Int J Adv Manuf Technol (2017) 92:393–406

DOI 10.1007/s00170-017-0137-0

ORIGINAL ARTICLE

Effect of mechanical vibration on Ra, Rq, Rz, and Rt

roughness parameters
Jefferson Dobes 1 & Jose Eduardo Silveira Leal 1 & Jonas Profeta 1 &
Marcos Morais de Sousa 1 & Francisco Paulo Lépore Neto 1 & Antonio Piratelli-Filho 2 &
Rosenda Valdés Arencibia 1

Received: 10 August 2016 / Accepted: 6 February 2017 / Published online: 24 February 2017
# Springer-Verlag London 2017

Abstract The effect of mechanical vibration on the roughness them were statistically significant with a 95% confidence level
parameters (Ra, Rq, Rz and Rt) obtained by an electromechan- for all parameters evaluated. In the stainless steel sample, fre-
ical surface roughness tester in two samples, a stainless steel quency produced the greatest effects on the values of rough-
with Ra equal to 0.369 μm and a carbon steel with Ra of ness parameters, except for Ra value, which was more influ-
5.342 μm, is discussed in this article. The tests were performed enced by the acceleration. It was also found that the effect of
by introducing vibration with different values of frequency and the interaction between frequency and acceleration produced
amplitude. The measurement uncertainty associated with all significant effects on the values of Rq and Rz. In the carbon
roughness parameters was estimated by applying the GUM steel sample, the interaction between frequency and accelera-
method proposed in the JCGM 101. The analysis of variance tion was the main generator of changes in the average values of
(ANOVA) technique was applied, and it was observed that the all parameters evaluated. For Ra, Rq, and Rz, the frequency
variables frequency and acceleration plus interaction between was also a statistically significant variable.

Keywords Contact measurement . ANOVA . Pareto chart .

Measurement uncertainty
* Jose Eduardo Silveira Leal
[email protected]

Jefferson Dobes 1 Introduction

[email protected]
Jonas Profeta The surface texture is a key factor that affects the functionality
[email protected] and reliability of components [1], and the surface roughness is
Marcos Morais de Sousa
a crucial parameter for evaluating the material surfaces since it
[email protected] directly affects the optical and mechanical properties of the
materials. It determines the characteristics of wear, friction,
Francisco Paulo Lépore Neto
[email protected] lubrication, corrosion, fatigue, heat transfer, optical properties,
fluid flow, and measuring surfaces, among others [2]. For this
Antonio Piratelli-Filho
reason, roughness is one of the most important factors in tri-
[email protected]
bology, and it can be used to evaluate the quality of a machin-
Rosenda Valdés Arencibia ing operation [3].
[email protected]
The techniques used for measuring the surface roughness
1
can be divided into two categories, which are contact and non-
Campus Santa Mônica, Federal University of Uberlândia, 2121, João
Naves de Ávila Avenue, Building 5F, Uberlândia, MG 38400-902,
contact methods. The contact method involves dragging a
Brazil measurement tip across the surface [2], whereas non-contact
2
Faculty of Technology, Department of Mechanical Engineering,
methods are based on imaging and microscopy principle [1].
Campus Darcy Ribeiro, University of Brasília, Asa Norte, In the contact measurement, a stylus tip comes into physi-
Brasília, DF 70910-900, Brazil cal contact with the surface to be evaluated and a piezoelectric
394 Int J Adv Manuf Technol (2017) 92:393–406

or inductive transducer converts the vertical movement into an investigate the effects of mechanical vibration on the average
electrical signal [4]. Non-contact measurement can be per- values of roughness parameters (Ra, Rq, Rz, and Rt) when
formed by measurement systems by interferometry, by induc- surface inspection is conducted post-process operation. The
tion (inductive sensor), and atomic force (atomic force micros- uncertainties associated with the roughness measurement are
copy (AFM)). also estimated.
The contact method is the most common technique for
characterizing surfaces [4, 5], and the electromechanical
roughness tester is the most used for roughness measurements.
2 Methodology
The contact roughness measurement results can be affected by
several factors, including the measurement instrument, the
The roughness of three samples was measured by an electro-
sample, the operator, the measurement strategy, and the envi-
mechanical roughness tester, digital model TR200, the
ronment. On the other hand, each of these factors is related to
HOMIS TIME (Fig. 2). This equipment has a resolution of
other sources as shown in Fig. 1.
1 nm and a nominal range of 160 μm. The calibration certif-
The manufacturing conditions of the factory floor environ-
icate declared an expanded uncertainty associated with the
ment (temperature gradients, dust, dirt, and mechanical vibra-
roughness tester calibration of 0.060 μm for a coverage factor
tion) can influence the roughness values and even invalidate
of 2.57 and a coverage probability of 95%.
them [4]. Thus, technical standards such as ISO 3274 [6]
The samples evaluated are made of aluminum, stainless
specify the operating conditions for measurement using stylus
steel, and carbon steel, respectively, through the turning pro-
instruments. The non-fulfillment of these nominal conditions
cess. The sampling length was determined according to the
leads to significant deviations of the values of roughness pa-
ISO 12085 standard [16] and was equal to 0.8 mm for the
rameters. For small roughness values in the order of a few
aluminum and stainless steel samples and 2.5 mm for the
tenths of micrometers, the amplitude of the mechanical vibra-
carbon steel sample. Five sampling lengths were considered,
tion in the environment can be the same order of magnitude as
resulting in an evaluation length equal to 4.0 and 12.5 mm,
that of the roughness.
respectively. In all cases, the Gaussian filter was applied on the
Thereby, the roughness measurement by stylus instruments
primary extracted profile. A diamond stylus tip with radius
should be performed in the absence of mechanical vibration.
equal to 2 μm was used. The values of the Ra, Rq, Rz, and
However, isolation and vibration control are costly and often
Rt parameters were collected. In all cases, five measurement
impractical, particularly on the factory floor. Thus, mechanical
cycles were performed.
vibration levels can be high in these places due to the presence
of equipment and machine tools.
Studies about surface roughness of machined parts using 2.1 Measurement of roughness in nominal conditions
direct or indirect measurement method have been reported. In
general, these studies are dedicated to identifying the effect of Initially, the surface roughness was measured at nominal con-
the cutting conditions on the surface finish of the machined ditions, wherein
parts [7–12] and to correlating the surface roughness and the
cutting vibration [3, 7, 8, 13–15]. a) The roughness tester and the samples were placed on an
The effect of mechanical vibration on roughness measuring inertial table in order to avoid the effect of mechanical
is not well reported in literature. However, this work aims to vibration that is transmitted through to the ground. It

