See

discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/225742829

Modeling Aspects of High Velocity Impact of

Particles in Cold Spraying by Explicit Finite
Element Analysis

Article in Journal of Thermal Spray Technology · May 2009

DOI: 10.1007/s11666-009-9325-2

CITATIONS READS

46 182

4 authors:

Wenya li Chao ZHANG

Northwestern Polytechnical University Yangzhou University
233 PUBLICATIONS 3,938 CITATIONS 91 PUBLICATIONS 1,270 CITATIONS

SEE PROFILE SEE PROFILE

Chang-Jiu Li Hanlin Liao

Xi'an Jiaotong University Université de Technologie de Belfort-Montbé…
364 PUBLICATIONS 5,708 CITATIONS 346 PUBLICATIONS 5,262 CITATIONS

SEE PROFILE SEE PROFILE

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

Thermal spray coatings View project

Marine anti-corrosion coatings View project

All content following this page was uploaded by Wenya li on 15 November 2015.

The user has requested enhancement of the downloaded file.

JTTEE5 18:921–933
DOI: 10.1007/s11666-009-9325-2
1059-9630/$19.00 Ó ASM International

Peer Reviewed
Modeling Aspects of High Velocity Impact
of Particles in Cold Spraying by Explicit
Finite Element Analysis
Wen-Ya Li, Chao Zhang, Chang-Jiu Li, and Hanlin Liao

(Submitted January 15, 2009; in revised form March 24, 2009)

In this study, an examination of cold spray particle impacting behavior using the ABAQUS/Explicit
program was conducted for typical copper material (OFHC). Various combinations of calculation set-
tings concerning element type, Arbitrary Lagrangian Eulerian adaptive meshing, contact interaction,
material damage, etc., were examined with the main focus on the element excessive distortion and its
effect on the resultant output. The effect of meshing size on the impact behavior was also clarified
compared to the previous results obtained by using the LS-DYNA code. Some fundamental aspects on
modeling of cold spray particle deformation are discussed.

Keywords cold spraying, copper particles, deformation

methods owing to its advantages over the conventional
behavior, explicit finite element analysis, high thermal spray processes to deposit a wide variety of
velocity impact metals, alloys, and composites (Ref 2). Although it has
great application potentials in aerospace, automobile
manufacture, chemical industry, etc., there are still some
important aspects to be well revealed including the actual
bonding mechanism of spray particles.
1. Introduction As for the bonding mechanism of cold spray particles,
the most prevailing hypothesis is that plastic deformation
Cold spraying is a coating technology on the basis of may disrupt thin surface films, such as oxides, and provide
aerodynamics and high-speed impact dynamics. In this intimate conformal contact under high local pressure, thus
process, spray particles (usually 5-50 lm in diameter) are permitting bonding to occur. And thus, this bonding pro-
accelerated to a high velocity (typically 300-1200 m/s) by a cess was considered to be comparable to that in process
high-speed gas flow that is generated through a conver- such as explosive welding or shock wave powder com-
gent-divergent de Laval type nozzle. A coating is formed paction (Ref 3-5). It was also reported that the formation
through the intensive plastic deformation of particles of jetting at the local intensively deformed zones could be
impacting on a substrate at a temperature well below the helpful in the cleaning of the crushed oxide films (Ref 3-8).
melting point of the spray material. This phenomenon has The above-mentioned bonding hypothesis is consistent
been discovered during the test of supersonic wind tunnel with the fact that a wide range of ductile materials, such as
in the middle of 1980s at the Institute for Theoretical and metals and alloys, have been deposited by cold spray and
Applied Mechanics of the Siberian Division of the the spray particles experience extensive deformation to
Russian Academy of Science in Novosibirsk and firstly form lens-like shapes. Non-ductile materials, such as
reported by the Russian scientists (Ref 1). It has been ceramics, can be only deposited if they are co-cold-
widely investigated by both numerical and experimental sprayed with a ductile (matrix) material or sprayed on a
ductile substrate to form a thin layer. This hypothesis can
also explain the observed critical velocity necessary to
Wen-Ya Li, Shaanxi Key Laboratory of Friction Welding achieve a successful deposition (Ref 7-9). However, there
Technologies, School of Materials Science and Engineering, are still some fundamental problems to be revealed, such
Northwestern Polytechnical University, 127, West Youyi Road,
XiÕan, Shaanxi 710072, P.R. China; Chao Zhang, State key as the metal jetting and shear instability at the contact
Laboratory for Mechanical Behavior of Materials, School of interfaces. On the other hand, the observed critical
Materials Science and Engineering, XiÕan Jiaotong University, velocity was demonstrated to be influenced by the particle
XiÕan, Shaanxi, P.R. China and LERMPS, Université de size (Ref 5), temperature (Ref 4-6, 9), and its oxidation
Technologie de Belfort-Montbéliard, Site de Sévenans, Belfort state (Ref 6, 10-12) besides the mechanical properties of
Cedex, France; and Chang-Jiu Li, State key Laboratory for given particle and substrate materials (Ref 1, 2). There-
Mechanical Behavior of Materials, School of Materials Science
and Engineering, XiÕan Jiaotong University, XiÕan, Shaanxi,
fore, it is of significant importance to study the deforma-
P.R. China; and Hanlin Liao, LERMPS, Université de tion behavior of cold spray particles upon impact.
Technologie de Belfort-Montbéliard, Site de Sévenans, Belfort A few studies focused on the impacting and deforma-
Cedex, France. Contact e-mail: [email protected]. tion behaviors of spray particles by experiments or

