Wear 255 (2003) 38–43

Abrasive wear analysis using factorial experiment design

J. Esteban Fernández, Ma del Rocı́o Fernández∗ , R. Vijande Diaz, R. Tucho Navarro
Faculty of Engineering, Oviedo University, Campus de Viesques s/n, 33203 Gijón, Spain

Abstract
This study describes multi-factor-based experiments that were applied to researching and investigating an abrasive wear system of
Ni-based alloy coatings both with and without WC reinforcement. The aim of the study was to evaluate the impact of factors such as the
load applied, WC reinforcement particles, abrasive grain size and environment on abrasive wear. Our results highlight how the grain size
of the abrasive and the WC reinforcement particles both have a major abrasive effect. This contrasts starkly with the impact of the load
applied and the environment, which have a far lesser effect and a very slight effect respectively. The relationship between these four factors
and wear loss is described, and the mechanics of how wear takes place is discussed.
© 2003 Elsevier Science B.V. All rights reserved.
Keywords: NiCrBSi; NiCrBSi + WC; Grain size of abrasive; Load; Reinforcement; Environment

1. Introduction the properties of the abrasive, test conditions, equipment

used, and so on and so forth, that make any direct com-
Although abrasion, or abrasive wear, is a prominent type parison or combination of research results totally unfeasi-
of wear and tear ubiquitous to all fields of industry, it is ble.
particularly relevant to the fields of agriculture, mining and In view of the above situation, a number of statistical
mineral processing [1,2]. In response to this problem, a num- methods have recently been implemented in wear studies [9].
ber of Ni- and Co-based alloys that exhibit high strength, All the methods share the advantage of facilitating research
hardness and excellent wear and corrosion resistance prop- into the effects of different factors and their interactions,
erties have recently been developed as coating materials for by limiting the number of tests that are performed. Factors
use in these fields [3,4]. can thus be studied not only individually but also in the
The whole phenomenon of abrasion is a complicated mat- combined effect that they exert.
ter that is influenced by a range of different factors. Such This study applies a factor-based research methodology
factors, as the properties of the materials coming into con- to an Ni-based alloy coating abrasive wear system with and
tact with each other, the service conditions or the environ- without hard WC reinforcement particles in order to gain a
ment all play their part in abrasive wear [5]. A comprehen- global understanding of the impact of four factors-applied
sive, overall understanding of abrasion is therefore called load, the addition of WC reinforcement, the abrasive grain
for, and there are indeed many studies on the subject, in- size and the environment.
cluding some on wear of coatings [6–8]. However, most re-
search has investigated only a single dimension of the re-
lationship between the different factors involved in wear 2. Experimental procedure
loss. Such research is valuable, accurate and detailed, but
fails to allow any overall evaluation of the cumulative ef- 2.1. Preparation of the NiCrBSi and NiCrBSi + WC
fect of all the factors and their interactions to be extrapo- coatings
lated from them. Nor can this general overview be achieved
by accretion of the results of the multifarious studies, as In this paper, two alloys were studied, as can be seen in
there are inevitably differences between the different re- Table 1, that have similar composition on their matrix, but
search experiments in terms of materials that are tested, one of them is reinforced with a second phase WC hard
particles. The alloys were plasma sprayed onto the substrate
∗ Corresponding author. Tel.: +34-985-18-19-16/10-40-21; of an AISI 1010 steel bar. After the coating, the layer has
fax: +34-985-10-20-60. been torch fused. The coated specimens, were 15 mm ×
E-mail address: rocio@correo.uniovi.es (M. del Rocı́o Fernández). 20 mm × 52 mm in size and the coat layer about 2–3 mm

0043-1648/03/$ – see front matter © 2003 Elsevier Science B.V. All rights reserved.
doi:10.1016/S0043-1648(03)00103-0
 J. Esteban Fernández et al. / Wear 255 (2003) 38–43 39

Table 1
The composition and properties of the coatings
Coating Commercial mark Chemical composition (%) Hardness (Hv0.3 )

WC Cr Fe Si B C Ni

NiCrBSi CASTOLIN CPS 1235 15.2 3.6 4 3.1 0.6 Balance 855 ± 20
NiCrBSi + WC METCO 36C 35 11 2.5 2.5 2.5 2.5 Balance Matrix: 840 ± 60; WC: 1290 ± 150

thick. The details of the spray processing parameters were Table 2

reported in previous work [10]. Levels of the four factors
Factor Low level High level