Fig. 1 Sources of errors in

contact roughness measurement
Int J Adv Manuf Technol (2017) 92:393–406 395

Fig. 2 Measurement system used

to measure the surface roughness. (4)
1 inertial table, 2 acrylic platform,
3 roughness tester, 4
electrodynamic vibration exciter, (3) (5)
and 5 accelerometer (2)

(1)

ensures that the frequency and the acceleration values of Simultaneously with the measurement of surface rough-
mechanical vibration can be considered equal to zero; ness, the amplitude of mechanical vibration was monitored.
b) The measurements were performed at a controlled room Therefore, a piezoelectric accelerometer coupled to a system
temperature of (23.0 ± 1.0) °C and at a relative humidity comprising a signal amplifier (brand PCB, model PCB-482-
of 55%. During the measurement, the air-conditioning A-20) was used, a signal analyzer (SPECTRAL DYNAMICS,
system was turned off in order to minimize the effect of SD 380), a power amplifier model 2712, a driver, a 1049
mechanical vibration generated by it. model signal generator manufactured by Brüel and Kjaer,
and a computer with a software specifically developed for
Figure 2a shows the roughness tester and the aluminum signal acquisition. Figure 3 shows the measuring system used
sample placed on the inertial table. for the experiments.
The piezoelectric accelerometer used is model PCB-352-
2.2 Pre-tests C-33, number of series 97,521, and it was fixed with wax on
the surface on the system’s center of mass (roughness tester,
The pre-tests consisted in measuring the roughness parameters sample, and acrylic platform) in order to ensure the highest
(Ra, Rq, Rz, and Rt) of an aluminum sample on the frequency incidence of mechanical vibration.
band from 0 to 500 Hz by using the measurement system The different levels of vibration were introduced by the
described above. The tests were performed by applying me- signal generator coupled to a power amplifier associated with
chanical vibration in the perpendicular direction to the probe the driver in order to have complete knowledge and control of
displacement for different values of amplitude. the frequency, amplitude, and acceleration of the mechanical

Fig. 3 Measuring system used to

perform the tests. 1 inertial table,
2 electrodynamic vibration
exciter, 3 acrylic platform, 4
electromechanical surface
roughness tester model TR200, 5
piezoelectric accelerometer, 6
sample, 7 computer 1, 8 power
amplifier, 9 signal generator, 10
signal amplifier, 11 signal
analyzer, 12 computer 2
396 Int J Adv Manuf Technol (2017) 92:393–406

Table 1 Factorial design (32) both over the driver, an acrylic device was designed and
Run Factor 1—frequency (Hz) 2
Factor 2—acceleration (m/s ) manufactured (Fig. 2b). This device was attached to the driver
by means of four screws. The mechanical vibration was ap-
1 −1 (60) −1 (0.6) plied in the perpendicular direction to the probe displacement.
2 0 (80) −1 (0.6) All measurements were carried out at a controlled room
3 +1 (100) −1 (0.6) temperature of (23.0 ± 2.0) °C. The instruments and samples
4 −1 (60) 0 (1.0) used in the measurement tests were exposed to this tempera-
5 0 (80) 0 (1.0) ture for approximately 12 h prior to the measurements. The
6 +1 (100) 0 (1.0) roughness tester was turned on 3 h before starting the mea-
7 −1 (60) +1 (1.6) surements. The temperature was monitored by the digital
8 0 (80) +1 (1.6) thermo-hygrometer with resolution of 0.1 °C and nominal
9 +1 (100) +1 (1.6) range of −20.0 to 60.0 °C. The calibration certificate of this
equipment declares an expanded uncertainty of 0.3 °C for a
coverage factor of 2.00 and a coverage probability of 95.45%.
vibration generated. The acceleration values were varied using
the signal generator. This equipment was adjusted to different 2.4 Analysis of variance
signals of amplitudes 250, 500, 750, and 1000 mV.
The statistical analysis was carried out by applying the analy-
2.3 Measurement of roughness at different nominal sis of variance (ANOVA) technique and the Pareto chart by
conditions using software STATISTICA®. The F test was applied to
verify the null hypothesis that all levels of both variables (fre-
From the results obtained in pre-tests and in order to investi- quency and acceleration) and their combinations produced the
gate the effect of mechanical vibration on the values of rough- same evaluated roughness parameter mean (H0 μ1 = μ2 = μ3).
ness parameters (Ra, Rq, Rz, and Rt), a factorial design (32)
was proposed. The investigated factors were the frequency 2.5 Measurement uncertainty evaluation
and the acceleration and both on the three levels, as shown
in Table 1. In order to estimate the measurement uncertainty, the concepts
Factor 1 (frequency) assumed the values of 60, 80, and and recommendations presented in the Joint Committee for
100 Hz, and factor 2 (acceleration) adopted the values of Guides in Metrology (JCGM) 100 [17] were followed. The
0.6, 1.0, and 1.6 m/s2. In accordance with the planning matrix, application of the Group on the Expression of Uncertainty in
9 tests were performed with 5 repetitions in a total of 45 tests Measurement (GUM) method proposed on this document re-
for each sample. quires a mathematical model of the measurement process;
The frequency values (60, 80, and 100 Hz) and the accel- e.g., the measurand (output variable) must be expressed as a
eration values (0.6, 1.0, and 1.6 m/s2) were obtained by a function of the input variables, as shown in Eq. (1).
driver, on which the roughness tester and the sample were
placed (Fig. 2b). In order to ensure the proper positioning of Y ¼ f ðW 1 ; W 2 ; :::; W N Þ ð1Þ