Journal of Thermal Spray Technology Volume 18(5-6) Mid-December 2009—921

 simulations. Because of the very short duration of particle increase the critical velocity (Ref 5). However, these views
Peer Reviewed

impact (tens of nanoseconds), it is very difficult to observe need further experimental study to be clarified.
the whole deformation process of particles. Only the In this study, numerical simulation of cold spray par-
deformed particles can be observed by microscopy. ticle impacting behavior using the ABAQUS program was
Therefore, the numerical method takes an important role in conducted for typical copper material (OFHC). Various
studying the particle deformation process. In some studies combinations of calculation settings concerning element
(Ref 3, 7), a coupled thermal-mechanical/hydrodynamic type, ALE adaptive meshing, contact interaction, material
(CTH) code, developed at Sandia National Laboratories damage, etc., were examined with the main focus on the
(Albuquerque, NM, USA) for modeling complex multi- element excessive distortion and its effect on the resultant
dimensional, multiproblems that are characterized by output. The effects of meshing size and particle size on the
large deformation and/or strong shocks, was used. This impact behavior were also clarified compared to the pre-
is based on a two-step Eulerian solution algorithm. vious results obtained using the LS-DYNA code.
Two explicit finite element analysis (FEA) softwares,
ABAQUS (Ref 4, 5, 8) and LS-DYNA (Ref 6, 9, 10) were
also used to investigate the cold spray particle impacting
behavior. These two explicit FEA programs are widely 2. Numerical Method
used to solve the engineering problems involving in non-
linear dynamic processes. The impacting behavior of a particle on a substrate was
The reported numerical results in literature (Ref 3-10) modeled using an explicit FEA program ABAQUS/
have indicated that the flattening ratio of particles increases Explicit with the Lagrangian formulation (Ref 14). Owing
with the increase in particle impact velocity, which is to the axisymmetric characteristic of normal impact pro-
comparable to the experiment results. Moreover, the cess, a 2D symmetric model was used as shown in Fig. 1.
temperature at localized contact interfaces rises remark- In this study, the substrate was taken as a cylinder. The
ably due to the possible adiabatic shear process. Through radius and height of the substrate were taken to be five
numerical simulations by ABAQUS, Assadi et al. (Ref 4, 5) times larger than the particle diameter (e.g., 20 lm). For
found that the instability of adiabatic shear flow occurs as the convenience of meshing, the computational domain
particle velocity becomes higher than a critical velocity. was partitioned into several regions as shown in Fig. 1(a).
Therefore, they took this velocity as the actual critical The meshing was conducted by using the 4-node quad
velocity for particle deposition in cold spray. However, element with coupled displacement and temperature,
owing to the excessive distortion of meshing at the local reduced integration and hourglass control as shown in
contact zones in simulation by the Lagrangian algorithm, Fig. 1(b). The nominal meshing size for the particle was
this calculated critical velocity is much dependent on the 0.4 lm (a meshing resolution of 1/50dp) for a 20-lm par-
meshing size (Ref 6, 9). In the previous study (Ref 9), when ticle. The interface region in the substrate has the element
the Arbitrary Lagrangian Eulerian (ALE) method avail- size of 1.2 times of that of particle (0.48 lm in case of a
able in LS-DYNA was used to avoid the problems associ- 20-lm particle) in order to avoid the large deformation of
ated with the severe mesh distortion, no steep change of substrate surface elements. Different meshing sizes were
plastic strain or temperature was observed, which was used examined in this study. Based on the previous simulation
to indicate the onset of the shear instability by Assadi et al. results (Ref 4-6, 8, 9), the elements at the particle surface
(Ref 4, 5). Accordingly, it seems that the steep change of were monitored and used for the output of results, which
plastic strain obtain by the Lagrangian method is attributed experience almost the maximum plastic strain during
to the abnormal deformation of elements and cannot be impacting. In the previous studies (Ref 4, 8), the impacting
directly associated with the actual shear instability upon process was assumed to be an adiabatic process; i.e., the
particle impacting. In addition, Schmidt et al. (Ref 5) also heat transfer is not considered based on a simple estima-
found the dependence of critical velocity on particle size by tion. However, a recent study by Schmidt et al. (Ref 5)
simulations using ABAQUS with the Lagrangian algo- and our preliminary simulations indicated that the heat
rithm. Although the experiments have shown the effect of conduction should be considered in most cases. There-
particle size on critical velocity (Ref 5), their simulation fore, in this study, both the Dynamic-Explicit procedure
results obtained by the Lagrangian algorithm are believed including adiabatic heating effects and the Dynamic-
to be unreasonable. The recent studies have shown the effect Temperature-Displacement-Explicit procedure available
of oxide films on particle critical velocity (Ref 6, 10, 12). in ABAQUS were used to clarify this effect.
Considering the fact that the small powders normally have The particle/substrate interaction was implemented by
higher oxygen contents than those of large powders using the surface-to-surface contact (Explicit) formulation
(Ref 13), it is expected that the oxide films on particles available in ABAQUS. In this study, the friction between
surfaces is the main factor influencing critical velocity the particle and substrate was also considered with dif-
besides the material mechanical properties. Schmidt et al. ferent coefficients of friction (COFs). The temperature
(Ref 5) also explained the possible reasons causing the rise is based on the commonly used empirical assumption
effect of particle size. They thought that high cooling rates, that 90% of plastic work and 100% friction work is dis-
high strain rates, high viscous shear strength, and high sipated as heat.
surface contaminations such as oxide shells of small parti- For particles and substrates, the material deformation
cles possibly hinder the particle deformation and thus was described by the Johnson and Cook (JC) plasticity

