Splats Formation, Interaction and Residual

Stress Evolution in Thermal Spray Coating

Using a Hybrid Computational Model

Abba A. Abubakar, Abul Fazal M. Arif,

Syed Sohail Akhtar & Javad Mostaghimi

Journal of Thermal Spray Technology

ISSN 1059-9630

J Therm Spray Tech

DOI 10.1007/s11666-019-00828-6

1 23
Your article is protected by copyright and
all rights are held exclusively by ASM
International. This e-offprint is for personal
use only and shall not be self-archived in
electronic repositories. If you wish to self-
archive your article, please use the accepted
manuscript version for posting on your own
website. You may further deposit the accepted
manuscript version in any repository,
provided it is only made publicly available 12
months after official publication or later and
provided acknowledgement is given to the
original source of publication and a link is
inserted to the published article on Springer's
website. The link must be accompanied by
the following text: "The final publication is
available at link.springer.com”.

1 23
 Author's personal copy
J Therm Spray Tech
https://doi.org/10.1007/s11666-019-00828-6

PEER REVIEWED

Splats Formation, Interaction and Residual Stress Evolution

in Thermal Spray Coating Using a Hybrid Computational Model
Abba A. Abubakar1 • Abul Fazal M. Arif1 • Syed Sohail Akhtar1 • Javad Mostaghimi2

Submitted: 29 June 2018 / in revised form: 21 December 2018

 ASM International 2019

Abstract Due to the multilayered pattern of coating Keywords finite element method  numerical model 
deposition, residual stresses are commonly developed in point cloud  residual stress  smooth particle
thermal spray coatings (TSCs). The large deformation, hydrodynamics  thermal barrier coating  thermal spray
complex interaction and material mismatch are the main coatings
contributing factors to residual stress formation. The con-
stitutive behavior and lifetime are directly dependent on the
nature and extent of the residual stress field. In the present Introduction
study, a computational approach for effective prediction of
residual stress evolution in TSCs has been proposed. The Thermal spray coating (TSC) is a type of advanced coat-
proposed approach is hybrid in the sense that it combines ings that are thermally deposited on a substrate surface by
‘‘point cloud’’ (PC) and finite elements (FE) to model the heating the coating material to a molten state (Ref 1). They
residual stresses. Sprayed droplets deposition and associ- are used extensively for various applications that require
ated deformations are captured on PC using smooth parti- improved substrate surface properties such as thermal
cle hydrodynamics, a popular meshless approach for resistance. Due to the presence of numerous defects (e.g.,
modeling of violent fluid flows. The conversion of pores, cracks, splat interfaces, second-phase particles, etc.)
deformed droplets from PC to FE domains is done using and complex interaction between process parameters,
several recent algorithms for point cloud processing. Then, nonlinear residual stress profile is developed along the
conventional FE schemes are used to model the heat width and thickness of the coatings (Ref 2). The process
transfer and structural deformation occurring during the parameters which greatly influence residual stress field are
process. The proposed approach has been found to be droplets size, droplets temperature, droplets impact veloc-
effective in predicting residual stress evolution in thermal ity, initial substrate temperature, the surface roughness (or
barrier coatings (TBCs). It can capture the effects of undulation) of the substrate, spray angle/path, spray rate,
microstructural defects (such as pores and cracks) and gun speed and standoff distance. The adhesion strength,
interaction of process parameters on residual stress cohesive strength, thermal shock resistance, thermal fati-
distribution. gue life, corrosion resistance, wear properties and service
life of coatings are affected by the residual stress field (Ref
3, 4).
Residual stress evolution in TSCs is highly complicated
as it is strongly dependent on material types, process
& Abul Fazal M. Arif parameters and process history; thus, stresses are devel-
afmarif@kfupm.edu.sa; afmarif1@gmail.com
oped at both deposition and post-deposition regimes.
1
Mechanical Engineering Department, King Fahd University Deposition stresses develop because of sudden solidifica-
of Petroleum and Minerals, Dhahran, Saudi Arabia tion of splats, peening action of droplets on pre-deposited
2
Mechanical Engineering Department, University of Toronto, splats, high thermal gradient or geometric discontinuities
Toronto, Canada developed during deposition. Post-deposition stresses

123
 Author's personal copy
J Therm Spray Tech

develop due to a mismatch in properties during cooling the The numerical modeling of thermal spray process has
coating to room temperature. Various factors such as sub- been very challenging due to various difficulties encoun-
strate geometry, substrate surface roughness, substrate tered during computation such as excessive mesh distor-
surface pretreatment and coating post-treatment have a tion, computational intensity and multiple droplets
significant influence on the residual stress state. The final interaction. Due to convergence difficulties associated with
residual stress distribution in the coating and substrate is a severe mesh distortion, Lagrangian grid-based schemes
superposition of stresses developed at various length and have been rarely used to model multiple droplets deposi-
time scales. tion occurring during the thermal spray process. Similarly,
Various analytical models for estimation of residual Eulerian schemes (such as phase field, level-set and the
stresses in coatings have been developed. But these models volume of fluid) are not suitable due to the coalescence of
are usually inaccurate due to the adoption of several nec- domain interfaces and high computational costs. Therefore,
essary assumptions during derivation (Ref 5). The analyt- the thermal spray process deposition should be modeled
ical model by Tsui and Clyne (Ref 5) combined the using an alternative scheme that is not only robust but also
quenching and post-deposition mismatch stresses. How- computationally inexpensive. Recently, meshless Lagran-
ever, the model cannot capture nonlinearity of residual gian methods (such as smooth particle hydrodynamics—
stress profile developed because of nonlinear deformation SPH and discrete gradient method—DGM) are commonly
that occurs during the spray process. Consequently, several used to model large deformation problems on point cloud
numerical models are commonly used to predict residual (PC) with less numerical difficulties and lower computa-
stresses and many researchers used the element birth-kill tional costs. Due to its level of advancement, SPH is more
(FEM) approach (Ref 6). This approach was suitable as it popular, robust and computationally inexpensive than
integrates both deposition and post-deposition stresses DGM, especially when applied to fluid problems. Conse-
during the computation. However, the stress results pre- quently, Zhang (Ref 14, 15) and Farrokhpanah et al. (Ref
dicted with the approach were found to be only qualita- 16) used SPH to model droplet impact and solidification
tively comparable to previous experimental measurements. during the thermal spray process. However, they could not
Wang et al. (Ref 7) integrated the birth–death approach couple the deposition process with residual stress evolution
with a micromechanics model in order to capture local due to tensile instability arising during the computations.
stress fluctuation near defects. Wu et al. (Ref 8) reported a Despite previous attempt (Ref 17), complete elimination of
modified element birth–death model where a group of soft the problem of tensile instability in SPH domains (sub-
hypothetical elements with suitable properties was added at jected to high tensile stresses) is currently not achievable.
appropriate locations in order to reduce mesh distortion and Thus, SPH cannot be used to effectively model the evo-
improve the accuracy. But, even this approach failed to lution of residual stresses in TSCs. In view of this, SPH
capture the local stress fluctuation. Elhoriny et al. (Ref 9) should only be used to model the dynamics of splats for-
and Berthelsen et al. (Ref 10) recently proposed that the mation during coating deposition.
coating elements should be activated block-by-block (in- In the present study, a hybrid computational approach is
stead of layer-by-layer) in order to capture the local stress used to model splat formation, the interaction between
fluctuation due to the complex interaction between process splats and evolution of residual stresses in thermal spray
parameters. But, even this model failed to capture local coating made of yttria-stabilized zirconia (YSZ) applied to
stress fluctuation occurring near the defect. For this reason, steel alloy. The hybrid approach combines process mod-
image-based finite element schemes were used to model eling on point cloud (PC) with a thermo-mechanical anal-
residual stresses based on real microstructure images cap- ysis of finite elements (FEs). Smooth particle
tured through optical micrographs, SEM or l-CT scan (Ref hydrodynamics (SPH) is used to model multiple droplets
11, 12, 13). Although this approach captured defects and impact, spread and interactions on the PC, while finite
the variation in residual stress field, the magnitudes of the element method (FEM) is used to model the heat transfer
residual stresses are expected to be inaccurate as deposition and structural deformation occurring during the spray
stresses developed during splats solidification were not process. This is the first contribution where coupling
considered during the analysis. Consequently, unrealisti- between SPH and FEM is used to evaluate residual stress
cally high-stress values were obtained. Therefore, there is formation in coatings. Unlike in previous models, latent
need to integrate residual stress computation with thermal heat release during splat solidification is considered during
spray process modeling to numerically capture important the heat transfer analysis in addition to conductive and
phenomena such as stress relaxation mechanisms, the convective heat transfer. Furthermore, both thermal-depo-
presence of microstructural heterogeneities, deposition sition (or quenching) and thermal-mismatch (post-deposi-
stresses and interaction of process parameters during the tion) stresses are combined to compute the final residual
computation. field. The predicted residual stress profile is validated by

