Surface & Coatings Technology 315 (2017) 408–416

Contents lists available at ScienceDirect

Surface & Coatings Technology

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

Fatigue behavior of thermal sprayed WC-CoCr- steel systems: Role of

process and deposition parameters
Andrew Vackel ⁎,1, Sanjay Sampath
Center for Thermal Spray Research, Stony Brook University, Stony Brook, NY 11794, United States

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

Article history: Thermal spray deposited WC-CoCr coatings are extensively used for surface protection of wear prone compo-
Received 1 November 2016 nents in a variety of applications. Although the primary purpose of the coating is wear and corrosion protection,
Revised 21 February 2017 many of the coated components are structural systems (aero landing gear, hydraulic cylinders, drive shafts etc.)
Accepted in revised form 23 February 2017
and as such experience cyclic loading during service and are potentially prone to fatigue failure. It is of interest to
Available online 27 February 2017
ensure that the coating and the application process does not deleteriously affect the fatigue strength of the parent
Keywords:
structural metal. It has long been appreciated that the relative fatigue life of a thermal sprayed component can be
Thermal spray affected by the residual stresses arising from coating deposition. The magnitude of these stresses can be managed
High velocity oxy-fuel by torch processing parameters and can also be influenced by deposition effects, particularly the deposition tem-
Fatigue perature. In this study, the effect of both torch operating parameters (particle states) and deposition conditions
Residual stress (notably substrate temperature) were investigated through rotating bending fatigue studies. The results indicate
WC-CoCr a strong influence of process parameters on relative fatigue life, including credit or debit to the substrate's fatigue
life measured via rotating bend beam studies. Damage progression within the substrate was further explored by
stripping the coating off part way through fatigue testing, revealing a delay in the onset of substrate damage with
more fatigue resistant coatings but no benefit with coatings with inadequate properties. The results indicate that
compressive residual stress and adequate load bearing capability of the coating (both controlled by torch and de-
position parameters) delay onset of substrate damage, enabling fatigue credit of the coated component.
Published by Elsevier B.V.

1. Introduction Characterization and optimization of these coatings' wear and corrosion

behavior have been extensively studied as a function of processing and
Thermal spray (TS) is a widely utilized coating technology that ser- resultant coating properties [1,5,6]. However, in the case of a coated
vices a variety of industries including aerospace, heavy machinery, component subject to cyclic loading, such as aero landing gear or hy-
and power generation, wherein a feedstock material (often in powder draulic cylinder rods, understanding the fatigue behavior of the coating
form) is melted and propelled towards a surface, producing a shower and substrate as an integrated system is critical for coating design.
of particles that impact and rapidly solidify on the surface, (known as Numerous academic fatigue studies of TS coated specimens have
splats), which form a coating through successive layers. Within the been carried out, however there appears to be no consensus on whether
wide menu of TS processes, High Velocity Oxy-Fuel (HVOF) deposition an enhancement [7–10] or debit [7,11–14] in fatigue life is generated
has received particular attention for its ability to produce very dense with coating application. Industry data is rarely available for such stud-
(N99% [1,2]) and well bonded coatings. HVOF processing also has the ca- ies. Since TS coating properties are highly dependent on the processing
pability to induce compressive residual stresses within the deposit due parameters used [5,15], an assessment with one coating process param-
to solid-state peening achieved by the high kinetic energy particle im- eter does not offer a complete picture on the fatigue of TS coated spec-
pact during deposition, reducing or neutralizing the tensile stresses gen- imens [13,16]. With parametric studies of coating processing, it has
erated by splat quenching [3,4]. As such, HVOF processing is often been reported that the residual stress state of the coating has a large in-
employed for the deposition of dense hard facing materials, such as fluence on the relative fatigue life of coated specimens [12,16,17], as
WC-CoCr, to protect wear prone components of various functions. well as coating properties such as coating toughness [10]. Residual
stress of thermal spray coatings have also been shown to have consider-
able influence in contact fatigue scenarios [18]. The presence of residual
⁎ Corresponding author at: Sandia National Laboratories, 1515 Eubank SE, 87123
Albuquerque, NM, United States.
stresses throughout a material prone to fatigue loading has been consid-
E-mail address: avackel@gmail.com (A. Vackel). ered a critical factor for the fatigue life of uncoated components [19,20],
1
Present address: Sandia National Laboratory, Albuquerque, NM. as the superposition of the applied stress and residual stress throughout

http://dx.doi.org/10.1016/j.surfcoat.2017.02.062
0257-8972/Published by Elsevier B.V.
 A. Vackel, S. Sampath / Surface & Coatings Technology 315 (2017) 408–416 409

