Professional Documents
Culture Documents
Abstract
The measurement data for evaluation of sphericity error can be obtained from inspection devices such as form measuring
instruments/set-ups. Due to misalignment and size-suppression inherent in these measurements, sphericity data obtained will be
distorted. Hence, the sphericity error is evaluated with reference to an assessment feature, referred to as a limacoid. Appropriate
methods based on the computational geometry have been developed to establish Minimum Circumscribed, Maximum Inscribed and
Minimum Zone Limacoids. The present methods start with the construction of 3-D hulls. A 3-D convex outer hull is established
using computational geometric concepts presently available. A heuristic method is followed in this paper to establish a 3-D inner
hull. Based on a new concept of 3-D equi-angular line, 3-D farthest or nearest equi-angular diagrams are constructed for establishing
the assessment limacoids. Algorithms proposed in the present work are implemented and validated with the simulated data and the
data available in the literature. 2002 Elsevier Science Ltd. All rights reserved.
Keywords: Sphericity error; Computational geometry; Equi-angular diagram; Minimum circumscribed limacoid; Maximum inscribed limacoid;
Minimum zone limacoids
0890-6955/02/$ - see front matter 2002 Elsevier Science Ltd. All rights reserved.
PII: S 0 8 9 0 - 6 9 5 5 ( 0 1 ) 0 0 1 2 4 - 9
406 G.L. Samuel, M.S. Shunmugam / International Journal of Machine Tools & Manufacture 42 (2002) 405–416
Nomenclature
ri, qi, bi Polar coordinate of a point
x0, y0, z0 Estimated coordinates of center of the assessment feature
r0 Radius of assessment sphere
ra Radius of the assumed sphere
ei Deviation of ith point from assessment feature
emax, emin Maximum and minimum deviations of the profile
⌬ Sphericity error
r⬘i Radius of the points on inverted profile
rmax, rmin Maximum and minimum radii
d Increment
k Constant, from 0.5 to 1.0
error from the form data obtained using roundness meas- in manufactured components. Computational geometry
uring instruments. Since the roundness profile is dis- based techniques show greater promise for solving the
torted by the misalignment between the axes of a compo- minimum zone problems encountered in the geometrical
nent and a rotating table and by radius-suppression evaluations. However, no attempt has been made by
inherent in the measurement, a limacon is used as an researchers in the past to adopt computational geometry
assessment feature [15]. Using a limacon as the assess- based assessment techniques for different methods of
ment feature, the Least Squares Method (LSM) and sev- evaluation of the sphericity error. In this paper, tech-
eral numerical methods have been employed to arrive at niques for evaluation of the sphericity error using form
the circularity error. Apart from Minimum Zone evalu- data have been presented. To take care of the distortion
ation, the concepts of Minimum Circumscribed and introduced during the sphericity measurement, a 3-D
Maximum Inscribed limacons are also used in circularity assessment feature referred to as a limacoid is estab-
evaluation. Considering the limacon as a curve rep- lished by extending the concept of a limacon. Apart from
resenting the locus of foot of the perpendiculars drawn using the existing computational geometric concept of a
from a pole to the tangents of a circle, the computational 3-D convex outer hull, a heuristic method for obtaining
geometric techniques have been successfully developed a 3-D convex inner hull has been proposed. A new con-
to arrive at the circularity error [17]. cept of 3-D equi-angular line is introduced and 3-D far-
In the case of sphericity, a few attempts have been thest and nearest equi-angular diagrams (EA diagrams)
made by previous researchers to develop methods for are developed to arrive at different assessment limacoids.
evaluating the sphericity error from the form data. The The results obtained using the simulated data and data
Least Squares Method that minimizes the sum of the reported in the literature are included in this paper.
squared deviations of the measured points from a fitted
feature has been suggested [15]. Though the Least
Squares Method is based on sound mathematical prin- 2. Evaluation of sphericity error
ciples, the error values obtained are not the minimum
[10,11]. A statistical technique for estimating the spher- Fig. 1 shows a schematic arrangement for the
icity from a few 2-D roundness values was also proposed measurement of spherical form. As only the deviation
[7]. Shunmugam has suggested a new simple approach from the true form is measured, there will be size sup-
called the Median technique that gives minimum value pression (radius suppression) in the form data. The data
of sphericity error [15]. Using discrete Chebyshev obtained from the measuring instruments/set-ups will be
approximations, Danish and Shunmugam have arrived at in the polar form, with the radial values in microns and
the minimum zone values [1]. Apart from the Minimum the angles in degrees. Also, the misalignment between
Zone principle, the concepts of Minimum Circumscribed component center and the axis of the measuring
and Maximum Inscribed features used in circularity instrument/setup cannot be totally eliminated. Due to the
evaluation can also be extended to the evaluation of the misalignment and size suppression, the form data
sphericity error. obtained will be distorted. Even when a truly spherical
In the past decade, the computational geometric tech- surface is measured, if the axis of the rotating table of
niques have gained enormous attention from the design- the measuring instrument does not coincide with the
ers of algorithms for solving geometric problems. These component center, distortion will be introduced in the
computational geometry based algorithms can be very measurement data. The surface obtained by such
well applied for the evaluation of the sphericity error measurement is approximated to a new surface referred
G.L. Samuel, M.S. Shunmugam / International Journal of Machine Tools & Manufacture 42 (2002) 405–416 407
Fig. 2. Spherical form data (ri in µm, qi and bi in degrees) for illustration only.