123
 Author's personal copy
J Therm Spray Tech

comparison with experimental results obtained from the heat of fusion, and u is a Dirac-delta function with nonzero
hole drilling method and previous numerical results avail- values outside the mushy zone.
able in the literature. Based on Hooke’s law, the final form of local-force
balance equation is given by:
o 2 ui
Mathematical Model q ¼ rij;j þ qbi ðEq 7Þ
ot2
The governing equations required are the flow (Navier– where ui is displacement field, rij is stress tensor, q is
2
Stokes’) equation for multiple droplets deposition (Eq 1 density, bi is body force per unit mass, and ootu2i is direc-
and 2), energy equation for heat transfer (Eq 4) and local- tional acceleration.
force (stress-equilibrium) equation (Eq 7) for thermal The stress tensor is related to the strain tensor (e)
stresses. Due to differences in length scales, the equations through the Hooke’s law as represented in Eq 8:
were developed such that only sequential coupling terms  
pl
are considered. Thus, the process model involves only rij ¼ Cijkl : eij  eth
ij  eij ðEq 8Þ
solving the flow equation, while for the thermo-mechanical pl
model, the stress terms are not considered in the energy eij ¼ eel th
ij þ eij þ eij ðEq 9Þ
equation. where eel th
ij is elastic strain tensor, eij is thermal strain tensor,
The Navier–Stokes’ equations for laminar, viscous and pl
and eij is plastic strain tensor.
semi-incompressible fluid flow are given as:
Due to the large temperature difference between the
oq coating and substrate materials, large deformation usually
þ ðqvi Þ;i ¼ 0 ðEq 1Þ
ot occurs coating deposition. Thus, based on large-strain
ovi formulation, the total strain is related to the displacement
q þ qvj vi;j ¼ qbi þ ðpdij þ sij Þ;j ðEq 2Þ
ot fields by:
where q is the density of droplet, vi is the fluid velocity 1 
eij ¼ ui;j þ uj;i þ uk;i  uk;j ðEq 10Þ
field, bi is body force per unit volume, p is pressure, sij is 2
viscous shear stress, and t is time. The thermal strain is developed due to the mismatch in the
Due to semi-incompressibility, a small change in pres- coefficient of thermal expansion between the coating and
sure density is defined according to the linear form of the substrate materials. It is usually expressed as:
popular Mie–Gruneisen equations of state as:
  ZT
q0 c0 g C0 g
p¼ 1  þ C0 q0 Em ðEq 3Þ eth
ij ¼ aðTÞ  dT ðEq 11Þ
ð1  sgÞ2 2
Tref
where q0 is the reference (or initial) density, Em is specific
Using the von Mises criterion, the flow rule for the plastic
internal energy, and C0 , c0 and g are all material constants
deformation can be represented by:
obtained by fitting to experimental data.
Based on Fourier’s law, the final form of the energy oQ 3 Sij
depij ¼ dk ¼ dk ðEq 12Þ
equation written in enthalpy form is given by: orij 2 req

q
oh
¼ kT;ii ðEq 4Þ where depij is an increment of the plastic strain tensor, Q is
ot yield function, Sij is deviatoric stress tensor, req is equiv-
where h is enthalpy, k is thermal conductivity, and T is alent stress and dk is a plastic multiplier which is a scalar
temperature. function of plastic strain.
The change in enthalpy is related to temperature and The equivalent stress based on von Mises yield criteria
specific heat capacity as: is usually expressed as a function of deviatoric stress tensor
as:
dhðTÞ ¼ Cp ðTÞdT ðEq 5Þ rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
3
The specific heat capacity (Cp ) is given by: req ¼ Sij : Sij ðEq 13Þ
2
Lf
Cp ðTÞ ¼ Cps ðTÞ þ  u þ Cpl ðEq 6Þ Due to its brittleness, the ceramic coating layer does not
Tm
undergo strain hardening or plastic deformation. Thus, it is
where Cps is specific heat capacity of the solid material, Cpl only required to restrain tensile stresses developed in the
is specific heat capacity of the liquid material, Lf is latent presumably microcracked ceramic layer to low values

123
 Author's personal copy
J Therm Spray Tech

using perfectly plastic material model [as used in recent

where vifluid is velocity of fluid and visolid is velocity of
works (Ref 8)]. For the substrate material (i.e., a metal), the
solid.
yield surface is assumed to evolve based on the Johnson–
For the thermo-mechanical model, convective boundary
Cook model as shown in Eq 14.
conditions are applied on all boundaries of the substrate
Q ¼ req  ryp  0 ðEq 14Þ and exposed surfaces of pre-deposited splats such that
temperature is fully dissipated in all directions. Symmet-
with initial yield surface, as:
rical boundary conditions are applied at boundaries OAFG
" !#
h  pl n i epl 
d  and OCDG. As done in the works of Klusemann et al. (Ref
ryp ¼ A þ B e 1 þ C ln pl
1  h^m 11), Nayebpashaee et al. (Ref 19) and Gupta (Ref 20), the
e0
d
boundaries ABEF and BCDE are constrained to deform
ðEq 15Þ uniformly in the y- and x-direction, respectively. To avoid
where epl is equivalent plastic strain variable, h^ is non- any rigid body motion by substrate geometry, one node at
the origin (i.e., vertex O) is constrained from moving in all
dimensional temperature, A, B, m, n and epl 0 are all material
directions. For simplicity, perfect bonding is assumed
constants to be obtained by fitting with experimental data.
between splat–substrate and splat–splat interfaces as used
Initial Conditions The initial parameters for the process
in previous research works (Ref 8, 9).
model are the diameter, impact velocity and temperature of
sprayed droplets. The process model is developed such that qi ¼ kTi;i ¼ hcv ðT  Ta Þ at all exposed surfaces
deposition is carried out in a layer-by-layer manner having ðEq 17Þ
a fixed number of droplets (i.e., 9 droplets in this case)
where hcv is convective heat transfer coefficient.
released at each step. For more realistic analysis, the pro-
cess parameters are defined to have random values which ni ui ¼ 0 at faces OAFG and OCDG ðEq 18Þ
fall within the range of statistical data given in the previous
where ni is unit normal vector
experiment by Xue et al. (Ref 18) (as demonstrated in
Fig. 1). Thus, the droplets have random initial diameters, ui ¼ 0 at one point O ðEq 19Þ
impact velocities and temperatures ranging from 29 to
Material Models Due to large temperature changes asso-
32 lm, 100-110 m/s and 2535-2550 C, respectively. The
ciated with the process, the stainless-steel substrate is
substrate dimensions are selected based on the volume
modeled using Johnson–Cook material model which
required for proper dissipation of the heat during the pro-
accounts for the influence of temperature and loading rate
cess. Any form of heat transfer during droplet flight is
on its yield strength. The initial substrate temperature
neglected. According to the usual practice, the substrate
(700 C) is taken as the reference temperature for stress
material is pre-heated to reduce deposition stresses and
computation. As used in recent works (Ref 8), the perfectly
improve coating adhesive strength; thus, the substrate layer
plastic model has been used to restrict tensile deposition
is at an initial temperature of 700 C.
stresses to lower values to consider stress relaxation
Boundary Conditions For the process model, the wall
(through microcracking) within the ceramic coatings when
boundary condition is applied at the interface between the
under high tensile stresses. Consequently, the ceramic
droplets and substrate or pre-deposited layer surfaces as:
coating has lower yield strength in tension as compared to
vifluid ¼ visolid at fluid - structure interface ðEq 16Þ compression. Thermal, fluid and structural properties used
in the current model are given in Table 1.

