Vous êtes sur la page 1sur 8

ARTICLE IN PRESS

International Journal of Machine Tools & Manufacture 47 (2007) 12291236 www.elsevier.com/locate/ijmactool

Evaluation of form data using computational geometric techniquesPart I: Circularity error


N. Venkaiah, M.S. Shunmugam
Manufacturing Engineering Section, Department of Mechanical Engineering, Indian Institute of Technology Madras, Chennai600036, India Received 12 March 2006; received in revised form 12 August 2006; accepted 15 August 2006 Available online 2 October 2006

Abstract The present work deals with evaluation of form error from the measured proles obtained using a form tester, namely roundness/ cylindricity measuring instrument. In Part I, details of circularity evaluation are presented. Due to eccentricity in component setting and radius-suppression inherent in the measurement, circularity error has to be evaluated with reference to a limacon. A computational geometry-based algorithm is proposed for establishing minimum circumscribed, maximum inscribed and minimum zone limacons. A new type of control hull for directly constructing equi-angular diagrams and a new procedure for updating are introduced. Validation has been done with bench-mark data set and corresponding results available in the literature. Being geometry-based algorithm, it is simple to follow and each iteration can be visualized and interpreted geometrically. On comparison with simplex search method, the proposed algorithm is found to be computationally efcient in terms of accuracy and time taken. The proposed methods can be easily implemented in computer-aided roundness measuring instruments. Extension of this work for evaluation of cylindricity error has been dealt in Part II. r 2006 Elsevier Ltd. All rights reserved.
Keywords: Circularity error; Form data; Limacon; Computational geometry; Control hull; Equi-angular diagrams

1. Introduction The geometric form of any manufactured feature always deviates from its nominal design to some degree, owing to the random and/or systematic errors. In order to satisfy certain functional requirements or assembly conditions, geometric tolerances are usually assigned to selected features. One such feature often used in engineering components is a cylindrical feature. In certain applications, it is enough to consider a transverse section of the cylinder and apply circularity tolerance. Verication of this tolerance is done using a roundness tester, which measures the deviations from an ideal circular trajectory established by the instrument. In case of cylindricity measurement, an additional straight datum is used and the measurements are carried out at few transverse sections of the cylinder. In such measurements, the size/radius of the component is always suppressed and the measurement data is referred to

Corresponding author. Tel.: +91 44 22574677; fax: +91 44 22570509.

E-mail address: shun@iitm.ac.in (M.S. Shunmugam). 0890-6955/$ - see front matter r 2006 Elsevier Ltd. All rights reserved. doi:10.1016/j.ijmachtools.2006.08.010

as form data. The form data is usually characterized by equi-spacing. The prole deviations obtained using roundness measuring instrument are magnied and plotted on a polar chart. The earlier manufacturers of the instruments provided templates with concentric circles to evaluate the circularity error. Whitehouse [1] and Chetwynd [2], however, showed that due to eccentricity between the rotational center of the instrument and center of the component, even a truly circular component results in a non-circular prole. Evaluation using a pair of concentric circles would show certain circularity error and an error value greater than the specied tolerance might lead to the rejection of the component. An exact mathematical representation of the measured prole obtained for a truly circular component for a given eccentricity, radius suppression and magnication was obtained and the rst-order approximation was shown to be a limacon [3]. Therefore, limacon would be the appropriate assessment feature for evaluation of circularity error. Geometrically, a limacon is obtained as a locus of the foot of the perpendiculars drawn from the origin (pole) to the tangents of a circle having its center offset from the

ARTICLE IN PRESS
1230 N. Venkaiah, M.S. Shunmugam / International Journal of Machine Tools & Manufacture 47 (2007) 12291236

Nomenclature ei i ri, yi ro xo, yo Cf Cn EA deviation of ith point from assessment limacon index for data points polar coordinates of ith point radius of circle for assessment limacon center coordinates of circle for assessment limacon farthest center nearest center equiangular line