a fatigued body leads to local stress intensifications. Such stress intensi- WOKA 3652 WC-CoCr powder (Oerlikon-Metco, Westbury, NY). Torch
fications are then prone to fatigue crack initiation and growth as de- parameters for these conditions are shown in Table 1.
scribed by Paris' Law [21]. Other investigations into the effect of For planar samples, a torch rastering speed of 1000 m/s was used with
substrate preparation prior to coating deposition, such as shot peening a step size between strokes of 3 mm and 5 mm for the DJ and JP,
or grit blasting, have also been studied as residual stress contributors respectively.
in attempts to mitigate fatigue debits incurred by coating application For cylindrical fatigue samples, the torch rastering speed was 30 mm/s
[11] or to isolate the coating's influence on fatigue life [7]. and 50 mm/s for the DJ and JP, respectively, in line with the center axis as
Manipulation of TS coating residual stress via processing has been the the samples were rotated at 630 rpm in order to keep the same pitch be-
subject of many previous studies, where torch operating parameters, tween coating passes as the planar samples and avoid barber-poling.
feedstock characteristics, and deposition rates have all been demonstrat- Shielding was used on the fatigue samples in order to eliminate over-
ed as having large roles in determining the magnitude of coating deposi- spray. Cooling air was directed at planar and cylindrical samples in differ-
tion stresses [3–5,22]. In addition to these deposition stresses, thermal ent manners but kept consistent for the specified geometry.
stress arising from differences in the coating and substrate's coefficient Planar geometry samples consisted of 25.4 × 225 mm steel beams
of thermal expansion (CTE) as the specimen cools back to ambient tem- that were mounted to an in-situ curvature measurement device in
perature contributes to the coating residual stress. As a deposition param- order to calculate deposition, thermal, and residual stresses, with
eter, variations in cooling and part geometry and size can cause back-side substrate temperature monitored by contact thermocouples.
significant differences in substrate temperature with the otherwise Such a device and calculations were first developed for thermal spray
same processing parameters, producing considerable thermal stress as by Kuroda and Clyne [24], with modelling further expanded upon by
the coated part cools back to ambient temperature. Summated with the Tsui and Clyne [25], and compared to other residual stress measure-
deposition stress to produce the final residual stress, the thermal stress ment methods by Sampath and Matejicek [26,27]. Detailed descriptions
component of a coating is a key contributor, particularly when large ther- of device use and calculations can be found in references [3,5,28]. Micro-
mal expansion mismatches between the substrate and coating exist and/ graphs were prepared from these beams and examined in back scatter
or high substrate temperatures are experienced during spraying. mode under SEM (Hitachi TM3000).
This paper seeks to address several unanswered questions on the fa- Rotating bend fatigue (RBF) specimens were machined out of cold
tigue behavior of HVOF WC-CoCr coated steel through studies of spray drawn 1018 steel to the nominal dimensions shown in Fig. 1. Rotating
conditions and depositions conditions. In this paper, “torch parameter in- bend testing was chosen as the fatigue method since a reversible, higher
fluence” will refer to differences in the coating process achievable by vary- stress amplitude would be directed to the surface and coating, amplifying
ing parameters such as the torch's fuel or oxygen flow and the subsequent the coating parameter influence on specimen fatigue life. 1018 steel was
results of thermal and kinetic energy transfer onto the feedstock material, selected since it has similar modulus and thermal expansion coefficient
independent of how the particles interact with a substrate. “Deposition to higher strength steels but a lower fatigue strength that allows the coat-
condition influence” will refer to how the particles are directed to and in- ing parameter influence on fatigue life to be further amplified. All samples
teract substrate material, particularly the substrate temperature. Some sa- were heat treated within a vacuum furnace at 900 °C for a dwell time of
lient questions that will be addressed through these studies include: 2 h and allowed to cool within the furnace to ambient temperature in
order to remove work hardening and reduce the cold drawn grain struc-
• Isolate the role of spray process conditions and deposition tempera- ture within the steel to produce a stress free steel with a more isotropic
tures on the fatigue response of the coated system. grain distribution.
• Impact of substrate surface preparation on fatigue response and the The planar steel beams and selected RBF samples were grit blasted
potential modification to the effect upon coating application. using an alumina grit. Samples were cleaned in an acetone bath with ul-
• Understanding fatigue damage accumulation within a coated speci- trasonic agitation for cleaning prior to coating deposition to ensure no
men and the role of process conditions on damage progression. trapped grit or dust would be included in the coating-substrate interface.
Surface temperature of the necks of the RBF samples during spraying
2. Experimental methods was monitored by an IR thermal camera (FLIR Systems, Wilsonville,
OR), with manipulation of the temperature being accomplished by ro-
Coatings for the investigation of torch parameter influence on fa- botically varying the time between coating strokes on the samples.
tigue life were produced using a DJ-2600 HVOF torch (Oerlikon- Cooling air was directed at the neck of the samples during spraying
Metco, Westbury, NY) with commercially available JK 120H WC-CoCr and cooling. Overspray of the samples was eliminated by securing fitted
powder (Kennametal-Stellite, Goshen IN). The “DJ-A” and “DJ-B” condi- caps machined to size onto the ends of the substrates doing the spray
tions were selected from an experimental design where in-situ curva- process. Coatings were sprayed to a nominal thickness of 165 μm, for a
ture-based stress measurements of coatings produced with different total diameter increase of 330 μm. No grinding or polishing of the coat-
levels of process gas flow, oxygen to fuel ratio, stand-off distance, and ing after deposition was done in order to not alter the roughness of the
powder feed rate were performed [23]. The DJ-A and DJ-B conditions coating after deposition.
were selected for use based on the large differences in formation and re- WC-CoCr coating stripping was performed using a solution of Potas-
sidual stresses between the coatings sprayed with these conditions. sium Sodium Tartrate (Rochelle Salt) and Sodium Carbonate in distilled
Coatings for the investigation of deposition condition influence on fa- water at concentrations of 50 g/l and 150 g/l, respectively. A D.C. voltage
tigue life were produced with a JP-5220 HVOF torch system (Praxair of 3–4 V was applied to the sample and an immersed steel electrode,
Surface Technologies, Indianapolis, IN) with commercially available producing a current of 1–1.3 A. The solution was continuously stirred

Table 1
Torch parameters of the conditions used for depositing the DJ-A, DJ-B, and LT/HT coatings along with measured average particle temperature and velocity using accuraspray.