Numerical Model

The computational approach used is hybrid in the sense

that it combines two different numerical methods, SPH and
FEM, to predict residual stress evolution in thermal spray
coatings (TSCs) as shown in Fig. 2.

SPH Formulation

According to SPH formulation, the flow or deformation of

Fig. 1 Release/impact of sprayed droplets on substrate in layer-wise continuum bodies is represented using sets of particles
fashion (vertex O shares the same edge with G)

123
 Author's personal copy
J Therm Spray Tech

Table 1 Material properties used for the simulation

Property YSZ (Ref 14, 21-23) 301 stainless steel (Ref 14, 21, 24)

Fluid properties
Density, q (kg/m2) 5890 7900
Viscosity, l (kg/ms) 0.008 …
Speed of sound, c0 (m/s) 3000 …
Dimensionless parameter, s 2.39 …
Dimensionless parameter, C0 0 …
Phase change properties
Latent heat, L (J/kg) 7.07 9 105 …
Specific heat capacity, cp (J/kg K) 713 @ 2535 C 477
580 @ 2522 C
Thermal conductivity k (W/m K) 2.32 @ 2535 C 14.9
2.0 @ 2522 C
Coefficient of thermal expansion (CTE), a 9.68 9 10-6 @ 20 C 16.6 9 10-6@ 100 C
(/C) 10.34 9 10 -6
@ 17.6 9 10-6@ 300 C
1000 C 18.6 9 10-6@ 500 C
11/22 9 10-6 @ 19.5 9 10-6@ 700 C
2527 C
Heat transfer coefficient of air, hcv (W/ 20 20
m2 K)
Structural properties
Elastic modulus/Poisson’s ratio, E (GPa)/v 63.40/0.21 @ 20 C 200/0.3 @ 700 C
28.33/0.21 @ 2000 C 0.2/0.4 @ 1400 C
26.67/0.21 @ 2522 C
3 9 10-3/0.4 @
2532 C
3 9 10-3/0.4 @
3000 C
Yield strength, ry (MPa) 30 (tensile) Based on Johnson–cook the following parameters: A = 310,
300 (compressive) B = 1000, C = 0.07, n = 0.65, e_ = 1/s and m = 1

Fig. 2 Hybrid computational

approach: (a) droplets impact,
spread and interaction using
SPH, (b) splats solidification
and residual stress evolution
using FEM

which have unique material properties (such as density, SPH formulation for the Lagrangian form of conserva-
viscosity, pressure). The solution across the domain is tion of mass and momentum equation can be expressed as
obtained using the kernel (or smoothing) functions. (Ref 25):

123
 Author's personal copy
J Therm Spray Tech

Dqi X n that fits the original surface with minimum error (or
¼ mj vij  ri Wij ðEq 20Þ deviation). The algorithm is generally faster and more
Dt j¼1
robust than other implicit surface reconstruction
! !
D
vi X n
pi pj Xn
si sj methods (such as RBF, least squares, distance function,
¼D mj 2 þ 2 ri Wij þ mj 2 þ 2 the multi-level partition of unity implicit). The only
Dt j¼1
qi qj j¼1
qi qj
drawback is that it needs consistent normal orientation

 ri Wij þ b for effective generation of stereolithographic (STL)
ðEq 21Þ mesh.
4. Quadratic Edge Decimation (QED) QED algorithm
where vij is the velocity of particle-i relative to that of
[developed by (Ref 33)] reduces the number of
particle-j, Wij is kernel (smoothing) function between
simplicial complexes in a given polytype iteratively
particle-i and particle-j, q is density, p is pressure, s is
with little compromise on shape, size and computa-
viscous stress tensor, and b is body force. The cubic spline
tional costs. The algorithm is used to reduce the
kernel function Wij , developed by Monaghan (Ref 26), is
density of the STL mesh to the required number of
adopted for analysis in the current work.
elements. Thus, two vertices along an edge of element
Using SPH particle approximation, the viscous stress
degenerate into a single vertex which reconnects with
tensor is usually represented as:
other simplified vertices to form a new edge. The new
Xn
mj X n
mj   position of the vertices and length of new edges for the
si ¼  li vij ri Wij  li ri Wij vij
j¼1 j
q q
j¼1 j
polytype are determined by optimizing the quadratic
! error metric iteratively.
2 Xn
mj
þ l vij  ri Wij I ðEq 22Þ 5. Least Square Subdivision Surface (LS3) LS3 algorithm
3 j¼1 qj i is used re-smoothen the surface of an object in such a
way that a coherent STL mesh results. The algorithm
The surface tension term is not included in the numerical
repositions vertices of the mesh such that the square of
model to reduce computational costs. Rather, the deposi-
distances from the points to its nearest neighbors is
tion model is calibrated to account for the effect of surface
least. Details about this algorithm and implementation
tension on splats size and spreading time using semiem-
should be checked elsewhere (Ref 34).
pirical equation (Ref 27).
6. STL-FE Mesh Conversion This is the last stage of the
geometric reconstruction step where the generated STL
Reconstruction of Splat Geometry from PC
(surface) mesh is converted into a solid 3D FE mesh.
In most of recent commercial CAD/FE packages (such
In this section, several algorithms used for the conversion
as ABAQUS, COMSOL and ANSYS), there exist
of point cloud (PC), obtained from SPH domain to high-
several options for this task. For instance, ABAQUS
quality finite element (FE) mesh, are discussed.
packages have an option of importing STL mesh and
1. Poisson-Disk Sampling (PS) PS algorithm [developed directly converting the 2D trigonal elements into 3D
by Corsini et al. (Ref 28)] is used to increase the tetrahedral elements using the in-built algorithm.
density of point cloud such that the point sets are Details for this can be found elsewhere (Ref 35).
sufficient for effective mesh generation. The algorithm
does not add points outside the domain of interest.
Finite Element Formulation
2. 3D Alpha-Shape 3D alpha-shape algorithm [first
developed by Edelsbrunner and Mücke (Ref 29) and
Using the converted FE mesh of splats, a thermo-me-
recently improved by Kamberov et al. (Ref 30)] is used
chanical analysis is to be conducted for the prediction of
to reconstruct the surface topology of objects from
the evolution of temperature and stresses. Using Galerkin’s
point cloud data. The algorithm requires the user to
finite element approximation and backward Euler time
provide an optimum a-value depending on the com-
stepping scheme, the variational form of the energy equa-
plexity of the shape. When the a-value is too small,
tion can be given in discrete form as:
unwanted cavities emerge on the surface, while larger Z Z
a-value results in a rough surface. d
½N T qðTÞ  fH gdV þ ½BT kðTÞ  ½BdV
3. Poisson Surface Reconstruction (PSR) PSR algorithm dt
V V
[developed by Kazhdan and Hoppe (Ref 31, 32)] is Z ðEq 23Þ
T
used for generation of an implicitly represented surface  ½N  hcv ðTÞ  ðT  Tair ÞdA ¼ 0
from the dense point cloud. Based on theory, Poisson’s Ah
equation is used to compute an implicit scalar function