2.2. Determining the factors to be studied R (reinforcement) NiCrBSi + WC NiCrBSi

L (load) 47 N 140 N
A (abrasive) 80 mesh 30 mesh
In view of the fact that Rabinowicz’s classic theory [11] E (environment) Dry Wet
claims that applied load and hardness of materials are the
most important factors affecting the abrasion process, these
two factors were considered first in this study. The applied
loads were set at two levels of 47 and 140 N, in the same Al2 O3 with a similar grain angularity were selected. The
order of magnitude as is specified in the ASTM G65 stan- factor of roundness [16] for both abrasive sizes were 0.68
dard [12]. The hardness of the alloys was tested with a mi- and 0.70, respectively. The effects dry and wet environments
crohardness SHIMAZU 2000V and are in the same order have remained unclear till now [17–19] and were therefore
of magnitude (Table 1) for both alloys and it was clearly considered as the fourth factor. The two levels of the four
superior for the reinforcement particles. factors selected are summarized in Table 2.
The microstructure of the matrix, analysed by means of
optical and scanning electron microscopy (SEM) was, quite 2.3. Factorial design of experiments
similar. In Fig. 1, can be appreciate the structure of different
phases near the substrate for both alloys. Depending on the As the experiments in this study are factor based [16], all
layer obtaining method, the Ni-based alloys form very dif- possible factor combinations can be experimentally studied.
ferent microstructures [13]. Conde et al. [14] have study the The most important experiment design is known as 2k , and
microstructure of two alloys very similar and it was neces- is widely used in experiments where k factors and each of
sary the XRD diffraction analyses to determine the presence their two levels are investigated. Both levels, i.e. low and
of boron in the layer, due to its not possible to detect by high, in this paper refer to the levels that are expected to
means of energy disperse X-ray analysis EDX. decrease or increase the response value. With this design,
In view of the high wear resistance of Ni-based alloy 2k tests are performed and (2k − 1) effects are investigated,
coatings used in this work, Al2 O3 , which is harder, was used including k principle effects,
as the abrasive instead of the more commonly used SiO2 , to

accelerate the wear process [15]. The grain size and shape k
have great effects on the given abrasive. Eighty and 30 mesh 2

Fig. 1. SEM microscopy of transversal section of the specimens. The structure is similar for both matrix and they the reinforcement particles are
distinguished clearly.
40 J. Esteban Fernández et al. / Wear 255 (2003) 38–43

Table 3
Comparison among the parameters of abrasive wear test in ASTM standards and in the present experiment
Parameters ASTM G65 (dry) ASTM G105 (wet) Present experiment

Rotation rate 200 rpm 245 rpm 245 rpm

Applied load A, B and C: 130 N; D: 45 N 222 ± 3.6 N High level: 140 N ± 2%; low level:
47 N ± 2%
Abrasive SiO2 , round, 50/70 mesh SiO2 , round, 50/70 mesh Al2 O3 , angular, 80/30 mesh
Flow of abrasive 300–400 g/min Sand/water = 1.5 Dry: 300–400 g/min
Wet: sand/water = 1.5
No. of rotation A: 6000 rev/4309 m; 1000 rev/interval × four intervals; 1000 rev/interval × four intervals;
B: 2000 rev/1436 m; total: 2236.7 m total: 2236.7 m
C: 100 rev/718 m; D:
6000 rev/4309 m
Diameter/width of wheel (mm) 228.6/12.7 178/13 178/13
Hardness of wheel Single wheel: shore A 58–62 Intervals 1 and 2: 50; interval 3: 60; Single wheel: shore A 58–62
interval 4: 70
Specimen (mm) (3.2–12.7) × 25 × 76 (6.4–15.9) × 25.4 × 57. 9 15 × 20 × 52

interactions between two factors, and ASTM G105 so the obtained results can be compared

 [12,20].
k The present experiment was designed as 16 tests based on
3 the two-level, four-factor factorial design (24 ) to evaluate the
interactions between tree factors and one interaction of k factors material (M), load (L), abrasive (A) and environment
factors. To estimate these effects, the contrast associated with (E), as shown in Table 4. The tests corresponding to each
the effects is first determined. The contrast for the effect condition were repeated for three times. Here, “+” and “−”
AB · · · K is given by: refer to the high and low level of each factors, as shown in
Table 2.
contrastAB···K = (a ± 1)(b ± 1) · · · (k ± 1) (1)