922—Volume 18(5-6) Mid-December 2009 Journal of Thermal Spray Technology

 Peer Reviewed
Fig. 1 (a) Computational domain and the initial and boundary conditions; (b) meshing arrangement (meshing resolution of 1/50dp for a
20 lm particle)

model, which accounts for strain hardening, strain rate particle. Therefore, the material damage model was also
hardening, and thermal softening effects. The stresses considered. The failure elements at the contact interface
were expressed according to the Von Mises plasticity can be deleted when some failure conditions are encoun-
model. The flow stress (r) of material is expressed as tered, such as the excessive deformation and high tem-
follows (Ref 14, 15): perature. It seems that the materials are extruded from the
interface to form the jet (of course, the true jet by just
r ¼ ðA þ Benp Þð1 þ C lnð_e
ÞÞð1  ðT
Þm Þ ðEq 1Þ element deletion cannot be obtained), which may also
where A, B, n, C, m are the constants dependent on corresponds to the shear instability. In addition, the ele-
materials. ep is the effective plastic strain (PEEQ). e_
is the ment deletion using the damage model can also lead to
effective plastic strain rate normalized with respect to a important influence on the deformed particle shape and
reference strain rate ð_ep =_eo Þ: T* is a homologous temper- the output. These results will be given in the following
ature defined as the following (Ref 14, 15). section. The Johnson and Cook damage model (referred
8 to as the ‘‘Johnson-Cook dynamic failure model’’ in
<0 for T < Tr ABAQUS) was used, which accounts for the effects of
T
¼ ðT  Tr Þ=ðTm  Tr Þ for Tr  T  Tm ðEq 2Þ hydrostatic pressure, strain rate, and temperature (Ref
:
1 for T > Tm 14). ABAQUS/Explicit also offers a more general imple-
mentation of the JC failure model as part of the family of
where Tm is the melting temperature and Tr is the refer- damage initiation criteria, which is the recommended
ence or transition temperature. technique for modeling progressive damage and failure of
The most important aspect during the simulations of materials. The JC dynamic failure model is based on the
cold spray particle impacting with the Lagrangian algo- value of the equivalent plastic strain at element integra-
rithm is the possible excessive distortion of elements near tion points, where failure is assumed to occur when the
the contact surfaces, especially at a higher particle impact damage parameter exceeds 1. The damage parameter, x,
velocity. The main measures to cope with this problem in is defined as
ABAQUS include ALE adaptive mesh controls, element
distortion control and damage of material. As reported by X
D
ep 
Assadi et al. (Ref 4), although the ALE adaptive meshing x¼ ðEq 3Þ

epf
can cope with the excessive element distortion at moder-
ate and high particle impact velocities, frequent remeshing where D
ep is an increment of the equivalent plastic strain,
resulted in nonconserving energy variation of the output
epf is the strain at failure, and the summation is performed
set, and in unphysical shape of the out-flowing jet of over all increments in the analysis. The strain at failure is
material at the interface. The preliminary results also assumed to be dependent on a nondimensional plastic
showed this phenomenon. In addition, the large length strain rate, e_
; a dimensionless pressure-deviatoric stress
ratio (nominally from 0 to 1) in element distortion control ratio, p/q (where p is the pressure stress and q is the Mises
can also cause the unreasonable shape of the deformed stress); and the nondimensional temperature, T*, defined