FE LO LI LS MC MI MZ NE V D

farthest edge edge of an outer control hull edge of an inner control hull least squares minimum circumscribing maximum inscribing Minimum zone nearest edge vertex on a control hull form (circularity) error

origin. Though the templates with concentric circles are no longer provided with roundness testers, the terms such as least squares circle (LSC), maximum inscribed circle (MIC), minimum circumscribed circle (MCC) and minimum zone circle (MZC) are quite liberally used by the industry while specifying the method used for evaluating the circularity error. ISO species that an ideal feature must be established from the measured prole such that deviation between it and the measured prole is the least possible value [4]. In practice, the ideal feature is taken to be a straight line for straightness, a plane for atness, etc. Some researchers simply extend this to circularity evaluation without any consideration to the method of measurement and use a circle as an assessment feature for the form data. Even the use of the circle is justied with a claim that the limacon is an approximation of a circle and therefore only the circle must be used for obtaining correct results. In some cases, even though the center coordinates of circle for establishing the limacon are obtained using least squares method (LSM), two concentric circles are drawn from this center enclosing the entire measured prole and circularity error is evaluated as radial separation between the two circles. The fact that the limacon is the basis for the LSM and the parameters evaluated actually correspond to the limacon is often ignored [5]. A number of attempts have been made in the past to evaluate circularity error from the form data obtained using roundness measuring instruments with limacon as the assessment feature. LSM is one such attempt based on a sound mathematical principle that minimizes the sum of the squared deviations of the measured points from the tted limacon feature [6,7]. This method is robust, but it does not follow the standards intently and will not guarantee the minimum zone (MZ) solution specied in the standards. The deviation values and the form error that are determined by LSM will be generally larger than the actual ones, and this may lead to rejection of good parts. To obtain the MZ solution in circularity evaluation, the numerical methods based on simplex linear programming [8] and simplex search [9] have been adopted. Shunmugam [6] suggested a new simple approach called the median technique, which gives minimum value of circularity error.

Using discrete Chebyshev approximations, Danish and Shunmugam [10] have arrived at the MZ values. In the past decade, computational geometric techniques have gained enormous attention from the designers of algorithms for solving geometric problems [11,12]. These techniques also show greater promise for solving the MZ problems encountered in the geometrical evaluations. They can be very well applied for the evaluation of the circularity error in the manufactured components. Though application of Voronoi diagrams for construction of circle has been reported in the literature, rst attempt to develop computational geometric techniques based on limacon was reported by Samuel and Shunmugam [13]. Two limacons obtained from the same center, enclosing the entire measured prole and having minimum separation, result in an MZ value. The concepts of function-oriented evaluation, namely methods based on minimum circumscribed (MC) and maximum inscribed (MI) limacons, are also applied for circularity evaluation. In order to obtain MC and MI limacons, convex outer and inner hulls have been used in the past. Construction of convex hulls considerably reduces the number of candidate points for establishing the assessment limacons. However, there is a need to reduce the candidate points further in cases where a large number of data points are to be evaluated. It is also observed that the algorithms used to construct the convex inner hull are not quite consistent [5,13]. Part I deals with the evaluation of circularity error bringing out a new concept of control hull which leads to consistent inner hulls and reduces the number of candidate points for establishing the MC, MI and MZ limacons. A new procedure for updating is also introduced. The results obtained for the circularity data reported in the literature are included in this part. In comparison with simplex search method, the proposed algorithm takes lesser time, leads to higher time saving for large datasets, and gives accurate results in absence of any convergence criterion. Part II of this paper deals with the cylindricity evaluation from the form data [14]. 2. Circularity error Fig. 1 shows a point on the roundness prole represented by {ri, yi}. The gure also shows a limacon used for the

ARTICLE IN PRESS
N. Venkaiah, M.S. Shunmugam / International Journal of Machine Tools & Manufacture 47 (2007) 12291236 1231

assessment, as a curve representing the locus of foot of the perpendiculars drawn from origin O (pole) to the tangents of a circle whose radius is ro and center is Oo(xo, yo) [13]. Fig. 2 shows the deviation of a given point Pi from the assessment limacon. For quantifying this deviation ei, another limacon passing through Pi may be assumed and the circle from which this limacon is obtained has the same center, but different radius. It can be clearly seen that the separation between these two limacons is a measure of the deviation of point Pi from the assessment limacon. Interestingly, the radial distance between the two circles from which the limacons have been generated is also equal to the deviation ei, but the point Pi falls outside the corresponding circle. By convention, deviation of a

