Comparison between non contact and contact ...

Second International Conference on Multidisciplinary Design Optimization and Applications, 2-5 September 2008, Gijon, Spain www.asmdo.com/conference2008/

Comparison between non contact and contact scanning systems for dimensional control Susana Martínez*, Eduardo Cuestaa, Joaquín Barreiro*, Braulio Álvareza, Pedro Fernándeza (*)

University of Leon, E.I.I.I. Campus de Vegazana. León. 24071. Spain [email protected]

(a)

Department of Manufacturing Engineering, University of Oviedo E.P.S.I.G. Campus de Gijón 33203, Gijón, Asturias, Spain [email protected]

Abstract In the last years, the presence of non contact scanning systems has increased continuously in the industry. The main reason is that these systems lead to an important reduction in the inspection time and consequently a reduction in manufacturing costs while maintaining quality levels. The advantages of these systems are well known, such as the high speed data acquisition and the high number of captured points. However there exist some disadvantages, like the poor (undefined) accuracy when comparing with traditional touch trigger probe inspection systems. For this reason, scanning systems are mainly used in Reverse Engineering, heritage conservation or multimedia applications (movies, video games, etc.). In metrological applications, their validity has not been tested, in terms of geometric and dimensional tolerance control and accuracy. This work deals with this problem, performing a comparison between two scanning systems. To carry out this comparison, a laser triangulation sensor (LTS) and a touch trigger probe, both mounted on a Coordinate Measuring Machine (CMM), have been used. The scope of this study is the measurement of surfaces representing canonical features: planes, spheres, cylinders (both outer and inner-holes) and conical surfaces (holes with countersink and counterbore). For this study, different test parts have been designed with different manufacturing features and good optical characteristics for the LTS. The test parts were designed taking into account previous studies about the influence of the machining process and surface finishing. The reconstruction of the feature surfaces and the corresponding comparison has been carried out using different CAD systems to analyze the convergence of results among them.

1

Introduction

Nowadays the search for more competitive products, with high quality and low cost has made that the inspection of parts becomes an important task in the life cycle of a product. The inspection and dimensional control tasks 1

Second International Conference on Multidisciplinary Design Optimization and Applications

ASMDO Conference

are usually performed in a CMM. There are several technologies in the inspection process with CMM: the contact digitizing (both, point to point or continuous) and the non contact digitizing. In contact scanning, CMMs maintain the leadership in inspection and quality control processes of industrial parts. Among the main reasons for that, the well known calibration process and measurement uncertainties stand out (Santolaria [3]). On the other hand, the measurement uncertainty is not guaranteed for non contact scanning technologies. Contact digitizing process with touch trigger probes are the most widely used because of their high performance with regard to their cost. The points acquisition can be performed either point to point or in continuous mode. The main disadvantage of these systems is the high operation time needed to obtain a large number of points in the inspection process. This time is even higher when the surfaces to be controlled are complex. In the last years, non contact digitizing processes, in particular those based on laser systems, have accomplished a good level of confidence among the users dedicated to Reverse Engineering (Seokbae et al. [4]). This is mainly due to the high speed of point acquisition and the consequent costs reduction. The fact of the measurement uncertainty is unknown becomes an important disadvantage of these systems, which supposes a drawback for using them in the scope of dimensional control activities. However, a great effort to increase the accuracy of laser systems has been performed in the last years and their application to inspection tasks is growing. Among the laser systems, those based on laser triangulation are the most common in metrological applications mainly due to their higher precision and lower cost with regard to other non-contact systems (structured light, conoscopic holography, image analysis, etc.). However, despite the mentioned advantages, these resources are still more expensive than traditional contact systems.

2

Comparison framework

The objective of this research is the study of non-contact digitizing systems for dimensional inspection of parts. In this sense, a comparison between a touch trigger probe and LTS system has been performed. The test parts have been designed so they include canonical surfaces or simple primitives (planes, cylinders, spheres, and cones). Contact digitizing offers better precision and repeatability, therefore this method has been taken as reference. The accuracy of both systems (contact vs. laser) has been contrasted by performing comparisons between surfaces reconstructed from both contact and laser obtained point-clouds.