Journal of Thermal Spray Technology Volume 18(5-6) Mid-December 2009—923

 earlier in the JC plasticity model. The dependencies are were investigated. It should be pointed out that the ele-
Peer Reviewed

assumed to be separable and are of the form: ment output includes mainly the effective plastic strain

 (PEEQ), stress and temperature, which is element-
p

epf ¼ d1 þ d2 exp d3 ½1 þ d4 lnð_e
Þð1 þ d5 T
Þ ðEq 4Þ averaged because the used elements are one-point inte-
q gration. It should also be declared that just the studied
where d1-d5 are failure parameters measured at or below setting factor is changed in the simulations to clarify its
the transition temperature, Tr. When this failure criterion effect on the output, while the other settings are in the
is met, the deviatoric stress components are set to zero and default state. Table 2 gives the default setting for the
remain zero for the rest of the analysis. Depending on the studied factors. It should also be pointed out that if ele-
selected values, the pressure stress may also be set to zero ment distortion control is not used, the program will finish
for the rest of calculation (if this is the case, it must be abnormally with the excessively distorted elements under
specified element deletion and the element will be other default settings. Accordingly, the distortion control
deleted) or it may be required to remain compressive for was used as a default setting. In addition, all the simula-
the rest of the calculation (if this is the case, it must be tions are conducted without considering the oxide films
chosen not to use element deletion). By default, the ele- and the possible adhesion between the particle and sub-
ments that meet the failure criterion are deleted. It can be strate, which will commonly cause the bounce of particle
specified the fracture energy per unit area, Gf, to be dis- after simulations. This is not the issue to be discussed in
sipated during the damage process directly, which means this study.
defining damage evolution based on energy dissipated in
ABAQUS. Instantaneous failure will occur if Gf is speci- 3.1 Effect of Solution Procedure
fied as 0. However, this choice is not recommended and As reported by Assadi et al. (Ref 4) and Grujicic et al.
should be used with care because it causes a sudden drop (Ref 8), the impacting process was assumed to be an
in the stress at the material point that can lead to dynamic adiabatic process; i.e., the heat transfer is not considered
instabilities. Therefore, the evolution in the damage was based on a simple estimation. However, the recent study
specified in exponential form in this study. Other detailed by Schmidt et al. (Ref 5) indicated that the heat conduc-
information can be found in the ABAQUS Analysis tion should be considered in most cases. There are two
UserÕs Manual (Ref 14). analysis procedures to deal with short dynamic processes
Because of the excellent cold sprayablity of copper, it in ABAQUS. One is the dynamic explicit procedure
has been widely studied experimentally. In this study, Cu (Dynamic-Explicit), which can be used to perform an
was also employed as the objective material based on its adiabatic stress analysis if inelastic dissipation is expected
easily available material data (e.g., Ref 15) used in simu- to generate heat in the material, as used by Assadi et al.
lations as shown in Table 1. The mechanical and thermal (Ref 4) and Grujicic et al. (Ref 8). Another one is a fully
material properties were assumed to be isotropic. It was coupled thermal-stress analysis procedure (Dynamic-
also assumed that the initial temperatures of particle and Temp-Disp-Explicit), which is performed when the
substrate (To) are 25 °C. The initial velocity of particle (vp) mechanical and thermal solutions affect each other
is dependent on the conditions to be study. Hereafter, it is strongly, and therefore, must be obtained simultaneously,
fixed at a moderate value of 500 m/s because the simula- as adopted by Schmidt et al. (Ref 5).
tions work well in most cases at a low impact velocity, e.g.,
300 m/s. In addition, the total solution time is set as 60 ns.

Table 2 Default settings of the studied factors

3. Results and Discussion in the simulations (Ref 14)
Setting factor Default Variation
According to our preliminary results by ABAQUS,
some important setting factors involving in the simulations Solution procedure Dynamic-Temp- Dynamic-Explicit
Disp-Explicit
Hourglass control Stiffness Enhanced, Relax
Table 1 Properties of OFHC copper used in simulations Stiffness, Viscous,
(Ref 15) combined with
stiffness-viscous
Density, kg/m3 8960 weight factor:
Thermal conductivity, W/(m °C) 386 0.5 and 0.9
Specific heat, J/(kg) (°C) 383 Element distortion Length ratio: 0.1 Length ratio:
Melting point, °C 1083 Control 0.4, 0.8
Expansion coefficient, 1/°C 0.00005 ALE adaptive NO (exclusively YES, Frequency:
Elastic modulus, GPa 124 meshing without distortion 100, 50, 10, 5, 1;
PoissonÕs ratio 0.34 control) Remeshing sweeps
JC plasticity: A, 90, 292, 0.31, 0.025, 1.09 per increment: 1, 5
MPa, B, MPa, n, C, m Interface friction NO Coefficient of friction
JC damage: d1, d2, d3, d4, d5 0.54, 4.89, 3.03, 0.014, 1.12 (COF): 0.2, 0.4
Reference temperature, °C 25 Material damage NO YES
Reference strain rate, 1/s 1 Meshing Size 1/50dp 1/20dp, 1/100dp