Condition Kerosene flow Hydrogen flow Air flow Oxygen flow Stand off distance Powder Feed rate Practice Practice velocity
[l/h] [slpm] [slpm] [slpm] [mm] [g/min] temperature [°C] [m/s]

DJ-A – 636 334 188 254 JK 120H 30 1849 609

DJ-B – 731 408 276 229 JK 120H 30 1918 731
LT, HT 24 – – 940 330 WOKA 65 1827 811
3652
410 A. Vackel, S. Sampath / Surface & Coatings Technology 315 (2017) 408–416

curvature measurement was used to monitor the deposition of WC-

CoCr. Fig. 2(a) shows the raw curvature-time data of conditions DJ-A
and DJ-B sprayed onto low carbon steel planar beams, with deposition
and cooling regimes noted, to a thickness of 230 μm and 200 μm, respec-
tively. A clear difference in the curvature evolution between the two
samples is seen, with positive curvature accumulated during deposition
with coating DJ-A and a slightly negative curvature accumulated for DJ-
B. This indicates the DJ-A coating having a tensile deposition stress, as
Figure 1. Rotating bend fatigue specimen nominal dimensions, with portion coated by calculated by the Stoney formula [32] inset in Fig. 2(a). In contrast, the
WC-CoCr indicated. DJ-B coating has a compressive deposition stress which indicates that
the peening effect of high kinetic energy particles has overcome the ten-
and keep ~ 60 °C during the stripping process, with periodic light sile quenching stresses occurring during coating deposition. Average
brushing of the coated area to expedite the stripping process. After coat- backside substrate temperature during deposition was measured to be
ing removal, the RBF samples were washed with distilled water and approximately 190 °C for DJ-A and 260 °C for DJ-B, contributing to a
dried prior to fatigue testing. Similar procedures have been used to ef- slightly higher thermal stress contribution in DJ-B. Final residual stress
fectively strip WC-CoCr coatings [29,30]. in the coating is a combined effect of process and thermal stresses
Fatigue testing was performed using a RBF-200 model rotating beam shown in Fig. 2(b).
fatigue testing machine (Fatigue Dynamics Inc., Dearborn MI, now Sys- In the second set of experiments a single spray torch parameter was
tem Integrators LLC, Glendale AZ). Samples were tested at a frequency used but two sets of coatings were processed to impart different ther-
of ~100 Hz at varying applied moments until failure or severe deforma- mal stresses via changing substrate temperatures. This was achieved
tion triggered the motor's cutoff switch while the cycle counter record- by changing the pause time, and hence cooling time, between coating
ed the number of cycles to failure. It should be noted that in order to deposition layers using robotic programming, resulting in two coating
assume a stress amplitude value for the surface of the RBF specimens, types; nominally Low Temperature (LT) and High Temperature (HT).
the increase in specimen diameter due to coating thickness is treated Both the LT and HT coatings accumulated similar positive curvature
as if the coating is identical to the substrate (i.e., coating modulus = (and hence tensile deposition stress), though at different rates with re-
205 GPa), which is not an ideal assumption [8–10]. Micrographs of fa- spect to time, shown in Fig. 3(a), with a thickness of 215 μm and 225 μm,
tigue fracture surfaces were taken using either an optical stereoscope respectively. The average back side substrate temperature stabilized at
or SEM in backscatter mode. approximately 350 °C for the High Temperature (HT) coating and
Vickers hardness of the coatings sprayed on planar substrates was 200 °C for the Low Temperature (LT). As such, a larger compressive
measured on mounted and polished cross sections at a load of 300 g for thermal stress is produced in the HT coating by the higher deposition
15 s. Indentation modulus measurements were taken on the top surfaces temperatures. Both the LT and HT coating had a similar deposition stress
of polished coatings using an instrumented indenter with a Berkovich tip value, making the residual stress primarily different by the thermal
(Micro-Materials Limited, Wrexham Technology Park, Wrexham, UK). stress component. Calculated values for the deposition, thermal, and
Elastic recovery of the coating upon unloading was used to calculate the combined residual stress calculated from the beam curvature measure-
indentation modulus, as described by Oliver and Pharr [31]. ments for the LT and HT coatings are shown in Fig. 3(b), using the Ston-
ey formula.
3. Results
3.2. Coating microstructure and properties
3.1. Quantification of residual stresses
Micrographs of the four coatings are shown in Fig. 4. The torch pa-
In order to quantify the differences in process induced residual stress rameter difference between DJ-A and DJ-B makes a noticeable differ-
that can be achieved by torch and deposition conditions, beam ence in the size and frequency of coating porosity, with DJ-B showing

Fig. 2. (a) Raw curvature-time data for the DJ-A and DJ-B conditions sprayed onto low carbon steel beams. Undulations during deposition represent the torch rastering over the beam, with
positive curvature accumulation for DJ-A and negative curvature accumulation for DJ-B. In-set shows the Stoney formula and appropriate variables for stress calculations. (b) Calculated
deposition, thermal, and residual stress values from beam curvature of the WC-CoCr coatings. Note that the deposition and thermal stress for DJ-A were nearly the same, making a near-
zero residual stress in the coating.
 A. Vackel, S. Sampath / Surface & Coatings Technology 315 (2017) 408–416 411

Fig. 3. (a) Smoothed curvature-time data for the High Temperature (HT) and Low Temperature (LT) coating conditions sprayed onto low carbon steel beams. Note that similar net
curvature change occurs for the deposition of the two coatings, though they occur over different time scales due to the robotically controlled pause time between coating deposition
passes. A large difference in curvature change between the two coatings is due to the different substrate temperatures during spraying; a direct result of the pause time between
deposition passes as well. (b) Calculated deposition, thermal, and residual stresses of the HT and LT coatings. A similar deposition stress but largely different thermal stresses summate
to a large residual stress difference in the coatings, with HT being in compression and LT being nearly neutral.