Angle B

Angle A

Laser Head

Figure 1: Laser head orientations Two types of sensor have been used: For contact scanning, a touch-trigger probe of 2 mm tip diameter and a length of 40 mm. 2


ASMDO Conference

For laser scanning, a LTS system from Metris model LC-50. Both systems have been mounted on the same CMM, a Global model of Brown&Sharpe (Figure 1). The CMM provides three controlled displacements along its axes (X, Y, Z), and is equipped with an indexable head (Renishaw PH10MQ) capable of rotating around two axes A (horizontal axis), and B (vertical Axis). This way the digitizing system could adopt the most suitable orientation. The uncertainty of the CMM in contact digitizing is given by the expression, according to ISO 10360-2: MPEE[µm]=2.2+3·L/1000 where L[mm] is the magnitude being measured, whereas for the LTS, the manufacturer only provides the parameter 2σ (repeatability measurement) less than 25µm. For the contact inspection, the control software used was PCDMIS v4.1, and for the non contact the software used for the point cloud acquisition was METRIS SCAN v4.0. The study has also considered the effects of using different software applications for surface reconstruction. Commercial applications have been chosen, in particular CATIA v5 and Geomagic v9 (Studio and Qualify). CATIA is one of the most common and powerful CAD/CAM applications. Geomagic Studio is an application used in Reverse Engineering for surface reconstruction, and Geomagic Qualify is used for inspection and quality control. In this sense, the ease of use, the computation time and precision in the reconstruction of surfaces have been analyzed for both applications.

3

Methodology

The following steps have been performed: Design of test parts Generation of inspection program for the contact system Selection of laser scanning parameters in CMM Generation of inspection program for the LTS Points acquisition (contact and laser) Treatment of the acquired point clouds Reconstruction of surfaces in CAD Comparison of reconstructed surfaces 3.1

Design of test parts

Three parts with different canonical features have been designed to study the effect of them in the laser scanning system, as shown in Figure 2: Part A, contains horizontal, vertical and inclined planes, as well as cylinders Part B, contains several slots, fillets and chamfers Part C, contains holes with countersinks and counterbores The manufacturing of the parts has been an important factor. According to previous studies (Cuesta et al. [1]), the part roughness is decisive for the quality of the captured points. In this sense, it is recommended to avoid specular, very shiny or dark surfaces. A high reflectivity surface, such a specular surface generates spurious points, which do not lie on the desired surface. Shiny surfaces causes LTS saturation, and the system cannot acquire it properly. On the other hand, dark surfaces do not reflect the minimum light needed for the point acquisition.

3


ASMDO Conference

For this reason the use of a typical block gauges test is not recommended. Therefore, a manufacturing process should be chosen for these test parts to achieve a good surface finish quality while maintaining adequate optical properties. In this work a wire electro-discharge machining (EDM) process has been used when possible.

(a) Part A

(b) Part B (c) Part C mounted in the fixture Figure 2: Reference parts used in the tests. In order to align the surface geometry obtained in both digitizing systems (contact and non contact), three precision spheres were located over the part surface, as can be seen in Figure 2. Moreover, a fixture has been used, as can be seen in Figure 2c. This fixture has tree additional spheres, with higher radii than those placed on the parts, in order to achieve an optimal alignment fitting. 3.2

Generation of inspection program for the contact system

A PCDMIS program was elaborated for touch trigger probe scanning of each test part. In order to improve the accuracy, a unique probe orientation has been used when possible (part C). For parts A and B, other orientations have been used for avoiding collisions between the stylus and the vertical surfaces of the parts. The PCDMIS programs are divided into several subroutines to capture the points over each feature (plane, cylinder …) individually. The number of selected points to capture each canonical feature was higher than the minimum required for the feature mathematical definition. For instance 28 points were used to define each plane when only four points were needed. The same strategy was applied to the rest of the features (cylinders, cones and spheres). In this way these point clouds are closer to the point clouds captured by laser scanning. Each point cloud obtained was used to reconstruct the simple surfaces with PCDMIS, CATIA and Geomagic applications. Surfaces reconstructed with PCDMIS will be considered as reference geometry for comparisons with those reconstructed with the laser system. 3.3

