed had a greater and B3, although having a much greater cooling rate. Moreover,
influence on the cooling rate compared to the other two parameters compared to the cellular microstructure of samplesB1 − B3, sam-
(i.e. laser power and powder feed rate). Thus, a new set of experi- ple B4 has a needle-like microstructure. This type of microstructure
◦
ments were conducted; changing the travel speed, while keeping and the very high cooling rate (C = 764 sC ) are signs of marten-
the laser power and powder feed rate constant during single-line site formation during solidification. To verify whether sample B4
depositions. is martensite or not, micro hardness tests were taken from these
Cooling rate and melt pool temperatures of the B samples are samples. Table 6 shows the micro hardness of the B samples, which
presented in Fig. 8 and Fig. 9, respectively. It is apparent from these were measured at similar locations on top of the clad cross sections.
figures that as the speed increases, the cooling rate increases dra- It is clear that B4 has a much greater hardness compared to the other
matically, whereas the melt pool temperature decreases. Although, three samples, and values >350HV are indications that martensite
the cooling rate and melt pool temperature are both affected by the is present. Since martensite has a higher hardness compared to
speed change, the changes in the cooling rate are more significant. austenite, the higher hardness in sample B4indicates that the pri-
More interestingly, cooling rate measurements reveal that as the mary phase in this sample is martensite opposed to austenite in
◦
effective energy (E) drops (see Table 2) from sample B1 to B4, the samplesB1 − B3. Since B4 has a very high cooling rate (C = 764 sC ),
cooling rate increases. On the other hand, although, the decrease the austenite is quenched to the unstable phase martensite, which
in E should result in a decrease in the melt pool temperature, an is not present in the other samples. Thus, the significance of cooling
increasing trend is observed for the melt pool temperature. This rate on the primary phase of solidification is once again justified.
increase of melt pool temperature is due to the decreasing powder Furthermore, as a consequence of the high cooling rate, B4 has a
deposition density (PDD) (see Table 2), which indicates less amount higher dilution. Therefore, due to the higher geometrical dilution
of mass to be heated up. Therefore, it is observed that the effective of the mild steel substrate, martensite is produced.
energy (E) is positively correlated with the cooling rate, whereas, it
is the powder deposition density (PDD) that defines the melt pool 4.2.1. Grain size variation with cooling rate
temperature. To better illustrate grain size variations in the microstructure,
In order to compare cooling rate and melt pool temperature a higher magnification micrograph of the B samples are presented
variations in more detail, the average cooling rate and melt pool in Fig. 11. There is significant correlation between the grain size
temperature values of the stable regions in each of the B samples and real-time cooling rate. According to Fig. 10 and Fig. 11, with
are listed in Table 5. Fig. 10 compares low magnification micro- successive increase of the cooling rate, grain boundaries shrink
graphs of the B samples. Sample B1 has a greater height compared drastically. Consequently, B1, which has the lowest cooling rate,
to the other samples, hence, only top of the sample B1 clad is shown. yields the largest grain size, and B4, which has the highest cooling
The solidification of the  SS 316L undergoes
  the following rate, has very fine grain structures.
transitions:(L) → ( + L) →  + ı + L →  + ı → (). Thus, The ASTM grain size number of each B sample is listed in Table 7.
the primary solidification occurs as austenite and since the line For each sample, three fields were measured. The subsequent grains
476 M.H. Farshidianfar et al. / Journal of Materials Processing Technology 231 (2016) 468–478

Fig. 10. Low magnification micrographs of B samples.

Fig. 11. High magnification micrographs of Bsamples.

Table 7
Grain size number of the B samples with changing speed.

Sample no. G-ASTM grain size (No.) NA -Grains/unit area (No./mm2 at1×) d̄ -Average grain diameter (␮m)

B1 9 3968.0 15.9
B2 10 7936.0 11.2
B3 12 31744.1 5.6
 M.H. Farshidianfar et al. / Journal of Materials Processing Technology 231 (2016) 468–478 477

rate control may be used to tailor the microstructure changes by

varying laser power and traveling speed.

5. Conclusion

Using a novel infrared thermal imaging technique, real-time

cooling rate and melt pool temperature of the LAM process were
captured online during solidification. A set of samples with identi-
cal effective energy and powder deposition density values were
produced. Although, the amount of heat and mass input for all
samples were the same, different microstructures were observed.
Studying real-time cooling rate values measured by the developed
system, positive correlations were observed between microstruc-
Fig. 12. Effect of traveling speed on cooling rate. ture changes and variations of the real-time cooling rate during
solidification. Microstructural analysis proved that both the size
of the solidification structure and mode of solidification can be
defined by the real-time cooling rate. Additionally, the geometrical
dilution of different samples was observed to be in direct relation
with the cooling rate. Finally, the change of the cooling rate and
melt pool temperature were studied as a function of the travelling
speed.

Acknowledgments

The authors would like to acknowledge the financial support of

the Natural Sciences and Engineering Research Council of Canada.