measured point lying outside the assessment limacon is taken to be positive and a point inside is considered to have negative deviation. The circularity error (D) is, therefore, obtained as absolute sum of the maximum and minimum deviations. 3. Control hull Crest and valley limacons pass through the extreme measured points. Therefore, it was considered appropriate to construct corresponding convex hulls. Outer hull was constructed as smallest convex hull enclosing all the measured points and inner hull was taken as largest empty convex hull. For each line connecting a vertex of the convex hull and the origin, a perpendicular line was drawn and equi-angular lines were constructed at the intersection of these perpendicular lines. It is observed that the algorithms [5,13] found in the literature for the convex inner hull are not consistent in constructing the same. While analyzing several alternatives for constructing unique inner convex hull, it is realized that the perpendicular lines drawn at each measured point yield a polygon, which can be directly used for constructing equi-angular diagrams. This polygon also results in the reduction of the number of candidate points for the construction of equiangular diagrams and hence the subsequent computational effort. To differentiate between this polygon and the conventional convex hull used in computational geometry literature, the polygon obtained in the present work is referred to as control hull. The new concept of forming a control hull followed in the present work is explained using Figs. 3(a) and 4(a). As shown in Fig. 3(a), at each point on the measured prole, a line is drawn perpendicular to the radial line joining the measured point and the origin, namely the instrument center. A smallest polygon is formed

Fig. 1. Circularity data and limacon (for illustration only).

Fig. 2. Deviation of a point from limacon.

ARTICLE IN PRESS
1232 N. Venkaiah, M.S. Shunmugam / International Journal of Machine Tools & Manufacture 47 (2007) 12291236

Fig. 3. Construction of farthest EA diagram: (a) outer control hull, (b) initial farthest EA edges, (c) updated hull and farthest EA edges, (d) farthest EA diagram and MC limacon.

by the intersection of these lines such that all the measured points are conned within it. Such a polygon is referred to as outer control hull. Similarly, largest empty polygon obtained by the intersection of the perpendicular lines as shown in Fig. 4(a) is referred to as inner control hull. Interestingly, outer and inner control hulls are also convex in nature.

4. MC limacon Fig. 1 shows hypothetical prole of a circular feature. Outer control hull of the measured points is constructed as explained earlier. The edges of the hull such as LO1, LO2, LO3, LO4 and LO5 are established by the measured points P1, P3, P6, P8 and P11, respectively, as shown in Fig. 3(a).

ARTICLE IN PRESS
N. Venkaiah, M.S. Shunmugam / International Journal of Machine Tools & Manufacture 47 (2007) 12291236 1233

Fig. 4. Construction of nearest EA diagram: (a) inner control hull, (b) initial nearest EA edges, (c) updated hull and nearest EA edges, (d) nearest EA diagram and MI limacon.

In the next step, for each pair of adjacent edges, equiangular (EA) lines, such as EA12, EA23 etc are constructed at V12, V23, etc. as shown in Fig. 3(b). If a line EA23 is considered, it intersects with EA12 and EA34 at Cn (C12) and Cf (C34), respectively. The intersection point Cf, which is the farthest from the vertex V23, is the farthest center (C23). A circle drawn with this farthest center and with suitable radius will be tangential to the respective edges of the hull. For example, a circle with center Cf (C23) can be drawn tangential to the lines LO2 and LO3. It should be remembered that the circle being referred to here represents the circle from which the limacon is established. The limacon constructed from this circle would pass through points P3 and P6. The portion of the line EA23 beyond Cf, away from V23 is the farthest EA edge (FE23) corresponding to the edges LO2 and LO3 of the hull. Following the procedure outlined here, the farthest EA edges corresponding to all pairs of edges of the hull are constructed as shown in Fig. 3(b). It is seen in Fig. 3(b) that a few edges have common farthest centers. For example, farthest edges FE23

and FE34 have C23 and C34 as their centers, and for any point in the region enclosed by these two edges, the edge LO3 is the farthest. The outer hull is therefore updated by dropping the edge LO3. In case of the conventional convex hull, a vertex is dropped while updating [5,13]. The edge LO5 is also dropped and a new hull is formed as shown in Fig. 3(c). New EA edges are formed for this updated hull following the same procedure. The new edges FE12, FE24 and FE41 pass through a common center and therefore the procedure comes to end at this stage. The complete diagram as shown in Fig. 3(d) is referred to as farthest EA diagram. By taking the centers of the complete EA diagram, a number of circles can be drawn for the given data set. Out of these circles, one having the least radius is chosen to construct the MC limacon. The smallest circle thus obtained is tangential to the edges LO1, LO2 and LO4 of the hull with center at C124. For the sake of understanding, the limacon established from this circle is also shown in Fig. 3(d), and this limacon passes through P1, P3, and P8.