Selection of laser scanning parameters in CMM

Previous studies (Rico et al. [2]), revealed the influence of some laser parameters over the accuracy of captured points. In consequence, the most suitable values were selected for carrying out the tests. These values optimize the digitizing process since the maximum number of points is acquired with the minimum dispersion. As can be seen on Figure 3, the LTS has a laser light source which emits a laser beam over the part surface. The image of the intersection of the laser beam with the surface is captured by a CCD sensor. Then, the image is processed and the 3D coordinates of the captured points are obtained by triangulation techniques. Tests were carried out in absence of any light. The values of the following parameters were kept constant: Point distance: 0.5 mm 4


ASMDO Conference

Stripe distance: 0.5 mm Overlap: 0.1 mm Laser intensity: 39 %

Resonant Scanning mirror

Laser Source

Image on the CCD

CCD Sensor

Laser beam

Lens & Filters Field of View Laser stripe

Scanning Direction

Figure 3: The LTS scanning system

The point distance is the distance between digitized points measured on the same laser stripe (Figure 3). The stripe distance is the distance between two consecutive digitized lines. These two parameters define a uniform grid in the point cloud with only a point in each cell of 0.5 x 0.5 mm. Therefore these parameters fix the point cloud density. Although this density could be excessive for primitive surfaces (planes, cylinders, etc.), this value has been considered convenient in order to avoid lack of information in fillets and edges with abrupt changes of direction. The overlap denotes the width of the area where two consecutive parallel sweepings overlap. It is applied when performing several parallel scans. Finally the laser intensity represents a 39% of the laser maximum power (1 mW). This intensity value provides a high density of captured points and is the maximum intensity that does not cause the CCD sensor to enter in saturation, and therefore minimize the number of wrong points acquired. Additionally, it is possible to apply filters for removing wrong points, such as the saturated points (points that appears too much illuminated on the CCD) and the low quality points (points whose image on the CCD is too weak). In this study, these filters were set at 95% to remove a very high percentage of low quality points, and all the saturated points. 3.4

Generation of inspection program for the LTS

The laser sensor native application, METRIS SCAN, was used to carry out the laser scanning, the calibration of the different head orientations and to define the scanning paths. From the point of view of the CMM head, two different strategies were used with the aim of analyzing their suitability. The first one is a global strategy, and consists on performing a complete scans of the part with the minimum orientations needed. In this case, it is possible to scan the parts only with two vertical orientations, position (A = 0º), and twist the B axis 180º for each scan to avoid occlusion phenomenon. The disadvantage of this global strategy is that it is impossible to obtain any point in the vertical surfaces, losing all the information about them. 5


ASMDO Conference

The second strategy (multi-oriented strategy) consists on digitizing each region of the part surface independently. The best orientation to digitize a surface is when the laser beam is perpendicular to the surface, so for each region the orientation that is nearest to the surface normal was selected. This head orientation selection is not always possible, due to the occlusion and accessibility problems. This second strategy increases the operational time of the test for several reasons. The pre-processing time is greater due to the necessary qualification of every used orientation. The scanning time also increases with the number of different scans that are necessary to digitize completely the part. Finally, the point clouds treatment is more complex due to the number of different point clouds. 3.5

Points acquisition

Once the digitized programs and strategies were verified, the tests were carried out consecutively to minimize the negative effect of different test conditions. The machine only was restarted to change the touch probe for the laser head while the test parts remained at the same location during all the digitizing operations. In the case of contact digitizing, a part reference system was considered. However, in the case of the laser digitizing, 3D coordinates of the captured points are related to the machine reference system. This fact does not represent an inconvenient for surface reconstruction. However it constitutes a problem for the later comparison between the reconstructed geometries. Therefore it is necessary to make an alignment of both reference systems. Several methods exist to perform these alignments (Yau et al. [6], Wolf et al. [5]). One of the most common and simple method consists on placing three reference spheres that are digitized at the same time that the pattern surfaces. The surfaces of the spheres are reconstructed and their centers are used to establish a unique reference system. In this work, a double set of references spheres has been used, three spheres were in the fixture system and the other three were fixed to the test parts. The objective is to digitize the spheres with both scanning systems and then to correlate the corresponding centres of each captured sphere. The point clouds acquired can be seen in Figure 4. Points on contact point cloud are distributed in a homogeneous way and concentrated on each feature whereas in the laser point cloud the points are distributed uniformly over the entire surface.