408 G.L. Samuel, M.S. Shunmugam / International Journal of Machine Tools & Manufacture 42 (2002) 405–416
is the boundary of the smallest convex domain contain- the common vertex V1 is deleted. The updated faces are
ing all the points of the set. A domain is said to be con- obtained either by joining V2 and V4 or V3 and V5. If
vex if, for any two points in the domain, the segment there is at least a pair of adjacent faces with a common
connecting them should be entirely contained in the farthest center, the edge common to adjacent faces is
domain [14]. In other words, the convex hull of points eliminated. For example, Fig. 4(c (ii)) shows that two
in three-dimensional space is the shape taken by plastic faces V1, V3, V4, and V1, V4, V5 have a common center
wrap stretched tightly around the points [12]. The algor- C1345 and faces V1, V2, V3 and V1, V2, V5 have another
ithms used in the present work for constructing the con- common center C1235. In this case also, the vertex V1
vex hull is based on the Divide-and-Conquer and Merge along with the edges V1, V4 and V1, V2 gets eliminated
techniques suggested by Preparata and Hong [13]. resulting in the formation of a new edge V3, V5 as shown
Fig. 3(a) shows the outer convex hull with vertices in Fig. 4(c (iii)). For other cases with more than four
V1, V2 etc. for the data points shown in Fig. 2. In the faces also, the same logic is applied and the updated
next step, the initial EA lines are established. The details faces are obtained. The updating procedure is equivalent
of the EA planes and lines are given in Appendix B. to the formation of faces in arriving at the convex hull.
Taking the vertices V2, V3 and V4, lines connecting these The first stage of updating is carried out by consider-
vertices to the origin O are obtained. Three planes PL2, ing the vertices and faces that have common centers.
PL3 and PL4 perpendicular to these lines V2O, V3O and However, a vertex belonging to the updated face should
V4O are established as shown by thin lines in Fig. 3(a). not be deleted in the process of updating the vertices and
For each pair of planes, a plane bisecting the angle the faces in the neighborhood. After the first stage of
between them (referred to as the EA plane) is estab- updating covering all potential vertices, the next stage
lished. There will be three such EA planes and the inter- of updating starts. The process of updating continues till
section of these three EA planes results in an EA line. the number of vertices is reduced to three.
For the three planes PL2, PL3 and PL4, the EA line The hull shown in Fig. 3(b) with the initial farthest
obtained is referred to as EA234. Taking any point on EA edges is updated following the procedure explained
this EA line, a sphere can be drawn tangential to the above. The vertices V1 and V3 get eliminated as the three
three planes PL2, PL3 and PL4. In the next step, the far- faces surrounding each of these vertices have a common
thest center of the EA line is determined. For example, center. The remaining vertices are V2, V4 and V5. The
EA234 will be intersected by three adjacent EA planes EA line for this face results in FE245, as shown in Fig.
(corresponding to PL2 and its adjacent plane, PL3 and 5. The centers obtained at all stages of updating are the
its adjacent plane, etc.). The farthest intersection point candidate centers for the spheres from which the circum-
among them gives the farthest center and the sphere scribed limacoid can be generated. A complete EA dia-
drawn with it as a center will also be tangential to the gram with Minimum Circumscribed Limacoid is shown
adjacent EA plane resulting in this intersection point. in Fig. 5. This Minimum Circumscribed Limacoid passes
The portion of the EA line beyond the farthest center is through four crest points and encloses all other points
known as the farthest edge, FE234. Following the above of the data set, in such a way that the sphere from which
procedure, all EA lines and farthest edges are con- it is generated has the least radius.
structed. The initial farthest EA edges are shown in
Fig. 3(b).
