Int J Adv Manuf Technol(2000)16:668–674©2000Springer-Verlag London
Limited
Automated Surface Roughness Measurement
C.Bradley
Department of Mechanical Engineering,University of Victoria,Victoria,Canada
A non-contact roughness sensor is described that is suited for integration with a computer-controlled coordinate measuring machine(CMM).The sensor employs afibre optic interfer-ometer,electronic control system and data-processing software. The combination of the sensor and computer controlled CMM allows surface texture assessment to be made during scheduled dimensional inspections of complex curved surface components, such as turbine blades.The sensor system will measure surface roughness parameters,for example R a,using a method that reflects standard procedures.The lightweight sensor head can be mounted on a touch probe arm and the associated articu-lated mounting head;this combination gives quasi5-axis positioning ability to the overall sensor.This is suitable for automated surfacefinish inspection of compound curved surface blades.The sensor and its control unit are integrated with the CMM controller and its operation can be controlled through standard part-program commands used by the CMM. Keywords:Automated inspection;Coordinate measuring machine;Fibre optic sensor;Surface texture
1.Introduction
This paper describes an optical surface texture sensor that is integrated with a computer numerically controlled(CNC) coordinate measuring machine(CMM).The combination of the two technologies permits the measurement of both dimen-sional and roughness metrics on a part placed on the deck of the CMM.
1.1Literature Review
The demand for incorporating sensor technology into the production environment is being driven by the simultaneous need to minimise manufacturing costs while maintaining a high standard of quality[1].In particular,new surface texture Correspondence and offprint requests to:Dr C.Bradley,Department of Mechanical Engineering,University of Victoria,PO Box3055, Victoria BC,V8W3P6,Canada.E-mail:cbrȰengr.uvic.ca sensors have predominantly been non-contact,employing optical,ultrasonic,and capacitance methods[2–5].The sensors typically measure a common surface roughness parameter,such as the average roughness amplitude or R a.This work presents an optical texture sensor,employingfibre optics,that is physi-cally compact and lightweight making it ideal for integration with a CMM.Previous research on usingfibre optics in surface roughness
measurement was performed by Spurgeon and Slater [6]and by Lin et al.[7].Both used a bundle of opticalfibres to deliver light to the surface and also to collect the reflected light and guide it to a photo detector.A correlation was found between the intensity of the reflected light(as measured by the photo detector)and the average roughness of the surface. However,both techniques suffered from changes in the reflected light intensity owing to varying reflectance properties of the surface under inspection.The measurements also lacked sufficient sensitivity and were only suitable for measurements on smooth surfaces up to R a=0.5␮m.North and Agarwal [8]circumvented both of these problems by using twofibre optic bundles,which illuminated the surface at two angles of incidence.The ratio of the two reflected light intensities removed the problem of surface reflectivity variation.The instrument showed good correlation with stylus measurements, up to R a=1␮m.Thefibre optic sensors described above do not provide surface profile data;they simply correlate the collected light intensity with the R a measured by a stylus.The fibre optic sensor described below alleviates the problems stated above.
1.2Background
There is a range of high-value,geometrically complex,and dimensionally precise components that are typically inspected on a programmable CMM.For example,turbine blade assemblies,machined on multi-
axis CNC machine tools,must meet dimensional and surfacefinish specifications before they are approved and incorporated into thefinal product.A surface roughness specification(usually R a or R z)is mandatory,owing to the effect that surface texture has on the efficiency of the airflow over the blade surface.As an example,Fig.1shows a turbine CAD model from which a CMM inspection part program is typically derived.The CMM controller uses the
Automated Surface Roughness Measurement
669
Fig.1.(a)Shaded image of turbine assembly;blades plus hub.(b) CAD model of turbine assembly composed of parametric curve and surface entities.
part program to move the touch probe around the component and perform the spatial measurement at each required location. Thefigure illustrates the geometric complexity of the part; each blade can have several machining patches,corresponding to different cutting tool orientations,generated during milling. The component is then removed from the CMM table and a sequence of surface roughness measurements is then made on all blade surfaces.The surface roughness is measured using a manually operated stylus instrument and the R a or R z values are noted for each position.Several R a values are collected from each patch along the length of the blade.The exact location of the measurement is not critical to the result;how-ever,several representative roughness values from each patch are necessary.Inspection efficiency for this type of component could be enhanced by automating the surface texture measure-ment as well as the dimensional measurements.A surface roughness sensor system that can be integrated with a CMM would be desirable and should have the following character-istics:
The sensor head must be physically compact and able to perform roughness measurements on compound curved sur-faces.