(a) Contact point cloud

3.6

(b) Multi-oriented Laser point cloud Figure 4: Points clouds for part C

Treatment of the acquired point clouds

The point clouds obtained by contact scanning do not need any treatment. As can be seen in Figure 4a, all the captured points in the PC-DMIS program correspond to nominal points extracted from the part CAD model, so wrong points cannot exist. Point clouds were exported to IGES format that can be later imported into any CAD software to build the geometry. Moreover all the primitive features were recreated in PC-DMIS from the 6


ASMDO Conference

acquired points in order to measure them. The measured values have been considered as reference for later comparisons. The laser point clouds obtained with the global strategy (vertical orientation, A = 0°) only require the manual elimination of some wrong points (spurious points). These wrong points are determined subjectively, the user has to eliminate them with the tools provided by METRIS SCAN. On the other hand, point clouds from the multi-oriented strategy required more post-processing treatment. The orientation of each scan was determined with regard to a particular surface. However, points of adjacent surfaces are acquired simultaneously. These points should be eliminated since they are not captured with the best orientation. Finally, these point clouds were exported to IGES format too. 3.7

Reconstruction of surfaces in CAD

Two different software packages were used for reconstructing each surface from point clouds: -

CATIA V5R18, modules: Digitized Shape Editor and Quick Surface Reconstruction

-

Geomagic Studio V9

The reconstruction process was first carried out using point clouds from contact scanning, and later repeated with those point clouds obtained with laser scanning. After each point cloud was imported in the CAD system, points were clustered in different clouds according to corresponding part features (planes, cylinders, cones). This step was faster and simpler in the case of contact point clouds due to the homogeneous distribution of the points over each digitized surface. However in the case of laser point clouds the main troubles were to determine exactly the feature boundary, and the number and distribution of the points over the surface. For instance, few points were acquired at the countersinks of the holes of the part C and with no homogeneous distribution (points were concentred in some areas, as can be seen in Figure 4b). Chamfers and the fillets of part B made difficult to distinguish between the points of the adjacent planes and the points of the feature. Each individual region is adjusted to the corresponding primitive region, applying the algorithms that provide each CAD system. Once each feature is reconstructed, these new features are joined in order to obtain a global reconstruction of the part geometry closer to the real. This joint is carried out by trimming boundary defects and stitching the feature surfaces along the intersections between individual features (Figure 5). In spite of the clear focus of the Geomagic Studio application towards the working with points-clouds, it has been observed that CATIA deals perfectly with primitive surfaces reconstruction, with the additional advantage of its great potential for surface design.

(a) CATIA V5 with a LTS point cloud (b) Geomagic V9 with a contact point cloud Figure 5: Reconstruction of surfaces in different software and different digitizing system

The surfaces reconstructed with CATIA and Geomagic Studio were exported to IGES format too. Then, the Geomagic Qualify V8 application was used to perform the surfaces comparison.

7


3.8

ASMDO Conference

Comparison between reconstructed surfaces

The Geomagic Qualify is a software package oriented to the virtual metrology, on point clouds. By means of this software it is possible to compare the deviations between the reconstructed surfaces obtained in the previous stage. In this comparison geometric tolerances (parallelism, perpendicularity, angularity, cylindricity) were analyzed. Also, the dimensional analysis was considered.

Figure 6: 3D Comparison between two reconstructed geometries

Figure 6 shows the three-dimensional differences between the surfaces reconstructed from the contact point clouds (reference) and those reconstructed from the “multi-oriented” laser point clouds. In both cases the CATIA application was used for the reconstruction of surfaces. It can be observed that the three-dimensional adjustment is very good, and that the maximum 3D deviation does not exceed 20 µm.