123
 Author's personal copy
J Therm Spray Tech

where qðTÞ is density, fH g is nodal enthalpy matrix, kðTÞ where ½M  is the mass matrix.
is thermal conductivity, ½N  is element interpolation (or For an implicit-static procedure, the stiffness matrix
shape) function matrix, ½B is shape function derivative directly relates to the displacement field based on Newton–
matrix, hcv ðTÞ is convection heat transfer coefficient on Raphson iteration scheme as:
external boundaries, and Tair is ambient temperature.
½K tþDt
i fDU gtþDt
i ¼ fPext gtþDt tþDt
iþ1 fPin gi ðEq 27Þ
Using Newton’s iteration to linearize Eq 23, we obtain
R T
the following discretized form of the equation as: where ½K  ¼ ½B ½J ½BdV is the stiffness matrix and
tþDt
½J  ¼ oDr V
½Kth  DTiþ1 ¼ fRT g ðEq 24Þ oDe is the Jacobian matrix.
At each time step, the temperature distribution obtained
where using Eq 24 is used to compute thermal strains and cor-
Z  tþDt responding internal stresses which are used to predict the
dH
½Kth  ¼ ½N T qðTÞ 
½N dV displacement or deformation of the bodies.
dT i
V
Z Z
T tþDt
þ ½B  ki ðTÞ  ½BdV  ½N T htcv ðTÞ  ½N dA; Numerical Implementation
V Ah
Z
Thermal spray process can be divided into sub-regimes,
fR t g ¼ ½N T  htcv ðTÞ  ðTitþDt  Tair ÞdA
i.e., multiple droplets impact and spreading (occurs within
Ah
Z 0-2 ls), solidification of splat layers (occurs within
1  
 ½N T qðTÞ  HitþDt  H t dV 20-500 ls) and post-deposition cooling to room tempera-
Dt ture (occurs within 1-50 s). As the time scale of these
V
Z regimes differs considerably, a sequentially coupled
 ½BT  kt ðTÞ  ½BdV approach is adopted during the implementation of the
V hybrid model. Thus, the flow (Navier–Stokes’) equations
are solved solely for prediction of splat formation process
and Dt is the time step.
during the first regime. Then, the reconstructed FE mesh of
Using Eq 24, the nodal temperature at each time step is
splat structures is used to solve for quenching stresses
found by solving the resulting algebraic form of equations.
developed during splat solidification (i.e., second regime).
The latent heat release during the solidification process is
 tþDt Lastly, the results of the second regime are read as an
accounted for at each time step in the term, ddHT , initial condition for computation of mismatch stresses
i
representing the effective heat capacity of the material. developed when the coated structure is finally cooled to
While the variational form of the stress-equilibrium room temperature. This coupling greatly improves the
equation (given in Eq 26) is given as: computational efficiency of the model without much
Z compromise on results quality.
d2 The numerical implementation was carried out in stan-
qðTÞ  ½N T  2 fU g  ½N dV
dt dard finite element packages (ABAQUS and MeshLab) and
V Z Z
self-developed codes. The required initial and boundary
¼  ½BT  ½r  dV þ ½N T  ftgdS conditions are first defined. Then, the velocity field and
Z
V S peening stress developed after droplets impact are solved
T using PC3D continuum particle elements based on SPH
þ qðTÞ  ½N  fggdV ðEq 25Þ
V
model (implemented in ABAQUS/Explicit). For consis-
tency, the SPH model is calibrated such that droplets
where fU g is nodal displacement vector, ½r is internal spreading stops after reaching a spreading factor that is
stress tensor, ftg is boundary traction vector, and fgg is equivalent to that of the previous semiempirical model (Ref
body force vector. 27). The geometry of the deposited splats (i.e., shape and
From Eq 25, it can be seen that the inertia of the system size) is then extracted (in form of PC) using python codes
can be can be generally expressed as the difference before PC-to-FE conversion which has been fully imple-
between external forces (Pext ) acting on the body and mented in the open-source package, MeshLab, using MLX
internal forces developed due to stress field (Pin ) leading to scripts. Firstly, point cloud file which contains the number
dynamic equilibrium as expressed in Eq 27, and coordinate of about 3000 points (from SPH results) is
d2 read and imported into Meshlab server. Then, the surface
½M   fU g ¼ fPext g  fPin g ðEq 26Þ topology of deposited splat structures is roughly
dt2

123
 Author's personal copy
J Therm Spray Tech

reconstructed using 3D alpha shape with an optimum alpha thermo-mechanical analysis of the coating process during
value of 5% of the bounding box diagonal length (i.e., the second and third regimes. The heat transfer and
about 0.0102353 mm). The population of original PC is implicit-static (post-deposition stress) model are imple-
populated to around 2 million points using the PS algo- mented in ABAQUS/Standard using implicit time stepping
rithm. Using the 2 million points, a smoother surface scheme, while the explicit-dynamic (deposition stress)
topology is generated for the second time using a 3D alpha model is implemented in ABAQUS/Explicit using a con-
shape (with a lower alpha value of 2%). The smoothened ditionally stable explicit time stepping scheme. The final
surface is used to generate denser PC having up to 8 mil- residual stress field is predicted by integrating the deposi-
lion points using the PS algorithm. Then, the denser PC is tion stresses with that of post-deposition cooldown mis-
used to generate high-quality implicit surface (or STL match stresses obtained using thermostatic-structural
mesh) using PSR algorithm (with octree depth of 7 and analyses.
solver divide of 8). With insufficient PC density, the
implicit surface generated by PSR algorithm may contain
few holes or distorted structures. In that case, the hole sites Experimental Procedure
need to be deleted and repaired using the ‘‘close hole’’
algorithm available in Meshlab. Before the final conversion For the purpose of validation of numerical results, residual
of STL to 3D FE mesh, QED algorithm is used to reduce stress measurement was conducted based on hole drilling
STL mesh density to the required level of refinement, while experiment (as shown in Fig. 4). A stainless-steel (SS301)
LS3 algorithm is used to re-position surface and improve substrate was first to cut into dimensions of 12 mm 9
elements aspect ratio. Upon import into ABAQUS/CAE, 2.5 mm 9 1.5 mm and grit-blasted with an alumina grit
the STL mesh is transformed into 3D (solid) FE mesh using having an average diameter of 0.16 mm. Then, a thin layer
algorithms embedded in the software (i.e., in edit-mesh tool of NiCrAl bond coat [of type Metco 443NS (Ref 36)] was
options). The average time required for the conversion is to improve the adhesion of YSZ top coat. The YSZ top coat
about 45 min per splat using 32 cores/64 GB RAM/2 GHz was deposited with conventional DC plasma torch using
Dell workstation. The current process of converting the PC spherical powder of type Oerlikon Metco 204NS-G (Ref
into FE mesh works well for all type of splats regardless of 37). The coating layer was deposited in 6 passes up to
how intricate their geometry might be as shown in later coating thickness of about 150 lm (as shown in Fig. 4c).
sections. A typical demonstration for the conversion is Before the start of hole drilling experiment, the surface of
demonstrated in Fig. 3. The 3D FE mesh is used for the coating sample was prepared according to the

