WEAR

ELSEVIER Wear 213 ( It)t)71 185-19t~

Organic coating removal by particle impact

M. Papini, J.K. Spelt *
Uaiv~'r.~ity ~I Toroato. D~Tulrtment r~l M~'cham~at aml hJ~htstrial Erjgitleerit~g. 5 King "x College Road. Toronto. Ontario. Canada MSS .'IGB
Reccivctl i i September It.lOft;accepted 2(I March 1997

Abstract

The collision of individual glas.,, heads wilh steel sa,l~ple.,,,'t~ated with an alkyd paint was examined in the context of coating removal using
the blasl cleaning process. It was fi~undthal coatiqg removal was best modeled "ts.,~umingthe development of a critical shear stress rather than
the creation of .stress wave.,,. For this particular .~y~tcm,only the normal component of incident velocity correlated with coating removal, and
a critical velocity existed where the coatin~ could he removed without damaging the suhstrate. A rneth~xlis proposed and tested for determining
this velocity, hasud tm a calculation t~f the veh~city needed for the glass bead to just penetrate the coating. © 1997 Elsevier Science S.A.

Kevwords: C~iltings:('olli.,ii~n.~:Blast cleaning:Impact: I)clamination:Paintrcmo,,al

1. Introduction
Organic coatings are often removed using solvents that
create waste which is difficult and expensive to dispose of.
For this reason, mechanical coating removal techniques such
as blast cleaning are of increasing interest, in the blast clean-
ing process, a stream of particles is directed towards the
coated substrate, and the coating is removed by mechanical
means. The fundamental mechanisms of coating, removal.
however, are not well understood, and there have been very
few attempts at modeling the process. As a result, blast clean-
ing is. for the most part, dominated by trial-and-error and
rules of thumb.
Although coating removal by blast cleaning may be
affected by multiparticle interactions, the spread of the par-
ticle streams, and the distribution of particle sizes and shapes,
a single-particle model serves as a logical iirst step to under-
standing the fundamentals of coating removal, This paper
attempts to explain, through experiment and theory, the fun-
damental mechanism of coating removal tbr single, spherical
glass particles impacting an alkyd-coated steel subslrate.
2. Literature review
Previous research related to blast cleaning is mostly of an
empirical nature and very specilic to particular systems I I-
3 i. Other useful literature can be found in the contact mechan,
* Corresponding author.
0043-1648197/$ I"L(X}~ It)97 ElsevierScienceS.A. All rights reserved
PI! S00,~.3- I b48{ 97 ](10()fi2-8
ic,~ literature, and the field of solid particle erosion of bare
substrates. An excellent review of the solid panicle erosion
of coatings can be tbund in the paper by Shipway and Hutch-
ings 141. The paper also outlines a method for describing
coating erosion in terms of a critical dose of particles of a
particular size and shape needed to remove a coating. A book
by Johnson provides an excellent introduction to contact
mechanics, containing solutions forelastic collisions between
particles and substrates 15 ], and for static contact analysis of
elastic layers on semi-infinite substrates 16]. Ramamunhy el
al. [7.81 have modeled the removal of chips of paint on
automobiles due to stone impacts using shock physics and
finite element models of wave propagation. Brach has devel-
oped a very simple method for characterizing collisions of
single particles with uncoated substrates in terms of energy
losses, the coefficient of restitution, and the ratio of normal
to tangential impulse 19]. Another paper by Hatchings [ 10l
provides valuable insights into the different mechanisms of
solid particle erosion of substrates using both irregular and
spherical panicles. Matthewson [ I ! ! has developed an
asymptotic solution for the static indentation of an elastic
layer on a semi-infinite substrate that includes the calculation
of" intcrfacial shear stress. Rossikhin and Shitikova [ 12 ] have
recently developed a wave propagation model of the impact
of a rigid sphere with an elastic layer based on Pay methods.
Lu I 13 ! has provided valuable observations of indentation
cracks in coating-~ubstrate systems, and has proposed a
method to determine the crack paths, Dioh and Williams i 14 !
have studied the impact behavior of paints and developed
186 M. P¢+pinLJ.K. Spelt I Wear 213 f 1997t 185-1~
methods for the experimental determination of material prop-
erties under impact conditions.
in order to characterize the erosion of both coatings and
bare substrates, many investigators have designed various
types of erosion rigs and gas guns. For example, Hatchings
and Winter I 151 have designed a gas gun that uses a reservoir
lull of either helium or nitrogen, and is triggered by the
bursting of two diaphragms. The single particles are placed
in polythene sabots which are stopped at the end of the barrel,
yielding speeds of up to 600 m s - ~. The performance of this
gun under different operating pressures and fluids was eval-
uated and compared to theory. Graham et al. 1161 built an
erosion test facility based on exploding aluminum foils that
propel single particles in the size range 100--8000 p.m at
velocities ranging from 250 to 7000 m s - ' . Hatchings et al.
! 171 have designed a gun with a rectangular bore capable of
launching 0.5 g particles at speeds of 300 m s ', The gun
uses a solenoid valve to release a compressed gas into u barrel,
and the rect.'mgular bore ensures that the particles are
launched without rotation. A theoretical and experi-
mental study of the effect of valve opening times on the
performance of gas guns can be found in the paper by White
and Fowles 1181.
3. Experimental method
Collisions between single particles and bare substrates are
normally characterized in terms of,+ +ergy losses and coeffi-
cients of restitution. A similar approach ;vas utilized in this
study to build an experimental database for collisions
between small glass spheres and coated substrates.
To perform single-particle impact experiments, a gas-gun-
h,gh-speed photographic setup was constructed. A single
glass sphere was loaded into a cylindrical urethane sabot
(6.3 mm diameter and 10 mm length) which was, in turn.
loaded into a 6.3 mm inner diameter, 50 cm long steel barrel
via a breech. The barrel-breech was attached to a solenoid
valve, which was. in turn connected to a compressed air
cylinder. The fast+acting solenoid valve (Model 73216
BN2MT(X). Hon,.ywcll) permitted a burst of compressed air
to accelerate the urethane sabot and glass sphere to the end
of the barrel where a urethane ring stopped the sabot, but
allowed the glass sphere to exit. A serie~ of holes drilled at
the end of the barrel relieved the pressure behind the sabot as
it stopped, and minimized the amount of air exiting the end
of the barrel with the glass sphere. By regulating the pressure
in the range 140-69(X1 kPa, particle velocities between 2(I
and 130 m s t were obtained. It was found that the speed of
the particles exiting the barrel was independent of the size,
density and shape of the particles when tested with angular
and spherical glass, plastic, and steel in the diameter range
0.1-1 ram. This was because the mass of the particles was
very small compared to that of the sabot.
Two cameras ( black and white C¢'D cameras+ not special
field-on-demand) were used to obtain images in perpendic-
ular planes, thereby allowing a verification that the rebound
and incident velocity vectors were in a single plane. A true
color RGB frame grabber ( Occulus TCX. Coreco Inc.) was
used with the two cameras connected to two of the three
channels. The frame grabber continuously grabbed frames at
the rate of 60 Hz, continuously overwriting its frame buffer.
Because the collisions were expected to last on the order of
hundreds of microseconds, it was only necessary to have the
setup determine which of the frames was the frame of interest
(i.e. the frame in which the collision occurred), and then the
continuous grab could be terminated with the frame of interest
in the frame grabber buffer. An infrared trigger mounted at
the end of the barrel (V|S ll, Optikon Inc.) sensed tile par-
ticles as they left the barrel, and sent a signal to an l/O hoard
(Opto 22 PB8) attached to the computer and to a delay
generator. The I/O board was connected to the frame grabber
and the frame grab sequence was terminated upon triggering.
The trigger also caused the delay unit to generate pulses to
trigger four high-speed flashes mounted near the target at
adjustable delays. Because the cameras are light integrators.
summing all the light that reaches their sensor in one frame,
a trajectory of the particle (four images) appeared on the
grabbed frame. The delays were timed so that two images of
the particle were just before collision, and two just after
collision. By using image analysis software ( Imi~ge Pro Plus.
Media Cybernetics Inc. ), the precise distance between suc-
cessive particle images ( i.e. the distance the particle traveled
between flashes) was measured. This di,~tance divided
by the delay between flashes gave the incidt.'nt and rebound
velocities of the particle.
The model coating system was typical of a class of rela-
tively soft coatings on rigid substrates. The system consisted
of Valspar Inc. alkyd enamel of various thicknesses on Bon-
derite I000 iron phosphate pretreated, polished cold-roiled
steel ( 0.66 mm thick) obtained from ACT Laboratories. The
specimen holder consisted of a 9 mm thick piece of steel on
which the specimens ( 30 × 30 mm ) were clamped. The spec.
imen holder was pivoted in a clcvicc which allowed rotation
to any incident impact angle. The clevice0 in turn, was
clamped to a long steel rod. permitting adjustments in the
height of the target.
In order to investigate the effects of particle impacts on
this paint, the photographic setup described above was used
to determine the incident and rebound velocities of single
glass spheres (0,64ram diameter) launched, at various
angles of attack, against the coated specimens. Four paint
thicknesses were used ( 20. 25, 40and 50 Ixm ) and the normal
component of the incoming velocity was varied between I0
and 120 m s +~. Experiments were also done on the bare steel
substrate for comparison purposes. In all cases, the particles
remained undamaged after impact.
The Young's modulus and Poisson's ratio of the coating,
,,leel substrate and the glass were, respectively, 2470 MPa,
0.4:21)0 GPa, 0.3; and 70 GPa, O, 17. The coating was found
to have material properties strongly dependent on strain rule,
and therefore, the Young's modulus was determined using a
 M. Papim, J.K. Spelt/iig,ar 213 ~19971 IbtS-I~tt 187