4

Results

As it was said previously, the dimensional analysis was performed using PC-DMIS in the case of contact scanning and this scanning was used as reference entity for the later comparison in Geomagic Qualify. The comparison was carried out both dimensionally and geometrically (GD&T) including parallelism, perpendicularity and angularity tolerances. This paper does not show all the results, but some of the most representative ones. Figure 7 shows the meaning of the titles that will be used for illustrating the whole survey. Caption

Source of values Dimensional analysis of PC-DMIS by contact

PC-DMIS CATIA

Surface reconstruction with CATIA from contact point clouds

GEOMAGIC

Surface reconstruction with GEOMAGIC from contact point clouds

GLOBAL CATIA

Surface reconstruction with CATIA from laser point clouds (global strategy)

GLOBAL GEOMAGIC

Surface reconstruction with GEOMAGIC from laser point clouds (global strategy)

MULTI-ORIENTED CATIA

Surface reconstruction with CATIA from laser point clouds (multioriented strategy)

MULTI-ORIENTED GEOMAGIC

Surface reconstruction with GEOMAGIC from laser point clouds (multi-oriented strategy)

Figure 7. Meaning of the captions used in results figures. 8

Second Intternational Connference on Muultidisciplinary Design D Optimizzation and Appllications

ASMDO Conference C

In Figuree 8, the analyysis of distancce between veertical planes of part A is shown. Theree are no valu ues for the global strrategy becausse no points can c be acquirred on verticaal planes wheen only the veertical head orientation o (A = 0º) is used. For the t other straategy, the surffaces reconstrruction obtainned from conntact point clo ouds (both CATIA and a GEOMAG GIC) achieves similar valuees to those useed as referencee (PC-DMIS).. In this case, deviations d are aboutt 2 µm, whichh could be connsidered negliggible. These deviations d incrrease to a great extent (15 µm) µ when values obbtained from laser point clouds are analysed. a In these point clouds, c a revversed tenden ncy in the inner/outeer slot occurs. While dv1 (inner ( slot) deeviation is possitive, dv2 (ouuter slot) deviation is negattive, being both simiilar in amounnt. The opposiite sign of these deviations is due to thhe difference bbetween an in ncomplete point clouud used for ann inner slot annd the good quuality point clo oud obtained in i an outer sloot.

dv2

MU ULTI-ORIENTE ED GEOMAGIC (Laser) MU ULTI-ORIENTE ED CATIA (Laserr) GE EOMAGIC (Contact)

dv1

CA ATIA (Contact) PC C-DMIS (Contactt)

14.965 14.970 14.975 14.9980 14.985 14.990 1 14.995 15.000 155.005 15.010 15.015

m mm

Figure 8: Comparison of CAD systeems for distancces between vertical v planess of Part A Figure 9 shows the anngularity dimeensions (in deegrees) of thee plane of chaamfers in parrt B. For the angularity dimensions labelled 2 and 5, the vaalues obtainedd in CATIA and a GEOMAG GIC differ to a great extentt from the referencee PC-DMIS values, v overalll when the GEOMAGIC G reconstruction r n is considereed. This effeect can be explainedd in the same way as beforee. Down cham mfers have beeen reconstructted using a lesss number of points; p the points off low quality are a due to a bad b scanning head h orientatiion that had to be selected for avoiding occlusion problemss. This trend remains r in moost of the valuues obtained in i CATIA, buut in this casee with a less deviations. d When muulti-oriented sttrategy is com mpared againstt global strateg gy, no more accuracy a is achhieved. Figure 100 illustrates annother analysiss using distannce between ho orizontal plannes of part B. IIt can be seen again that deviationns from contact point cloudds are small, less than 5 µm. µ However,, deviations ffrom laser point clouds increase, reaching a mean m value of 15 µm. When comparing multi-orientedd and global strategies as before, b no differencee between theem has been foound.

9

Second Intternational Connference on Muultidisciplinary Design D Optimizzation and Appllications

ASMDO Conference C

Ang g2 Ang g5 Ang g6 GLOB BAL GEOMAGIIC (Laser) MUL LTI-ORIENTED GEOMAGIC G (Laaser) GLOB BAL CATIA (Laaser) MUL LTI-ORIENTED CATIA C (Laser) GEOM MAGIC (Contactt) CATIIA (Contact) PC-D DMIS (Contact)