where the small letters a, b, . . . , k refer to the high level of

the corresponding factors A, B, . . . , K in a combination of 3. Results and discussion
treatments, and 1 is replaced with (1) in the final expression.
The negative sign is used if the factor is included in that The results of the abrasive wear tests are shown in Table 5.
effect and the positive sign in the contrary case. Thus, the The relatively high variation in wear loss might be attributed
effect of AB · · · K is estimated by:
1
EAB···K = (contrastAB···K ) (2) Table 4
2k−1 Experiment design with the factorial design 24
The estimated values of these effects are represented in graph No. of test Factor Treatments
form in normal distribution paper. The non-significant ef- R L A E
fects can be assimilated to a straight line, while the signif-
icant ones will have an own cause of variation and clearly 1 − − − − (1)
2 + − − − r
differ from this alignment. 3 − + − − l
4 + + − − rl
2.4. Test design 5 − − + − a
6 + − + − ra
7 − + + − la
The wear tests were undertaken using a standard abrasive 8 + + + − rla
wear tester (ASTM G105), which was prepared to allow 9 − − − + e
a dry or wet sand abrasive wear test. Before testing, the 10 + − − + re
specimens were ground using #600 SiC paper to remove the 11 − + − + le
rough surface layer. Prior to and after each test, the specimen 12 + + − + rle
13 − − + + ae
was ultrasonically cleaned in acetone, blown dry with warm 14 + − + + rae
air, and then weighed to determine the weight loss by using 15 − + + + lae
an electronic balance having an accuracy of 0.0001 g. Except 16 + + + + rlae
for the parameters selected for the investigation, as shown Note: The symbols “+” and “−” refer to the high and low level for each
in Table 2, the others test conditions were set, as shown in factor, as shown in Table 2; (1) refers to a treatment where all the factors
Table 3, taking some parameters of the standards ASTM G65 are of low levels.
 J. Esteban Fernández et al. / Wear 255 (2003) 38–43 41

Table 5 wear. In stark contrast to these findings, our results point to

Results of the abrasion tests the fact that abrasive size and reinforcement exert a greater
Treatment Wear loss (g) effect on abrasive wear than the load that is applied. Simi-
− − − − 0.158 ± 0.024 lar results were also reported by Spuzic et al. [9] when they
+ − − − 0.227 ± 0.086 applied fractional design of experiments to evaluate the ef-
− + − − 0.522 ± 0.105 fects of force, temperature, material and sliding velocity on
+ + − − 0.404 ± 0.111 rolling–sliding abrasion. Their results showed that force had
− − + − 0.422 ± 0.072
less effect than the other three factors. It is interesting to note
+ − + − 1.443 ± 0.085
− + + − 0.645 ± 0.171 that Spuzic et al. applied a statistical method that is concep-
+ + + − 2.668 ± 0.351 tually similar to that of the present work. The vast amount of
− − − + 0.106 ± 0.107 work generated by the traditional physical approach to ex-
+ − − + 0.152 ± 0.071 periment design that has been applied in previous research
− + − + 0.072 ± 0.015
makes it difficult to evaluate the effects of various factors si-
+ + − + 0.163 ± 0.052
− − + + 0.565 ± 0.070 multaneously. This is the main reason why load has always
+ − + + 2.108 ± 0.149 been considered first in wear research, whilst other factors,
− + + + 0.575 ± 0.154 e.g. abrasive grain size, which may be more important, have
+ + + + 3.304 ± 0.460 not been given the attention they deserve. The advantage of
the statistical method is obvious.
to the high heterogeneity/variability of the metallurgical and
3.2. The effects of the four factors on the abrasion
mechanical properties of the coatings [15].
The two-level effects (with and without WC) and interac-
3.1. Evaluation of the effects of the factors and their
tions of each two factors of R (reinforcement), L (load), A
interactions
(abrasive gain size) and E (environment) on abrasive wear
are shown in Fig. 3. Average wear loss in this figure refers
Eqs. (1) and (2) were applied to the result values in Table 5 to the average value of one level of each of the two factors
in order to calculate the effects of the factors in Section 2.3, and is calculated by:
and were then plotted onto graphs, as shown in Fig. 2. This
figure clearly highlights how factor A (abrasive grain size), 1
ȲXi Xj = (x1 + 1)(x2 + 1) · · · (xi−1 + 1)xim (xi+1 + 1)
R (reinforcement), and interaction between AR all deviate 2k−2
markedly from the linear distribution and, therefore, have a × · · · (xj−1 + 1)xjn (xj+1 + 1) · · · (xk + 1) (3)
significant effects on abrasive wear. In contrast, L (normal
load) is seen to exert little effect and E (environment) is even where