924—Volume 18(5-6) Mid-December 2009 Journal of Thermal Spray Technology

 Figure 2 shows the simulation results obtained by these from the interface to inner particle and substrate is found

Peer Reviewed
two solution procedures. Both the procedures finished (Fig. 2e). It is also noticed that a higher maximum PEEQ
normally in 60 ns. It is seen that although they yield the is obtained at the interface using the Dynamic-Explicit
extremely similar shapes of the deformed particle and procedure than using the Dynamic-Temp-Disp-Explicit
substrate, much higher temperature and temperature procedure. The large differences in the PEEQ and tem-
localization at the interface are clearly observed using the perature during the impacting process can also be clearly
Dynamic-Explicit procedure (Fig. 2b). With the Dynamic- observed from Fig. 2(c) and (f). Therefore, the present
Temp-Disp-Explicit procedure, obvious heat conduction results clarified that heat conduction should be considered

Fig. 2 Simulated contours of PEEQ (a, d) and temperature (b, e) and temporal evolutions of temperature, stress and strain (c, f) using
the Dynamic-Explicit (a-c) and Dynamic-Temp-Disp-Explicit (d-f) procedures

Journal of Thermal Spray Technology Volume 18(5-6) Mid-December 2009—925

 in the simulations of cold spray particles impacts. As reduced-integration elements without introducing exces-
Peer Reviewed

documented in the ABAQUS manual (Ref 14), although sive constraints on the elementÕs physical response. Sev-
the Dynamic-Explicit procedure is computationally effi- eral methods are available in ABAQUS for suppressing
cient for the analysis of large models with relatively short the hourglass modes, including the techniques based on
response times and for the analysis of extremely discon- Stiffness, Relax Stiffness, Enhanced, Viscous, and Com-
tinuous events, fully coupled thermal-stress analysis is bined Stiffness-Viscous. According to the preliminary
needed when the stress analysis is dependent on the simulations, the Stiffness, Enhanced, and Combined
temperature distribution and the temperature distribution methods are important for modeling cold spray particles
depends on the stress solution. For such cases, the thermal impacts. With the Stiffness technique, the program can
and mechanical solutions must be obtained simultaneously finish normally as shown in Fig. 2. But for the other
rather than sequentially. techniques, the program aborted due to the excessive
distortion of elements causing the exaggerated tempera-
ture fields as shown in Fig. 3 (just three cases). The further
3.2 Effect of Hourglass Control
study showed that those hourglass control techniques can
It is known that hourglass control is a very work when combined with other settings, such as ALE
useful method to minimize the problems involved in the adaptive meshing and material damage.

Fig. 3 Simulated contours of PEEQ (a-c) and temperature (d-f) with the techniques of Enhanced (a, d, fails at around 13 ns), Combined
of a stiffness-viscous weight ratio of 0.5 (b, e, fails at around 11 ns) and combined of a stiffness-viscous weight ratio of 0.9 (c, f, fails at
around 10 ns)

926—Volume 18(5-6) Mid-December 2009 Journal of Thermal Spray Technology

3.3 Effect of Element Distortion Control are enforced by using a penalty approach, and one can con-

Peer Reviewed
trol the associated distortion length ratio. Distortion control
Without element distortion control, analyses may fail
cannot be used when the elements are included in an adap-
prematurely when the mesh is coarse relative to strain gra-
tive mesh domain (mutually exclusive in ABAQUS) (Ref 14).
dients and the amount of compression. ABAQUS/Explicit
In this study, the analysis failed prematurely at around 17 ns
offers distortion control to prevent solid elements from
without element distortion control. Hence a distortion
inverting or distorting excessively for these cases. Constraints
length ratio of 0.1 was used as the default setting. Figure 4

Fig. 4 Simulated contours of PEEQ (a, d) and temperature (b, e) and temporal evolutions of temperature, stress and strain (c) with the
distortion length ratios of 0.4 (a-c) and 0.8 (d, e, fail at about 40 ns)