a denser microstructure than DJ-A. The LT and HT coatings show a sim- greater hardness and indentation modulus, though the difference is
ilar porosity character relative to each other. within the standard deviation of the measurements. This points to the
Hardness measurements of the DJ-A and DJ-B coatings revealed a time between coating deposition passes having a much smaller effect
large difference between the two coatings, with measured values of on the HT and LT coating mechanical properties (if any) as compared
1170 and 1330 kgf/mm2, respectively. It is clear to see that the differ- to the influence particle state can have, as is the case for the DJ-A and
ence in torch operating parameters and thus in-flight particle state DJ-B coatings.
had a significant influence on coating hardness for DJ-A and DJ-B, in
agreement with literature findings [5,23]. Detailed property and perfor- 3.3. Torch parameter influence on fatigue life
mance measurements of DJ-A and DJ-B have been previously published
as coatings D1 and D9, respectively, in ref. [23]. To assess the mechanical The DJ-A and DJ-B conditions were applied to the steel RBF speci-
property influence from deposition parameters, hardness and indenta- mens and were fatigue tested until failure. In order to avoid the com-
tion moduli of the HT and LT WC-CoCr coatings were measured and plexities in fatigue behavior arising from surface compressive residual
are shown in Fig. 5. Here it can be seen that only a slight difference in stress, roughness, and grain distortion that can be induced by grit
coating properties can be observed with the HT coating having the blasting, the coatings were applied without prior grit blasting of the

Fig. 4. Back scatter SEM images of the coating microstructures cut from the steel beams used to measure stress. Comparison of the DJ samples shows larger and more frequent porosity in
DJ-A than in DJ-B. LT and HT show similar porosity character, differing only in substrate temperature.
412 A. Vackel, S. Sampath / Surface & Coatings Technology 315 (2017) 408–416

respectively. It is clear to see that there is a large difference between

the coated samples' fatigue life depending on which coating is applied
to the steel and if grit blasting was used.
In the cases of substrates that were not grit blasted, the HT coated
samples (S + HT) produced a large increase in fatigue life vs. the uncoated
steel. The LT coated samples yielded a fatigue debit relative to the uncoat-
ed steel, indicating a high sensitivity of the coating application and surface
deposition temperature on expected specimen fatigue life.
From Fig. 7(b), it is evident that grit blasting slightly increased the fa-
tigue life relative to the annealed steel samples, which can be attributed
to the compressive stresses and grain distortion induced into the surface
of the sample from the blasting process. The samples with the LT coating
applied to grit blasted substrates did not show significant fatigue life dif-
ference from the grit blasted only samples, indicating that the grit
blasting was able to improve the fatigue life of the LT coating and can-
celled the negative effect of the LT coating in the non-grit blasted sam-
Fig. 5. Indentation modulus and cross section hardness of the HT and LT WC-CoCr coatings ples. The samples with grit blasting and the HT coating showed an
on planar beams. increase in fatigue life relative to the grit blasted only specimens, with
a higher fatigue life than the HT coated samples that were not grit blast-
substrate. No coating delamination was observed during fatigue testing ed, further improving fatigue life. Thus it can be said that grit blasting
until final failure. Substrate temperature was measured during deposi- enhanced the fatigue life of the coated and non-coated specimens in
tion by IR camera to reach 215 °C for the DJ-A coating and 325 °C for all cases relative to the non-grit blasted counterparts.
the DJ-B coating application. The stress vs. number of cycles to failure
curve (S-N) for steel (S), DJ-A coated, and DJ-B coated samples are 3.5. Fractographic observation of fatigue failures
shown in Fig. 6. It can be seen that the DJ-B coating adds a significant en-
hancement to the fatigue life of the coated specimen, while the DJ-A A typical fracture surface of the non-grit blasted and coated samples is
coating has a slight debit to the fatigue life compared to the uncoated shown in Fig. 8(a), noted within is the fatigue crack initiation and final
specimens. fracture site. Coating delamination was observed on many specimens ad-
jacent to the final fracture site, though no adhesive coating failure was ob-
3.4. Thermal stress influence on fatigue life served adjacent to fatigue crack initiation sites, confirming an adequate
coating bond strength during fatigue testing, even without grit blasting.
As with the previous example, a set of RBF samples were coated Fig. 8(b) shows a more detailed image of the crack initiation site, where
without grit blasting using the LT and HT conditions under a single cohesive coating failure was frequently found in the adjacent area.
torch parameter condition. As noted earlier, these two conditions im- Fracture surfaces of the sample grit blasted prior to LT WC-CoCr coat-
pose different levels of net residual stress in the coating without a signif- ing deposition are shown in Fig. 9(a). The coating shows a severe amount
icant change to the microstructure, Vickers hardness, or indentation of in-plane cracks, as well as complete cohesive delamination adjacent to
modulus. No coating delamination was observed while spraying onto the fatigue crack initiation site in the substrate. The fracture surface of a
the substrates or during fatigue testing, which confirmed adequate coating on a non-grit blasted substrate is shown in Fig. 9(b), with an
bonding. A set of specimens where the steel was grit blasted prior to inset highlighting a crack that passed through the coating and substrate.
coating deposition was also tested with the same coatings. The S-N for Cohesive cracking within the coating is again visible, yet this instance
the bare substrate, LT coated, and HT coated samples, without and shows a crack path passing through the substrate and into the coating.
with grit blasting prior to spraying, are shown in Fig. 7(a) and (b),
3.6. Partial fatiguing