Fig. 4 Spectra of frequency of Test 1 - White noise

the four conditions evaluated in
pre-tests (signals denoted by vib1
(250 mV), vib2 (500 mV), vib3
(750 mV), and vib4 (1000 mV))
Arms [m/s²]
Int J Adv Manuf Technol (2017) 92:393–406 397

Fig. 5 Image of the aluminum

sample obtained by SEM after
roughness measurement with a
TR200 HOMIS
electromechanical roughness
tester with a stylus tip of 2 μm of
radius, 10-μm scale

where Y represents the output variable and W1, W2, …, WN are The coverage probability of the combined standard un-
the input variables. certainty is approximately 68.27% only. For this reason,
The law of propagation of uncertainty can be applied in the International Committee for Weights and Measures
Eq. (1) in order to determine the combined standard uncertain- (CIPM) proposes the description of measurement uncer-
ty (uc) related to the measurand, so that Eq. (2) is obtained. tainty using an interval that represents the expected values
for the measurement error, with a known probability. The

2 expanded uncertainty (U) was used by the CIPM to de-
N ∂f
u2c ðyÞ ¼ ∑ u2 ðwi Þ ð2Þ scribe such interval, as shown in Eq. (3).
i¼1 ∂wi
N −1 N ∂f ∂ f     U ¼ k⋅uc ð3Þ
þ2 ∑ ∑ uðwi Þ⋅u w j ⋅r wi ; w j
i¼1 j¼iþ1 ∂wi ∂w j
where uc is the combined standard uncertainty associated
In Eq. (2), y is the estimate of the output variable Y; wi with the measurand and k is the coverage factor.
is the estimate of variable Wi; u2(wi) is the variance asso- The coverage factor k relates to the probability distribution
ciated with wi, for i varying from 1 to N; N is the number of the values obtained during measurement, which usually
of variables that affect the values of the measurand Y; presents a normal distribution. For situations where this prob-
u(wi) is the uncertainty associated with the error source ability is not normally distributed, the central limit theorem
represented by the estimate wi; and r(wi, wj) is the corre- together with the Student’s t distribution can provide a cover-
lation coefficient between the estimates wi and wj. age factor based on the effective degree of freedom of the

Fig. 6 Average values of the 0.8 0.8

roughness parameters (Ra, Rq,
Rz, and Rt) of a stainless steel 0.6 0.6
Rq (µm)

sample for the different

Ra (µm)

frequencies and accelerations and 0.4 0.4

the expanded uncertainty (95%) 0.2 0.2

0.0 0.0
60 80 100 60 80 100
Frequency (Hz) Frequency (Hz)

6.0 10.0
5.0 8.0
Rz (µm)

Rt (µm)

4.0
6.0
3.0
4.0
2.0
1.0 2.0
0.0 0.0
60 80 100 60 80 100
Frequency (Hz) Frequency (Hz)

a = 0.000 m/s2 a = 0.620 m/s2 a = 0.977 m/s2 a = 1.610 m/s2

398 Int J Adv Manuf Technol (2017) 92:393–406

Table 2 ANOVA results for parameter Ra (stainless steel sample) Table 4 ANOVA results for parameter Rz (stainless steel sample)

Factor ANOVA Ra (stainless steel sample); R2 = 0.27327 Factor ANOVA Rz (stainless steel sample); R2 = 0.62565

SS df MS F p SS df MS F p

(1) Frequency 0.007053 1 0.007053 2.991101 0.091247 (1) Frequency 36.93415 1 36.93415 45.79110 0.000000
(2) Acceleration 0.020856 1 0.020856 8.844399 0.004907 (2) Acceleration 0.53387 1 0.53387 0.66189 0.420593
Interaction (1) and (2) 0.008446 1 0.008446 3.581709 0.065494 Interaction (1) and (2) 17.80196 1 17.80196 22.07094 0.000029
Error 0.096682 41 0.002358 Error 33.06975 41 0.80658
Total SS 0.133038 44 Total SS 88.33972 44

standard measurement uncertainty. The effective degree of random measurement error, the plastic deformation of the sur-
freedom νeff is obtained using the Welch-Satterthwaite formu- face, and the tip geometry. Others factors such as the thermal
lation, as shown in Eq. (4). expansion of the instrument due to temperature variations in
the room, the vibration of the instrument, and the effect of
u4c ðyÞ filters should also be considered. For those more complexity
veff ¼ ð4Þ
N u4 ð w Þ parameters, such as Ra, Rq, Rsk, and Rku, the author sug-
i
∑ i
i¼1 vi gested that the measurement uncertainty should be equal to
twice the standard error of the mean of at least 12 measure-
According to Leach 2014 [4], to estimate the measurement ments considering 95% of confidence.
uncertainty associated with some roughness parameters is a The mathematical models used to calculate all evaluated
rather difficult task due to the large number of factors that can parameters are shown from Eqs. (5), (6), (7), and (8).
contribute to the uncertainty. It is usually attempted only by
the laboratories close to the top of the traceability chain. 11
It can be added that each roughness parameter is obtained Ra ¼ ∫ jZ ðxÞjdx ð5Þ
l 0
by applying a particular equation, and a different mathemati-
cal model must be used for evaluating the uncertainty associ- sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1
ated with each one. In some cases, the application of the law of 1
Rq ¼ ∫ Z 2 ðxÞdx ð6Þ
propagation of uncertainty can be difficult because the math- l 0
ematical models are complex. In other cases, some parameters
do not followed a normal distribution, and the GUM method In Eqs. (5) and (6), Z(x) represents the values of the ordi-
cannot be applied. As an alternative, the Monte Carlo method nates of the profile points in the sampling length 1.
proposed in JCGM 100 [17] must be used.
An important contribution was given by [4], who proposed Rz ¼ Zps þ Zvs ð7Þ
a simple model to estimate the uncertainty associated with Rp,
Rv, Rz, Rc, and Rt parameters. This model takes into account In Eq. (7), Zps represents the maximum height of profile
five main factors, which contribute to the uncertainty in the peaks and Zvs the largest of the depths of profile valleys, and
measurement of vertical displacement. These factors are the both are determined in the sampling length.
actual height value at a given point in the measurement, the
imperfections in the profile of the datum slideway or skid, the Rt ¼ Zpe þ Zve ð8Þ