Fig. 3 PC-to-FE mesh

conversion for splats of intricate
shapes

123
 Author's personal copy
J Therm Spray Tech

Fig. 4 (a) Drilling of YSZ

sample using HD setup,
(b) bonded and soldered
062UM-type strain gauge,
(c) cross section view of coating
sample (optical micrograph)

procedure given in bulletin B-129-8 (Ref 38) of the manual Results and Discussion
by Vishay Micro-measurement Inc. The top surface of the
sample was first cleaned with a degreaser, M-prep condi- In the present study, a hybrid computational approach for
tioner-A and M-prep neutralizer 5-A. This helps in estab- the numerical modeling of residual stress evolution in
lishing a strong bond between the strain gage and surface thermal spray coatings (TSCs) is tested on thermal barrier
of the sample. Then, CEA-XX-062UM-120-type strain coating (TBC) system made of yttria-stabilized zirconia
gage (having mean diameter of 5.13 mm) was bonded to (YSZ) deposited on steel alloy. The predicted residual
the surface using M-bond 200 adhesive as explained in stress profile has been validated by comparing with results
bulletin B-127-4 (Ref 39). The terminals of the strain gage obtained from the analytical model by Tsui and Clyne (Ref
were connected to the P3 strain gage indicator using lead 5) and hole drilling experiments. Using this new compu-
wires and three-wire quarter bridge circuit. The HD device tational approach, the effect of influential process param-
was installed and aligned properly above the strain gage eters (such as droplet size, impact velocity, temperature,
according to the RS-200 Milling Guide (Ref 40). Diamond- impact angle, spray path and substrate surface roughness)
coated tungsten carbide cutter with a diameter of 1.4 mm on the residual stress profile can be properly investigated.
was used to drill through the strain gauge using an air
turbine. It is essential to limit the drilling rotational speed Multilayer Deposition with SPH
and feed rate to minimize the influence of cutting forces
and sample overheating on strain readings. After the For a complete assessment of efficiency and accuracy of
establishment of the zero depth, the hole drilling was the proposed approach, several tests are carried out based
started, and strain readings were taken in increments of on single droplet deposition on a flat substrate. Figure 5
20 lm and feed rate of 5 lm/min until attaining a total shows that the results predicted using the current SPH
depth of 600 lm. Before taking readings, it was necessary model is reasonably comparable to those of previous
to wait for at least 10 min for the ceramic sample to cool experiments (Ref 44), and numerical results obtained using
down and strain readings to stabilize. The multilayered Coupled–Eulerian–Lagrangian (CEL) approach (Ref 21).
nature of coated samples necessitates the use of finite The spreading time of the droplet which has an initial
element analysis to compute the calibration coefficients impact velocity of 190 m/s is comparable to that of the
required for the estimation of residual stress profile. This CEL model, i.e., 0.39 ls. The results also show that higher
task was carried in ABAQUS/Standard environment, and initial impact velocities result in higher spread factor with
the popular differential method was used for residual stress slight deviation observed between the SPH results and
calculation as previously done by Gadow et al. (Ref 41) those of the CEL.
and Buchmann et al. (Ref 42, 43). Using statistically defined initial conditions, multiple
droplets deposition is modeled on PC using SPH. The

123
 Author's personal copy
J Therm Spray Tech

computation is carried out layer-wise up to the fifth layer of 79.2 and 4.2 lm, respectively. The size and shape of
deposition. For brevity, only results of the second and deposited splats are found to be strongly influenced by their
fourth layer will be presented here. Figure 6(a) and interaction with nearby droplets, thereby resulting in
(b) shows the variation of the velocity component in the z- complex splat geometries. Consequently, the simulated
direction during droplets deposition for the second and coating microstructure developed in current work has a
fourth layers (each consisting of nine droplets). Due to the variety of discontinuities or imperfections (such as pores,
low impact energy associated with APS process, peening cracks and other complex geometrical features) typically
stresses are not considered in the present study. The aver- encountered in thermal spray coatings.
age spreading time for the droplets is found to be about
1 ls which is comparable to times reported in previous Thermo-Mechanical Analysis
studies (Ref 14, 45). The average spread factor, diameter
and thickness of the simulated splats are found to be 3.47, Simulated Coating Microstructure

Upon completion of multiple droplets deposition in layers,

the geometry of each splat is reconstructed using the var-
ious algorithms described in Numerical Model section. The
geometry is first represented in the form of STL (or sur-
face) mesh which contains high-quality trigonal elements.
Then, the STL mesh is transformed into a 3D linear
tetrahedral mesh using edit-mesh tool options of ABA-
QUS/CAE. Through stacking of multiple splats, we
developed a numerical coating microstructure with good
overlap at splats interface as shown in Fig. 7. We found
that meshing splat bodies in ABAQUS results in higher
computational efficiency and minimized numerical issues
(arising due to poor elements) as compared to when it is
done elsewhere. Also, grid-independence test was carried
out to ensure that the level of mesh refinement is sufficient
for numerical convergence at a moderate computational
cost. The average size of elements within each layer is
Fig. 5 Validation of SPH model for droplet deposition demonstrated in Table 2 (in the form of average edge
lengths). The criteria for mesh quality are that ABAQUS

Fig. 6 Velocity of droplets in impact direction (mm/s) during deposition of (a) second layer and (b) fourth layer

123
 Author's personal copy
J Therm Spray Tech

accepts any tetrahedral element that has a shape factor of at difference in terms of results quality. The approach adopted
least 3 9 10-3, face corner angle within 5 and 170 and here is more suitable and flexible as compared to the
aspect ratio of at most 10. Upon preliminary checks, we popular image-based finite element approach. Apart from
found that the mesh generated with current approach pas- having finite elements of higher quality, it is developed
sed the overall criteria despite the presence of discontinu- based on standard codes available in the literature. Thus, it
ities (as demonstrated in Table 2). Even deleting the very does not involve tedious noise cleaning and re-processing
few elements that failed some of the criteria makes no of microstructure images. The approach also captures the
process dependence and layer-wise build-up of temperature
and residual stress during the spray process.

Temperature Distribution

Figure 8(a) and (b) shows the temperature distribution

developed after the deposition of the second and fourth
layers, respectively. The figures show that the newly
deposited splat layer (in gray) starts to solidify after its
contact with the pre-solidified coating layers. The predic-
tion of temperature distribution for the first and second
layer is more computationally demanding due to the steep
gradient developed near the interface. It is also found that
the solidification rate of a given splat is immensely affected
by its interaction with its nearest neighbors. Consequently,
partial re-melting of pre-solidified splats (due to contact
with new splats) is observed in certain regions. Thus, the
numerical results predicted here show that temperature
distribution is not only complex but highly dependent on
many factors such as geometrical complexity, a number of
layers deposited, solidification rate, quality of contact and
other factors which should be considered for effective
prediction of residual stress evolution.
As shown in Fig. 9, five points are selected for quanti-
tative analysis of temperature and stress profiles developed
Fig. 7 3D tetrahedral FE mesh of splats deposited in five layers
along the thickness of the coating layer. Figure 10 shows

Table 2 Information on tetrahedral element size and quality

Layer name First Second Third Fourth Fifth

# Splats 9 9 9 9 9
# Elements 124,591 90,499 60,418 96,119 60,432
# Nodes 31,314 19,566 15,965 20,335 16,317
Shape factor \ 0.0001: 0.6085/2 9 10-2/0 0.5721/1 9 10-5/1 0.5973/4 9 10-6/4 0.5668/1 9 10-10/6155 0.6/1.1 9 10-4/0
average/worst/# Elem.
Min angle \ 5: average/worst/ 36.17/7.97/0 35.93/4.83/0 35.69/3.42/1 36.12/5.00/1 35.81/4.35/1
#Elem.
Max angle [ 170: 91.56/150.90/0 91.95/168.65/0 92.39/170.21/1 91.64/163.76/0 92.03/169.27/0
average/worst/#Elem.
Aspect ratio [ 10: 1.75/6.14/0 1.77/8.93/0 1.80/10.23/1 1.76/8.79/0 1.78/7.43/0
average/worst/# Elem.
Average min. edge length (lm) 1.65 1.82 2.13 1.76 2.19
Average max. edge length (lm) 2.81 3.23 3.66 3.00 3.76
#Elem. number of elements that failed the criteria