In order to assess how the coating and substrate deteriorate during

fatigue testing, coated samples from the LT and HT set were fatigue test-
ed for portions of the expected fatigue life and then halted from testing
prior to fracture. The WC-CoCr coating was then removed by the strip-
ping processes described in the experimental methods and resumed fa-
tigue testing until failure as a bare substrate. No grit blasting of the
substrates was done prior to coating application in these cases. With
the LT and HT coated samples requiring very different stress amplitudes
to break within the high cycle fatigue regime, a direct comparison of
pre-fatiguing was not possible. Instead the stress amplitude of the
stripped specimens was held at ~ 250 MPa for comparison of damage
from the pre-fatiguing. The results of this testing are shown in Fig.
10(a) for a sample with the LT coating and Fig. 10(b) for a sample
with the HT coating, both with the relevant data from Fig. 7 shown in
grey. Sample a, with the LT WC-CoCr coating applied, was fatigued to
approximately 50% of the expected linearly measured fatigue life at
that stress. After pausing fatigue testing and stripping the coating off,
sample a’ was fatigued at the same load until failure, falling short of
the expected fatigue life for the uncoated steel samples, indicating sig-
Fig. 6. Stress amplitude vs. cycles to failure for 1018 steel (S) with DJ-A and DJ-B WC-CoCr
coatings applied (Referred to as S + DJA and S + DJB, respectively, in order to avoid
nificant damage to the substrate had occurred during prior fatigue test-
confusion). The DJ-B coating offers significant improvement to the specimen fatigue life, ing while the coating was present. Sample b with the HT WC-CoCr
whereas the DJ-A coating produces a slight deficit. Arrows indicate runout. coating applied was fatigue tested to the similar portion of expected
 A. Vackel, S. Sampath / Surface & Coatings Technology 315 (2017) 408–416 413

Fig. 7. (a) Stress amplitude vs. cycles to failure for 1018 steel (S) with Low and High Temperature WC-CoCr coated steel (S + LT and S + HT, respectively). (b) Stress amplitude vs. cycles to
failure of RF specimens that were grit blasted (S + GB) with and without LT and HT coating deposition. Arrows indicate runout. The HT coating shows fatigue improvement in all cases,
whereas the LT coating either offers a deficit or minimal change from the substrate behavior.

fatigue life as sample a (50% linear), though at a higher load and longer coating, stripped, and fatigue tested until failure. No difference in fatigue
time due to the HT coated sample's prolonged life. After pausing and life was observed between the bare steel substrates (S) and samples
stripping the coating off, sample b’ was fatigue tested to failure, reaching that underwent coating, stripping, and fatigue testing.
a life equivalent to that of the baseline annealed 1018 steel material,
suggesting that no significant damage accumulated within the substrate 4. Discussion
during the prior fatigue testing.
In order to confirm that the spray process itself did not alter the The results presented above points to the importance of process in-
substrate's fatigue life, specimens were sprayed with the HT and LT duced residual stress states of coatings on the fatigue response of a WC-
CoCr coated steel system. In one instance, the stress states were modu-
lated through changing particle kinetic and thermal energies via process
conditions (DJ-A and DJ-B coatings). In this instance, the rotating bend
fatigue data suggests that coatings processed with compressive deposi-
tion stress through high particle kinetic energy resulted in fatigue credit
of the system while coatings sprayed under tensile evolving stresses
showed a slight decrease in specimen fatigue life. The higher peening
stress associated with DJ-B also aided in the densification and hardness
of the coating, which can also be contributing factors to an enhanced fa-
tigue life. By changing deposition temperatures for constant spray con-
ditions, different levels of thermal stresses can be imposed onto the
coating-substrate system. This was demonstrated through the analysis
of the LT and HT coatings. Here the coating densities, hardness, and in-
dentation modulus were similar but the final stress states of the coated
samples were significantly different.
Though direct comparison of coating residual stress as measured by
beam curvature cannot be directly translated to cylindrical geometry,
calculation of deposition, thermal, and residual stress in coatings on
both flat and cylindrical geometries have been established by Tsui and
Clyne [25,33]. Fig. 11 shows the magnitudes of the expected longitudi-
nal thermal stress component to the residual stress for the LT and HT
coatings, calculated using the equations described in [33] and a deposi-
tion temperature of 50 °C and 150 °C for the LT and HT coatings, respec-
tively. A coating modulus of 1.5 times that of steel (~300 GPa) is used to
demonstrate an estimate of the WC-CoCr coated steel system, with
lower coating moduli (1 and 0.5 times that of steel) also shown to dem-
onstrate how a coating with lower stiffness would influence the thermal
stress magnitude. Depending on the modulus of the coating after depo-
sition, the magnitude of the stress imposed in both the coating and sub-
strate can vary, with a stiffer coating producing higher stresses. This
demonstrates the high sensitivity of the coating's stress state upon the
thermal stress, with a significantly lower stress induced in the substrate
as compared to the coating for this scenario. It is important to note that a
coating material with a different CTE and mechanical properties would
result in different expected coating-substrate stress profiles, such as
Fig. 8. (a) Light micrograph of typical fatigue fracture surface with fatigue crack initiation,
final fracture site, and coating de-bonding from final fracture indicate. SEM of the
other common wear resistant thermal spray coatings like Cr3C2-NiCr
highlighted region of initiation site shown in (b) with cohesive coating failure and which has a more closely matching CTE to steel and therefore a less
fatigue crack initiation site. Both images are of a single S + HT sample. compressive thermal stress, potentially leading to differences in fatigue
414 A. Vackel, S. Sampath / Surface & Coatings Technology 315 (2017) 408–416

Fig. 9. SEM images of fatigue fracture surface of coated samples (a) grit blasted prior to coating deposition, with fatigue crack initiation site and cohesive coating cracks indicated and (b)
non-grit blasted prior to coating deposition with similar cohesive coating cracks, as well as a substrate crack. Inset image highlights the observed crack passing through the coating-
substrate interface, though not a result of crack driving into the substrate from the coating. Images are those sprayed with the LT condition.