Appendix A. Supplementary data

Fig. 13. Effect of traveling speed on melt pool temperature.
Supplementary data associated with this article can be found, in
the online version, at http://dx.doi.org/10.1016/j.jmatprotec.2016.
per unit area (NA ) and average grain diameter (d̄) of each grain size 01.017.
number were also obtained using the tables provided in ASTM stan-
dard E 112 from the experimentally measured G-ASTM grain size References
number.
The results show that the number of grains per unit increases Chalmers, B., 1963. The structure of ingot. J. Aust. Inst. Metals 8, 255–263, Available
at: http://scholar.google.com/scholar?q=#0 [Accessed June 12, 2014].
with the cooling rate. On the other hand, the grain diameter Davies, G.J., 1973. Solidification And Casting. Essex: Applied Science Publishers LTD.
decrease with the increasing cooling. However, more measure- Delgado, J., Ciurana, J., Rodríguez, C.A., 2011. Influence of process parameters on
ments at different cooling rates are required to establish an analytic part quality and mechanical properties for DMLS and SLM with iron-based
materials. Int. J. Adv. Manuf. Technol. 60 (5-8), 601–610, Available at: http://
relation between the cooling rate and the grain size. Generally, the link.springer.com/10.1007/s00170-011-3643-5 [Accessed July 13, 2014].
cooling rate has direct effect on: (1) the volumetric percentage of Doubenskaia, M., Bertrand, P., Smurov, I., 2006. Pyrometry in laser surface
austenite and ferrite phases inside the microstructure, (2) forma- treatment. Surf. Coat. Technol. 201 (5), 1955–1961, Available at: http://
linkinghub.elsevier.com/retrieve/pii/S0257897206003768 [Accessed July 15,
tion and growth of these phases, and (3) grain boundaries and grain 2014].
size. Doubenskaia, M., Bertrand, P., Smurov, I.Y., 2004. In: Veiko, V.P. (Ed.), Temperature
Monitoring of Nd:YAG Laser Cladding (CW and PP) by Advanced Pyrometry
and CCD-Camera based Diagnostic Tool. Laser-Assisted Micro- and
4.3. Cooling rate and microstructure variations with travelling Nanotechnologies, pp. 212–219, Available at: http://proceedings.
speed spiedigitallibrary.org/proceeding.aspx?articleid=843139 [Accessed July 14,
2014].
Elmer, J.W., Allen, S.M., Eagar, T.W., 1989a. Microstructural development during
It was realized that cooling rate variations resulted in solidification of stainless steel alloys. Metall. Trans. A 20A (October),
microstructural evolutions, however, the source of these variations 1989–2117.
Elmer, J.W., Allen, S.M., Eagar, T.W., 1989b. The influence of cooling rate on the
and the effects of speed on the cooling rate were not discussed. ferrite content of stainless steel alloys. In: Proceedings of the 2nd International
It is quite revealing to combine the input and output data of Conference on Trends in Welding Research, Gatlinburg: ASM International, pp.
the LAM process in Table 2 and Table 5. In samples B1 − B4 only 165–170.
Farshidianfar, M.H., Khajepour, A., Gerlich, A., 2015. Real-time control of
the travelling speed is changing, with a constant laser power and microstructure in Laser Additive Manufacturing. Int. J. Adv. Manuf. Technol.
powder feed rate. The cooling rate and melt pool temperature are Felice, R., 2008. Pyrometry for liquid metals. Adv. Mater. Process. (July), 31–33.
graphically sketched as a function of the traveling speed in Fig. 12 Flemings, M.C., 1974. Solidification processing, McGraw-Hill. Available at: http://
books.google.ca/books/about/Solidification processing.
and Fig. 13, respectively.
html?id=Jpg4FBvb8hwC&pgis=1 [Accessed June 12, 2014].
The increasing speed has created an increasing trend for cooling Griffith, M., et al., 1999. Understanding thermal behavior in the LENS process.
rate and a decreasing trend for the melt pool temperature. Appar- Mater. Des. 20 (2-3), 107–113, Available at:
http://linkinghub.elsevier.com/retrieve/pii/S0261306999000163.
ently, the traveling speed has a linear relation with the cooling rate
Gu, D., Shen, Y., 2008. Processing conditions and microstructural features of porous
and melt pool temperature, if the data from the B1 sample were 316L stainless steel components by DMLS. Appl. Surf. Sci. 255 (5), 1880–1887,
to be neglected. In Fig. 13 the cooling rate is nearly doubled by Available at:
doubling the traveling speed. The traveling speed directly affects http://www.sciencedirect.com/science/article/pii/S0169433208015456
[Accessed July 13, 2014].
the cooling rate and hence indirectly affects the clad microstruc- Heyn, E., 1903. Short reports from the metallurgical laboratory of the royal
ture. Taken together, these results suggest that real time cooling mechanical and testing institute of charlottenburg. Metallographist 5, 37–64.
478 M.H. Farshidianfar et al. / Journal of Materials Processing Technology 231 (2016) 468–478