time-of-flight ultrasonic method at 5 MHz [ t9l. it was rea- 4. Results a n d discussion

soned that the combination of a high measuring frequency
and low strain amplitude would produce an elastic (as 4.1. Obs('rvation.~ t!f c'~xtt'k path and impact site
opposed to viscoelastic) coating response approximating that
created by the very high strain rates of impact, The Poisson's Fig. I ( a ) shows a 3 3 m s ~ velocity (at normal inci-
ratio was not measured for this coating, but was assumed to dence ) impact site on the paint photographed with a scanning
be approximately 0.4, based on values for similar coatings electron microscope. The coating did not delaminate, but the
examined by Ramamunhy el al. 18]. The ma'~s of a single impact created an indentation that was similar to the plough-
panicle was measured on a microbalance as the average of ing or cratering observed by Hutchings in the collisions of
50 randomly selected particles and was 0.364 rag. The radius spheres with bare steel substrates [10l.
of the particle was also measured taking the average of 50 At higher velocities, the coating delaminated approxi-
measurements, and was found to be 0.32 ram. with a standard mate!y axisymmetrically. Fig. l ( b ) and ( c ) , for incident
deviation of approximat,zly 0.03 ram. velocities of 50 and 90 m s - i (90 ° incident angle), show the

. ~, • ,.. , . ~ .

. .. . . ::".~ ?,~.,~

• • . .~ ~.~~

..

• .~. . . . . . ,..~ .~ • -. . . . . . . .

,.,.i .
:.,: ,~ ~ ....

fig. I. I m p a c t ...,i~e,~o n 41b ~ m t h i c k c o a t i n g . ~ 0 ° i n c i d e n ! ; i n g l e . i n c i d e n t v e | c ~ i l i e . , , o f : ~.', ) 3 3 rn s ~. ( h ~ 5 0 m ~ ~. ,I ¢ ~ t~) m ~ '.

188 M, PtqJhri, J.K. Spelt / Wear 21.¢ (1997) 185-199
blister-like nature of
the impact site. with the coating delam-
inated below the blister. This characteristic feature may have
been due to the delaminated paint chip adhering to the glass
particle as it rebounded and thus being partially pulled offthe
substrate. Alternately, the blister may have resulted from the
lateral displacement and buckling of the coating as the panicle
penetrated to the steel. A hole in the paint layer is visible in
the center of Fig. ! (b) and (c ], indicating that the glass
sphere had penetrated through the paint layer. An interesting
feature of the 50 m s ~ collision (Fig. I ( b ) ) is the radial
cracking pattern, which was typical only tbr a small range of
velocities (45-55 m s- ~for the 40 p..n coating). Impact sites
lbr angles other than 90 ° showed similar features, but were
slightly skewed into an oval shape for impacts at low angles
of attack (high tangential component of incident velocity
vector).
At higher velocities, it was clear that delamination had
occurred: however, at lower incoming velocities, it was not
so obvious. In order to determine it' the coating had delami-
nated, special adhesive tape was applied to the impact site
and pulled off in the manner described in the ASTM standard
adhesion scratch test 1201. ira chip came off. it was assumed
that delamination had occurred, In all cases, the tape consis-
tently removed the paint to the bare substrate, i.e. the coating
always failed by delaminating from the pretreated steel. The
area removed in each collision was measured using a video
camera on a microscope, a frame-grabber and digital image
analysis software,
Fig. 2 shows the steel substrate below the impact sites of
Fig. i (b) and (c) after the paint chips were removed with
the adhesive tape, as described above. The paint has been
completely removed in both cases, except for a small spot of
crushed paint directly below the impact in the 50 m s - i case
Fig. 2. Slecl sub.~lrale b¢lc~w impact sit¢ after ,~'rnoval of 40 p,m coating using adhesive tape. 90 ° incident angle a! incident veloci|ies of: (a) 50 m ,~ '. {b)
(Fig. 2(a) ). Similar cracking patterns were observed by Lu
in static indentation tests on ZnO coated Si substrates [ 13J.
Fig. 2(b) shows that at higher velocities ( > 6 0 m .~- a), the
remaining dot of crushed paint on the substrate is replaced
by a dimple of permanent deformation in the steel. X-ray
photoelectron spectroscopy (XPS) analysis of the back sur-
face of the removed paint chip revealed that there was no
phosphate present. This, together with the fact that there was
no observed paint residue left on the substrates in Fig. 2,
indicated that delamination occurred at the interface between
the iron phosphate pretreatment and the paint, and not within
either the paint or the iron phosphate l,yer.
A detailed inspection of the 50 m s- ' ( normal incidence)
impact site (Fig. 2 ( a ) ) revealed no visible damage to the
substrate, indicating that it was possible to remove paint with-
out damaging the substrate ( this will be discussed further in
Section 4.3 ).
4.2. D o m i n a n c e oJ'normal c;~k,~'ts allrl comptzrisolt to
B r a c h 'x m o d e /
Brach has shown that the energy losses in a collision due
to normal and tangential effects are, respectively 19 J,
I
"IT. = I( I - e 2) sin-cel~_ m r : (I)
"'-[[~]
/',.- " (~)
~ 2- ]' mr:
cos-'o¢ ~ (2 )
where o~ is the angle of attack, defined as the angle between
the incident velocity vector and the surface, ,~ is lhe ratio of
the particle radius squared to the radius of gyration of the
particle radius squared ( 5 / 2 for a sphere), e is the coefficient