life. Calculation of an expected deposition stress contribution using the specimen then allows for calculation of the local stresses throughout
Tsui-Clyne model for the RBF specimens cannot be assumed to be the the specimen
same as the planar specimens due to differences in relative surface
speed, expected angular particle impact, and different substrate surface σ s ¼ Es KR ð2Þ
temperatures experienced by depositing splats. However, calculations
indicate approximately a 10 to 1 ratio for the magnitude of deposition σ c ¼ Ec KR ð3Þ
stresses experienced by the coating and substrate, respectively, within
the dimensions of these RBF samples. This again underlines the critical Where R is the radius throughout the specimen. Fig. 12 shows the
importance of processing on coating residual stress. calculated stress profile under a single moment using these equations
The resulting stress profile within the substrate and coating upon and the same relative values for the coating modulus as in Fig. 11. It is
application of a bending moment to the RBF sample is a superposition clear to see that coating addition lowers the mean stress amplitude
of both the residual stress and the applied stresses, determining the within the substrate by the coating accommodating a portion of the
local stress intensity that can initiate and drive fatigue cracks. The mag- load, depending on the modulus of the coating. However, there may
nitude of the applied stress throughout the specimen will depend on the also be some stress concentrations within the substrate due to non-ho-
coating's thickness and modulus, altering the experienced stresses mogeneity (cracks, poor adhesion) within the coating. It is also impor-
within the substrate as compared to an uncoated sample under the tant to note that the load carrying capability of the coating will likely
same load. Specimen curvature calculated using statics equations of be lower under a tensile stress compared to a compressive stress due
pure bending of a coated cylinder (an assumption for the thinnest diam- to the large amount of microstructural defects within the coatings
eter of the RBF specimen) take into account the moduli difference and (pores, micro-cracks, inter-lamellar boundaries etc.). This will also
dimensions of the coating and substrate under a linear elastic assump- play a role in crack initiation and propagation, where tensile fatigue
tion as such stresses typically drive crack growth. Finally, as the stiffness of a coating
decreases through damage accumulation, such as thinning from wear,
64 M
K¼   ð1Þ corrosion, or fatigue cracking, more load is distributed back to the sub-
4
Es πDs þ Ec π Dsþc 4 −Ds 4 strate and thus raises both the mean and local stress within the sub-
strate [34].
Where K is the curvature of the specimen, M is the applied moment, The expected residual and applied stress profiles and coating prop-
E is the modulus of the coating or substrate, and D is the diameter of the erties can now be utilized to explain the fatigue performance of these
substrate or substrate with coating. Calculating the curvature of the coated specimens. As evident for samples b’ in Fig. 10(b) testing a

Fig. 10. Stress amplitude vs. cycles to failure for coated samples that were partially fatigue tested, stripped, and tested as uncoated until failure for a sample with (a) LT coating and (b) HT
coating. Related samples from Figure 3 are included in grey as reference.
 A. Vackel, S. Sampath / Surface & Coatings Technology 315 (2017) 408–416 415

experienced by the substrate as well. It would be expected that the

DJ-B and DJ-A coatings would behave in a similar manner, with the
DJ-B coating carrying more load through its denser microstructure and
experiencing a slower rate of damage accumulation from its more com-
pressive residual stress state and higher hardness as compared to the
more porous, more tensile, and less hard DJ-A coating. Thus it is the
load carrying ability and damage accumulation rate of a coating that de-
termine fatigue life enhancement or debit. Additional residual stresses
due to grit blasting or shot peening would also superimpose to the stress
profile, offering a benefit to the substrate and thus to the coated
specimen's fatigue life as seen in Fig. 7. These stresses produced via
grit blasting are largely left unaltered by the coating process heat,
which is evident by the samples that were coated and stripped prior
to fatigue testing. It is also suggested from the stripped coating samples
without any grit blasting that the high kinetic energy particles that ini-
tially impact the substrate surface do not impart a significant fatigue life
benefit like grit blasting or shot peening. Thus it can be said that the
state of the substrate is largely left unaltered by the coating processes
Fig. 11. Calculated longitudinal thermal stresses of a cylinder using the Tsui-Clyne model in this scenario. This may not be the case for more thermally sensitive
[33] for the Low Temperature (LT) and High Temperature (HT) coatings with coating or softer substrates, such as tempered aluminum or copper for example.
modulus at 0.5, 1.0, and 1.5 times that of the steel substrate. Deposition temperatures of Experimental studies, as well as modelling [35], have reported on
50 °C and 150 °C were used for the LT and HT coatings, respectively, and ambient
temperature of 25 °C Substrate: E = 205 GPa, ν = 0.29, Radius = 3.12 mm, α =
the behavior of crack advancement through two well bonded materials
12 μm/m °C Coating: ν = 0.23, Thickness = 0.165 mm, α = 6 μm/m °C of the same modulus, with crack advancement in the softer material oc-
curring when the crack is initiated in the harder material, and crack
blunting when an initiated crack within the softer material approaches
significant portion of the expected HT coated sample's fatigue life did the interface of a harder material. In the case of a WC-CoCr coating per-
not accumulate significant damage within the substrate, indicating fectly bonded to a 1018 steel substrate, it is possible that crack advance-
that the stiffness of the coating adequately reduced the stress amplitude ment from the coating into the substrate can occur. However, the
experienced by the substrate, delaying damage accumulation. The weaker TS coating-substrate interface, which relies on mechanical
higher hardness and more compressive residual stress in the coating, interlocking and the anisotropic nature of thermal spray coatings inher-
such as the case for DJ-B and HT, contributes to the integrity of the coat- ent in their deposition method [36] offers different crack pathways than
ing being maintained for a longer period of fatigue testing and delays the through-thickness direction, producing the in plane and interfacial
substrate damage. The opposite case was observed with the LT coating branching directions the coating crack takes in Fig. 9(b). The large
in samples a and a’, with the higher amount of coating tensile stress amount of in-plane (cohesive) coating cracks observed on the fracture
not being able to maintain the coating integrity, with the possibility of surfaces of the majority of the RBF samples agree with this preferred
a lower coating stiffness not adequately alleviating the stress crack pathway, with primary fatigue crack initiation occurring in the
substrate. It can be hypothesized that the coating can act to delay fatigue
damage initiation and propagation within the substrate if adequate
stress reduction is supplied to the substrate. The hardness and residual
stress of the coating determines its ability to resist damage due to cyclic
loading. However, the presence and growth of the cracks within the
coating and their orientation do not offer a pathway for crack propaga-
tion directly into the substrate. Instead, the failure of the specimens is
due fatigue crack initiation and growth within the substrate.

