Professional Documents
Culture Documents
Search
Collections
Journals
About
Contact us
My IOPscience
This article has been downloaded from IOPscience. Please scroll down to see the full text article. 2006 Meas. Sci. Technol. 17 1496 (http://iopscience.iop.org/0957-0233/17/6/030) View the table of contents for this issue, or go to the journal homepage for more
Download details: IP Address: 115.111.10.50 The article was downloaded on 19/03/2011 at 18:38
doi:10.1088/0957-0233/17/6/030
Received 3 November 2005, in nal form 3 March 2006 Published 8 May 2006 Online at stacks.iop.org/MST/17/1496 Abstract On-line roughness measurement of a surface with one-dimensional manufacturing marks is difcult to implement. For example, a contact stylus-type inspection method often does not perform very well or fails without any prior knowledge of the mark distribution on the surface. In this paper, we propose an on-line surface roughness measurement method based on laser light scattering, which is very effective for roughness measurement of one-dimensional manufacturing surfaces. The surface roughness is obtained from the spatial distribution of the scattered light intensity. The measurement setup has a very simple conguration, which consists of a CCD sensor, a collimated diode laser and an expander. The orientation of the spatial distribution of the scattered light intensity from the surface, which depends on the surface orientation, is detected by the CCD sensor, and then the mark direction can be readily determined from image processing. After that the root-mean-square (RMS) height of the surface roughness is extracted by means of image processing of the scattered light distribution in the direction parallel to the manufacturing mark, rather than in the direction perpendicular to the mark which is often followed by other measurement probes. The experimental tests show that the non-contact method has great potential for on-line surface roughness measurement.
Keywords: on-line measurement, surface roughness, light scattering,
one-dimensional manufacturing surface (Some gures in this article are in colour only in the electronic version)
1. Introduction
Surface roughness measurement is one of the essential quality control processes which are carried out to ensure that the quality of a manufactured part can conform to its specied standard. It has been performed using numerous methods, such as contact stylus-type inspection, microscopy, ultrasonic and optical methodology, etc [1]. However, among them only a few methods are capable of being used in on-line measurement processes. Comprehensive reviews of these methods have been given in [16]. In this paper, we concentrate on the light
0957-0233/06/061496+07$30.00
scattering methods and their capabilities of on-line surface roughness measurement. The typical methods for surface roughness measurement based on light scattering are the total integrated scattering (TIS) and angle-resolved scattering (ARS) methods [6, 7]. TIS is dened as the ratio of the light intensity scattered into a hemisphere to the light intensity reected by the specimen surface. The root-mean-square (RMS) roughness height of a specimen surface has a denite relationship between the ratio and the incident light wavelength [6]. The ARS method is often used to analyse the in-plane angular 1496
intensity distribution of scattered light for surface roughness determination. In general, it is more convenient to employ TIS for surface roughness measurement than ARS, because the instrumentation is simple and the conversion from TIS to RMS roughness height is straightforward unlike ARS, the data analysis of which is more difcult. However, one of the advantages of roughness measurement by ARS is that ARS can be employed on surfaces covered with multilayer dielectric lms and/or transparent or opaque metal lms as long as the thicknesses and optical constants of these lms are known. Another advantage is that much more statistical information can be extracted from ARS than from the TIS method. Due to the advantages mentioned above, many ARS surface roughness measurement methods have been investigated [811]. However, the theoretical expression of ARS against RMS roughness height involves states of polarization of incident light and light diffraction. It is therefore impossible to precisely establish it in theory. The derivation often takes a lot of assumptions and simplications so that the theoretical expression sometimes only has qualitative meaning, and is difcult to use to quantitatively determine the surface roughness [12]. In practical online surface roughness measurement, the statistical methods representing scattered light intensity distribution against surface roughness are often used. One of the methods is based on the phenomenon that the intensity distribution of the scattered light in the plane formed by the incident light beam and surface normal depends on the scattered light angle against