Table 3 ANOVA results for parameter Rq (stainless steel sample) Table 5 ANOVA results for parameter Rt (stainless steel sample)

Factor ANOVA Rq (stainless steel sample); R2 = 0.37443 Factor ANOVA Rt (stainless steel sample); R2 = 0.19556

SS df MS F p SS df MS F p

(1) Frequency 0.120714 1 0.120714 16.46288 0.000217 (1) Frequency 14.2196 1 14.21959 5.803485 0.020571
(2) Acceleration 0.004613 1 0.004613 0.62909 0.432255 (2) Acceleration 2.9066 1 2.90661 1.186283 0.282441
Interaction (1) and (2) 0.089780 1 0.089780 12.24416 0.001139 Interaction (1) and (2) 17.80196 1 7.29511 2.977377 0.091967
Error 0.300631 41 0.007332 Error 100.4574 41 2.45018
Total SS 0.515738 44 Total SS 124.8787 44
Int J Adv Manuf Technol (2017) 92:393–406 399

Pareto Chart of Standardized Effects; Variable: Ra Pareto Chart of Standardized Effects; Variable: Rq
Ra Rq

(2)Acceleration(L)
2,9739 (1)Frequency(L) 4,05

Interaction Interaction
(1) and (2) 1,89254 (1) and (2) 3,499166

(1)Frequency(L) 1,72948 (2)Acceleration(L) ,7931531

p=,05 p=,05
Standardized Effect Estimate (Absolute Value) Standardized Effect Estimate (Absolute Value)

Pareto Chart of Standardized Effects; Variable: Rz Pareto Chart of Standardized Effects; Variable: Rt
Rz Rt

(1)Frequency(L) 6,76691 (1)Frequency(L)

2,409

Interaction 4,697972 Interaction

(1) and (2)
(1) and (2)
1,725508

(2)Acceleration(L) -,813566
(2)Acceleration(L) 1,089166

p=,05
Standardized Effect Estimate (Absolute p=,05
Value) Standardized Effect Estimate (Absolute Value)

Fig. 7 Pareto charts of the factorial design 32 of stainless steel sample (confidence level of 95%)

In Eq. (8), Zpe corresponds to the largest peak height of the tester calibration, Δr is the correction associated with the sty-
profile and Zve the largest of the depth of profile valley, and lus tip radius, α is the coefficient of thermal expansion of the
both are determined in the evaluation length. workpiece, and ΔT is the difference between the measurement
These four parameters depend on the values of the ordi- temperature and the reference temperature (20 °C).
nates of the profile points. In this work, the values of the In this work, the following considerations should be taken
ordinates of the profile points (input variables) present from into account:
Eqs. (5), (6), (7), and (8) may be influenced by resolution of
the roughness tester (R) [17], uncertainty associated with the a) The variability of the each parameter was included in
roughness tester calibration [17], radius stylus tip (r) [1, 4, 18], Eq. (9), considering the n measurement cycles performed;
and difference between the measurement temperature and the b) The amplitude of mechanical vibration was not con-
reference temperature (20 °C). sidered as an individual input variable, since its con-
Considering the five measurement cycles performed, the tribution is included in the variability of the values
mathematical model proposed to evaluate the measurement indicated by the roughness tester. The GUM [17]
uncertainty of four evaluated parameters is given in Eq. (9). points out that the effect of a particular variable of
influence should not be considered twice.
P ¼ L þ ΔR þ ΔC þ Δr þ LαΔT ð9Þ c) The effect related to the temperature variation during mea-
surement was disregarded because the measurement was
where P is the measured roughness parameter, L is the average performed in a brief time.
of values indicated by the roughness tester considering the d) The contribution of the plastic deformation of surface was
five measurement cycles performed, ΔR is the correction as- disregarded because it was difficult to calculate.
sociated with the resolution of the roughness tester, ΔC is the According to [4], the value of 20 nm can be adopted for
correction associated with the uncertainty of the roughness a metal surface measured with a 2-μm-radius stylus and a
400 Int J Adv Manuf Technol (2017) 92:393–406

Normal Prob. Plot; Raw Residuals Normal Prob. Plot; Raw Residuals
Ra Rq
3,0 3,0
2,5 ,99 2,5 ,99

E x p e c te d N o rm a l V a lu e
Expected Normal Value
2,0 2,0
1,5 ,95 1,5 ,95
1,0 1,0
0,5 ,75 ,75
0,5
0,0 ,55 0,0 ,55
-0,5 ,35 -0,5 ,35
-1,0 ,15 -1,0 ,15
-1,5 ,05 -1,5 ,05
-2,0 -2,0
,01 -2,5 ,01
-2,5
-3,0 -3,0
-0,15 -0,10 -0,05 0,00 0,05 0,10 0,15 -0,20 -0,15 -0,10 -0,05 0,00 0,05 0,10 0,15 0,20 0,25
Residual Residual

Normal Prob. Plot; Raw Residuals Normal Prob. Plot; Raw Residuals
Rz Rt
3,0 3,0
2,5 ,99 2,5 ,99
Expected Normal Value

Expected Normal Value

2,0 2,0
1,5 ,95 1,5 ,95
1,0 1,0
0,5 ,75 0,5 ,75
0,0 ,55 0,0 ,55
-0,5 ,35 -0,5 ,35
-1,0 ,15 -1,0 ,15
-1,5 ,05 -1,5 ,05
-2,0 -2,0
-2,5 ,01 ,01
-2,5
-3,0 -3,0
-2,0 -1,5 -1,0 -0,5 0,0 0,5 1,0 1,5 2,0 2,5 -4 -3 -2 -1 0 1 2 3 4 5
Residual Residual

Fig. 8 The normal probability plots of residues for the parameters Ra, Rq, Rz, and Rt, respectively, evaluated in the stainless steel sample

Fig. 9 Values of the roughness 8.0 10.0

Ra, Rq, Rz, and Rt of carbon steel 9.0
sample for the different 7.0
8.0
Ra (µm)