123
 Author's personal copy
J Therm Spray Tech

Fig. 8 Evolution of temperature (C) during splats solidification for: (a) second layer and (b) fourth layer

gradient developed along the thickness of the coating

decreases with an increase in coating thickness, resulting in
a decrease in the solidification rate with the deposition of
more splat layers. Figure 10(c) and (d) shows the variation
in temperature during the deposition of the fifth layer at
points C and E. It can be seen that the temperature of the
coating layer drops from the melting point to lower values
during deposition, while that of the substrate rises. There is
a sharp drop in temperature in some regions within the
coating layer due to the presence of pores or cracks.
Consequently, the temperature profile varies from point to
point because of the influence of complex geometrical
features on temperature distribution.
Figure 11(a) shows the variation in temperature along
the width and length of the coating. The thermal gradient
developed along those directions is far lower as compared
to that along the thickness direction. Consequently, it is
expected to have less stress variation along that direction.
Figure 11(b) shows that the thermal histories of three dif-
Fig. 9 Five points selected for the analysis of results
ferent points (A, B and E) located at different layers are not
only different but highly random due to the complex nature
the variation in temperature along the thickness direction of the process. As observed from the figure, the tempera-
after solidification of the fifth splat layer. As observed ture at those points overshoots to higher values after a
experimentally, the coating layer has a variable thickness certain time due to several cycles of reheating caused by
due to undulation resulting from the interaction of multiple the deposition of new splat layers. It is expected for the
splat layers (shown in Fig. 10a). Thus, the simulated coating layer to develop complex residual stress field
microstructure has a larger thickness (around 15-30 lm) at within the coating microstructure due to the strong
points A, C and D as compared to points B and E (5- dependence of temperature with location and process
10 lm). Figure 10(b) shows that the high-temperature history.

123
 Author's personal copy
J Therm Spray Tech

Fig. 10 (a) Temperature profile

developed after solidification of
5th layer at different locations,
(b) temperature profile
developed after solidification of
1st–5th layer, (c) evolution of
temperature during deposition
of 5th layer at C and
(d) evolution of temperature
during deposition of 5th layer at
E

Fig. 11 (a) Temperature profile

along coating width and length;
(b) temporal of variation of
temperature from deposition to
end of cooldown at different
points

Residual Stress Evolution solidification, deposited splats have near-zero stress values
since they are in the molten state. However, during solid-
The evolution of residual stress was computed using tem- ification, all solidified regions of the coating develop ten-
perature solution obtained at each time step. Fig- sile quenching stress with a low magnitude of about
ure 12(a) and (b) shows the corresponding quenching 30 MPa. This low magnitude is attributed to the restriction
stresses developed after the solidification of the second and of stress values with the use of the perfectly plastic material
fourth layers, respectively. At the beginning of model. This is commonly done to consider stress relaxation

123
 Author's personal copy
J Therm Spray Tech

Fig. 12 Evolution of quenching (von Mises) stress (MPa) during deposition of: (a) second and (b) fourth layers

by microcracking and interfacial sliding within the ceramic Fig. 15). The equi-biaxial nature of the stress profile is
coating layer [as found in previous works (Ref attributed to the nature of coating deposition. The coating
6, 8, 46, 47)]. However, unlike with the previous models, is usually deposited in a layer-by-layer (or pass-by-pass)
quenching stresses predicted here display an uneven field manner such that temperature gradient is more predomi-
distribution having a fluctuation of stress values because of nant along the thickness direction. However, as observed
process dependence and presence of discontinuities. Sim- experimentally, minor differences occur between the two
ilar observations can be made with the fifth layer as orthogonal stress profiles due to the presence of disconti-
demonstrated in Fig. 13(a). Moreover, the quenching nuities (or imperfections) within the coating layer. Fig-
stresses influence the final stress state of the coating layer ure 14(a) and (b) shows that the quenching (deposition)
despite its small magnitude. As demonstrated by stress developed after splat solidification is tensile with low
Fig. 13(b), the von Mises residual stress field shows that values attributed to stress relaxation by microcracking and
very high residual stresses of about 300 MPa are developed interfacial sliding as previously stated when discussing
at various critical regions within the coating microstruc- Fig. 12. Despite its low magnitude, the quenching stress
ture. The residual stress field is dominated by post-depo- influences the nature of the final residual stress profile,
sition stresses due to the large difference in the coefficient especially at the near-interface region of the coating layer.
of thermal expansions (CTEs) of coating and substrate It is well established in the literature (Ref 6) that cera-
layers. mic coatings develop very high compressive post-deposi-
For the analysis of results, the emphasis is given to in- tion (mismatch) stress because of their low CTE as
plane normal stresses due to their significance on coating compared to the underlying substrate. Consequently,
lifetime. Figure 14(a) and (b) shows the variation of biaxial compressive residual stress is developed in the ceramic
in-plane deposition and residual stresses developed along coating layer due to the dominance of mismatch (com-
the coating thickness. As observed in previous results by pressive) stresses over the low quenching (tensile) stresses.
Buchmann et al. (Ref 42), the in-plane stresses (acting in x- Thus, the residual stress profiles (shown in Fig. 14a and b)
and y-direction) are equi-biaxial, in the sense that they are display equi-biaxial compressive stress values that increase
almost equal in terms of both magnitudes despite acting in with depth of coating layer. Also from the figures, it can
perpendicular directions. This has also been observed from also be seen that the residual stress profiles are fairly
the results of our hole drilling experiment (shown in comparable to results obtained with Tsui and Clyne

123
 Author's personal copy
J Therm Spray Tech

Fig. 13 von Mises stress field

(MPa) developed in five-layer
coating: (a) quenching stress
and (b) residual stress

analytical model (Ref 5) for a coating layer deposited in 5 compressive residual stress developed near the coating
steps. The inputs to the analytical model were: elastic interface improves the adhesion strength of the coating
properties given in Table 1, intrinsic (quenching) stress of layer. As found in previous studies (Ref 5, 8),
30 MPa and same geometrical dimensions as used in Fig. 14(a) and (b) shows that the compressive residual
numerical simulations. As an elastic material model with stress developed within the coating layer is balanced by
temperature-independent material properties was used to corresponding tensile stresses on the substrate side due to
derive the analytical model, it is observed that deviation mismatch of structural properties at the coating–substrate
exists between the results obtained analytically and interface. Specifically, Fig. 14 shows that the in-plane
numerically especially at regions close to the interface. The residual stresses developed at the interface (in x/y direc-
analytical model overestimates the residual stress profile tions) are - 110/- 130 MPa and 310/370 MPa on coating
due to the assumptions made during derivation. Thus, for and substrate sides, respectively. The large difference
more realistic residual stress prediction, it is necessary to between the stresses developed at the interface (in terms of
consider the nonlinear material behavior of the coating and magnitude and direction) signifies the influence of residual
substrate layers. For the ceramic coating layer, the presence stresses on the adhesion strength/lifetime of coatings. The
of pores and microcracks considerable affects the residual equi-biaxial compressive stresses (shown in Fig. 14) are
stress profile. For metallic coatings, plastic deformation comparable to the results of our hole drilling experiment
influences the stress profile more than pores and cracks. In and those of previous studies (Ref 5, 8, 42, 48, 49). As
all cases, it is essential to model the steel substrate layer demonstrated in Fig. 15, the current numerical results fall
with a robust material model that considers plastic defor- within the expected range of stress values measured
mation with proper strain hardening criterion. The high experimentally with incremental hole drilling method.