the normal direction, and there is more angular light scattering for a rougher surface [13]. The surface roughness in the algorithm is determined by computing the variance of the in-plane light scattering angle. Another method has been carried out with a machine vision system, where the surface roughness is characterized by the frequency distribution of the grey level occurrence in a scattered light intensity image [1417]. The scattered light frequency distribution is actually the histogram of the light scattering intensity image obtained by a CCD camera. The surface roughness in this method is evaluated by the ratio of the root-mean-square value of the histogram and its standard deviation. Other algorithms, such as using the grey level co-occurrence matrix [18], are also used. The third method uses the core-ring ratio of the scattered light distribution at a particular point on the observation screen, where the light scattering image can be identied as combinations of an inner region and an outer region [19, 20]. The two inner and outer regions are called the core component and ring component separately. The surface roughness is related to the ratio of the areas of the two regions. In the three methods above, the rst method is implemented by in-plane light scattering. That means the variance of the scattering light angle is computed in the light plane formed by the incident light, surface normal and mirror reected light. Thus the measurement system needs to be congured delicately so that the sensor or sensor arrays are located in the same plane. The last two methods are implemented in the observation plane; therefore they are more robust. However, all three methods are often effective for surfaces with uniform roughness distribution and under conditions of non-coherent light illumination. For one-dimensional manufacturing surfaces, such as the surfaces of steel parts
manufactured by grinding, milling, turning, etc they often fail without any knowledge of the surface micro manufacturing mark directions. For example, the ratio value obtained in the second method is sensitive to measuring noises for laser light illumination and one-dimensional manufacturing surface. This has been observed in our experiments. The ratio of the maximum ratio value to the minimum ratio value can reach more than 10. This is because the light scattering intensity distributions of coherent laser light are more complicated due to light diffraction and light speckle noise. In this paper, we will investigate a simple and effective light scattering method of on-line surface roughness measurement with laser light illumination for one-dimensional manufacturing surfaces of metal parts. The surface roughness is extracted from the laser light scattering intensity distribution in the direction parallel to the surface micro manufacturing mark direction, rather than the in-plane light scattering position which has been extensively investigated for the uniform roughened surface or in its perpendicular direction where, like in the contact stylus-type inspection, the measurement probe has to be moved. The on-line surface roughness measurement method is based on our theoretical and experimental research results which show that the laser light scattering intensity distribution in the direction parallel to the surface micro manufacturing mark direction follows a Gaussian distribution, the parameters of which depend on the surface RMS roughness height.
n
incident light
d i
sensor
r
h=0
ns
d r
reflected light
dAs
r
X
angle on the surface facet; G is the geometric attenuation factor, which models the effects of masking and shadowing and takes values between 0 and 1. The denitions of other geometrical parameters are shown in gure 1. In equation (2), the rst term is the diffuse Lambertian component, and the second term indicates the scattered specular component. In the formula derivation, the model assumes that the normals of surface facets have a Gaussian probability distribution 1 2 exp 2 , (3) 2 2 where indicates the surface RMS roughness slope. The relation of to RMS roughness height, as shown in gure 2, is given by [24] (4) h = 0.5T tan( 2 ), () = where T is the surface correlation distance.
(PSD) and the RMS roughness height h for isotropic specimens, is used [7, 22]:
2 h = 2 fmax
PSD(f )f df
fmin
f =
fx2 + fy2
ARS(r , r ) = =
1 dPr Pi dr