Rq (µm)

frequencies and accelerations

6.0 7.0
6.0
5.0
5.0
4.0 4.0
60 80 100 60 80 100
Frequency (Hz) Frequency (Hz)

30.0 34.0

28.0 32.0
Rz (µm)

26.0
Rt (µm)

30.0
24.0
28.0
22.0
20.0 26.0
60 80 100 60 80 100
Frequency (Hz) Frequency (Hz)

a = 0.000 m/s2 a = 0.620 m/s2 a = 0.977 m/s2 a = 1.610 m/s2

Int J Adv Manuf Technol (2017) 92:393–406 401

Table 6 ANOVA results for Ra parameter (carbon steel sample) Table 8 ANOVA results for Rz parameter (carbon steel sample)

Factor ANOVA Ra (carbon steel sample); R2 = 0.27327 Factor ANOVA Rz (carbon steel sample); R2 = 0.62565

SS df MS F p SS df MS F p

(1) Frequency 1.034163 1 1.034163 11.08015 0.001852 (1) Frequency 0.23408 1 0.23408 0.30924 0.581171
(2) Acceleration 0.031948 1 0.031948 0.34229 0.561714 (2) Acceleration 10.13845 1 10.13845 13.39344 0.000714
Interaction (1) and (2) 2.305884 1 2.305884 24.70551 0.000012 Interaction (1) and (2) 47.18592 1 47.18592 62.33511 0.000000
Error 3.826727 41 0.093335 Error 31.03584 41 0.75697
Total SS 7.198722 44 Total SS 88.59430 44

0.75mN stylus force. However, in this work, it was ob- To evaluate the standard uncertainty associated with the
served that the damage suffered by the aluminum surface resolution (R) of the roughness tester, the stylus tip radius
was significantly different from that of the carbon steel or (r), and the coefficient of the thermal expansion of the
the stainless surfaces. sample (α), rectangular distributions were adopted with
infinite degrees of freedom in Eqs. (13), (14), (15), and
The combined standard uncertainty was determined by the (16), respectively.
application of the law of propagation of uncertainty in Eq. (9)
and by calculating the partial derivatives. In this manner, pffiffiffi
uðRÞ ¼ R=2⋅ 3 ð13Þ
Eq. (10) was obtained.
  pffiffiffi
u2C ðPÞ ¼ u2 L þ u2 ðΔRÞ þ u2 ðΔC Þ þ u2 ðΔrÞ uðrÞ ¼ 0:02⋅r= 3 ð14Þ
pffiffiffi
þ ðL⋅ΔT Þ2 ⋅u2 ðΔαÞ þ ðL⋅ΔαÞ2 ⋅u2 ðΔT Þ ð10Þ uðαÞ ¼ 0:1⋅α= 3 ð15Þ

To determine the standard uncertainty associated with the

The standard uncertainty associated with the effect of the
variability of the values indicated by the roughness tester and
difference between the measurement temperature and the ref-
the uncertainty associated with the roughness tester calibra-
erence temperature of 20 °C (u(ΔT)) can be determined by
tion, the Student’s t distributions were adopted, as indicated in
Eq. (16).
Eqs. (11) and (12).
  pffiffiffi s

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi


ffi
u ð LÞ ¼ s L = n ð11Þ ΔT 20 2 ΔRT 2 ΔC T 2
uðΔT Þ ¼ pffiffiffi þ pffiffiffi þ ð16Þ
3 2⋅ 3 kT
uðC Þ ¼ U ðC Þ=k C ð12Þ
  where ΔT20 is the difference between the room temperature
where s L is the experimental standard deviation, n is the and 20 °C, ΔRT is the correction associated to the thermom-
number of the measuring cycles performed, U(C) is the ex- eter resolution, ΔCT is the correction due to uncertainty asso-
panded uncertainty associated to the calibration of roughness ciated with the thermometer calibration, and kT is the corre-
tester, and kC is the correspondent coverage factor. spondent coverage factor.

Table 7 ANOVA results for Rq parameter (carbon steel sample) Table 9 ANOVA results for Rt parameter (carbon steel sample)

Factor ANOVA Rq (carbon steel sample); R2 = 0.37443 Factor ANOVA Rt (carbon steel sample); R2 = 0.19556

SS df MS F p SS df MS F p

(1) Frequency 0.988267 1 0.988267 9.82877 0.003172 (1) Frequency 0.85345 1 0.853453 1.592259 0.214135
(2) Acceleration 0.076507 1 0.076507 0.76090 0.388126 (2) Acceleration 0.05985 1 0.059853 0.111666 0.739958
Interaction (1) and (2) 4.105992 1 4.105992 40.83597 0.000000 Interaction (1) and (2) 3.36200 1 3.362000 6.272372 0.016333
Error 4.122485 41 0.100548 Error 21.97606 41 0.536001
Total SS 9.293253 44 Total SS 26.25136 44
402 Int J Adv Manuf Technol (2017) 92:393–406

Pareto Chart of Standardized Effects; Variable: Ra Pareto Chart of Standardized Effects; Variable: Rq
Ra Rq

Interaction Interaction
(1) and (2) -4,97 (1) and (2) -6,39

(1)Frequency(L) (1)Frequency(L) 3,135087

3,328685

(2)Acceleration(L) ,5850598 (2)Acceleration(L) ,872297

p=,05 p=,05
Standardized Effect Estimate (Absolute Value) Standardized Effect Estimat (Absolute Value)

Pareto Chart of Standardized Effects; Variable: Rz Pareto Chart of Standardized Effects; Variable: Rt
Rz Rt

Interaction Interaction
(1) and (2) -7,895 (1) and (2) -2,50447

(2)Acceleration(L)
3,659705 (1)Frequency(L) -1,26185

(1)Frequency(L) (2)Acceleration(L)
-,55609 ,3341652

p=,05 p=,05
Standardized Effect Estimate (Absolute Value) Standardized Effect Estimate (Absolute Value)

Fig. 10 Pareto charts of the factorial design 32 of carbon steel sample (confidence level of 95%)