The sensor head must be lightweight and suitable for attach-ment on a CMM probe articulating a Renishaw PH10probe head).
The sensor’s operating parameters must be compatible with a stylus profilometer and have a measurement range of0.10ϽR aϽ20with a0.10␮m resolution.
The component must remainfixed in the same position,on the CMM,during both dimensional and surface roughness measure-ments.
The roughness sensor probe head,data processing software and electronic control unit must be transparent to the CMM controller. The motion of the sensor head must be under the control of the CMM using a standard inspection part program.
2.Sensor Operation
The details concerning the operation of the sensor head as an interferometric cavity have been previously reported[9,10]. The essential component of the overall sensor is the head, illustrated in Fig.
2,which is comprised of a single modefibre attached to a cylindrical graded-index(GRIN)lens.The lens focuses the incoming light wave(␸),from thefibre,onto the part surface at a stand-off distance of approximately5.0mm from the lens front face.The front face of the lens is coated with a partially reflective material and divides the incoming light wave into two components:
1.Wave A1,having10%of the original intensity,that is
reflected from the lens front face and which travels back down thefibre.
2.Wave A2,the transmitted portion that is focused onto the
target surface and then collected by the GRIN lens and re-focused back down thefibre.
The intervening surface profile difference,(d2−d1),is meas-ured from the phase difference of the two reflected waves by the principle of
interferometry.
Fig.2.Cut-away view of the sensor head details showing mounting block and GRIN lens.
670  C.
Bradley
Fig.3.A complete surface texture profile obtained by translating the sensor over one sampling length by means of the CMM system.
The sensor,measuring a surface over a sampling length S between the start position A and the end position A Ј,is depicted in Fig.3.A digitised surface profile is generated that consists of a set of surface profile samples {(x i ,z i ):i =1,%,N }.The figure shows the mean reference line placed through the digitised data set.The sensor speed can be accurately set by the CMM controller through the range 0.1–1.0mm s Ϫ1;a slow speed is necessary to prevent unwanted vibration of the sensor head.For a sensor data sampling rate of 1kHz and a sampling length of 1.4mm,the control software will acquire 14000data points.This is a larger data set than required,therefore,the control software permits data decimation (by factors of 10or 100)after scanning is complete.The selection of the sampling length depends on the type of surface under investigation.
The relationship between relative phase shift,laser wavelength and path length difference is given by Eq.(1)and illustrated in Fig.4.The figure shows the sensor acquiring two consecutive measurements of surface profile,at Point 1and Point 2,respect-ively.As shown,the surface height change between the locations is (d 2−d 1).It is assumed in this example that (d 2−d 1)is less than half of the wavelength of the laser,or 0.4␮m.The change in optical phase shift between the two positions is:
⌬␾=(2␲␭−1)2(d 2−d 1)(1)
where,
(d 2−d 1)=vertical distance between the positions of points
1and 2
⌬␾=phase difference between the waves reflected from
the lens coating and the surface
␭=wavelength of the laser (800nm or 0.8␮m)The interference intensity versus optical phase shift change (Eq.(1))is also shown in Fig.4.The variation of intensity,I ,for the displacement (d 2−d 1)is shown.If points 1and
2
Fig.4.Determination of surface height from the sensor’s intensity profile for two points less than 0.85␮m vertically apart.
are positioned as shown on the intensity function,then the movements of the corresponding interference intensity values are C and E .The phase shift measured by the sensor elec-tronics,⌬␸1,is also indicated.Therefore,phase shift or surface height variation can be accurately determined provided the total distance change (from lens to surface and back to lens)does not exceed 0.4␮m.This value is obtained by rearranging Eq.(1)and using the laser wavelength of ␭=0.8␮m.The maximum intensity (I max )is attained when there is zero or 2␲phase shift between A 1and A 2.These are the so-called maximum brightness fringes in standard interferometer nomenclature.Monitoring of I ,between the brightness fringes (fringe 0,fringe 1,fringe 2,etc.)allows the determination of surface profile.