16 2 cos i cos2 r Q(r , r )PSD(fx , fy ), (1) 4 where it is preferable to use plane polarized light as the incident light, whose polarization direction is either parallel or perpendicular to the plane of incidence, if the angle of the incidence light is greater than 10 . Pi is the incident light power, dPr /dr is the power scattered per unit solid angle from the surface, is the wavelength, i is the angle of incidence and r is the scattering angle. The optical factor Q depends on the incident and scattering angles, the complex refractive index of the surface material and the state of polarization of the incident and scattered light. The bandwidth limitation of the spectral density function PSD(fx , fy ) is detailed in [22]. The meanings of geometrical parameters are illustrated in gure 1. All geometrical parameters in the formula need to be accurately calibrated. Thus the conguration of the measurement setup is complicated and difcult to employ in on-line measurement. Another reasonable geometrical model, which describes the relationship between angle-resolved scattered light radiance and surface roughness, assumes that the scattered light from a roughened surface is composed of two components: a diffuse component independent of the angles of reection and a specular reection from mirror-like surface facetsa function of the angles of reection. The formulation of the model is given by [23] 1 dLr = cos i Li dr gF (i , r , r , n, )G(i , r , r ) 2 + exp 2 , (2) cos r 2 where Li and Lr respectively indicate the incident irradiance to, and scattered radiance from the facet dAs ; represents the fractions of incident energy reected by the diffuse mechanism; g is a constant related to the surface roughness; F is the Fresnel reectance, a function of the index of refraction n, the coefcient of absorption and the local illumination 1498
where f = dLr /Li dr , t0 = cos i and t = gF (i , r , r , n, )G(i , r , r )/ cos r . Equation (5) is only true when the roughened surface is an isotropic one and the geometric parameters i , r , r are small enough. For non-isotropic roughened surfaces and one-dimensional manufacturing surfaces, the BRDF relation becomes more complicated. Figure 3 exemplies the situation with the scattered light intensity distributions of two specimens. The picture of the scattered light intensity
r
(a)
r
(b)
r r
(c)
150
150
100 80
60
40 30
100
100
100
60 40 20
40
20
50
50 0
50
20
10
400
800
1200 0
400
800
1200 0
400
800
1200 0
400
800
1200 0
400
800
0 1200 0
400
800
1200
300
200
140 120
120 100 80
100 80 60
50 40 30 20 10
200
150
100 80 60
100
100
60 40
40 20
500
1000
1500
50
500
1000
1500
40
500
1000
1500
20
500
1000
1500
500
1000
1500
500
1000
1500
(a) 0.05 m
(b) 0.1 m
(c) 0.2 m
(d) 0.4 m
(e) 0.8 m
( f ) 1.6 m
Figure 4. The variances of light scattering intensity distributions parallel to and perpendicular to the mark direction of the grinding surface specimens with different surface roughnesses. The units in the gure are pixels on the abscissa axis and image grey value on the ordinate axis.
distribution was taken by the experimental setup shown in gure 5. The specimen surface in gure 3(a) is manufactured by a at lapping method. The surface microstructure has the micro at lapping mark #, so that the scattered light intensity distribution has a cross pattern. The pictures of the scattered light intensity in gures 3(b) and (c) were taken from the same specimen, fabricated by a grinding method. The specimen has one-dimensional micro grinding marks on the surface. The scattered light intensity distribution indicated by gure 3(b) is obtained when the one-dimensional mark direction is parallel to the incident plane. The off-plane light scattering pattern width is much longer than the inplane pattern width. This is because there is much stronger light diffraction in the perpendicular direction of the mark, which spreads out the scattered light. This can be testied by changing the mark direction against the incident plane. As shown in gure 3(c), when the mark direction is rotated to a certain angle, the pattern shape of the scattered light intensity distribution is also changed with the same angle. However, the intensity distribution parallel to and perpendicular to the mark direction remains constant when the collimated incident laser light spot is big enough to cover sufcient marks and the light incident angle is near to the surface normal. Our experimental result also shows that the shape of the intensity distribution perpendicular to the mark direction changes dramatically with increasing surface roughness, but the pattern shape parallel to the mark direction looks the same and only its width and
amplitude are varied gradually, as shown in gure 4 which is obtained from the experimental result shown in gure 6. Furthermore, there is one important characteristic which is that the pattern of the scattered light intensity distribution parallel to the mark direction still obeys the Gaussian distribution expressed in gure 5, but a little change needs to be imposed: y = y0 + k(h , l) exp 2 2 2 , (6)
where l indicates the mark direction, y is the scattered light intensity received by sensors and = (r , r ) is the angle measured in the l direction. The ratio of and k(h , l) depends on the surface RMS roughness height rms = f ( /k(h , l)). (7) 1499
-10
(a) 0.05 m
(b) 0.1 m
(c) 0.2 m
(d) 0.4 m
(e) 0.8 m
( f ) 1.6 m
Figure 6. Light scattering intensity distributions in space from the grinding surface specimens with different one-dimensional manufacturing surface roughnesses.