of restitution (defined as the ratio of the normal rebound
velocity to the normal incident velocity ). and/.t,, is the critical
impulse ratio which detines the transition from sliding to
rolling during a collision and is detined as [ 9 I.
I
#.,,: =
( I +A)( I +e) t a n o¢
The impulse ratio, #, is t h e r a t i o o f l a n g e l l l i , ' d t o n o r m a l
impulse generated during the ilnpact, P~/P,,, and is ineasured
by taking the ratio of the change in momentum in the tangen-
tial direction to the change in momentt, m in the nornlal direc-
tion ( using the appropriate components of measured incident
and rebound velocities). It is irnporlant to note that the
impulse ratio is equiwdent to a dynamic friction coeflicient
only in the case where the mechanism of tangenlial impulse
is due to friction. For a more complete discussion of impulse
ratio, and the role of friction, see Rel\ [gJ. An underlying
assumption is thai, regardless of the mechanism of the tan-
gential impulse, there can be no impulse ratio higher than ih,
critical impulse ratio./x~., so that if a series of in]pact experi-
ments is performed at tixed incident velocity..rnd the meas-
ured impulse ratio is plotted vs. the .'ingle of incidence, there
should be a point where the transitffm from sliding to rolling
occurs, alter wllich the data points lie within a band detined
by Eq. (3) at e = 0 and e = I 191. Brach used experimental
data on uncoated specimens to support his analysis, and in
particular, this hypothesis of a critical impulse ralio 191.
There is. however, no reason why the analysis should not
also hold for coated substrates, and Fig. 3 shows typical data
gathered in the present study. The data points are averages of
multiple impacts performed at each angle of incidence. As
mentioned above, the impulse ratios were calculated by tak-
ing the ratio of the cl~ange in mornentum in the tangential
direction to the change in momentum in the normal direction.
using the experimentally measured component.~ of values
obtained at each angle of incidence, and incident velocity aqd
rebound velocity in the tangential and normal directions. The
o.25 i .... !---
i
l i
i ,
,~ 0,15 ..... ~.
~ O.L
O.05
0 ~
0 to
Fig. 3. Vari;itit)n of impulse r,'itk~ wilh iiIl~lt' lll';.lli~.L~.'k I'l, 85 m .',
I ~) i :,1~ sh4~wn for e = 0 ~l,~d (' =- I.
M, !~ q~ini, J.K. Sp~,ll I Wear 213 ~I qq7 ~I,'¢5~-/ t~q 189
transition between roiling and sliding appears to occur at 45 °
angle of incidence, after which the data points roughly Ibllow
the "'envelope" dctined by the theoretical critical impulse
ratios at e = 0 and e = I. The other experimental data show
similar trends and all have a large amount of scatter in the
(3) impulse ,'alto. This relatively large scatter has been experi-
enced by other~ [ 9.211 and is probably due to surface irreg-
ularities and the I~ct ti]at the impulse ratio is very sensitive
io error,~ in Ihe measurement of the impact and rebound
velocities.
The dependence of the paint area removed on the energy
Io.,,.~ ass~ciated with the normal and tangential incoming
,,'elocitics is shown in Figs. 4. and 5. The amount of paint
removed d~cs not depend very much on energy loss due to
tangential effects ( Fig, 5 ). bill does increase with the enemy
loss due Io normal effects (Fig. 4). This relatively strong
corrclalion between the area loss and the energy losses due
to normal efft2cts alld tl~e relalivcly weak correlation between
the area Io.,,s and the energy losses due to tangential effects
st,~zesls that normal effects dominate paint removal. Fur-
thermo,'e. ~t threshold value of energy loss due to normal
effects exi.,,tx for each coating thickness (Fig. 4). whereas.
there is no st,oh threshold value for energy loss due to
tangential effects ( Fig, 5).
Fig. 6 is a plot of the incident normal velocity component
vs. the amount ~1 coaling removed for all experiments per-
f~rmcd at an~les of incidence between 18 and 90 °. This illus-
trates that a critical amount of incideni nonnal velocity is
required, for each coating thickness, to initiate removal of the
coating, and. furthem~ore, that this critical value is independ-
ent of the tin,genital component. The dependence on normal
efl'cct,~ is furthe," illustrated by Fig. 7 which is a plot of the
energy losses normalized by the total incident kinetic energy.
vs. angle of attack, in all cases, the enerey loss in the normal
direction accot, nts for most of" the energy loss, This system
exhibits, what is termed in the lield of solid particle erosion.
defor,natio,} (}r ploughing wear. a,s opposed to cutting wear
[ - i
c -
F
"!., ]-,
° - i
-
I
i ° "-~-
i , . . - - ~ . . . . . . •
20 30 40 50 60 70 80 90
Incident Annie of Attack, at ( d ~ . )
I 1~1al incident ~.'t,..hlt,'ily ~..olli~i~,~ll~, with 40 ~111 Ihick ~.'olJ|itl~. Crlti,,:.'LI illlptli.~¢ r:ltio ~ t 19r,tq.'ll
190 M. Pupini, J.K. Spelt/We(zr 213 (1997) 1i~5-199

1.8
& 0
| a'~ &

m
,"" 1,4

4, •
!.2 ~,

~ 0.8 •

0.6 u_~ ••~$ •.

~ o.. ."
< 0.4 oo ~=••:'o o
_ O •o '~q~o
0.2 =o •

0 0,$ 1.5 2 2.5 3

Norm=l ZnerID' I o n , T L" (mJ)

Fig. 4. Area of coating removed w . energy loss due to normal direction etTccts for coating thickness: A 50 ~m, 4) 40 Is,re. [] 2.5 p.m. C) 20 p.m.

1,8

i.6 +

1.4 .
A A
A •
• Ai A A
h, A

= I b

~s o.B i o
+
- i t
• ~ • u•
0.6, 41' • ,, 4••
n
< = UO % 4 4
0.4 a -I 0

0 D O
0.2.

0 .1331nG4 ~ ~ 0 0 6~6 Mbt & $ 4 44•46 ~SO& • ~ •

0 0.05 O. I O. 15 0.2 0.2.5

Tanl¢enlii| Enerl~ Locq, TL+(mJ)

Fi E. 5. Arca of coatin~ rumnovcd vs. cncr~y Ios~ duc Io I~t)gcnliol direction ct'fect.~ for coatin~ thicknc.~s: ,~. 50 pro. 4 40 p.m. I-t 25 u.m. C) 20 I-Lm.

1.8
A ¢'
1.6

13 i
~ 1,4 -
[] & A
4
~ 1.2 . A .