Ang g3 Ang g4 Ang g1 0.000

0.025

0.050

0.075

0.100

0.125

0.150

0.175

0.200

0.2255

0.250

Figuure 9: Compaarison of CAD D systems for angular dimennsions of Partt B

dh7 dh6 dh5 dh4 GLOBAL GEOMAGIC (Laser) G M MULTI-ORIENTE ED GEOMAGIC C (Laser) G GLOBAL CATIA A (Laser) M MULTI-ORIENTE ED CATIA (Laseer) G GEOMAGIC (Con ntact) C CATIA (Contact) P PC-DMIS (Contacct)

dh3 dh2 dh1 19.970

19.9775

19.980

19.985

19.990

1 19.995

20.000

20.0005

Figure 100: Distances between b horizontal planes for f part B. 10

20.010 mm


5

ASMDO Conference

Conclusions

The applicability of laser systems for dimensional control of industrial parts has been analyzed. Meanwhile, the influence of the CAD system used for the surface reconstruction and for the dimensional and geometrical control has been studied. Taking into account the tests, several conclusions can be extracted about the suitability of laser digitizing for dimensional control and about how different CAD systems reconstruct canonical surfaces from the same point cloud. In the case of laser digitizing, neither of the two CAD systems available allow for a straightforward surface reconstruction. Chamfers, fillets and countersinks caused troubles in the fitting of canonical surfaces, such as cylinders and cones. The two cylindrical surfaces (concave and convex) of the part A can be reconstructed easily and with a minimum error between the surface and the point cloud. This is due to the visible area of these cylinders was near 180º of angular extension and therefore a high number of points was acquired both for contact and laser. For the fillets of the part B, these surfaces only cover from 0º to 90º, so a low number of points generates errors in radii and in the centre location, avoiding an accuracy cylinder creation. In laser scanning, there are two opposite effects. First, the dispersion of points is higher than in contact and, on the other hand, although there are more points the impact in the GD&T accuracy due to the first effect is higher than the second. About the laser scanning strategies, the multi-oriented strategy is recommended only when vertical surfaces are being digitized. This strategy not only does not provide a higher accuracy but also spends much more operation time. Comparing with the global strategy, the multi-oriented one requires more time for probe qualification (more orientations needed), more time for point acquisition and more time for point cloud post-processing. From the point of view of the CAD system used, CATIA system was more useful than GEOMAGIC for the reconstruction of canonical surfaces. This is not true for freeform surfaces, which need posterior surveys in order to evaluate levels of accuracy, programming time for surface reconstruction, ease of use, etc.

6

Acknowledgements

This paper is part of the results of a Research Project supported by the Spanish Comisión Interministerial de Ciencia y Tecnología (about automated planning with laser scanning sensor, ref. DPI2004-03517). Also, this work was supported by the Spanish R2-TAF initiative through a researcher mobility grant.

References [1] Cuesta, E. and García, A. and Álvarez, B. and Valiño, G. and Rico, J.C. Análisis de la influencia del acabado superficial en la calidad del digitalizado con sensores láser por triangulación. In Proceedings of the 2nd Manufacturing Engineering Society International Conference (CISIF-MESIC´07) (CD-ROM). 2007 [2] Rico, J.C. and Fernández, P. and Cuesta, E. and Valiño, G. and Mateos, S. Evaluación de los parámetros de influencia en el digitalizado por laser de barrido sobre una máquina de medir por coordenadas (MMC). In Proceedings of the 3er Congreso Españo de Metrología (CEM). Zaragoza, 2005. [3] Santolaria, J. Diseño y Calibración de brazos articulados de medición por coordenadas e integración de sensor láser por triangulación para medición sin contacto. Thesis. Zaragoza, 2006 [4] Seokbae, S. and Hyunpung, P. and Kwan, H.L. Automated laser scanning system for reverse engineering and inspection. In International Journal of Machine Tools & Manufacturing. 42, 889-897, 2002. [5] Wolf, K. and Roller, D. and Schäfer, D. An approach to computer-aided quiality control base don 3D coordinate metrology. In Journal of Materials Processing Technology. 107, 96-110, 2000. [6] Yau, H-T. and Chen, C-Y, and Wilhelm, R.G. Registration and integration of multiple laser scanned data for reverse engineering of complex 3D models. In International Journal of Production Research. 38, no 2, 269285, 2000 11