123
 Author's personal copy
J Therm Spray Tech

Fig. 14 Biaxial in-plane stress profiles at A: (a) stress in x-direction

and (b) stress in y-direction

Also, the overall trend of numerical and experimental stress

profile is similar. As obtained numerically, the experi- Fig. 15 Comparison of residual stress profile (at A) with experimen-
mental results revealed that compressive residual stresses tal results: (a) stress in x-direction and (b) stress in y-direction
(which increase with depth) are developed within the
confirmed by Fig. 16(b), it can be observed that the various
ceramic coating layer. The substrate layer develops high
tensile stresses, especially near the interface. Also, the stress components add up to result in higher von Mises
nonlinear stress profile is developed due to the oscillation stress variation along the coating thickness. The developed
shear stresses are found to be very low with the xy-shear
in strain readings caused by the presence of defects. The
high compressive stresses measured experimentally near residual stress being the highest (shown in Fig. 16c). The
maximum principal stress, shown in Fig. 16(d), seems
the coating surface are due to absurd variation in initial
qualitatively like the normal stress component in the z-
strain readings possibly caused by the high surface
roughness or variable thickness of the coating layer. direction. A similar observation is made at other locations.
Figure 17(a) shows that, despite the low-temperature gra-
Figure 16 shows the variation in other stress compo-
dient along the width/length of the coating layer, there
nents along the thickness of the coating and substrate layers
at point A. The normal (out-of-plane) residual stress acting exists fluctuation of stresses to within same range of values
due to obvious presence of discontinuities at random sites
in the z-direction is considerably high and tensile, espe-
within the coating microstructure. The expansion and
cially at regions near the interface (as shown in Fig. 16a).
The difference between the in-plane stresses developed at contraction of the substrate are tracked (at base) during
splats solidification (i.e., second regime) and post-deposi-
the interface on coating and substrate sides is small as
tion cooldown (i.e., third regime), respectively (as shown
compared to that of the equi-biaxial stresses explained
previously. Consequently, the coating is weaker with in Fig. 17b). As expected, substrate expansion during
splats solidification is not very significant due to
loadings in z-direction than in other directions. As further

123
 Author's personal copy
J Therm Spray Tech

Fig. 16 Other stress profiles at A: (a) in-plane normal stress, (b) von Mises stress, (c) shear stress and (d) maximum principal stress

localization of heat near the interface. However, during meshless methods (such as SPH) are found to be more
post-deposition cooldown stage, the substrate contracts suitable for the numerical modeling of the thermal
considerably due to rapid decrease in temperature spray process, while the evolution of temperature and
throughout the substrate layer. residual stress in the multilayered coating are best
modeled using grid-based numerical schemes (such as
FEM) due to high stability.
Conclusions • The results also show that the use of realistic process
parameters leads to a more reliable estimation of
Residual stress is considered to have a significant influence residual stresses. As observed experimentally, we have
on constitutive behavior, lifetime and integrity of thermal predicted a residual stress profile that is nonlinear along
spray coatings including TBCs. Based on the results pre- the length and thickness of the coating. Specifically, it
sented, the following conclusions can be made: is found that the discontinuities such as pores and
cracks lead to local stress relaxation in various regions
• The proposed hybrid approach has proven to be an
within the coating layer. The relaxation of the residual
efficient computational tool for the numerical predic-
stress field weakens several regions within the coating
tion of residual stresses. Due to the large deformation
which can serve as a source of damage initiation while
associated with multiple droplets deposition, the
in active service.

123
 Author's personal copy
J Therm Spray Tech

Spray Coatings: Current Status and Way Forward, J. Therm.

Spray Technol., 2017, 26(6), p 1115-1145
7. L. Wang, Y. Wang, X.G. Sun, J.Q. He, Z.Y. Pan, and C.H. Wang,
Finite Element Simulation of Residual Stress of Double-Ceramic-
Layer La2Zr2O7/8YSZ Thermal Barrier Coatings Using Birth and
Death Element Technique, Comput. Mater. Sci., 2012, 2012(53),
p 117-127
8. L. Wu, J. Zhu, and H. Xie, Numerical and Experimental Inves-
tigation of Residual Stress in Thermal Barrier Coatings During
APS Process, J. Therm. Spray Technol., 2014, 23(4), p 653-665
9. M. Elhoriny, M. Wenzelburger, A. Killinger, and R. Gadow,
Finite Element Simulation of Residual Stress Development in
Thermally Sprayed Coatings, J. Therm. Spray Technol., 2017,
26(4), p 735-744
10. R. Berthelsen, D. Tomath, R. Denzer, and A. Menzel, Finite
Element Simulation of Coating-Induced Heat Transfer: Appli-
cation to Thermal Spraying Processes, Meccanica, 2016, 51(2),
p 291-307
11. B. Klusemann, R. Denzer, and B. Svendsen, Microstructure-
Based Modeling of Residual Stresses in WC-12Co-Sprayed
Coatings, J. Therm. Spray Technol., 2011, 21(1), p 96-107
12. J. Mostaghimi, S. Chandra, R. Ghafouri-Azar, and A. Dolatabadi,
Modeling Thermal Spray Coating Processes: A Powerful Tool in
Design and Optimization, Surf. Coat. Technol., 2003, 2003(163-
164), p 1-11
13. C. Li, X. Zhang, Y. Chen, J. Carr, S. Jacques, J. Behnsen, M. di
Michiel, P. Xiao, and R. Cernik, Understanding the Residual
Stress Distribution through the Thickness of Atmosphere Plasma
Sprayed (APS) Thermal Barrier Coatings (TBCs) by High Energy
Synchrotron XRD; Digital Image Correlation (DIC) and Image
Based Modelling, Acta Mater., 2017, 132, p 1-12
14. M.Y. Zhang, H. Zhang, and L.L. Zheng, Simulation of Droplet
Spreading, Splashing and Solidification Using Smoothed Particle
Hydrodynamics Method, Int. J. Heat Mass Transf., 2008, 51(13),
p 3410-3419
15. M. Zhang, Simulation of Surface Tension in 2D and 3D with
Smoothed Particle Hydrodynamics Method, J. Comput. Phys.,
Fig. 17 (a) Residual stress profile (acting in the x-direction) from C- 2010, 229(19), p 7238-7259
to-B at 10 lm above interface, (b) expansion/contraction of substrate 16. A. Farrokhpanah, J. Mostaghimi, and M. Bussmann, Free-Surface
layer (in positive z-direction) during deposition and post-deposition Enthalpy Method for Transient Convection/Diffusion Phase
stages Change, arXiv Prepr. arXiv1701.00463 (2017)
17. J.J. Monaghan, On the Problem of Penetration in Particle Meth-
Acknowledgments The authors would like to acknowledge the ods, J. Comput. Phys., 1989, 82(1), p 1-15
support provided by King Fahd University of Petroleum and Minerals 18. M. Xue, S. Chandra, J. Mostaghimi, and C. Moreau, A Stochastic
(KFUPM) in funding this work through project FT161016. Coating Model to Predict the Microstructure of Plasma Sprayed
Zirconia Coatings, Model. Simul. Mater. Sci. Eng., 2008,
2008(16), p 065006
19. N. Nayebpashaee, S.H. Seyedein, M.R. Aboutalebi, H. Sar-
References poolaky, and S.M.M. Hadavi, Finite Element Simulation of
Residual Stress and Failure Mechanism in Plasma Sprayed
1. F. Hermanek, Thermal Spray Terminology and Company Origins, Thermal Barrier Coatings Using Actual Microstructure as the
ASM International, Materials Park, 2001 Representative Volume, Surf. Coat. Technol., 2016, 291, p 103-
2. R.B. Heimann, Plasma-Spray Coating: Principles and Applica- 114
tions, Wiley, New York, 2008 20. M. Gupta, K. Skogsberg, and P. Nylén, Influence of Topcoat-
3. P. Araujo, D. Chicot, M. Staia, and J. Lesage, Residual Stresses Bondcoat Interface Roughness on Stresses and Lifetime in
and Adhesion of Thermal Spray Coatings, Surf. Eng., 2005, Thermal Barrier Coatings, J. Therm. Spray Technol., 2014, 23(1),
21(1), p 35-40 p 170-181
4. Z. Xue, A.G. Evans, and J.W. Hutchinson, Delamination Sus- 21. Z. Zhu, S. Kamnis, and S. Gu, Numerical Study of Molten and
ceptibility of Coatings Under High Thermal Flux, J. Appl. Mech., Semi-Molten Ceramic Impingement by Using Coupled Eulerian
2009, 76(4), p 041008 and Lagrangian Method, Acta Mater., 2015, 90, p 77-87
5. Y.C. Tsui and T.W. Clyne, An Analytical Model for Predicting 22. J.W. Adams, R. Ruh, and K.S. Mazdiyasni, Young’s Modulus,
Residual Stresses in Progressively Deposited Coatings Part 1: Flexural Strength, and Fracture of Yttria-Stabilized Zirconia
Planar Geometry, Thin Solid Films, 1997, 306(1), p 23-33 Versus Temperature, J. Am. Ceram. Soc., 1997, 80(4), p 903-908
6. A.A. Abubakar, A.F.M. Arif, K.S. Al-Athel, S.S. Akhtar, and J. 23. H. Hayashi, T. Saitou, N. Maruyama, H. Inaba, K. Kawamura,
Mostaghimi, Modeling Residual Stress Development in Thermal and M. Mori, Thermal Expansion Coefficient of Yttria Stabilized