1
• •o
B I
4k
o° 4 ' ' *
3 0s
•4. 4
o.6 o u -'..'*+. 4'
P.
OA
°o° 4~0 • O0 0
O 0°
0.2
" S
0 IZX)~CJmsIl~,mdlk4m~ 4- db ~, •

0 20 40 60 80 I00 120 ]40

Incident Normal Compoueut of Velocity, v, (m/s)
Fig. 6 . Arca ol'¢oating r c n u r , r c d a s it I+unctiono f in¢idc,]t normal vclo¢ily ¢ol'rl~t)llelll |'tlr c o a l [ l l ~ tJli~.'kll¢~.~: A .%0 p.111, • 40 ~m. [] ...5 ~ m , O 20 ~llt.
 M. Papini. J.K. Speh / Wear 213 ¢1997~ I1s'5-199 191

1.00
!
0.90 45- o

o
, t
0.80
.°~
0.70
qr.
Normal Direction - ~e
~ 0.60

~a 0.50
:!
i 0.40
E m,
,p-
o

~ 0.30 41, n .

- la~qll~ . Tangental Dir~ti~n

0.20

0.10 •
_~a." :tP':t
-4, a . '- ILhl~a~ J
..aT'~ 4,.
0,~ "

0 10 20 30 40 50 60 70 80 90
Incident Angle of A l t a e k , o (degrees)
Fig. 7. Energy loss due to normal :rod tangential direction cl'fecl.~ normalized hy total incident kinetic energy as a function o f incident angle o f attack. Normal
energy loss for coating thickness and total incident velocity:/', 50 v-m thick coating, 80 m .~ ': - 40/,tm, 85 m s '." © 40 vtrn, 56 rn s- ' ; [] 25 v.m, 42 m s *;
O 20 v,m. 57 m s *. Tangential energy loss fl~r coating thickness and total incident velocily: & 50 bun. 811 m s " i, --411 ~m, 85 m s '; @ .tO i.tm, 56 m s ~2
• 25van, 42ms ~ : 0 2 0 1 1 m . 5 7 m s ~.

19, I0]. This type of deformation wear is also clmracterized 4.2, I. Role of umgentiai effects
by the dominance of normal effects, with the maximum The onset of coating removal was, as demonstrated above,
amount of damage (energy loss and area loss) occurring at independent of tangential effects. However, the role of tan-
high angles of incidence (ot--*90°). This is demonstrated gential effects in the magnitude of coating removal has not
further in Figs. 8 and 9, which show the area of paint removed yet been discussed. Fig. 6 shows that the area of coating
as a function of the incident angle of attack 1o1"various total removed increases as the normal component of incident
incident velocity ranges. The area losses are averages of mul- velocity increases. These data. however, are for a large num-
tiple impacts performed at each angle. The maximum coating ber of different impact conditions ( i.e. angle of attack, inci-
removal occurs at higher angles of alto(ok, and removal does dent velocity). In order to fully understand the role of
not begin until the normal componer;t of velocity is suffi- tangential effects on the area of coating removed, a series of
ciently large. experiments was performed in which the incident angle and

!.8

].6

'¢E 1.4 - : A

v
El
I 1.2 '

B
A A

@ .
° 8
"a o.s • ° O

E
0.6 •
4P =
e
0.4

0.2

0 ' ~ O11,0, ~.4B

0 I0 20 30 40 50 60 70 80 90
incident A n g k o f Attack, ~t (delgo)
Fig. 8. Average area o f coating reu'toved :ts a function o f incident angle o f attack l'or different ranges of incident total velocity and coating thickness. Data
points represent averages of area removal measurements at each angle. [] 5 5 - 6 2 m s 1.41) v.m: $ 8 I-iN] m s *, 411 I~m: A 78-81 m s t. 511 V,in, O 102-
1 ( ) g m s - t 5()p,m.
192 M. Pal, tat, J.K. Spelt/Wear 213 (1997) 185-19~

1.8

!.6

~ 1.4

I 1.2 -
B
I
,e-
,,a o.e •

E
~ 0.6

~0,4' 0__ ,

04,

0,2 "

• & A
0 o
4~
0 ........
0 10 20 70 80 gO

Incident Angle of'Attack, cc (dog,)

Fig. 9. Area of paint removed as a function of incident angle o f attack tbr different ranges of incident total velocity and coating thickness. Data points represent
averages of area removal measurements at each angle. • 38--44 m s- ', 25 p.m; ~ 25-27 ,n s ~. 25 p.m: A 102-110 m s ~, 25 o.m: O 54-62 m s ', 21) ~ m :
• 2 5 - 3 0 m s " i, 20 I.tm.

total incident velocity were varied so as to always give the
same normal component of incident velocity. The average
and standard deviation of the area removed was calculated
for a number of experiments performed at each condition,
and the results are shown in Table I. The area removed, for
each coating thickness, is virtually independent of tangential
effects (the scatter in the data is approximately constant).
This result is somewhat counterintuitive, as one might expect
that an increase in tangential velocity should have some effect
on the amount of coating removed, As mentioned in Sec-
tion 4. I, the only tangential effect was that, at lower impact
angles ( tangential component of incident velocity high), the
coating removal area changed from being axisymmetric to
being slightly oval shaped, even though the total area
removed was not affected by this change in shape.
The above arguments lead to the conclusion that, for this
particular system, normal effects dominate the onset of coat-
ing removal (i.e. the threshold value of normal incoming
velocity and energy loss required to remove the coating), and
that, subsequently, tangential effects have little or no influ-
ence on the magnitude of coating removed. Thus. for the
purposes of investigating the onset of coating removal, and
for simplification of analytical models, the tangential effects
can be ignored.
4.3, Mechanism qf coaling removal
4.3. I. Spall&g by stress waves
In order to determine whether the mechanism of coating
removal was dominated by dynamic (stress wave propaga-
tion) effects or quasi-static effects, a series of experimental
observations was made. It was initially suspected that
removal was the result of a spalling mechanism, in which the
collision of the particle with the coating caused acompressive
spherical stress wave to propagate through the coating, Most
of the wave would then be reflected at the coating-substmte
interface (because the substrate has a much higher acoustic
impedance) as a compressive wave, then reflect at the top
surface of the coating, and back down towards the coating-
substrate interface as a tensile wave which could delaminate
the coating. It was also possible that the portion of the com-
pressive wave that was transferred to the steel substrate could

Table I
Comparisonof coating area removedkeeping normalcomponentof veh~citycon.,~tant,whilevaryingtangentialcomponentof velncity

Coating thickness incident angle Normal component Tangential N u m b e r of data Average tenting St:mdard devialion
(lu.m) (°) of incident velt~:ity component of points area removed id" coating area
Im s ' ) incident velocity (mm 2) removed ( ram: )
(ms ')