Hofmeister, W., Griffith, M., 2001. Solidification in direct metal deposition by LENS Pinkerton, A.J., Li, L., 2004. Multiple-layer cladding of stainless steel using a
processing. JOM 53 (9), 30–34, Available at: high-powered diode laser: an experimental investigation of the process
http://link.springer.com/10.1007/s11837-001-0066-z [Accessed July 16, 2014]. characteristics and material properties. Thin Solid Films 453-454, 471–476,
Hua, T., et al., 2008. Research on molten pool temperature in the process of laser Available at: http://www.sciencedirect.com/science/article/pii/
rapid forming. J. Mater. Process. Technol. 198 (1-3), 454–462, Available at: S0040609003017449 [Accessed May 25, 2015].
http://linkinghub.elsevier.com/retrieve/pii/S0924013607007157 [Accessed Ready, J.F., 2001. LIA handbook of laser materials processing, 2nd ed. Laser
July 12, 2014]. Institute of America, Florida.
Kobryn, P., Semiatin, S., 2003. Microstructure and texture evolution during Sadiq, H., et al., 2011. Effect of high temperature oxidation on the emissivity of
solidification processing of Ti–6Al–4 V. J. Mater. Process. Technol. 135 (2-3), steel. In: 1st World Congress on Advances in Structural Engineering and
330–339, Available at: Mechanics. Korea Advanced Institute of Science & Technology, Seoul, pp.
http://www.sciencedirect.com/science/article/pii/S0924013602008658 3839–3849, Available at: http://researchbank.swinburne.edu.au/vital/access/
[Accessed June 24, 2014]. manager/Repository/swin:25847 [Accessed May 23, 2014].
Kou, S., 2003. Available at. In: Welding Metallurgy Second. John Wiley and Sons, Shen, Y.F., Gu, D.D., Wu, P., 2008. Development of porous 316L stainless steel with
Hoboken http://books.google.com/ controllable microcellular features using selective laser melting. Mater. Sci.
books?hl=en&lr=&id=N8gICBzzgRwC&oi=fnd&pg=PR7&dq=WELDING. Technol. 24 (12), 1501–1505, Available at: http://www.maneyonline.com/doi/
+METALLURGY&ots=KbMCYOBY2l&sig=bUKTMZRhqD6LE6kzQIivLlTfJK4. abs/10.1179/174328408X287691 [Accessed July 13, 2014].
Li, R., et al., 2009. 316L Stainless steel with gradient porosity fabricated by selective Shifeng, W., et al., 2014. Effect of molten pool boundaries on the mechanical
laser melting. J. Mater. Eng. Perform. 19 (5), 666–671, Available at: properties of selective laser melting parts. J. Mater. Process. Technol. 214 (11),
http://link.springer.com/10.1007/s11665-009-9535-2 [Accessed July 11, 2014]. 2660–2667, http://dx.doi.org/10.1016/j.jmatprotec.2014.06.002.
M. L. Griffith, M. T. Ensz, J. D. Puskar, C. V. Robino, J. A. Brooks, J. A. Philliber, J. & E. Song, B., Dong, S., Coddet, C., 2014. ScienceDirect rapid in situ fabrication of Fe/SiC
Smugeresky, W.H.H., 2000. Understanding the microstructure and properties bulk nanocomposites by selective laser melting directly from a mixed powder
of components fabricated by laser engineered net shaping. In Proceedings of of microsized Fe and SiC. Scr. Mater. 75, 90–93, http://dx.doi.org/10.1016/j.
the Solid Freeform and Additive Fabrication Conference. San Francisco, pp. scriptamat.2013.11.031.
9–20. Toyserkani, E., Khajepour, A., Corbin, S., 2005. Laser Cladding. CRC Press, Florida.
Mitzner, S., et al., 2012. Grain Refinement of Freeform Fabricated Ti-6Al-4 V Alloy Zhang, Y., et al., 2012. Numerical and experimental investigation of multilayer
Using Beam/Arc Modulation. In: 23rd Annual International Solid Freeform SS410 thin wall built by laser direct metal deposition. J. Mater. Process.
Fabrication Symposium, 6-8 Aug 2012; Austin, TX; United States http:// Technol. 212 (1), 106–112, Available at: http://linkinghub.elsevier.com/
utwired.engr.utexas.edu/lff/symposium/proceedingsArchive/pubs/ retrieve/pii/S0924013611002378 [Accessed July 14, 2014].
Manuscripts/2012/2012-42-Mitzner.pdf. Zhao, X., et al., 2015. Fabrication and characterization of AISI 420 stainless steel
Pavlov, M., Novichenko, D., Doubenskaia, M., 2011. Optical diagnostics of using selective laser melting fabrication and characterization of AISI 420
deposition of metal matrix composites by laser cladding. Phy. Proced. 12, stainless steel using selective laser melting. Mat. Manuf. Process. 30 (11),
674–682, Available at: http://www.sciencedirect.com/science/article/pii/ 1283–1289.
S1875389211001635 [Accessed July 14, 2014].