Journal of Thermal Spray Technology Volume 18(5-6) Mid-December 2009—927

 gives the results with higher distortion length ratios. It is models in ABAQUS. There are several friction models in
Peer Reviewed

found that the maximum PEEQ and temperature values ABAQUS, such as the basic Coulomb friction model,
obtained with the length ratio of 0.4 are lowered com- shear stress versus elastic slip while sticking, exponential
pared to those obtained with the length ratio of 0.1 decay friction model (Ref 14). The simple Coulomb fric-
(Fig. 2d, e and f). A little change in jet shapes is also tion model was used in this study. Figure 6 shows the
observed. However, for the length ratio of 0.8 (Fig. 4d simulation results obtained with the constant COFs of 0.2
and e), the program failed at about 40 ns due to the and 0.4. It is found that the maximum PEEQ and tem-
much constraint to element deformation. As documented perature at the interface are lowered when friction is
in the ABAQUS manual (Ref 14), care must be used in considered compared to the frictionless state (Fig. 2). The
interpreting results since the distortion control con- jet formation is also limited due to the friction shear stress.
straints may inhibit legitimate deformation modes and However, the value of COF has a little effect on the
lock up the mesh. In addition, distortion control cannot resultant output in the range from 0.2 to 0.4 as shown in
prevent elements from being distorted due to temporal Fig. 6. In addition, it is clearly observed from that friction
instabilities, hourglass instabilities, or physically unreal- takes a small role in increasing the internal energy during
istic deformation. Therefore, the distortion length ratio the particle impacting process compared to the plastic
is, if required, usually of a small value; e.g., 0.1. dissipation. Therefore, a small COF is appropriate in
modeling cold spray particles impacts. It should be point
out that with the basic Coulomb friction model, which
3.4 Effect of ALE Adaptive Meshing does not specify a limiting shear stress, the interface shear
The ALE adaptive meshing technique in ABAQUS stress may be overestimated in the case with extremely
combines the features of pure Lagrangian analysis and high contact pressure.
pure Eulerian analysis. It is a tool that makes it possible to As mentioned above, one can make the program with
maintain a high-quality mesh throughout an analysis, even the meshing size of 1/100dp be conducted without abortion
when large deformation or loss of material occurs, by by the combination of other solution techniques, such as
allowing the mesh to move independently of the material ALE adaptive meshing and material damage (Fig. 7c).
(Ref 14). However, in some extreme cases, it does not Through the preliminary study, the program can finish
work well. Based on the preliminary results, just frequency normally using the ALE adaptive meshing with the fre-
and remeshing sweeps per increment were considered quency-remeshing sweeps per increment of 1-5 and the
with other settings as default in this study. Figure 5 shows enhanced hourglass control as shown in Fig. 8.
the simulation results obtained with different combina-
tions of frequency and remeshing sweeps per increment, 3.6 Effect of Material Damage
with which the program finished normally. It is seen that
with decreasing the frequency, i.e., increasing the reme- Although the abovementioned calculation techniques
shing times in given increments, the maximum PEEQ and can be performed successfully to yield comparable output,
temperature are decreased and the jet shape appears more sometimes they do not work well; e.g., when the meshing
smooth, in other words, more unrealistic deformation of size is relatively fine which can result in more precise data.
the periphery (Fig. 5a, b, d, and e) compared to the result Therefore, the material failure model was examined for
without ALE adaptive meshing (Fig. 2). When increasing the first time in this study according to the suggestion
the remeshing sweeps per increment (Fig. 5c and f); i.e., documented in the ABAQUS manual. Here the slave
more remeshing is performed per increment, the maxi- surface in the interaction definition must be node region to
mum PEEQ and temperature and contours seem to be be possibly eroded; while the Status variable must be
changed a little as compared to Fig. 5(b) and (e). How- selected in the field output requests. Figure 7 shows the
ever, these two factors are interactional, because in other typical simulation results obtained with the JC failure
combinations of frequency and remeshing sweeps per model. It is found that the elements of large deformation
increment the program aborted due to excessive distortion in the particle at the interface have been deleted from the
of some element, such as 100-1 (frequency-remeshing calculation. Compared to the deformed shape without
sweeps per increment), 5-1, 100-5, and 10-1 (fail at about material damage (Fig. 2), the obtained particle shape with
12 ns). In these cases, ALE adaptive meshing may be material damage seems more reasonable. It is expected to
combined with other techniques, such as material damage, obtain a better shape of the deformed particle with a finer
to make it work. Nevertheless, ALE adaptive meshing is meshing size. However, when the meshing size is 1/100dp,
not strongly recommended to use in modeling cold spray the result (Fig. 7c) is not satisfactory owing to the unre-
particles impacts owing to the unrealistic jet as indicated alistic deformation of substrate elements at the interface,
by Assadi et al. (Ref 4). which cannot be set with material damage as the master
surface in a 2D model (Ref 14). As a result, the 3D
modeling may be very interesting in modeling the failure
3.5 Effect of Interface Friction
of both particle and substrate, where the erosion contact is
In the previous study on particle/substrate interaction permitted. The unrealistic deformation of substrate ele-
(Ref 4-10), no friction between the particle and substrate ments at the interface can be avoided with the 3D model,
was considered. Therefore, friction was included for the which has been examined and will present in another
first time in this study by using the available friction contribution.