Now consider the situation illustrated in Fig.5,where the sensor moves over the surface,from point 1to point 2,and the surface height varies by more than 0.4␮m.To accommo-date larger surface height variations,the sensor tracks the number of brightness fringes that pass a given reference point,denoted by D in the figures.Therefore,as the sensor moves from point 1to point 2,through a vertical drop of (d 2−d 1)=0.85␮m,the sensor tracks 2fringes or 0.80␮m of vertical drop.Reference to Fig.5shows that “fringe 1”and “fringe 2”have passed the reference point D relative to the starting point given in Fig.4.The electronic sensor counts the fringes that pass the reference point and then measures the intensity in the last fringe.As shown,the remaining phase ⌬␸2is
Automated Surface Roughness Measurement
671
Fig.5.The fringe-counting method for determining the vertical height between two points greater than 0.85␮m apart.
monitored on the photo detector by the increase in intensity
from I Ј1to I Ј
2.This measurement corresponds to the remaining 0.05␮m of vertical profile drop.Employing the fringe tracking method,larger variations in surface topography can also be measured.
Therefore,the sensor provides two output signals to the data acquisition system and supervisory control software:
1.A subfringe signal capable of resolving surface detail to one hundredth of an interference fringe,or 4nm.
2.A fringe counting signal that monitors the larger scale variations in surface topography.The two signals are processed by the software to generate a surface profile of the part,over the sampling length,from which surface roughness amplitude parameters,such as R a and R z can be calculated.
3.Integration of the Sensor System with a CMM
The major integration issues inherent in combining the fibre optic sensor with the CMM (Mitutoyo BH1
0M)are highlighted in Fig.6,which illustrates the main components and the data flow inter-connections.
3.1
Mounting the Sensor Head on the CMM
The mounting of the sensor head on the CMM,the electronic control unit and the software user interface are shown in Fig.7.The physical dimensions of the sensor probe head
are:
image pro plusFig.6.Schematic diagram showing the inter-connection of the main components.Sensor head length 8.0mm Sensor head diameter    5.0mm Fibre optic length up to 4.0m Fibre optic diameter    2.5mm
The sensor is attached to a touch probe head by means of an extension arm.The extension fits the probe-articulating head at one end,and provides a mounting barrel for the GRIN lens at the other.The combination of sensor and articulating head allows flexibility for positioning the sensor relative to the surface of an object.The articulating head allows positioning in 7.5°increments in both of the rotational axes.Overall,this approximates to 5-axis computer control of the sensor head location.This capability is crucial to successful operation of the fibre optic sensor system.The axis of the sensor’s GRIN lens mus
t be maintained at a perpendicular orientation to the object surface during the surface scanning operation.If the object is a compound curved surface,the sensor head must have 5-axis positioning capability.Each position of the articul-ating head is qualified during the measurement procedure rela-tive to a reference sphere placed on the probe tip holder rack.Qualification of the sensor probe position is a standard pro-cedure when using an articulating probe head on a computer controlled CMM.
3.2Synchronisation of the CMM and Sensor System
The commencement of a surface roughness data acquisition scan is synchronised with the motion of the CMM’s servo-motors.The synchronisation method is outlined below and illustrated in Fig.8:
The CMM inspection program positions the sensor head at a measurement location above the object’s surface at the correct stand-off.