less significant than L. 0 for low level
These results clearly contradict pre-test expectation. Ac- m, n =
1 for high level
cording to the classic theory [11], which is backed up by
most of the research documented in the literature, load and Xi and Xj refer to any two of the k factors (1 ≤ i, j ≤ k,
hardness are the two major factors in research on abrasive i
= j), and the small case letters x1 , x2 , . . . , xk are the high
level of the corresponding factors in a treatment. Here, k =
4 for the present work.
Fig. 3a–c demonstrate that NiCrBSi wear loss is patently
higher than that of NiCrBSi + WC, suggesting that WC
does indeed have a reinforcement effect. SEM microscopy
of the worn surfaces of NiCrBSi and NiCrBSi + WC are
shown in Fig. 3. The worn surface of NiCrBSi shows ob-
vious evidence of cutting and ploughing (Fig. 3a). How-
ever, no obvious plastic deformation can be found on the
worn surface of NiCrBSi +WC (Fig. 3b). Although there
are some slight traces of ploughing to be seen the matrix
areas, these cease when WC particles are encountered, indi-
cating that the WC particles effectively stopped the abrasive
from cutting or ploughing into the surface layer during the
wear process, thereby noticeably lowering wear loss. The
resistance of WC to the abrasive could be attributed to its
Fig. 2. Normal probability of the effects of the factors and their interac- higher hardness. Further observation of Fig. 3b also provide
tions. evidence of cutting and cracking of the WC. This is due to
42 J. Esteban Fernández et al. / Wear 255 (2003) 38–43

Fig. 3. Two-level effects and interactions of R (reinforcement), L (load), A (abrasive gain size) and E (environment) on the abrasive wear.

the hardness of the WC in the coating (Table 1) being lower

than that of abrasive Al2 O3 (HV1800) [21]. It is reasonable
to believe that wear resistance of the coatings will increase
if the hardness of the WC particles is improved.
Fig. 3b, d and f indicate that grain size of the abrasive
has the greatest effects on abrasion. Wear loss obviously
increases as grain size increases from 80 to 30 mesh. The
weight loss gradient is much higher for NiCrBSi compared
with NiCrBSi + WC, demonstrating considerable interac-
tion between R and A, as indicated in Fig. 1. This suggests
that the extent of the impact of abrasive grain size depends
to a large extent on the nature of the materials, or specifi- Fig. 4. SEM microscopy of worn surface of NiCrBSi at
cally in this study, on the coating reinforcements. wet/30 mesh/140 N. The sliding direction is from right to left.

Fig. 5. SEM microscopy of worn surface of the specimens at dry/30 mesh/140 N. The sliding direction is from right to left.
 J. Esteban Fernández et al. / Wear 255 (2003) 38–43 43