ARTICLE IN PRESS
1234 N. Venkaiah, M.S. Shunmugam / International Journal of Machine Tools & Manufacture 47 (2007) 12291236

5. MI limacon Fig. 4(a) shows the inner control hull of the measured points, constructed as explained earlier. This hull is used to construct the nearest EA diagram. The procedure for constructing the nearest EA diagram is similar to that of the farthest EA diagram, except that the nearest intersection points are considered as end points of the nearest EA edges instead of farthest intersection points. Fig. 4(b) shows the nearest EA edges for the initial hull. The nearest EA edge between LI2 and LI3 is obtained by considering three EA lines namely, EA12, EA23 and EA34. The nearest intersection point is taken as the nearest center Cn (C23), and the EA line toward the vertex V23 from this center is the nearest EA edge NE23. The hull is updated by dropping the nearest edges on the hull from the common nearest centers. For example, the EA centers C23 and C34 are common, and hence the edge LI3 is dropped. Similarly, the edge LI1 is dropped, and a new hull is formed as shown in Fig. 4(c). The complete nearest EA diagram and the MI circle and corresponding limacon are also shown in Fig. 4(d). It can be noted that the MI limacon thus constructed passes through P5, P10 and P12 with its center at C245. 6. MZ limacons Two concentric limacons with least possible separation between them and enclosing all the measured points are said to be MZ limacons. The MZ limacons are controlled by at least four points of the dataset. This can occur when three points lie on one limacon and the fourth point on the other limacon concentric to it or when two points are on each limacon. The farthest and nearest EA (shown as

dashed lines) diagrams are superimposed, as shown in Fig. 5. The intersection points of these two diagrams are the candidate centers [5]. The smallest possible circumscribing circle and largest possible inscribing circle with center at the intersection points of farthest and nearest EA diagrams are obtained and corresponding limacons are established. Such limacons will contain all the measured points between them. In case an intersection point coincides with a common center of either the farthest or nearest EA edges, one of the limacons would be controlled by three points. Otherwise, both the limacons are controlled by two points each. The radial distance between these concentric limacons is found, and a pair having minimum radial separation gives the MZ limacons. The radial distance between these limacons is the circularity error. It is interesting to note that the difference between the radius values (ro) of the outer and inner circles corresponding to the respective limacons also denotes the MZ error. The center Oo obtained by the intersection of the farthest and nearest EA diagrams as shown in Fig. 5 represents the center corresponding to the circles for MZ limacons. It may be noted in Fig. 5 that the outer limacon passes through two points P3 and P8, whereas the inner limacon passes through P5 and P10. 7. Results and discussion The proposed algorithms were applied on different sizes of data sets. Inclusion of large sets in this paper requires more space and even in such cases, only a few signicant points nally inuence the control hulls. Also, the points controlling the hull are not equi-spaced even for equi-spaced data set. Therefore, it is considered prudent to include in this paper a bench-mark data set taken from the literature instead of a large and new data set and to validate the proposed algorithms with the results published for the bench-mark data set. The bench-mark data set [6] shown in Table 1 is used in the present work and salient features of the proposed algorithm are brought out. Table 2 shows the results obtained on the basis of MC, MI and MZ methods and the corresponding roundness error (D) values. The outer control hull is formed by four points only and it needed one update to get the complete farthest EA diagram. The MC circle is tangential to LO1, LO3 and LO4 and the limacon obtained from this circle passes through the points 1, 4 and 6. The inner control hull is formed by six points necessitating two updates to get the complete nearest EA diagram. The MI circle is tangential to LI3, LI4 and LI5 and the limacon obtained from this circle
Table 1 Circularity data [6] Point i 1 0 4 2 45 4 3 90 3 4 135 5 5 180 2 6 225 3 7 270 1 8 315 2