928—Volume 18(5-6) Mid-December 2009 Journal of Thermal Spray Technology

3.7 Effect of Meshing Size relative meshing size. Figure 9 shows the simulation

Peer Reviewed
results obtained for the meshing sizes of 1/20dp and
As has been recognized by Assadi et al. (Ref 4) using
1/100dp. But the program failed prematurely at about 4 ns
ABAQUS/Explicit and in our previous study (Ref 6, 9)
due to the excessive element distortion (Fig. 9d and e). In
using the LS-DYNA code which is also a popular explicit
comparison with the result obtained for the meshing size
finite element method, the meshing density takes an
of 1/50dp (Fig. 2), the maximum PEEQ and temperature
important role in the computation and influences signifi-
decrease for the meshing size of 1/20dp (Fig. 9a, b and c).
cantly the resultant output. However, Assadi et al. (Ref 4)
It is understandable taking into account the fact that the
did not explore this effect and present the effective mea-
output is element-averaged and the used elements are
sures to cope with it. Therefore, this research will give
one-point integration. From this point, it is expected that
some validation and explanation on the effect of the

Fig. 5 Simulated contours of PEEQ (a-c) and temperature (d-f) with different ALE adaptive mesh combinations of frequency and
remeshing sweeps per increment: (a, d) frequency-remeshing sweeps per increment 50-1, (b, e) 1-1, and (c, f) 1-5

Journal of Thermal Spray Technology Volume 18(5-6) Mid-December 2009—929

Peer Reviewed

Fig. 6 Simulated contours of PEEQ (a, c) and temperature (b, d) and temporal evolutions of energies (e) with the constant contact
surface COFs of 0.2 (a, b) and 0.4 (c-e). Note that ALLFD: friction dissipation, ALLPD: plastic dissipation, ALLIE: internal energy,
ALLKE: kinetic energy

the finer meshing, the more approaching to the realistic velocity is taken as the so-called critical velocity by
data. Assadi et al. (Ref 4), it is seen that the critical velocity
Based on the finding by Assadi et al. (Ref 4), the decreases with decreasing the meshing size for the same
onset velocities of excessive element distortion under material. It is also found that the instability velocities for
different meshing sizes are given in Fig. 10. For com- different materials tend to be 300-400 m/s, even for the
parison, the simulation results on Cu obtained by high strength In718 alloy, which is much lower than the
LS-DYNA are also plotted in Fig. 10. If this onset reported critical velocities (Ref 5, 8). According to

930—Volume 18(5-6) Mid-December 2009 Journal of Thermal Spray Technology

 Peer Reviewed
Fig. 7 Simulated contours of PEEQ (a, c) and temperature (b) with the material damage model for the meshing sizes of 1/50dp (a, b)
and 1/100dp (c)

Fig. 8 Simulated contours of PEEQ (a) and temperature (b) for the meshing size of 1/100dp using the ALE adaptive meshing with the
frequency-remeshing sweeps per increment of 1-5 and the enhanced hourglass control

the previous study (Ref 6), a critical velocity of about up-to-date findings by different research groups. Kang
300 m/s for Cu of low oxygen content was found. Schmidt et al. (Ref 12) later reported the similar oxidation
et al. (Ref 5) also reported an adhesion velocity of about dependency of aluminum powders and proved our model.
250 m/s for 20 mm Cu spherical projectile. Therefore, However, much work is necessarily required to reveal the
the dominant effect of particle oxidation state was pro- realistic adiabatic shear instability by numerical and
posed in the previous study (Ref 6) explaining well the experimental methods.

Journal of Thermal Spray Technology Volume 18(5-6) Mid-December 2009—931

Peer Reviewed

Fig. 9 Simulated contours of PEEQ (a, d) and temperature (b, e) and temporal evolutions of temperature, stress and strain (c) with the
meshing sizes of 1/20dp (a-c) and 1/100dp (d, e, fail at about 4 ns)

4. Concluding Remarks uated. The results showed that the satisfactory output
could be obtained with the appropriate settings, but
Through the systematic investigation of the effects of careful regulation of the setting factors is necessary. No
the simulation settings on the output in modeling the mater which solution technique is used, the main objective
impacting behavior of cold spray particles, the feasibility is to model the physical processes occurred in cold spray
and capability of numerical simulation of particle particles impacts, such as metal jetting, adiabatic shear
impacting by the ABAQUS/Explicit program were eval- instability, oxide film effect, and final adhesion between