3 Results and discussions magnitude of the surface deterioration was revealed when this
sample was observed by a scanning electron microscopy
3.1 Results of pre-tests (SEM) (Fig. 5). In this figure, it is possible to observe the
marks printed by the turning tool in a vertical position, while
Figure 4 shows the frequency spectrum of the vibration signal the deformation caused by the stylus tip on the sample surface
resulting from the four vibration levels generated by the dy- is observed in a horizontal direction. The peaks were de-
namic driver and captured by the accelerometer during the formed giving way to scratches; this can be explained by the
pre-tests. In this figure, the signals with amplitudes of 250, combination of ductile material (aluminum) with a diamond
500, 750, and 1000 mV were denoted by vib1, vib2, vib3, and stylus tip of 2-μm radius. The scratches on the surface for
vib4, respectively. The energy of the vibration signal was ductile materials, such as some steels, silver, gold, lead, and
higher in frequency range, from 50 to 150 Hz. The average elastomers, can be observed after measurement [1, 19]. Hence,
amplitude values in this frequency range were, respectively, the aluminum sample was not evaluated during the execution
0.04, 0.08, 0.13, and 0.17 m/s2, and for all signals, the maxi- of the factorial design.
mum values of amplitudes were found in the range of frequen-
cies from 60 to 80 Hz.
The possible sources of vibration that may influence the 3.2 Measurement results of the stainless steel sample
values of roughness in the experiment conditions were attrib-
uted to traffic, air currents, and mechanical equipment, such as Figure 6 shows the average values of roughness parameters
rotary motors, air conditioners, and compressors that when put (Ra, Rq, Rz, and Rt) obtained in different conditions with three
into operation produce mechanical vibration in frequency levels of frequency (Hz) and three levels of acceleration (m/s2)
range of 80 Hz. for the stainless steel sample. The results obtained during the
After execution of the pre-tests, marks left by the stylus tip measurement of roughness in nominal conditions (accelera-
on the surface of the aluminum sample were observed. The tion and frequency equal to zero) and the expanded
Int J Adv Manuf Technol (2017) 92:393–406 403

Normal Prob. Plot; Raw Residuals Normal Prob. Plot; Raw Residuals
Ra Rq
3,0 3,0
2,5 2,5
,99 ,99
2,0 2,0

Expected Normal Value

Expected Normal Value

1,5 ,95 1,5 ,95

1,0 1,0
,75 ,75
0,5 0,5
0,0 ,55 0,0 ,55
-0,5 ,35 -0,5 ,35
-1,0 ,15 -1,0 ,15
-1,5 -1,5 ,05
,05
-2,0 -2,0
,01 ,01
-2,5 -2,5
-3,0 -3,0
-0,8 -0,6 -0,4 -0,2 0,0 0,2 0,4 0,6 0,8 -0,8 -0,6 -0,4 -0,2 0,0 0,2 0,4 0,6 0,8
Residual Residual

Normal Prob. Plot; Raw Residuals Normal Prob. Plot; Raw Residuals
Rz Rt
3,0 3,0
2,5 ,99 2,5 ,99

Expected Normal Value

2,0 2,0
Expected Normal Value

1,5 ,95 1,5 ,95

1,0 1,0
,75 0,5 ,75
0,5
,55 0,0 ,55
0,0
,35 -0,5 ,35
-0,5
-1,0 ,15 -1,0 ,15
-1,5 -1,5 ,05
,05
-2,0 -2,0
,01 -2,5 ,01
-2,5
-3,0 -3,0
-2,0 -1,5 -1,0 -0,5 0,0 0,5 1,0 1,5 2,0 -2,0 -1,5 -1,0 -0,5 0,0 0,5 1,0 1,5
Residual Residual

Fig. 11 The normal probability plot of residues for the parameters Ra, Rq, Rz, and Rt, respectively, measured in a carbon steel sample

uncertainty values (95%) associated with the measurement of For constant acceleration of 0.977 m/s2, as the frequency
all roughness parameters are also shown in this figure. increases, the expanded uncertainty values exhibit a rising
The average of all roughness parameters exhibited lower trend for all parameters. In turn, in the acceleration of
values when evaluated in standard conditions and were 1.610 m/s2, an increasing frequency causes the opposite effect,
0.369, 0.482, 2.564, and 2.760 μm to Ra, Rq, Rz, and Rt, leading to a reduction in uncertainty values associated with all
respectively. For constant frequency values (60, 80, and evaluated roughness parameters. For acceleration of 0.620 m/
100 Hz), the average values of Ra parameters showed a s2, no trend was observed.
trend to increase as the acceleration was augmented. The The variability of the values indicated by the roughness
differences in percentage increase of the values obtained tester was the input quantity that contributed the most to the
from the acceleration of 1.610 m/s2 and the values found final uncertainty for Rq, Rz, and Rt parameters, followed by
when there was no acceleration were of 26% to 60 Hz, 39% the standard uncertainty of roughness tester calibration. For
to 80 Hz, and 42% to 100 Hz for Ra. In turn, for Rq were Ra, the opposite was observed, when the major contribution to
23, 44, and 41% to 60, 80, and 100 Hz, respectively. For the final uncertainty was attributed to the standard uncertainty
Rz, the percentage increase with respect to the nominal of roughness tester calibration followed by the variability of
conditions were 62% to 60 Hz, 65% to 80 Hz, and 58% the values obtained in repeated measurements.
to 100 Hz, while for Rt were 118% to 60 Hz, 172% to The significance of the effects of variables (frequency and
80 Hz, and 118% to 100 Hz. The Ra and Rq parameters acceleration) and possible interactions between them was
were less influenced by the mechanical vibration than the assessed by ANOVA and based on the Pareto charts. The
Rz and Rt parameters. This can be explained because in the results are shown in Tables 2, 3, 4, and 5 for Ra, Rq, Rz, and
latter case, the presence of isolated peaks and/or valleys Rt, respectively. From Tables 2, 3, 4, and 5, it was concluded
can significantly affect their values, while the Ra and Rq that with the 95% confidence level, the frequency of the me-
represent average roughness and for this reason were little chanical vibration was significant for Rq, Rz, and Rt parame-
influenced by the presence of outliers. ters evaluated with a positive effect, i.e., for higher-frequency
404 Int J Adv Manuf Technol (2017) 92:393–406