As shown in Fig. 3a, d and e, wear loss increases with the References
applied load. Effects are, however, much lower compared to
those of reinforcement and abrasive grain size. Fig. 3c, e and [1] T.S. Eyre, Wear characteristics of metals, Tribol. Int. 10 (1976) 203–
f also point to how the effects of the environment are slight. 212.
[2] J.H. Tylczak, A. Oregon, Abrasive wear, in: Friction, Lubrication,
When the environment is changed from dry to wet, wear
and Wear Technology, ASM Handbook, vol. 18, 1995, 184–
loss exhibits a contradictory tendency to either increase or 190.
decrease slightly. The worn surface of NiCrBSi in a wet en- [3] B.A. Kushner, E.R. Novinski, Thermal Spray Coatings, ASM
vironment is shown in Fig. 4. Except for a little more plas- Handbook, vol. 18, ASM, 1992, pp. 829–833.
tic deformation corresponding to the slightly greater weight [4] P. Crook, Friction and Wear of Hardfacing Alloy, ASM Handbook,
vol. 18, ASM, 1992, pp. 758–765.
loss, as shown in Fig. 3c, wet wearing is similar to dry wear-
[5] K.H. Zum Gahr, Relation between abrasive wear rate and
ing (Fig. 5a). microstructure of metals, in: Proceedings of International Conference
The effects of the environment have yet to be clarified. on Wear of Materials, Reston, 1983, pp. 266–274.
Whilst some researchers claim that a wet environment pro- [6] V.H. Hidalgo, J.B. Varela, A.C. Menendez, S.P. Martinez,
motes wear due to hydrogen attack [17] or higher sand load- High temperature erosion wear of flame and plasma-sprayed
nickel–chromium coatings under simulated coal-fired boiler
ing at the contact zone [22], others affirm that water reduces
atmospheres, Wear 247 (2000) 214–222.
the tendency for abrasive particle embedment and has lubri- [7] S. Hogmark, S. Jacobson, M. Larsson, Design and evaluation of
cation effects, leading to lower wear loss than in dry envi- tribological coatings, Wear 246 (2000) 20–33.
ronments [18,23]. We tend towards the belief that various [8] J. Mateos, J.M. Cuetos, E. Fernández, R. Vijande, Tribological
mechanisms, as stated above, may act simultaneously during behaviour of plasma-sprayed WC coatings with and without laser
remelting, Wear 239 (2000) 274–281.
abrasive wear in wet environment, and that the extent of the
[9] S. Spuzic, M. Zec, K. Abhary, R. Ghomashchi, I. Reid, Fractional
different effects depends on test conditions. If mechanisms design of experiments applied to a wear simulation, Wear 212 (1997)
for promoting wear and for reducing wear counteract, the 131–139.
environment will have a limited effects, as this work indeed [10] A. Martin, J. Rodrı́guez, J.E. Fernández, R. Vijande, Sliding wear
showed. behaviour of plasma sprayed WC–NiCrBSi coatings at different
temperatures, Wear (2001) 1017–1022.
[11] E.D. Rabinowicz, Friction and Wear of Work Hardening in the Design
of Wear Resistant Materials, Wiley, New York, 1965, p. 168.
4. Conclusions [12] ASTM G65-94, Standard test method for measuring abrasion
using the dry sand/rubber wheel apparatus, in: Annual Book
A statistical method, i.e. the factorial experiment design, of ASTM Standards, vol. 3.02, ASTM, Philadelphia, PA, 1995,
was adopted to investigate the effects of reinforcement, load, p. 232.
[13] Q. Li, D. Zhang, T. Lei, C. Chen, W. Chen, Comparison of
abrasive grain size and environment on abrasive wear of laser-clad and furnace-melted Ni-based alloy microstructures, Surf.
Ni-based coating alloy. A wealth of information was ob- Coat. Technol. 137 (2001) 122–135.
tained based on the limited number of tests. The results can [14] Z.A.F. Conde, J. De Damborenea, Cladding of Ni–Cr–B–Si coatings
be summarized as follows: with a high power diode laser, Mater. Sci. Eng. 334 (2002) 233–
238.
1. Abrasive grain size exerted the greatest effect on abrasive [15] A.C. Bozzi, J.D. Baisoli del Mello, Wear resistance and wear
wear, followed by reinforcement. The load applied had mechanisms of WC–12% Co thermal sprayed coatings in three-body
a much lower effect, and the environment was similarly abrasion, Wear 233–235 (1999) 575–587.
[16] H. Berns, S. Koch, Influence of abrasive particles on wear mechanism
found to have minor effect. and wear resistance in sliding abrasion tests at elevated temperatures,
2. The addition of WC reinforcement particles improved Wear 233–235 (1999) 424–430.
wear resistance of the NiCrBSi alloy coating. Increasing [17] J.P. Tu, M.S. Liu, Wet abrasive wear of ordered Fe3 Al alloys, Wear
abrasive grain size led to obviously greater wear, espe- 209 (1997) 31–36.
cially for NiCrBSi without WC. Wear loss increased with [18] S. Wirojanupatump, P.H. Shipway, A direct comparison of wet and
dry abrasion behaviour of mild steel, Wear 233–235 (1999) 655–665.
load applied, but showed unclear tendencies as regards [19] D.C. Montgomery, Diseño y Análisis de Experimentos (translated
the influence of the environment. by L.J. Delgado), Grupo Editorial Iberomaérica, México, 1991,
3. The principle wear mechanism was micro-cutting and pp. 175–284 (in Spanish).
micro-ploughing in both NiCrBSi and NiCrBSi + WC [20] ASTM G105-89, Standard test method for conducting wet
coatings, except for fracturing accompanied by WC par- sand/rubber wheel abrasive tests, in: Annual Book of ASTM
Standards, vol. 3.02, ASTM, Philadelphia, PA, 1995, p. 418.
ticles. Increasing the hardness of the reinforcement could [21] K.H. Zum Gahr, Wear by hard particles, in: I.M. Hutchings (Ed.),
improve wear resistance. New Directions in Tribology, Mechanical Engineering Publication
Ltd., Burry, St. Edmunds, 1997, pp. 483–494.
[22] R.L. Deuis, C. Subramanian, J.M. Yellup, Three-body abrasive wear
Acknowledgements of composite coatings in dry and wet environments, Wear 214 (1998)
112–130.
[23] S. Wirojanupatump, P.H. Shipway, Abrasion of mild steel in wet and
The authors want to acknowledge the research support dry conditions with the rubber and steel wheel abrasion apparatus,
from CICYT to the project: CC-98-MAT0939-C02-01. Wear 239 (2000) 91–101.