20 q() 31 0 I0 0.273 O.03fi

20 38 3! 40 II 0.261 0.030
40 90 59 0 12 0.577 0.05 t)
40 46 59 57 12 0.545 0.060
40 33 59 91 9 0.559 t).lt)l
 M. Paph~i. J.K. Spelt~Wear 213 ~lq97J 185-1q9
travel to the back surface of the substrate and reflect as a
tensile wave which would then travel to the interface and
delaminate the coating. In order to investigate the hypothesis
that paint removal was due to a type of spalling, a series of
experiments was perlbrmed.
A ! 2.5 mm thick piece of steel was clamped to the back of
the painted (40 p,m thick coaling) steel sheet with three dif-
ferent coupling layers: an ultrasonic coupling fluid, honey,
and a thin layer of epoxy, so that the clamped interface would
trans|'er any waves without much attenuation. If a wave
reflected from the backside of the 12.5 mm plate as a tensile
wave and caused a dehunination as explained above, the size
of the delamination should be decreased relative to the steel
sheet alone because the spherical wave would have been
dispersed and attenuated by the much thic~;er subslratc
(apprr~ximately 19 times as thick as the original sample). The
paint areas removed using the ('oupling gel, honey, and
epoxy, were, respectively, 0.659.0.6,~3 and 0.613 mm -~.The
area removed from the painted steel sheet alone ( without the
12.5 mm plate coupled to it) was 0.690 mm z. The differences
between these values are small, and certainly well within
experimental error, which is approximately + 30% as shown
in Fig. 4. This means that compressive waves reflecting as
tensile waves from the back of the substrate can be eliminated
as a mechanism of coating removal, Nevertheless, a spalling-
type mechanism could still possibly result if compressive
waves reflected at the substrate-paint interlace and back
down from the free surface a! the top ofthe paint layer towards
the interface as tensile waves, The other conclusion that can
be drawn from this experiment is that the 0.66 mm thick steel
substrate can be modeled effectively as semi-infinite,
Further insights into the mechanism of removal may be
gained from Fig, 6, a plot of the paint area removed vs. the
incident normal velocity component for all experiments per-
formed on this coating-substrate system. An interesting fea-
ture is that the area removed tends to reach a plateau at higher
incoming normal velocities. Although this observation is
intuitively correct, it is contrary to the spalling hypothesis,
where one would expect the area removed to continue to
increase, as the intensity and energy of the wave increases
with the incident normal velocity,
An additional assessment of whether quasi-static or inter-
tial (wave) effects dominate can be obtained by calculating
the amount of time it takes for stress waves to traverse the
coating thickness, and comparing this to the amount of time
that the sphere is in contact with the coating during the col-
lision. Assuming that the coating has a density, O, of
1200 kg m -~ and a Young,'s modulus, E, of 2470 MPa, the
wave speed is ( E / p ) ~:: = 1435 m s - '. For a 50 I~m coating.
it takes 0.03 p,s to traverse the coating. Experimental obser-
vations have shown thai the contact lime is on the order of
50 Ixs. Because the stress waves will traverse the thickness
of the coating approximately 1700 times during the contact
time, the stress field will be dominated by static, rather than
inertial effects.
193
Fi$, IO, Impact near a free edge for a 40 ixm thick coaling, Nole the at(ached
pain1 in tile shape of the original free edge.
The other possible mechanism for coating removal is the
attainment of a quasi-static critical delamination shear stress.
In this model, the collision is treated as a quasi-static inden-
tation problem, under the assumption that the coating will
delaminate once a critical shear stress is reached. An exper-
imental observation that appears to support this mechanism
is seen in Fig. 10, which shows an impact site very close to
the cut edge of the paint layer (though relatively far away
from the edge of the steel substrate). Here, a 0.64 mm glass
sphere was launched at the edge of a 40 Ixm thick paint layer
sample to investigate the effect of a free surface in the vicinity
of an impact site. The coating was delaminated in the usual
manner, except that in the vicinity of the free surface at the
edge, a small ridge of paint (shown by arrow in Fig. 10)
remained intact in exactly the shape of the original edge of
the paint layer. Because the shear stress on a free boundary
must be zero, equilibrium considerations imply that the shear
stress along the paint-substrate interface must also fall to zero
very quickly at the edge, thereby explaining why a narrow
strip of paint remains attached in the immediate vicinity of
the free surface of the edge. Under the spalling i~ypolhesis,
one would expect the tree surface of the edge to have little
effect on the reflected tensile wave, and for all of the paint to
be removed.
4.3.2, Dehmtination by shear stresses
The above experimental and analytical observations led to
the conclusion that the coating removal was due to a quasi-
static interracial shear effect. An estimate of the stress distri-
bution within the coating layer was obtained by modifying a
model due to Matthewson [ I ! ], who developed expressions
for the stresses and strains generated by the indentation of an
elastic layer on a rigid substrate by a rigid particle, The
194 M. Papini. £K. Spelt/Wear 213 (1997) 185-100

:p o

P(a) ,= 2m'.dr (7)

o

Matthewson's result is [ i I ]
8h I G~rhv(4+ v)aet d I ( I -- v) 7rt,,4
"h P(") .---6 -~a( aK) i22 -- hR (8)
zT.
- ...- r .-
Eq. (8) assumes an indentation profile given by Eq. (6).
Fig. I I. Geometry of indentation of coating (thickness h) on rigid substrate
By a simple modification of the model (to be published else-
by a sphere of radius R.
where), the authors have found that the restriction of small
stresses in the coating (assumed very thin) were averaged a / R can be relaxed somewhat for deeper indentations by
through the thickness of the coating as adding a term to the indentor (particle) profile, Eq. (6), as,
r 2 _ a 2 r4 _ a 4
2vG ~- (9)
~,.(r) = - - (~+~+E) +2G~. fir)
1-2v 2R 8R "~

2vG This improved approximation to the actual spherical inden-

~-'~,(r) i _ 2 v (~+Z;o + ~ ) + 2G~o, tor profile, for a / R = 0 , 5 , has a maximum error of 2%,
(4) whereas Eq. (6) has a maximum error of almost 14%. Using
2vG Eq. (9), the analysis proceeds in a manner similar to that in
~ : ( r ) = l - 2 v ( ~ + ~,--S+~) + 2 G ~ . Matthewson's paper [ 111, and Eqs. (4) and (5) can be used
with
~( r) = G B ( r )
-vr'[g(6v-l)(,,-l)h" 2v ]
and the corresponding average strains are
B(r) (2v_I)R. ~- 3R~(2v-l) 2 +R(2~,--I) r
Id B i ,d
E(r) =2~ r (r)+~h-~rC("J.
+fll~ 1 - v) (O<r_<a)
I B(r) I .~C(r) (to)
=- +-.- --, (5)
~(r) 2 r 3 r (/6(I - v)r~
B(r', (r>a)
r:(r) =---~-
f(r) +
ah
")="K,tV s ; ;
c(,-) =
I ( B(,.) + - - g f ) (0<r<a)
where or,. oro. or- are the average normal stresses in the r-, 0-
and :.-directions as shown in Fig. ! l, ~, is the interfacial shear (II)
stress. ~,, e,. and 7. are the corresponding strains averaged 2-v
C(r) . . . . B(r) (r>a,~
through the thickness of the coating layer, h is the coating 4h
thickness, v and G are the Poisson's ratio and shear modulus
of the coating, respectively, cSh is a constant representing the to find the stresses and strains in the coating, and the shear
pile-up of material at the edge of the contact area (Fig. I I ), stress along the interface with
and B(r) and C(r) are functions which depend on the inden- I .,[ hi, 4(6v - I)(v-I)11_.
tor profile. The profile of the indentor.f(r), in Matthewson's a= h(4+ v)(K'I-KI) "~.~al' 3 ( 2 v - I)2R ~
paper, was approximated by the first term of a parabolic series
12a/'
expansion as
E2 - - ( i 2
l(3a2+2R2)(6v - I ) ] 4(6v--l)(v-l)h ~
.f(r) = ~ (6) + 4R3(2v- I ) J 9 ( 2 v - I)2R ~
2R
( 6 e -- I ) ( 2R" + a 2) h~[
wilere a is the contact radius, and R is the radi(Js of the Y (12)
indenting sphere. This expression was used to calculate the
average strain in the z-direction ( Fig. I I ), and it assumes that and
the ratio a / R <_0.2. Assuming thai the profile given in Eq. (6)
4h2( - 7 v + 6 v 2+ I) K'
is used, expressions for & B(r) and C(r) can be found in /3=
Ref. I 111. 3 ( 2 v - I )"R~I ' £ ~ ( 4 + v)~7
An expression for the indentation force. P. can be obtained ( 3 a 2 + 2 R 2 ) ( 6 v - 1)
by integrating the average stress in the z-direction over the + (13)
4 ( 2 v - I )R31 '
contact area.
 M. Papini. J.K. S p d t I Wet." 213 ¢1997~ 185-199 195