3.2. Maximum Inscribed Limacoid (Valley Limacoid)
After constructing the initial EA diagram, the hull is
updated by eliminating the vertices with adjacent faces
having common farthest EA centers. Different cases that The points controlling the inscribed limacoid lie on
may be encountered and updating of the hull in these the inner hull and, therefore, the inner hull of the given
cases are shown in Fig. 4. Fig. 4(a) shows three points set of data points is determined before constructing the
of a face V1, V2, and V3. Radial lines are established nearest EA diagram. The inner hull of a set of points is
through these points and the planes PL1, PL2 and PL3 the boundary of the largest empty subset of a given set
are established tangential to the radial lines through V1, of points.
V2 and V3, respectively. The EA line for these planes is A new heuristic algorithm for finding the inner hull
established following the procedure explained in Appen- is proposed in the present work. This method involves
dix B. Fig. 4(b (i)) shows three faces formed by four transformation of the original data with reference to the
vertices V1, V2, V3 and V4, and all have a common far- measurement center O of the sphere as shown in Fig.
thest center. Any two faces can also have a common 6(a). With this center, a suitable diameter of the sphere
farthest center as shown in Fig. 4(b (ii)). In both cases, is selected such that all the measured points are within
the vertex V1 is deleted and the hull is updated by for- this assumed sphere. The transformation is done by
ming a new face V2, V3, V4, as shown in Fig. 4(b (iii)). arriving at radial deviations between the measured points
In the case of a vertex with four faces, if all the faces and the corresponding points on the assumed sphere. For
have a common farthest center as shown in Fig. 4(c (i)), example, the point P⬘1 is obtained by taking along the
G.L. Samuel, M.S. Shunmugam / International Journal of Machine Tools & Manufacture 42 (2002) 405–416 409
corresponding radial line the distance a from the center ra⫽rmax⫹d (4)
O, that represents the deviation P1 S1.
Assuming a very large radius ra can lead to a consider- where d=k(rmax⫺rmin) and k can be selected between 0.5
able distortion of the data. A simple heuristic rule to and 1.0. rmax and rmin are the maximum and minimum
select the radius of this sphere is radii with respect to the selected center.
410 G.L. Samuel, M.S. Shunmugam / International Journal of Machine Tools & Manufacture 42 (2002) 405–416
The 3-D convex outer hull for the transformed data is by finding the EA center nearest to the face. The hull is
determined as explained earlier. The points of the orig- updated in a similar way as explained earlier. The verti-
inal data set corresponding to the vertices of the outer ces V3 and V4 are eliminated as they have common near-
hull of the transformed data are taken as the vertices of est EA centers. The nearest EA line NE125 is established
the inner hull as shown in Fig. 6(b) schematically. The for the remaining three vertices of the hull V1, V2 and
procedure for constructing the nearest EA diagram is V5. The complete nearest EA diagram and the Maximum
similar to that of the farthest EA diagram, except that Inscribed Limacoid are shown in Fig. 6(c). The
the nearest common centers are considered as end points Maximum Inscribed Limacoid passes through four
of the nearest EA lines. Fig. 6(b) shows the initial near- points and all other points lie outside. The sphere from
est EA lines for the faces of the inner hull. For example, which it is generated has a maximum radius.
the nearest EA line NE123 for face V1, V2, V3 is obtained
G.L. Samuel, M.S. Shunmugam / International Journal of Machine Tools & Manufacture 42 (2002) 405–416 411
Table 2
Form data set 2a
Sl ri µm qi deg bi deg
No.
Table 3
Sphericity error using form data
Set 1 x0 (µm) 1.29c 1.59c 1.41b, 1.29c 1.292893 1.5857864 1.41421 0.681218
y0 (µm) 0.296c 0.17c 0.24a, 0.296c 0.292893 0.171573 0.24264 0.806218
z0 (µm) 0.71c 0.728c 0.58b, 0.71c 0.707107 0.707107 0.58578 0.412132
r0 (µm) 4.29c 1.62c 3.00b, 2.85c 4.292893 1.62132 3.00000 2.884615
⌬ (µm) 2.88c 2.97c 2.82b, 2.88c 2.878680 2.964466 2.82843 3.397595
Control 1,14,20,23c 3,6,19,24c 1,6,14,19,24b, 1,14,20,23 3,6,19,24 1,6,14,19,24 –
points 1,6,14,20,23c
Set 2 x0 (µm) – – – 1.86667 0.63818 1.11808 ⫺0.53055
y0 (µm) – – – 0.50951 0.22282 0.41494 0.13803
z0 (µm) – – – ⫺0.35705 ⫺0.60176 ⫺0.17266 0.12931
r0 (µm) – – – 9.25327 1.78505 5.37178 4.76377
⌬ (µm) – – – 8.391589 7.85256 7.66012 8.48576
Control – – – 2,7,16,45 20,25,39,48 7,16,20,39,45 –
points
a
Assessment feature having equal magnitude of maximum deviations on either side.