known. In the following section, we will use this experiment as an example to introduce how to extract the surface roughness from the scattered intensity images taken by the measurement probe, which is listed as follows: (1) Filter the image taken by the CCD sensor. Due to the image noises caused by light speckle and the electronic disturbance of the CCD, the images are rst ltered by a 2D Gaussian low pass lter. The lter size is 8 8 pixels and covariance is 4 pixels. (2) Find the light diffraction direction in the scattered light intensity distribution. The diffraction direction is perpendicular to the one-dimensional mark direction. It can be readily searched out by nding the peak positions of the light intensity distribution through scanning the images. Then t the peak positions with a linear equation to get the diffraction direction. (3) Average the image grey value along the diffraction direction. In order to show the effect of the Gaussian low pass lter, in gure 7 we list two average results. The thick solid curve, which is formed by point overlapping, is obtained directly from the original image without using the lter. The thin solid curve is yielded after the image is implemented with a Gaussian lter. It is apparent that the noises are reduced signicantly. (4) Fit the average results with equation (8): y = y0 + k(h , l) exp (x x0 )2 2 2x , (8)
where y is the grey value; x is the grey value of coordinates in image and y0 , x0 , k and x are unknowns to be determined. It is worth pointing out that in order to
10 8
/k
6 4
20 10 0
200
400
600
800
1000
1200
pixel
Ra ( m)
0 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
Figure 9. The relation between RMS roughness (Ra) and the light scattering.
80 70 60 50 40 30 20 10 0
gray value
Ra = 0.1m
Ra = 0.05m Ra = 0.2m
pixel
0 200 400 600 800 1000 1200
simplify the computation, it is not necessary to transform the image pixel coordinates into their corresponding zenith and azimuth angles. As shown in gure 8, the six x-mark curves, the xmark number of which is reduced four times in order that the curve can be shown clearly, are the average results in the diffraction direction after the images are tted by a Gaussian low pass lter. The solid curves are the tting results. From the excellent resemblance between the curve of each original light scattering intensity distribution and its tting result, it is apparent that the light distribution parallel to the manufacturing mark direction follows a Gaussian distribution very well and depends on the surface roughness. (5) Compute the ratio /k(h , l). (6) Establish the relationship between the RMS roughness height value of the surface roughness and the ratio. This can be performed by drawing the relation curve between the surface RMS roughness height and the ratio. From the curve shown in gure 9, it is apparent that the curve, which indicates the relation of the RMS roughness height value of a one-dimensional grinding surface to the light scattering intensity distribution from the surface, has two linear areas. One of the linear areas indicates a one-dimensional grinding surface RMS roughness height from 0.05 m to 0.4 m, and another from 0.4 m to 1.6 m. The two areas have a good linear relation from the sample tested. The reason for the surface RMS roughness height curve having a curve inection at about RMS 0.4 m may be interpreted by electromagnetic wave scattering from the roughened surface with equation (5) [12, 24]. That is, at this point the laser
light wavelength is near to the surface RMS roughness height, and the light scattering, which is dependent on the light wave interaction with the surface, becomes more complicated. Especially when h / < 0.025, t (h ) approaches zero, and the scattered light from the surface consists of only a small portion of diffuse Lambertian component and mirror reection. This means that when the surface is smooth enough to its RMS roughness height h < 0.025 (in our case h < 0.017 m), the surface roughness cannot be characterized. More discussion about the relationship between the surface roughness and light scattering distribution from the point of view of geometrical optics can be found in the literature [24]. As shown in gure 8, we also nd that the Gaussian curve will atten out when the RMS roughness height h goes beyond a point more than 1.6 m. Our experimental result on the grinding sample surface shows that the point is about 3.2 m.
From the experimental results using the measuring standard grinding surfaces, it is apparent that the RMS height of the surface roughness has a good linear relationship to the image signature of the light scattering intensity distribution. However, for surfaces made of different materials and fabricated by different methods, the relation curve may be different, since the conditions of the electromagnetic wave of the light interacting with the surface have been changed. More effort on this aspect will be required in the future. For on-line surface roughness measurement of the surface of a specic material and the manufacturing method, the curve needs to be calibrated. More characterization of on-line measurement of different manufacturing surfaces will be investigated in the future.