Table 2
Comp it)son of predicted velodly to cause penetration of coaling Io steel substrale, and measured vdoCily at onset of coating removal

Coal)hi:' thickness ( ~m I Predicted normal componenl o f vdocity m Experimenlal nt~rmal component of velocity at Difference ira , )
penetrate coating i m s ' 1 which coating begins Io be removed (Fig. 6)
( m s ')

20 20 20 I
25 25 22 13
40 39 40 4
50 50 45 11

where /,(x) and K~(x) are the first-order modified Bessel

7,,,,.-" 2ol
,l/,<,_2,>,;)
functions and,

/=,ltV:~],-_- ) (14) where Eq. (97 has been used to describe the relationship
between the contact radius, a, and the penetration depth at

K:
.[/" ( I . - . ) 4
#¢,[ V6--~.~.Z) (i 5)
t'= 0. l,d is the dynamic hardness, am~, is the maximum con-
tact radius reached during the indentation, and m, R, and r
are. respectively, the mass, radius and incident velocity of the
dl panic le. The dynamic hardness, p,,. was determined by adjust-
/'- t16) ing it in Eq. (20) until the predicted maximum contact radius.
da
a ....... matched the contact radius measured on actual impact
dK ,;ires ( neglecting any spring-back due to elastic effects). The
K'= m (17)
da value of p,, for this coating was found to be approximately
Using Eq. (7), the indentation lbrce, assuming that 190 MPa.
Eq. (97 is the indentation prolile, is The Mly-plastic model was used to calculate the velocity
necessary to just penetrate the coating to the bare substrate.
GTrhv(4 + u ) a a d Grr( u - I )a" Table 2 shows a comparison of these predicted velocities and
P(a)= 6(2v-I) 7aa(aK) 3hR~(2v-I) the experimental observations of the normal velocity com-
ponent required to begin coating removal (from Fig. 6).
[vG~rh(6u7 i ) Grr(v-i) 1~
There is a good correlation between the predicted coating
penetration velocity and the measured velocity at which the

+
[ 16G~'uh~( 6 v - 171 u - I )
9( 2 v _ 7 )",~'<
coating begins to be removed for four different coating thick-
nesses, in other words, coating removal begins only when the
coating is fully penetrated for this coating/substrate/particle
2Grruh( 6 v - I)] 4G~,h~lla system. The reason for this may be that the onset of delami-
+ giVR ,1:+ 3 ( 2 v - I ) (18) nation requires the creation of a free edge that is formed when
The pile-up of coating at the edge of the indentation the particle reaches the substrate.
( Fig. I I ). Bh, is the same as tiae one derived by Matthewson If the plastic deformation in the coating is assumed to be
I I I 1, and is given by, restricted to just under the penetrating particle, then the mod-
)lied Matthewson model should yield realistic values of inter-
u (4+ ~,)h a facial shear stress at some distance from the impact site. If
8tl--- "oe(K'a + K) (19)
I-v 12 a the fully-plastic indentation model is used to calculate the
For the range o f a i R ratios of the present study, the mod- penetration depth for a given incident velocity, the modified
ified analysis based on the indenlor prolile ofEq. ( 9 ), yielded Matthewso, analysis, Eqs. ( lO)-(19), gives an estimate of
threes that were 20-25% higher than the original Matthewson the interfacial shear stress for this incident normal velocity.
analysis I I l l . at a given distance from the point of impact. This was done
The modified Matthewson analysis. Eqs. ( 10)-(19), is for a distance of 0.4 mm from the impact site at 20 m/s
strictly only valid for linear elastic indentations. A more real- incident normal velocity, and the results are shown in Fig. 12.
istic model of the kinematics of the incident portion of the The model predicts that the interracial shear stress increases
collision assumes that the coating behaves locally in a fully with increasing coating thickness. This trend is consistent
plastic manner during the impact, resisting indentation with with the experimental results of Fig. 6, which show that the
a constant dynamic flow pressure (or dynamic hardness. area removed increases with coating thickness for a given
defined as the force divided by the contact area) 151. For a incident normal velocity and particle diameter. Under the
sphere, setting the incoming kinetic energy equal to the work hypothesis of a tensile wave reflection, this would not be
done indenting the coating gives, expected: one would expect that the thicker coating would
196 M. Papini. .I.K. Speh / Wear 213 f 1997) 185-199

have smaller amounts of coating removed, since a spherical proceeded in a similar manner for all the coating thicknesses
wave would disperse more alter reflection from the interlace. from this point. This is further illustrated in Fig. 14, which
it is important to note that the model is only valid for pene- shows that the coefficient of restitution tended to reach the
trations less than the coating thickness, and, strictly, cannot value measured on bare steel at higher incoming energies.
be applied after penetration to the bare substrate. Neverthe- The coefficient of restitution can be estimated by using the
less. the experimental data suggest that the trend of Fig. 12 fully-plastic indentation model (Eq. ( 2 0 ) ) to describe the
continues after penetration of the coating to the bare substrate. incident portion of the impact, and the modified Matthewson
A cerlain amount of energy is lost in penetrating the coat- model (Eqs. ( 10)-( 191 ) to describe the elastic rebound of
ing. :rod then the collision proceeds as if the remainingenergy the particle. Johnson ! 51 used a similar approach with bare
I incide,~t energy minus energy to penetrate the coating to the substrates. Doing this Ibr incident kinetic energies less than
substratr ) is incident on a bare substrate. This is demonstrated or equal to that lbr penetration to the substrate (range for
in Fig. 13. which shows that for all coating thicknesses, e which Eqs. ( 10)-( 191 ) are valid), yields values of e that
tended to reach an approximately constant value of 0.35 when agree very well with the data of Fig. 14. This further illus-
the coating began to be removed, meaning that the collision trates the validity of the present model, and shows that the

].00E+06

/
1.00E+05 . .,,.....,.f...._....----~"'~
.,..I
m
~-~ I.OOE+04

I.OOE*03 -
m

~ 1.0OE+O2

1.00E+01

1.00E+00
20 25 30 35 40 45 50 55 60
Coating Thicl~aess (lam)
I;ig. ] 2. Prediction o f intcrlacial .,,hear ,,.trcnn ~..,..coatil|~ I]]iCkllt.',',;~.Ill r ==().--I-Illln Ior fixed iuciden! norm;,I v e l o c i t y o f 20 n'n/.,,; and spherical particle wilh R = 0.32
toni. ( "a I,.'u laliort.,, biL~cd ~m m o d i f i e d lllotlc ] ( l'.q~,. ( I (~)--( 19 ~.