932—Volume 18(5-6) Mid-December 2009 Journal of Thermal Spray Technology

 1200 2. T. Stoltenhoff, H. Kreye, and H.J. Richter, An Analysis of the

Peer Reviewed
Cold Spray Process and its Coatings, J. Therm. Spray Technol.,
2002, 11, p 542-550
1000 3. R.C. Dykhuizen, M.F. Smith, D.L. Gilmore, R.A. Neiser,
Instability velocity (m/s)

X. Jiang, and S. Sampath, Impact of High Velocity Cold Spray

Particles, J. Therm. Spray Technol., 1999, 8, p 559-564
800 4. H. Assadi, F. Gärtner, T. Stoltenhoff, and H. Kreye, Bonding
Mechanism in Cold Gas Spraying, Acta Mater., 2003, 51(15),
600 p 4379-4394
5. T. Schmidt, F. Gärtner, H. Assadi, and H. Kreye, Development
of a Generalized Parameter Window for Cold Spray Deposition,
400 Acta Mater., 2006, 54, p 729-742
Cu 6. C.J. Li, W.Y. Li, and H.L. Liao, Examination of the Critical
Cu, LS-DYNA Velocity for Deposition of Particles in Cold Spraying, J. Therm.
200 Spray. Technol., 2006, 15(2), p 212-222
316L
7. M. Grujicic, J.R. Saylor, D.E. Beasley, W.S. DeRosset, and
In718 D. Helfritch, Computational Analysis of the Interfacial Bonding
0
0.0 0.2 0.4 0.6 0.8 1.0 between Feed-Powder Particles and the Substrate in the Cold-
Gas Dynamic-Spray Process, Appl. Surf. Sci., 2003, 219, p 211-227
Meshing size (µm) 8. M. Grujicic, C.L. Zhao, W.S. DeRosset, and D. Helfritch, Adi-
abatic Shear Instability Based Mechanism for Particles/Substrate
Fig. 10 Effect of meshing size on the onset velocity of element Bonding in the Cold-Gas Dynamic-Spray Process, Mater. Design,
instability for different materials 2004, 25, p 681-688
9. W.Y. Li, H.L. Liao, C.J. Li, G. Li, C. Coddet, and X.F. Wang, On
the particle and substrate. The future work on the effect of High Velocity Impact of Micro-sized Metallic Particles in Cold
particle size and the 3D modeling may be necessary to Spraying, Appl. Surf. Sci., 2006, 253(5), p 2852-2862
investigate more complex impacting conditions, such as 10. W.Y. Li, H.L. Liao, C.J. Li, H.-S. Bang, and C. Coddet,
Numerical Simulation of Deformation Behavior of Al Particles
the erosion of the particle and substrate, off-normal Impacting on Al Substrate and Effect of Surface Oxide Films on
impact, and multi-particle impact. Interfacial Bonding in Cold Spraying, Appl. Surf. Sci., 2007,
253(11), p 5084-5091
11. W.Y. Li, C. Zhang, H.T. Wang, X.P. Guo, H.L. Liao, C.-J. Li, and
C. Coddet, Significant Influences of Metal Reactivity and Oxide
Acknowledgments Film of Powder Particles on Coating Deposition Characteristics
This work was partially supported by the NPU Talent in Cold Spraying, Appl. Surf. Sci., 2007, 253(7), p 3557-3562
12. K. Kang, S. Yoon, Y. Ji, and C. Lee, Oxidation Dependency of
Development Foundation, the Scientific and Technologi- Critical Velocity for Aluminum Feedstock Deposition in Kinetic
cal Innovation Foundation for Youth NPU Teachers, and Spraying Process, Mater. Sci. Eng. A, 2008, 486, p 300-307
the Ao-Xiang Star Project. 13. C.J. Li and W.Y. Li, Effect of Sprayed Powder Particle Size on
the Oxidation Behaviour of MCrAlY Materials during HVOF
Deposition, Surf. Coat. Technol., 2003, 162(1), p 31-41
14. Abaqus Analysis UserÕs Manual, ABAQUS 6.7 HTML Docu-
References mentation, Dassault Systèmes, 2007
15. G.R. Johnson and W.H. Cook, Fracture Characteristics of Three
1. A. Papyrin, Cold Spray Technology, Adv. Mater. Proc., 2001, Metals Subjected to Various Strains, Strain Rates, Temperatures
159(9), p 49-51 and Pressures, Eng. Fract. Mech., 1985, 21(1), p 31-48

Journal of Thermal Spray Technology Volume 18(5-6) Mid-December 2009—933

View publication stats