b
Danish and Shunmugam ([1]).
c
Shunmugam ([15]).
the surface measured. The present method for minimum It can be inferred that the line OP1 and OoT1 are perpen-
zone evaluation always guarantees a minimum value for dicular to the tangent plane established at point T1 of the
a given set of data points. sphere. OoT1 is also a radial line for the sphere. For the
These algorithms are computationally quite robust second point T2 on the same sphere, there will be another
giving unique solutions and require a short time for set of lines OP2 and OoT2, perpendicular to the second
execution. The proposed algorithms can handle data with tangent plane. By geometry, it can be shown that the
both uniform and non-uniform spacing. The present plane bisecting the angle between the first and second
algorithms cannot only be applied to measurement data tangent planes will pass through the center of the sphere.
corresponding to a full sphere but also to the partial Alternatively, if a plane is established at P1 perpendicu-
sphere. lar to the line joining P1 and the pole O and another
plane is established in a similar way at a different point
P2 on the limacoid, then the angle bisector of these two
Appendix A. Limacoid and its properties planes will also pass through the center of the sphere
from which the limacoid is established.
A limacoid has a pole O and an axis OOo. In Fig. 8(i),
a base sphere is drawn with OOo as its diameter. A point Appendix B. Equi-angular planes and lines
Pi on the limacoid is at a distance of ri from the pole
where the distance between the point Pi and P⬘i located Fig. 9 shows two planes PL1 and PL2 intersecting. The
on the base of a sphere is equal to a constant value ro. equi-angular plane (EA plane) is drawn such that it
Similarly, other points on the limacoid are obtained by bisects the angle between the two planes PL1 and PL2.
laying out the constant value ro from the corresponding For any point X on the EA plane, the perpendicular dis-
points on the sphere. The surface of the limacoid passes tance d to the planes PL1 and PL2 will be equal.
through all these points. Three planes PL1, PL2 and PL3 intersecting each other
Alternatively, a sphere of radius ro is drawn with Oo are shown in Fig. 10. The equi-angular plane EA12
as its center, as shown in Fig. 8(ii). A plane is estab- bisects the angle between the planes PL1 and PL2. Simi-
lished passing through a point T on the sphere and tan- larly the EA planes EA23 and EA13 are established.
gential to it. From the pole O, a perpendicular line to These three EA planes EA12, EA23 and EA13 intersect
this tangent plane is established. The foot of this perpen- along a line. This line will be referred to as the EA line
dicular gives Pi that lies on the limacoid. The distance hereafter. Any point X on this EA line will be at an
of this point from the pole is given by ri. The limacoid equal distance from the three planes and this line will
is obtained as a surface that passes through the foot of pass through the point of intersection of the three planes
the perpendicular to the tangent planes of the sphere. PL1, PL2 and PL3. A sphere can be drawn with X as a
Fig. 8(iii) shows two points P1 and P2 on the limacoid center such that it is tangential to the planes PL1, PL2
and the corresponding points T1 and T2 on the sphere. and PL3.
G.L. Samuel, M.S. Shunmugam / International Journal of Machine Tools & Manufacture 42 (2002) 405–416 415
Fig. 11 shows the three planes PL1, PL2 and PL3 along
with the adjacent planes PL4, PL5 and PL6. The line
EA123 corresponds to the three planes PL1, PL2 and PL3
as explained earlier. This line EA123 is intersected by
three other EA planes, namely EA14, EA25 and EA36
established taking the adjacent planes PL4, PL5 and PL6.
Three intersection points are obtained. The intersection
point Cf farthest from the planes PL1, PL2 and PL3 is
the farthest center and intersection point Cn nearest to
the planes is the nearest center. For example, if the far-
thest center Cf is obtained by intersection of the EA line
EA123 with the plane EA14, a largest possible sphere can
be drawn tangential to the four planes PL1, PL2, PL3 and
PL4 with Cf as a center. The line beyond Cf is known
as farthest EA line FE123. Any point taken along the line
EA123 away from Cf can yield a sphere tangential to the
Fig. 8. Geometry of a limacoid. three planes PL1, PL2 and PL3. The portion of the line
from Cn towards the planes PL1, PL2, PL3 is known as
nearest EA line NE123 and point Cn would result in the
smallest sphere.
416 G.L. Samuel, M.S. Shunmugam / International Journal of Machine Tools & Manufacture 42 (2002) 405–416