0.9

0.8

~0.Ta

0.65
",l~ ii
o &
• •
(I.5
o D~
• 41,
o
"G (1.4

-o.
o

g 4,
0 e

".
0 •

).I 0 o
~, ~

ql, ,
¢v'D ~

o'~
,e
.~ • '~ e.O

•
"i ~.
A
A
q" 4%
0.2 a ,, * • •

0"1

O
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
Area or Coating Removed (mm z)
I'ig. 1 3 V a r i ; l l i ~ u I,f ct~ctlicicnl o l ru'nlilttlio.. ~.. w i t h ill'¢;i o f c o : f l i n g rell]ezvcd I'~lr c o a l i n g Ihickt]cns: ,:,, - 5() IJ.Ill. • I 4 I ) g i l l " uJ " 25 ~1"111"(~) " 20 ~Izl"
 M. t'~tl#,i..I.K. Spctl / i~'~tt 213 t l~97~ 185 199 1')7

0.9

0.8 , -

0.7- a~"

ii - -
0.6 u ~, ":-~i.~ •

° = -:
a

"~ q.4 "~1~"~ ~ * ~ . ~',,i1% . *

4~4k 41 . ~ •
0,4, 41' ~ o e. 4 ,k
o.s. ,~ ®~".~., ,, : ¢ • & II,

0.2 [] ,~ ,%~,
a ,G

0.1 '

0 -

0 0.5 I 1.5 2 2.5 3

Nonmal Inciden! Kinetic Energy (m J)
l:ig. 14. V a r i a l i o n of cn¢l'licicnl o f Iesli;.ulioll. t'. xvifll i n c i d e n t k i n e t i c encr,c'.~ ill n o t n | a l d n w . ' t i o n f o r coa.~in~ tl'tickl|¢sn: .~. 5() Is.In. • 41~ ~.m. i 25 p m .
C) 20 p.m, ~ bare steel.

value of Young's modulus of the coaling obtained ultr;.lson- I.cl ! d he the incideill kinetic energy l~w the particular data
ically is a reasonable estimate of the true value trader impact plaint ()f inle]'cst, and let Ip eqtlal the incident kinetic energy
conditions. of a .~phere that .just penetrates the coatin~ to the substrate.
Figs. 4 and 5 show energy losses for the coated system as (a) Use the fully-plastic inodel [Eq. (2())1 to calculate I,..
a whole, including energy losses due to local plastic defor- the incident kinetic energy to penetrate the coating, corre-
mation i,] the bare steel substrate directly below the impact sponding to Ihe onset of coatiltg remo,;al.
site. An estimate of the amount of energy lost to the coating ( b ) Subtract I,, from I,~ to estimate the incident kinetic energy
alone can b," obtained using the normal direction enel~.y loss that is available to deform the steel nubstrate. As illustrated
for experiments performed on both coated and bare sub- in Fig. ! 3. once the coating has been penetraled, impact pro-
strates, as ~ function of incident norrnal direction eqergy ceeds its if the substrute were bare.
( Fig. 15 ). The following steps demonstrate how to use these i c) Use Ihc linear curve fit of the energy loss vs, incoming
plots for estimating energy losses to the coating due to rmrrnal cnerg) ~ for bare steel (bottom curve of Fig. 15, or figure
effects at high incident normal velocilies. All energies and similar to it for other coaling thicknesses) with the result of
energy losses in the following steps :ire due to normal effects. (b) io estimate the energy loss It) the steel substrate.

3.0

2.5

.= T, ° ~ 0.89751 - 1.091 x I 0 '2

2.O

i
m

1.5
o

~ 1.0

|
0.5

T I '~ = 0 . 8 3 0 9 1 - 6.980x 10:

c:, 'P"
0 0.5 1.0 1.5 2.0 3~0 3.5
Inciden! normal titration e n e r g y , I ( m J)
lql~. 15 IEt~cr~ l . s n d u ¢ IO llor111~li dirl:t:lit|ll eft,.'cln ~.',. l]Orllliil iilu'iiiL'llt ~ { n e l i e L'llt.'r~). li~t~U:ili,.Wi "., nlio'~~.n iirk• l i n e a r l,.2~i,..~.'.,i,tH1 l..'uix.c IllS. ~ .4(I' ~i.i~i c~.1,:.|liii~. : !
b~m:SIL.'cl.
198 M. Papini. J.K. Spelt / Wear 213 (1997) 185-199
1.8
1,6
1.4
!t
1.2
? tilb ""
! • _
0,8
d D
0.6
0.4
q,o
0
0.2
0 O.I
Fig. 16. Are.'J o f c o a t i n g r e m o v e d vs. e n e r g y lost to c o a t i n g tbr c o a t i n g thickness: A 50 Ixm, × 40 I~m. •
(d) Use the top curve of Fig. 15 with Id tO find the energy
loss for the coated system (which is made up of energy loss
to the coating and energy loss to the substrate). Subtract the
energy loss to the steel substrate obtained in (c) from this
energy loss. This is an estimate of the amount of the energy
lost to the coating.
Figures similar to Fig. 15 were obtained for all coating
thicknesses, and the above steps were performed on all data
obtained in this study. The resulting estimates of energy loss
to the coating due to normal effects can be found in Fig. 16.
As seen before in Figs. 4, 8 and 9, there is a threshold value
of the normal energy loss corresponding to the onset of coat-
ing removal. The threshold increases with coating thickness,
and beyond it, the area removed increases rapidly with energy
loss. The energy losses due to tangential effects for coated
substrates (Fig. 5) are approximately equ:al to the energy
losses to the coating itself due to tangential effects. This is
because the energy losses due to tangential effects were found
to be very low for impacts on bare substrates when the impact
angle was varied from 17 to °0° for incident total velocities
of 50 and 90 m s- i (less ~.han 10% of total energy lost in
most cases).
As the incident normal :inetic energy is increased beyond
that required to penetrate the coating to the substrate, the area
of coating removed increases. It is hypothesized that this may
be due to the deformation of the particle after it contacts the
steel substrate. Because the steel is approximately three times
stiffer than the glass, it is possible that, upon contacting the
steel, the glass particles experienced significant lateral defor-
mation which resulted in increased shear stresses at the coat-
ing-substrate interface. An estimate of the extent of this
deformation can be obtained by using a Hertzian analysis
[ 22 !. Although the substrate will deform plastically at higher
impact velocities, it is noted that coating removal does occur
A
&
..........
dl &JS, X)~ x
'~
b
xl,ii ~
~,,=x x
..... .........x~.-~ x~ ,
x'. ,.
Qe ~ - ._
. ~. x . . . . .
x ~
0.2 0,3 0.4 0.5 0.6
Normll EielT.y ~ to Coating (mJ)
25 p,m. [~ 20 p,m.
for a range of velocities that do not cause plastic deformation
of the steel (Fig. 2). Hence, the use of an elastic (Hertzian)
analysis is expected to yield realistic results, at least for inter-
mediate velocities. For example, according to Table 2, a
39 m s- Lincident normal velocity is required to just penetrate
a 40 p.m coating, corresponding to 0.27 mJ total incident
energy (particle velocity is zero at substrate contact). Now,
assume that a second particle is incident at 50 m s - ~normal
velocity, corresponding to 0.46 mJ of incident normal energy.
The particle will again use 0.27 mJ of incident energy in
penetrating the coating and the remaining incident energy
will be 0.19 mJ. Assuming all of the remaining incident
energy is available to deform the substrate and particle, the
Hertzian analysis gives a maximum deflection of the steel
substrate of approximately 3 p.m, and the particle diameter
at the point of impact decreases approximately 8 ~m. The
volume of coating displaced by this particle deformation was
approximately 40% of the original coating volumc displaced
by the penetration to the substrate. Therefore, significant
amounts of lateral deformation of the coating are expected to
occur as the particle deforms (flattens), even at the relatively
low velocity of 50 m s - I This additional deformation of the
coating may account for the continuing increase in the area
removed. The analysis of the stresses along the interface after
coating penetration is an important subject for further study.
As noted above, an interesting result in the context of
coating removal using blast cleaning is that it may be possible
to remove the coating without causing visible permanent
damage to the substrate. An examination of the impact sites
for all experiments (optical microscope at 160× ) showed
the existence of a range of incident normal velocities, begin-
ning at the coating penetration velocity, where coating
removal occurred with no visible damage to the substrate.
For the 50 p.m coating, this range was about 50-80 m s- ~,
 M. Papini. J,K. Spelt / Wear 213 ( 1997~ Ibl5-1t) t)
for the 40 p,m coating about 40-65 m s - *, and about 21-
36 m s-~ for the 25 p,m coating (no conclusions could be
drawn from the 20 m s- ~ experiments due to a lack of data
in the proper range). The lack of damage to the substrate in
these cases is thought to be due to the coating protectiqg the
substrate. At higher velocities, the particle does not lose
enough energy penetrating the coating, and sufficient incident
energy remains to deform the substrate. It is important to note
that this result, although interesting, is not applicable to
streams of particles such as in blast cleaning, because, after
removing the coating, subsequent impacts on the bare steel
would damage the substrate.
5. Conclusions
For this alkyd-steel-glass bead system, both the onset and
magnitude of coating delamination was controlled by normal
impact effects, and a maximum in coating removal occuned
at normal incidence. A quasi-static deformation analysis was
sufficient to predict a critical amount of normal energy
required to penetrate a coating of given thickness. This critical
normal kinetic energy was found to correlate with the onset
of coating removal, and was independent of total incident
energy and angle of attack. A range of incident normal kinetic
energies was found where the coating was removed without
damaging the steel substrate.
The removal of the coating was consistent with the gen-
eration of a quasi-static critical interfacial shear stress. The
critical shear stress appeared to be reached alter the coating
was completely penetrated. Thereafter, it was assumed that
the particle continued the collision with the bare substrate,
causing significant lateral deformation as the particle flat-
tened, leading to a further increase in interfacial shear stress.
The deformation of the coating after penetration and the
resulting interfacial shear stress remains a subject of further
study.
Acknowledgements
The authors wish to thank the Manufacturing Research
Corporation of Ontario and the Natural Sciences and Engi-
neering Research Council of Canada for financial support,
and Valspar Inc., particularly Mr Denis Johnston, for provid-
ing the samples and valuable technical advice.
199
References
I l ] J. Ocslreich. D J . L Monette, An investigalion of Envirostrip starch
media coaling removal from 2024-T3 aluminum alloys. Proceedings
of the 1994 DOD/Industry Advanced CoatingsRemoval Conference,
New Orleans. LA, 17-19 May. 1994, pp. 113-152.
12 ] G. Ncumeister, J. Kozol. Type V plastic media blasting of monolithic
graphite epoxy composites. Proceedings of the 1994 DOD/lndustry
Advanced Ctlalings Removal Conference. New Orleans. LA, 17-19
May. 1994, pp. 409-424.
131 T. Mickle. J, Talia, Paint removal on composites by particle blasting,
Proceedings of the 1993 DOD/lndustry Advanced Coalings Removal
Conference. Phoenix, Arizona, 25-27 May, 1993, pp. i 10---129.
141 P.H. Shipway. I.M+ Hulchings. Measurement of coating durability by
solid parllcIe erosion. Surface Coatings and Technology 71 ( 1995 ) I-
S.
[5i K,L. Johnson, Contact Mechanics, Cambridge University Press,
Cambridge. !985, pp. 351-373.
16i K.L. Johnson, Contact Mechanic~, Cambridge University Press,
Cambridge. 1985. pp. 136-142.
171 A.C. Ramamunlly. W.I. Lorenzen, S.J. Bless, Stone impact damage to
aulomoti~'e paint linishes: an ir~r~nluction to impact physics and impact
induced corrosion, Progress in Organic Coatings 25 (1994) 43-71,
IS1 A.C. Ramamurthy, W.I. Lorenzen, A,M. Rajendran, D. Grove,
Proceedings of the ACS, PMSE Division 67 (1992) 43-44.
191 R.M. Broth, Impact dynamics with applications to solid particle
erosion, International Journal of Impact Engineering 7 ( I ) 11988)
37-53.
I ]0 ] I+M.Hutchings, Mechani sins of,he erosion of metals by solid particles.
in: W.F. Adler lEd. ) Erosion: Prevention and Uselul Applications,
ASTM STP 604, American Society lot Testing and Materials, 1979,
pp. 59-76.
i I I I M,J. Matthewson. Ant-symmetric contact on thin compliant coatings,
Journal of the Mechanics and Physics of Solids 29 (2) ( 1981 ) 89-
113.
1121 Y.A. Rossikhin, M.V. Shhikova, The impact of a rigid sphere with an
elastic layer of linite thickness, Acla Mechanica 112 ( 1995 ) 83-93.
113 [ M, Lu, Observations of indentation cracks in brittle coating/substrate
systems, Journal of Applied Physics 60 (8) (1989) 3916--3918.
1141 N.N. Dioh, J.G. Williams, The impact behavior of paints, Journal of
Materials Science 29 (1994) 6091-6096,
1151 l.M. Hutchings, R.E. Wimcr, A simple small-bore laboratory gas-gun,
Jottmal of Physics E: Scien,ific Instruments, 8 ( 19751 85-86.
[161 M.E. Graham, J.D. Carlyle, T.L. Menna, Facility for high-speed
panicle impact testing, The Review of Sciendlic [nstrumenls 46 (9)
1975) 1221-1225.
[ 17] I.M. Hmchings, M.C, Rochester, J.J. Comus, A rectangular bore gas
gun. Jounal of Physics E: Scientitic instruments l0 { 1977) 455-457,
[ 181 R.W. White. R. Fowles, Effect of valve opening lime on gas gun
performance, The Review of Scientilic Instruments 39 (9) (1975)
1296-1297.
[ ] 91 A. Hamade, A preliminary assessment of the mechanical properties of
organic coatings, M. Eng. Report, Department of Mechanical and
Industrial Engineering. University of Toronto, Toronto, Canada, 1996.
12(} I ASTM D3359-93, Standard les! methods for measuring adhesion by
tape test. Annual Book of ASTM Standards, American Society for
Testing and Materials.
121 ] B. Ratner, E.E. Styller, Characteristics of impact fete,ion and wear of
polymeric materials, Wear 73 ( 1981 ) 213-234.
]221 K.L. Johnson, Contact Mechanics, Cambridge University Press,
Cambridge, [985, pp. 84-106.