5. Conclusion

Manipulation of residual stresses and coating properties in TS WC-

CoCr deposited by HVOF was achieved and quantified on planar steel
samples through torch operating parameters and deposition tempera-
ture. In-situ beam curvature measurements were used to monitor
both the stress evolution during layer by layer coating deposition and
during the post-deposition cooling sequence. Steel rotating bend fatigue
specimens were tested with and without coatings applied following the
torch and substrate condition variation as described above. It was found
that coatings with higher compressive residual stress, hardness, and mi-
crostructural density enhanced fatigue life of the coated specimens.
Coatings with tensile residual stress and somewhat weaker properties
(hardness/stiffness) incurred a fatigue debit in the system.
Coated specimens that were fatigue tested for a portion of their ex-
pected fatigue life, stripped of their coating, and retested as bare sub-
strates provided insights as to the coating quality on the specimen's
fatigue behavior. Samples with tensile coating stresses and lower hard-
Fig. 12. Calculated applied stress profile from pure bending moment (6.78 Nm) of a
ness weakened the system resulting in fatigue deficit of the substrate
cylinder with coating modulus at 0.5, 1.0, and 1.5 times that of the steel substrate. while coatings with strong compressive residual stress and better pro-
Dimensions used were the same as in Fig. 11. cess induced deposit integrity showed fatigue credit to the underlying
416 A. Vackel, S. Sampath / Surface & Coatings Technology 315 (2017) 408–416