Fig. 5. Superimposed farthest and nearest EA diagrams with MZ limacon.

yi (deg) ri (mm)

ARTICLE IN PRESS
N. Venkaiah, M.S. Shunmugam / International Journal of Machine Tools & Manufacture 47 (2007) 12291236 Table 2 Results of circularity evaluation Iteration Limacon MC edge (point i) Initial 1 2 Final Parameters xo (mm) yo (mm) ro (mm) D (mm) LO1(1), LO2(2), LO3(4), LO4(6) LO1, LO3 LO1, LO3, LO4 (1, 4, 6) 0.0000 1.4142 4.0 2.4142 MI edge (point i) LI1(1), LI2(2), LI3(3), LI4(5), LI5(7), LI6(8) LI2, LI3, LI5, LI6 LI3, LI5, LI6 LI3, LI4, LI5 (3, 5, 7) 0.0000 1.0000 2.0 2.2929 MZ edge (point i) LO1, LI3, LO3, LI4 (1, 3, 4, 5) 0.1213 1.1213 3.0 (1.8787, 4.1213) 2.2426 LSM 1235

0.1464 1.2071 3.0000 2.4571

passes through the points 3, 5 and 7. The MZ circles are obtained by the intersection of farthest and nearest EA diagrams. The outer circle (max ro ro+D/2 4.1213 mm) is tangential to LO1 and LO3 and the limacon obtained from this circle passes through the points 1 and 4. The inner circle (min ro roD/2 1.8787 mm) is tangential to LI3 and LI4 and the limacon obtained from this circle passes through the points 3 and 5. The parameters of the assessment limacons and corresponding error values have been included in the Table 2. For the purpose of comparison, the values based on the LS method are also included in Table 2. The equations for estimating LS parameters are given in Appendix A. The LS method yields higher values of circularity error. The evaluation of the circularity error by the computational geometric techniques yields accurate results. The results obtained with the present method are in agreement with those reported results [13]. However, search methods reported in the literature for the evaluation of circularity error based on MZ approach require some value for convergence, as search is made in parameter space, namely {xo,yo,ro}. The algorithms proposed in the present work do not require any specied convergence value and the results depend only on the points in the data set that are used for establishing the assessment limacons. The values of the center position and size obtained by LS method show considerable statistical stability, because all the measured points are included in the computation of these values. In case of MZ and function-oriented evaluation (MC and MI), the center position and size are determined by the extreme points. In the rst instance, the features established by the extreme points may seem to have little practical signicance. It should be remembered here that in any method used for establishing the assessment/reference features, the value of the circularity error is nally computed on the basis of extreme points only. Therefore, proper care must be exercised in all cases with regard to the unstable extreme points and uncertainty in their measurements.

In the proposed algorithms, the outer control hull and the inner control hull of the measured points are taken so that most of the points that do not inuence the nal results are eliminated, and the construction of EA diagrams becomes less cumbersome. It is also observed that the number of points on the outer/inner control hull is considerably lesser than that on its conventional counterpart, namely outer/inner convex hull, thereby minimizing the computational effort. The computer programs for the proposed algorithms were written in C++ and run on a Pentium (R) IV, 256 MB RAM, 2.4 GHz machine. For the bench-mark data set, LS method took 0.002 s, while simplex search method took 112 iterations and 0.062 s using a convergence criterion of 109 on standard deviation to get the nal result of MZ. The proposed method took 0.032 s for MZ evaluation of the same data set, whereas MC and MI evaluations took 0.015 s and 0.016 s, respectively. These algorithms were tested for various data in the literature [6,13] and considerable saving in time is observed for large data sets. MC, MI and MZ evaluations of 360-points data set by the proposed methods took 0.128, 0.144 and 0.286 s, respectively.

8. Conclusions The form data from roundness measuring instruments has to be treated differently from the data obtained from coordinate measuring machines. There is a radius-suppression in the form data and a prole-distortion is introduced due to eccentricity in setting the component on the table of roundness measuring instrument. Hence, a limacon has to be used as an assessment feature. In this work, computational geometric techniques have been developed to t MC, MI and MZ limacons. These techniques are simple and elegant, because the geometrical aspects can be visualized much better than the mathematical aspects of the numerical techniques. Since the details are explained with a simple data set, researchers can easily implement and test their algorithms.