[10]
[11]
[12] [13] [14] [15] [16] [17] [18] [19] [20] [21]
Acknowledgments
The authors would like to thank Andrew Nelson and Dr Stephen Dewey, Dr Nick Silk from Corus for funding the work.
References
[1] Whitehouse D J 2003 Handbook of Surface and Nanometrology (Bristol: Institute of Physics Publishing) [2] Whitehouse D J 1998 Stylus contact method for surface metrology in the ascendancy Meas. Control 31 4850 [3] Poon C Y and Bhushan B 1995 Comparison of surface roughness measurements by stylus proler, AFM and non-contact optical proler Wear 190 7688 [4] Pahk H J, Stout K and Blunt L 2000 A comparative study on the three-dimensional surface topography for the polished surface of femoral head Int. J. Adv. Manuf. Technol. 16 56470 [5] Jolic K I, Nagarajah C R and Thompson W 1994 Non-contact optically based measurement of surface roughness of ceramics Meas. Sci. Technol. 5 67184 [6] Bennet J M and Mattsson L 1993 Introduction to Surface Roughness and Scattering (Washington, DC: Optical Society of America) [7] Bennett J M 1992 Recent development in surface roughness characterization Meas. Sci. Technol. 3 111927 [8] Ohlidal M, Uncovsky M, Ohlidal I and Franta D 1999 Determination of the basic parameters characterizing the roughness of metal surfaces by laser light scattering J. Mod. Opt. 46 27993 [9] Azarova V V, Dmitriev V G, Lokhov Yu N and Malitski K N 2002 Measuring the roughness of high-precision quartz
[22]
substrates and laser mirrors by angle-resolved scattering J. Opt. Technol. 69 1258 Cheng C and Qi D 2002 Extraction of height probability density of random rough surfaces from the central d-peak of angle-resolved light scattering using the optical inversion algorithm Appl. Phys. Lett. 81 21246 Fang K, Adame R, Yang H N, Wang G C and Lu T M 1995 Measurement of roughness exponent for scale-invariant rough surfaces using angle resolved light scattering Appl. Phys. Lett. 66 20779 Beckmann P and Spizzochino A 1963 The Scattering of Electromagnetic Waves from Rough Surfaces (New York: Pergamon) Brodmann R 1984 An optical instrument for measuring the surface roughness in production control Ann. CIRP 33 4036 Luk F, Huynh V and North W 1989 Measurement of surface roughness by a machine vision J. Phys. E: Sci. Instrum. 22 97780 Gupta M and Raman S 2001 Machine vision assisted characterization of machine surfaces Int. J. Prod. Res. 39 75984 Li X, Wang L and Cai N 2004 Machine-vision-based surface nish inspection for cutting tool replacement in production Int. J. Prod. Res. 42 227987 Kiran M B, Ramamoorthy B and Radhakrishnan V 1998 Evaluation of surface roughness by vision system Int. J. Mach. Tools Manuf. 38 68590 Gadelmawla E S 2004 A vision system for surface roughness characterization using the gray level co-occurrence matrix NDT & E Int. 37 57788 Cheng L 1988 Core-ring-ratio method for surface measurement Proc. SPIE 927 15460 Cheng L and Siu G G 1990 Measurement of surface roughness with core-ring ratio method using incoherent light Meas. Sci. Technol. 1 114956 Nicodemus F E, Richmond J C, Hisa J J, Ginsberg I W and Limperis T 1977 Geometrical Considerations and Nomenclature for Reectance (NBS Monograph 160) (Washington, DC: National Bureau of Standards) Duparre A, Ferre-Borrull J, Gliech S, Notni G, Steinert J and Bennett J M 2002 Surface characterization techniques for determining the root-mean-square roughness and power spectral densities of optical components Appl. Opt. 41 15470 Torrance K E and Sparrow E M 1967 Theory for off-specular reection from roughness surface J. Opt. Soc. Am. 57 110514 Nayar S K, Ikeuchi K and Kanade T 1991 Surface reection: physical and geometrical perspectives IEEE Trans. Pattern Anal. Mach. Intell. 13 61134 Meister G, Rothkirch A, Spitzer H and Bienlein J 2001 Width of the specular peak perpendicular to the principal plane for rough surfaces Appl. Opt. 40 607280
1502