Fig. 12 Average values of Ra,

Stainless steel sample Carbon steel sample
Rq, Rz, and Rt parameters 0.6 8.0
evaluated on stainless steel (on y = 0.0326x + 0.4192 y = 0.3726x + 5.6663
the left) and on carbon steel (on R² = 0.9482 7.0 R² = 0.9921

Ra (µm)

Ra (µm)
the right) samples in the 0.5
frequency of 80 Hz as a function 6.0
of acceleration. The best fit line
obtained by the application of the 0.4 5.0
least squares method and the 0.62 0.977 1.61 0.62 0.977 1.61
correspondent correlation Acceleration m/s2 Acceleration m/s2
coefficient
Stainless steel sample Carbon steel sample
0.8 10.0
y = 0.0602x + 0.5163
y = 0.4222x + 6.2984

Rq (µm)

Rq (µm)
0.7 R² = 0.9818 9.0
R² = 0.9997
8.0
0.6
7.0
0.5 6.0
0.620 0.977 1.610 0.620 0.977 1.610
Acceleration m/s2 Acceleration m/s2

Stainless steel sample Carbon steel sample

5.0 30.0
y = 0.5851x + 2.4717 y = 1.3403x + 22.508
4.5 R² = 0.9985 28.0

Rz (µm)
R² = 0.9616
Rz (µm)

4.0 26.0
3.5 24.0
3.0 22.0
0.620 0.977 1.610 0.620 0.977 1.610
Acceleration m/s2 Acceleration m/s2

Stainless steel sample Carbon steel sample

10.0 34.0
9.0 y = 1.4659x + 3.2669 y = 0.4117x + 30.002
R² = 0.5449
Rt (µm)

R² = 0.9633 32.0
Rt (µm)

8.0
7.0
6.0 30.0
5.0
4.0 28.0
0.620 0.977 1.610 0.620 0.977 1.610
Acceleration m/s2 Acceleration m/s2

values. Larger values of the roughness parameters Rq, Rz, and and frequency equal to zero) and the expanded uncertainty
Rt were obtained. On the other hand, acceleration influences values (95%) are also presented. From this figure, it was con-
the Ra values with a positive effect. cluded that for the sample of carbon steel, the average values of
The interactions between the variables, frequency and ac- four roughness parameters were also smaller when they were
celeration, were significant for the Rq and Rz parameters with a evaluated in standard conditions and were 5.342, 5.884,
positive effect associated with them in both cases. These ef- 21.072, and 29.902 μm to Ra, Rq, Rz, and Rt, respectively.
fects are seen more clearly in the Pareto charts shown in Fig. 7. However, for constant values of frequency 60 and 80 Hz, all
Figure 8 shows a normal probability plot of residues for the parameters increase when the acceleration varies from 0.620 to
four evaluated parameters on stainless steel samples. From 1.610 m/s2, while for 100 Hz, the opposite effect was observed.
Fig. 8, it was concluded that the residues have a distribution The average values of the Ra parameter tend to increase to
near the normal distribution probability for all parameters. the extent that acceleration increases for constant frequency
values. The differences in percentage increase of the values
3.3 Measurement results of the carbon steel sample obtained from the acceleration 1.610 m/s2, and the values
found when there was no acceleration were 21% to 60 Hz,
Figure 9 shows the average roughness values (Ra, Rq, Rz, and 27% to 80 Hz, and 16% to 100 Hz. In turn, for Rq, the differ-
Rt) for carbon steel sample. The results obtained during the ences in percentage increase from values obtained in the ac-
measurement of roughness in nominal conditions (acceleration celeration of 1610 m/s2 and the values found when there was
Int J Adv Manuf Technol (2017) 92:393–406 405