ARTICLE IN PRESS
1236 N. Venkaiah, M.S. Shunmugam / International Journal of Machine Tools & Manufacture 47 (2007) 12291236

For tting the limacons, EA diagrams are established on the basis of a new concept of control hulls which are simpler to construct and the number of points to be dealt with are reduced considerably. Also, control hulls can be directly used to construct the EA diagrams and, as a result, computational effort is reduced. Following this, a new updating procedure involving edges of control hull has been introduced in this paper. The methods proposed for the circularity error using MC, MI and MZ approaches have been tested successfully on the data available in the literature. The present methods for MC, MI and MZ always guarantee the accurate results for a given set of data points and do not require any initial solution and convergence criteria. These algorithms are computationally quite robust giving unique solutions and require a shorter time for execution. In all cases tested, the results are obtained within a second. The MZ evaluation done using proposed computational geometric techniques strictly follows ISO standards. The proposed computational geometric techniques for circularity evaluation fulll the need for fast and efcient algorithms for processing data obtained from roundness measuring instruments and practitioners can easily implement them in computer-aided roundness measuring instruments for ready industrial applications. The present methods also have an advantage that they can be extended to cylindricity evaluation. Appendix A. Least squares method for tting a limacon The deviation of a point (ri, yi) on the measured prole from the limacon represented as ro xo cos yi yo sin yi is given by (1) ei ri ro xo cos yi yo sin yi . P 2 On minimizing s ei w.r.t. ro ; xo and yo and simplifying the normal equations considering equi-spaced and symmetric data points, the following least squares solution is obtained: 1X ro ri , N xo 2X ri cos yi , N

yo

2X ri sin yi , N

(2)

where N represents the total number of measured points.

References
[1] D.J. Whitehouse, A best t reference line for use in partial arcs, Journal of Physics E: Scientic Instruments 6 (1973) 921924. [2] D.G. Chetwynd, Roundness measurement using limacons, Precision Engineering 1 (3) (1979) 137141. [3] D.G. Chetwynd, P.H. Phillipson, An investigation of reference criteria used in roundness measurement, Physics E: Scientic Instruments 13 (1980) 530538. [4] International Organization for Standardization, Geneva, ISO 11011983, Technical drawings: Tolerancing of form, orientation, location, and runoutGeneralities, denitions, symbols, indications on drawing. [5] U. Roy, X. Zhang, Development and application of Voronoi diagrams in the assessment of roundness error in an industrial environment, Computers & Industrial Engineering 26 (1) (1994) 1126. [6] M.S. Shunmugam, On assessment of geometric errors, International Journal of Production Research 24 (2) (1986) 413425. [7] M.S. Shunmugam, Criteria for computer-aided form evaluation, Transactions on ASME, Journal of Engineering for Industry 113 (1991) 233240. [8] D.G. Chetwynd, Application of linear programming to engineering metrology, Proceedings of the Institution of Mechanical Engineers 199 (B2) (1985) 93100. [9] T.S.R. Murthy, S.Z. Abdin, Minimum zone evaluation of surfaces, International Journal of Machine Tools Design and Research 2 (1980) 123136. [10] P.B. Danish, M.S. Shunmugam, An algorithm for form error evaluation using the theory of discrete and linear Chebyshev approximations, Computer Methods in Applied Mechanics and Engineering 92 (1991) 309324. [11] J. ORourke, Computational Geometry in C, Cambridge University Press, Cambridge, 1997. [12] F.P. Preparata, M.I. Shamos, Computational Geometry: An Introduction, Springer, New York, 1985. [13] G.L. Samuel, M.S. Shunmugam, Evaluation of circularity from coordinate and form data using computational geometric techniques, Precision Engineering 24 (2000) 251263. [14] N. Venkaiah, M.S. Shunmugam, Evaluation of form data using computational geometric techniques, Part I: Cylindricity error, International Journal of Machine Tools & Manufacture, in press, doi:10.1016/j.ijmachtools.2006.08.011.

Vous aimerez peut-être aussi