The articulated head’s two angular positions are adjusted to position the sensor head perpendicular to the surface,see Fig.9.The sensor head is moved along the evaluation length (a distance of approximately 5.00mm)as data is gathered by the electronic control unit.During the scan,the position feedback sensors,on each axis of the CMM,provide axial position data to the CMM control unit (see Fig.8).The feedback data
is
Fig.7.Photograph of the sensor (mounted on the CMM arm)with the control unit and personal computer data interface.
672  C.
Bradley
Fig.8.Velocity profile of the CMM,in one axis,and the corresponding scale feedback related to the data acquisition timing of the
sensor.
Fig.9.Position of the sensor head relative to the blade surface.
available as an a-quad-b signal (digital pulse signals specifying motor movement and direction)where each pulse corresponds to the smallest distance increment that the CMM can move (for example,1␮m).
This signal is accessed,for all the motor-scale combinations,and input to the sensor data acquisition board to serve as an axial distance measuring signal.Figure 8illustrates the CMM scale signal that is used to measure precisely the distance over which roughness data is acquired.The a-quad-b pulse train is shown for one of the measurement segments,R a 5.Figure 8also shows the entire trapezoidal velocity profile for the CMM motion.The a-quad-b signal is employed to define each of the measurement segments,R a 1to R a 6,in this manner.
An RS 232synchronisation signal,triggered by the CMM inspection part program,initiates the data collection process in the electronic control unit.Once the CMM arm (moving in one axis only)has attained a constant velocity,the data acqui-sition begins.Using the a-quad-b signal,the data acquisition software measures off the 6sampling lengths as shown.Each sampling length is a distance equivalent to (0.8×␭c ).The R a is calculated for five sampling lengths (R a 1,R a 2,%,R a 5)and an average surface roughness value,R ave ,is calculated over the first five sample lengths.This average value is then checked against R a 6,calculated over the sixth sampling length.After the predetermined distance
has been moved by the sensor,the CMM decelerates as shown in the diagram.
Using this technique,the data acquisition card can calculate an R a value that corresponds to accepted standards.
4.Determination of Surface Profile Amplitude Parameters
The necessary steps for processing the sensor data and generat-ing the surface texture amplitude parameters are described below.The entire data acquisition and processing system has been implemented on a Windows based National Instruments LabView system.The data acquisition card is of the DIO-96type and all the data processing algorithms,outlined below,have been implemented in the LabView environment.
4.1Sensor Control and Data Acquisition
The software user interface allows the data collection para-meters to be set before commencing a surface roughness scan.The parameters are the translation speed of the CMM arm over the object surface,the sampling length,start location,stand-off distance of the sensor head above the object and the data sampling rate of the electronic control unit.The interface also synchronises the commenceme
nt of sensor data acquisition with the movement of the CMM arm as discussed above.In particular,the software ensures that no data is collected during the acceleration and deceleration phases of the CMM arm.The data produced by the sensor electronic control unit has two components:
1.A digital signal representing the distance from the lens to the surface expressed as a fraction of the current interfer-ometer fringe.
2.A digital signal that tracks the current fringe number from the commencement of the scanning operation.The combination of both data sets represents the actual height value between the lens and the surface under inspection.The control software converts each data pair into a value expressed in micrometres based on the known value of the laser wave-length (0.80␮m),and each value is stored in the computer memory until the acquisition process is complete.The data is then written to a data file and graphed on the control software interface.
4.2Surface Profile Data Processing
Two initial processing steps are performed before the surface roughness amplitude parameters are calculated.First,the data is filtered to remove high-frequency noise and second;a refer-ence line is determined from which the amplitude parameters are subsequently determined.Figure 10(a )illustrates
the unfil-tered sensor signal obtained when it remains stationary over the same surface location for approximately 0.5s.This sensor noise can be attributed to the air bearing flutter of the CMM gantry and it exhibits a normal distribution about the expected straight line.The noise is removed from the original data before further calculations are performed.The texture profile

版权声明：本站内容均来自互联网，仅供演示用，请勿用于商业和其他非法用途。如果侵犯了您的权益请与我们联系QQ:729038198，我们将在24小时内删除。