123
 Author's personal copy
J Therm Spray Tech

Zirconia for Various Yttria Contents, Solid State Ion, 2005, 38. Vishay-Micromeasurements, Instruction Bulletin B-129-8: Sur-
176(5), p 613-619 face Preparation for Strain Gage Bonding (2014), http://www.
24. F. Cverna, Worldwide Guide to Equivalent Irons and Steels, ASM vishaypg.com/docs/11129/11129_b1.pdf. Accessed 12 Mar 2018
international, Materials Park, 2006 39. Vishay-Micromeasurements, Instruction Bulletin B-127-14:
25. M. Zhang, H. Zhang, and L. Zheng, Application of Smoothed Strain Gage Installations with M-Bond 200 Adhesive (2014),
Particle Hydrodynamics Method to Free Surface and Solidifica- http://www.vishaypg.com/doc?11129. Accessed 12 Mar 2018
tion Problems, Numer. Heat Transf. Part A Appl., 2007, 52(4), 40. T. Valente, C. Bartuli, M. Sebastiani, and A. Loreto, Imple-
p 299-314 mentation and Development of the Incremental Hole Drilling
26. J.J. Monaghan, Smoothed Particle Hydrodynamics, Annu. Rev. Method for the Measurement of Residual Stress in Thermal Spray
Astron. Astrophys., 1992, 30, p 543-574 Coatings, J. Therm. Spray Technol., 2005, 14(4), p 462-470
27. S.D. Aziz and S. Chandra, Impact, Recoil and Splashing of 41. M. Escribano and R. Gadow, in Residual Stress Measurement and
Molten Metal Droplets, Int. J. Heat Mass Transf., 2000, 43(16), Modeling for Ceramic Layer Composites, 27th International
p 2841-2857 Cocoa Beach Conference on Advanced Ceramics and Compos-
28. M. Corsini, P. Cignoni, and R. Scopigno, Efficient and Flexible ites: A ed. by W.M. Kriven and H.-T. Lin (USA, 2003), pp. 615-
Sampling with Blue Noise Properties of Triangular Meshes, IEEE 622
Trans. Vis. Comput. Graph., 2012, 18(6), p 914-924 42. M. Buchmann, R. Gadow, and J. Tabellion, Experimental and
29. H. Edelsbrunner and E.P. Mücke, Three-Dimensional Alpha Numerical Residual Stress Analysis of Layer Coated Composites,
Shapes, ACM Trans. Graph., 1994, 13(1), p 43-72 Mater. Sci. Eng. A, 2000, 288(2), p 154-159
30. G. Kamberov, G. Kamberova, and A. Jain, in 3D Shape from 43. K. Berreth, M. Buchmann, R. Gadow, and J. Tabellion, in
Unorganized 3D Point Clouds BT: Advances in Visual Com- Evaluation of Residual Stresses in Thermal Sprayed Coatings,
puting: First International Symposium, ISVC 2005, Lake Tahoe, Proceedings of International Thermal Spray Conference,
NV, USA, 5–7 December 2005. Proceedings, ed. by G. Bebis, R. Düsseldorf (1999)
Boyle, D. Koracin, and B. Parvin (Springer, Berlin, 2005), 44. M. Vardelle, A. Vardelle, A.C. Leger, P. Fauchais, and D. Gobin,
pp. 621-629 Influence of Particle Parameters at Impact on Splat Formation
31. H. Hoppe, in Poisson Surface Reconstruction and Its Applica- and Solidification in Plasma Spraying Processes, J. Therm. Spray
tions, Proceedings of the 2008 ACM Symposium on Solid and Technol., 1995, 4(1), p 50-58
Physical Modeling (ACM, 2008), p. 10 45. S. Alavi and M. Passandideh-Fard, Numerical Simulation of
32. M. Kazhdan and H. Hoppe, Screened Poisson Surface Recon- Droplet Impact and Solidification Including Thermal Shrinkage
struction, ACM Trans. Graph., 2013, 32(3), p 29 in a Thermal Spray Process, Front. Heat Mass Transf., 2011, 2,
33. M. Garland and Y. Zhou, Quadric-Based Simplification in Any p 023007
Dimension, ACM Trans. Graph., 2005, 24(2), p 209-239 46. J. Matejicek and S. Sampath, In Situ Measurement of Residual
34. S. Boyé, G. Guennebaud, and C. Schlick, Least Squares Subdi- Stresses and Elastic Moduli in Thermal Sprayed Coatings Part 1:
vision Surfaces, Comput. Graph. Forum, 2010, 29(7), Apparatus and Analysis, Acta Mater., 2003, 51(3), p 863-872
2021–2028. https://doi.org/10.1111/j.1467-8659.2010.01788.x 47. S. Kuroda, T. Dendo, and S. Kitahara, Quenching Stress in
35. ABAQUS, ABAQUS Documentation, Dassault Systèmes, Provi- Plasma Sprayed Coatings and Its Correlation with the Deposit
dence, 2013 Microstructure, J. Therm. Spray Technol., 1995, 4(1), p 75-84
36. Oerlikon Metco, Nickel Chromium–Aluminum Thermal Spray 48. H. Liu, F. Azarmi, M. Bussmann, J. Mostaghimi, and T.W.
Powders (2017), p. 1-4, https://www.oerlikon.com/ecomaXL/ Coyle, Experiments and Simulation of Rapid Solidification of Air
files/metco/oerlikon_DSMTS-0091.5_NiCrAl.pdf&download=1. Plasma Sprayed Alloy 625 on Stainless Steel, Surf. Coat. Tech-
Accessed 20 Feb 2018 nol., 2010, 204(9-10), p 1521-1527
37. Oerlikon Metco, 8% Yttria Stabilized Zirconia Agglomerated and 49. J. Liu, R. Bolot, and S. Costil, Residual Stresses and Final
HOSP Thermal Spray Powders (2017), pp. 1-6, https://www.oer Deformation of an Alumina Coating: Modeling and Measure-
likon.com/ecomaXL/files/metco/oerlikon_DSMTS-0001.10_ ment, Surf. Coat. Technol., 2015, 2014(268), p 241-246
8YO_ZrO_HOSP.pdf&download=1. Accessed 20 Feb 2018

123