no acceleration were 25, 29, and 17% to 60, 80, and 100 Hz, this trend was not observed for the Rt parameter, and this
respectively. For Rz, the percentage increase with respect to effect can be justified by the likely loss of contact between
the nominal conditions were 31% to 60 Hz, 27% to 80 Hz, and the stylus tip and the surface of this sample.
15% to 100 Hz, while for Rt, the values were 6% to 60 Hz, 4%
to 80 Hz, and 2% to 100 Hz.
In this case, in particular, the parameters of roughness 4 Conclusions
Ra, Rq, and Rz underwent similar changes for the frequen-
cies considered as the acceleration increased; however, this It was observed, in this work, that in the frequency range from
effect was not observed for the Rt parameter, which was 60 to 80 Hz, the greatest signal energy values of mechanical
less significant and manifested a decreasing trend when the vibration were identified, and that in frequency of 80 Hz, the
acceleration passed from 0.620 to 1.610 m/s2. This fact can greatest effects on the roughness values were observed for the
be explained by the loss of contact between the stylus tip sample with better surface finish.
and the sample surface. The interaction between the variables, frequency and accel-
From Fig. 9, it follows that for the constant acceleration eration, was statistically significant for the parameters Ra, Rq,
values, as the frequency increases from 60 to 80 Hz, the ex- Rz, and Rt, which were evaluated in two samples, one of
panded uncertainty values show an increasing trend, which is stainless steel with an Ra equal to 0.369 μm and the other of
more significant for acceleration values of 0.620 and 0.977 m/ carbon steel with an Ra of 5.342 μm.
s2. On the other hand, for the acceleration of 1.610 m/s2, the In the stainless steel sample with better surface finish, the
increased frequency causes no definite trend leading to a re- average of roughness parameters Ra and Rq were less influ-
duction in uncertainty values associated with roughness pa- enced by the mechanical vibration than the Rz and Rt param-
rameters evaluated at a frequency of 80 Hz and at an increase eters. For the latter, the observed differences were significant
at 100 Hz. For all evaluated parameters, the major contribu- and even surpassed 100%. The results found were 118% for
tion to the final uncertainty was attributed to the variability of 60 Hz, 172% for 80 Hz, and 118% for 100 Hz as the acceler-
the values obtained in repeated measurements. ation passed from 0 to 1.610 m/s2. The major contribution to
Tables 6, 8, and 9 show the ANOVA results for the sample the final uncertainty of Ra was attributed to the standard un-
of carbon steel. certainty of the roughness tester calibration, and the variability
From Tables 6, 7, 8, and 9, it follows that the frequency of of values indicated by the roughness tester had the highest
the mechanical vibration is significant for Ra and Rq rough- contribution to the final measurement uncertainty for Rq, Rz,
ness parameters with a positive effect associated with them in and Rt parameters.
both cases as shown in the Pareto charts in Fig. 10. In turn, the
acceleration influences only the Rz values with positive effect.
The interaction between the variables, frequency and acceler- Acknowledgements The authors would like to thanks CAPES, CNPq,
ation, was significant for all evaluated parameters with a neg- and FAPEMIG, Brazilian Financing Agency, for supporting the develop-
ative effect. ment of this research.
Figure 11 shows the normal probability plot of the residuals
for the four parameters evaluated in the sample of carbon steel.
In all cases, the residues have a close distribution to the normal References
distribution.
1. Chand M, Mehta A, Sharma R, Ojha VN, Chaudhary KP (2011)
3.4 Final analysis Roughness measurement using optical profiler with self-reference
laser and stylus instrument—a comparative study. Indian Journal of
Pure & Applied Physics 49(May):335–339
Figure 12 shows the values of all roughness parameters eval- 2. Bhushan B (1996) Tribology and mechanics of magnetic storage
uated on stainless steel and carbon steel samples in the fre- devices, 2nd Editio edn. Springer-Verlag, New York
quency of 80 Hz as a function of acceleration. In the frequency 3. Jang DY, Choi YG, Kim HG, Hsiao HA (1996) Study of the cor-
of 80 Hz, the energy of the vibration signal was higher relation between surface roughness and cutting vibration to develop
an on-line roughness measuring technique in hard turning. Int J
(Fig. 4), and thus, this condition represents the most critical
Mach Tools Manuf 26:453–464
condition investigated. In the frequency of 80 Hz, as the ac- 4. Leach RK (2014) The measurement of surface texture using stylus
celeration increased, all the parameters exhibited a linear trend instruments. Measurement Good Practice Guide No. 37.
for the stainless steel sample, with a correlation coefficient of Engineering Measurement Division National Physical Laboratory.
94.82, 98.18, 99.85, and 96.33% for Ra, Rq, Rz, and Rt, re- London: HMSO/National Physical Laboratory. ISSN 1368–6550.
112p
spectively. For the carbon steel sample, the Ra, Rq, and Rz 5. Haitjema H (1998) Uncertainty analysis of roughness standard cal-
parameters also showed a linear trend with a correlation coef- ibration using stylus instruments. Precis Eng 22(97):110–119. doi:
ficient of 99.21, 99.97, and 96.16%, respectively; however, 10.1016/S0141-6359(97)00090-1
406 Int J Adv Manuf Technol (2017) 92:393–406

6. ISO (1996) ISO 3274:1996, Geometrical product specifications 14. Lin SC, Chang MF (1998) A study on the effects of vibra-
(GPS)—surface texture: profile method—nominal characteristics tions on the surface finish using a surface topography sim-
of contact (stylus) instruments. ISO TC 213. 13p ulation model for turning. Int J Mach Tools Manuf 38:763–
7. Abburi NR, Dixit US (2006) A knowledge-based system for the 782. doi:10.1016/S0890-6955(97)00073-4
prediction of surface roughness in turning process. Robot Comput 15. Risbood KA, Dixit US, Sahasrabudhe AD (2003) Prediction
Integr Manuf 22(4):363–372. doi:10.1016/j.rcim.2005.08.002 of surface roughness and dimensional deviation by measur-
8. Abouelatta OB, Mádl J (2001) Surface roughness prediction based ing cutting forces and vibrations in turning process. J Mater
on cutting parameters and tool vibrations in turning operations. J Process Technol 132(1–3):203–214. doi:10.1016/S0924-
Mater Process Technol 118(1–3):269–277. doi:10.1016/S0924- 0136(02)00920-2
0136(01)00959-1 16. ISO (2013) ISO 12085:1996—geometrical product specifica-
9. Asiltürk I, Akkus H (2011) Determining the effect of cutting pa- tions (GPS)—surface texture: profile method—motif parame-
rameters on surface roughness in hard turning using the Taguchi ters. ISO/TC 213. 17p
method. Measurement: Journal of the International Measurement 17. Joint Committee for Guides in Metrology (JCGM) (2008).
Confederation 44(9):1697–1704 Evaluation of measurement data: guide to the expression of
10. Rodrigues LLR, Kantharaj AN, Kantharaj B, Freitas WRC, Murthy uncertainty in measurement, (September), Geneva,
BRN (2012) Effect of cutting parameters on surface roughness and Switzerland. 120. doi:10.1373/clinchem.2003.030528.
cutting force in turning mild steel. Res J Recent Sci 1(10):19–26 18. Bhushan B (2000) Chapter 2 Surface roughness analysis and
11. Thomas M, Beauchamp Y, Youssef AY, Masounave J (1996) Effect measurement techniques. Modern tribology handbook. Volume
of tool vibrations on surface roughness during lathe dry turning 1. ISBN 0–8493–8403-6. CRC Press, LLC. Boca Raton,
process. Comput Ind Eng 31(3–4):637–644. doi:10.1016/S0360- Flórida, EUA
8352(96)00235-5
19. Vorburger TV, Rhee HG, Renegar TB, Song JF, Zheng A (2007)
12. Wang X, Feng CX (2002) Development of empirical models for
Comparison of optical and stylus methods for measurement of sur-
surface roughness prediction in finish turning. Int J Adv Manuf
face texture. Int J Adv Manuf Technol 33(1–2):110–118. doi:10.
Technol 20(5):348–356. doi:10.1007/s001700200162
1007/s00170-007-0953-8
13. Selvam MS (1975) Tool vibration and its influence on surface
roughness in turning. 147–157. Wear 35:149–157. doi:10.1103/
PhysRev.46.157