steel. Fracture surface analysis revealed fatigue crack initiation within [12] R. McGrann, D. Greving, J. Shadley, E. Rybicki, T. Kruecke, B. Bodger, The effect of
coating residual stress on the fatigue life of thermal spray-coated steel and alumi-
substrates, with the adjacent coating containing considerable cohesive num, Surf. Coat. Technol. 108 (1998) 59–64.
damage. Analytical calculations of the longitudinal thermal stress differ- [13] M. Costa, M. Venditti, H.J.C. Voorwald, M. Cioffi, T. Cruz, Effect of WC–10% Co–4% Cr
ences between the LT and HT coatings assuming cylindrical geometry coating on the Ti–6Al–4V alloy fatigue strength, Mater. Sci. Eng. A 507 (2009)
29–36.
were made using the Tsui-Clyne model, revealing the considerable in- [14] Y. Itoh, M. Saitoh, K. Takaki, K. Fujiyama, Effect of high-temperature protective coat-
fluence of expected differences in the residual stress profile within the ings on fatigue lives of nickel-based superalloys, Fatigue Fract. Eng. Mater. Struct. 24
coating and substrate. Statics calculations for pure bending of a coated (2001) 843–854.
[15] A. Valarezo, W.B. Choi, W. Chi, A. Gouldstone, S. Sampath, Process control and char-
cylinder were also made in order to demonstrate the expected load acterization of NiCr coatings by HVOF-DJ2700 system: a process map approach, J.
bearing capacity of the coating and subsequent stress reduction within Therm. Spray Technol. 19 (2010) 852–865.
the substrate. Finally, a proposed mechanism for failure in these sam- [16] J.A.M. de Camargo, H.J. Cornelis, V.M.O.H. Cioffi, M.Y.P. Costa, Coating residual stress
effects on fatigue performance of 7050-T7451 aluminum alloy, Surf. Coat. Technol.
ples expects that the residual stress and coating integrity resulting
201 (2007) 9448–9455.
from processing are key factors in determining the fatigue response of [17] A. Vackel, G. Dwivedi, S. Sampath, Structurally integrated, damage-tolerant, thermal
the system. This study thus provides a frame work for process design spray coatings, JOM 67 (2015) 1540–1553.
for optimal performance of thermal spray coated components subjected [18] R. Ahmed, M. Hadfield, Mechanisms of fatigue failure in thermal spray coatings, J.
Therm. Spray Technol. 11 (2002) 333–349.
to fatigue loading. [19] S. Kikuchi, Y. Nakahara, J. Komotori, Fatigue properties of gas nitrided austenitic
stainless steel pre-treated with fine particle peening, Int. J. Fatigue 32 (2010)
Acknowledgments 403–410.
[20] S. Hassani-Gangaraj, A. Moridi, M. Guagliano, A. Ghidini, M. Boniardi, The effect of
nitriding, severe shot peening and their combination on the fatigue behavior and
The authors would like to acknowledge the financial support to the micro-structure of a low-alloy steel, Int. J. Fatigue 62 (2014) 67–76.
Center through the Consortium for Thermal Spray Technology and the [21] P. Paris, F. Erdogan, A critical analysis of crack propagation laws, J. Fluids Eng. 85
(1963) 528–533.
financial support of Messier-Bugatti-Dowty Co. for A.Vackel's Ph.D. [22] K. Shinoda, J. Colmenares-Angulo, A. Valarezo, S. Sampath, Effect of deposition rate
work on the general subject of HVOF sprayed hard coatings. Special on the stress evolution of plasma-sprayed yttria-stabilized zirconia, J. Therm.
thanks also goes to Gregory Smith and Albert Marciniec for their assis- Spray Technol. 21 (2012) 1224–1233.
[23] A. Vackel, G. Dwivedi, S. Sampath, Influence of in-flight particle states on the particle
tance with sample spraying and fatigue testing.
peening intensity and related properties of HVOF WC-CoCr coatings, Thermal Spray
Sandia National Laboratories is a multi-mission laboratory managed Bulletin 8 (2015) 32–41.
and operated by Sandia Corporation, a wholly owned subsidiary of [24] S. Kuroda, T. Clyne, The quenching stress in thermally sprayed coatings, Thin Solid
Films 200 (1991) 49–66.
Lockheed Martin Corporation, for the U.S. Department of Energy's Na-
[25] Y. Tsui, T. Clyne, An analytical model for predicting residual stresses in progressively
tional Nuclear Security Administration under contract DE-AC04- deposited coatings part 1: planar geometry, Thin Solid Films 306 (1997) 23–33.
94AL85000. [26] J. Matejicek, S. Sampath, In situ measurement of residual stresses and elastic moduli
in thermal sprayed coatings: part 1: apparatus and analysis, Acta Mater. 51 (2003)
863–872.
References [27] O. Kesler, J. Matejicek, S. Sampath, S. Suresh, T. Gnaeupel-Herold, P. Brand, H. Prask,
Measurement of residual stress in plasma-sprayed metallic, ceramic and composite
[1] A. Ghabchi, T. Varis, E. Turunen, T. Suhonen, X. Liu, S.-P. Hannula, Behavior of HVOF coatings, Mater. Sci. Eng. A 257 (1998) 215–224.
WC-10Co4Cr coatings with different carbide size in fine and coarse particle abra- [28] T. Suhonen, T. Varis, S. Dosta, M. Torrell, J. Guilemany, Residual stress development
sion, J. Therm. Spray Technol. 19 (2010) 368–377. in cold sprayed Al, Cu and Ti coatings, Acta Mater. 61 (2013) 6329–6337.
[2] J. Kawakita, S. Kuroda, T. Kodama, Evaluation of through-porosity of HVOF sprayed [29] B.D. Sartwell, K.O. Legg, J. Schell, J. Sauer, P. Natishan, Validation of HVOF WC/Co
coating, Surf. Coat. Technol. 166 (2003) 17–23. Thermal Spray Coatings as a Replacement for Hard Chrome Plating on Aircraft Land-
[3] A. Valarezo, S. Sampath, An integrated assessment of process-microstructure-prop- ing Gear(DTIC Document) Naval Research Lab, Washington DC, United States, 2004.
erty relationships for thermal-sprayed NiCr coatings, J. Therm. Spray Technol. 20 [30] R. Menini, N.B. Salah, R. Nciri, Stripping methods studies for HVOF WC-10Co-4Cr
(2011) 1244–1258. coating removal, J. Mater. Eng. Perform. 13 (2004) 185–194.
[4] S. Kuroda, Y. Tashiro, H. Yumoto, S. Taira, H. Fukanuma, S. Tobe, Peening action and [31] W.C. Oliver, G.M. Pharr, An improved technique for determining hardness and elas-
residual stresses in high-velocity oxygen fuel thermal spraying of 316L stainless tic modulus using load and displacement sensing indentation experiments, J. Mater.
steel, J. Therm. Spray Technol. 10 (2001) 367–374. Res. 7 (1992) 1564–1583.
[5] T. Varis, T. Suhonen, A. Ghabchi, A. Valarezo, S. Sampath, X. Liu, S.-P. Hannula, For- [32] G.G. Stoney, The tension of metallic films deposited by electrolysis, Proceedings of
mation mechanisms, structure, and properties of HVOF-sprayed WC-CoCr coatings: the Royal Society of London, Series A, Containing Papers of a Mathematical and
an approach toward process maps, J. Therm. Spray Technol. 23 (2014) 1009–1018. Physical Character, 82, 1909, pp. 172–175.
[6] D. Toma, W. Brandl, G. Marginean, Wear and corrosion behaviour of thermally [33] Y. Tsui, T. Clyne, An analytical model for predicting residual stresses in progressively
sprayed cermet coatings, Surf. Coat. Technol. 138 (2001) 149–158. deposited coatings part 2: cylindrical geometry, Thin Solid Films 306 (1997) 34–51.
[7] O. Kovářík, J. Siegl, Z. Procházka, Fatigue behavior of bodies with thermally sprayed [34] A. Vackel, T. Nakamura, S. Sampath, Mechanical behavior of spray-coated metallic
metallic and ceramic deposits, J. Therm. Spray Technol. 17 (2008) 525–532. laminates, J. Therm. Spray Technol. 25 (2016) 1009–1019.
[8] A. Ibrahim, C. Berndt, The effect of high-velocity oxygen fuel, thermally sprayed [35] Y. Sugimura, P. Lim, C. Shih, S. Suresh, Fracture normal to a bimaterial interface: ef-
WC–Co coatings on the high-cycle fatigue of aluminium alloy and steel, J. Mater. fects of plasticity on crack-tip shielding and amplification, Acta Metall. Mater. 43
Sci. 33 (1998) 3095–3100. (1995) 1157–1169.
[9] A. Ibrahim, C. Berndt, Fatigue and deformation of HVOF sprayed WC–Co coatings [36] S. Usmani, S. Sampath, D.L. Houck, D. Lee, Effect of carbide grain size on the sliding
and hard chrome plating, Mater. Sci. Eng. A 456 (2007) 114–119. and abrasive wear behavior of thermally sprayed WC-Co coatings, Tribol. Trans. 40
[10] A. Ibrahim, R. Lima, C. Berndt, B. Marple, Fatigue and mechanical properties of nano- (1997) 470–478.
structured and conventional titania (TiO2) thermal spray coatings, Surf. Coat.
Technol. 201 (2007) 7589–7596.
[11] H. Voorwald, R. Souza, W. Pigatin, M. Cioffi, Evaluation of WC–17Co and WC–10Co–
4Cr thermal spray coatings by HVOF on the fatigue and corrosion strength of AISI
4340 steel, Surf. Coat. Technol. 190 (2005) 155–164.