International Journal of Machine Tools & Manufacture 42 (2002) 473–478

Interpolation algorithm based on central angle division

*
Wei Hu
CAD Center, School of Mechanical Science and Engineering, Huazhong University of Science & Technology, Wuhan, 430074 Hubei Province,
PR China

Received 18 December 2000; received in revised form 17 September 2001; accepted 19 September 2001

Abstract

Second-order interpolation is essential in the manufacturing of curve contours. Especially, according to different part shape,
corresponding interpolation algorithms are required. However, the problems of how to determine the parameters of conic sections,
how to reduce computational time and how to minimize the contour error according to desired accuracy have not been completely
solved. On the basis of the principle of central angle division of an arc, this paper develops an elliptic interpolation algorithm for
elliptic gears. And the idea is extended to general second-order interpolation. Although their representing equations are different,
the proposed recursive algorithm flows for different conic sections are similar, which is very easy for computer to satisfy the
interpolation requirements.  2002 Elsevier Science Ltd. All rights reserved.

Keywords: Numerical control; Central angle division; Elliptic interpolation

1. Introduction tation of a curve to linear segments needs to generate

huge number of straight line segments.
The interpolators used by most CNC machines can Some new interpolators can eliminate the above prob-
process only straight line and circular arc motion com- lems. For example, Abramovich and Crawford [3], and
mands. With the increasing needs to machine fewer but Qiu et al. [4] gave interpolators that were capable of
more complex parts, industries pay high attention to the generating conic sections. Shpitalni et al. [5] proposed
complex curve contours. Conic sections usually occur in an interpolator for 2D curves in non-parametric form and
manufacturing, such as machining pistons, elliptic gears, an interpolator for 3D curves in parametric form. On
saddles, moulds of focus lamps, etc. It is necessary to the basis of his method, Lo [2] developed a feedback
develop efficient interpolation algorithms for conic sec- interpolator for CNC machine tools. Some scholars
tions. To machine these kinds of curves, current com- devote to research free-form curve interpolators [6,7].
puter-aided manufacturing (CAM) systems approximate Farouki et al. [7] proved that the PH (Pythagorean-
the curves by many linear segments and then download hodograph) curve algorithms were more precise and
these linear segments to the CNC machines such as digi- flexible than other methods for free-form curves. Since
tal differential analyzer (DDA) [1]. This approach may this algorithm approximates the part contour curves by
cause the following problems [2]: (1) the linearization quintic PH curves, it is less economical and efficient for
and segmentation cause discontinuities in the first second-order curves than circular interpolator (e.g. using
derivatives along the cutter path so that it deteriorates a fifth order curve to approximates a second-order curve
the smoothness of the curves and surfaces; (2) this kind is not economical in the computational times and time).
of method increases the computational time and the dif- In fact, many real problems of second-order curves are
ficulty of programming, especially when good quality needed to be solved in manufacturing, such as machining
surface smoothness is required, because the segmen- elliptic gears, saddles, moulds of focus lamps, etc.
In this paper, an elliptic interpolation method based
on central angle division is presented and compared with
a typical circular interpolator. The principle of the typi-
* Tel.: +86-27-8754-3973 fax: +86-27-8754-7405. cal circular interpolation is discussed first; then, the
E-mail address: [email protected] (W. Hu). cases of the proposed elliptic interpolation are analyzed.

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 3 7 - 7
474 W. Hu / International Journal of Machine Tools & Manufacture 42 (2002) 473–478

And finally, the proposed elliptic interpolation method respectively. Based on central angle, we can describe a
is extended to hyperbolic and parabolic interpolation. circle by the following equation, i.e.

2. The principle of circular interpolation based on

再 x=Rcosq
y=Rsinq
(2)
central angle division

2.1. Principle of central angle division If the feedrate of tool at point i is fi, the approximate
value of central angle can be described as following equ-
It is important to define the algorithms of interpolation ation,
for tool positioning. Interpolation is the process of find- pipi+1 |pipi+1| fi
ing coordinate values of reference points along tool path dqi⫽ ⬇ ⫽ (3)
R R R
between the starting and ending points specified in the
program [8]. Then, the tool proceeds along the path with By using parametric equation of the circle, the point
the given feedrate. The method of time division is to i+1 can be represented as following form:
find a series of interpolation points along the given path
(Xi+1,Yi+1)⫽(R⫻cosqi+1, R⫻sinqi+1)⫽(Xicosdq (4)
and demands that the distance between two adjacent
points is equal to the feedrate. Time division of circular ⫺Yisindq, Yicosdq⫹Xisindq)
arc based on central angle division realizes circular
interpolation by using linear interpolation in terms of the
relation of circular arc and central angle [9]. In this If the radius of the circular arc is not too small, the
paper, the author extends this method to elliptic interp- feedrate (fi) in each sampling period is very small in
olation, hyperbolic interpolation and parabolic interp- contrast to the radius (R). So dqi is also very small.
olation. In order to find the approximately parametric According to differential transformation
equation between the curve segment and the central
angle, the program needs to determine the feature para-
meters (such as the coordinates of the center of the arc,
再 sindq⬇dq
cosdq=(sindq)⬘⬇(dq)⬘=1
(5)

the starting and ending position of the arc, together with

the radius of curvature of the reference points), which Substituting Eq. (5) into Eq. (4) yields

再
define the curves. As a result, the second-order interp-
olation algorithms are developed. Xi+1=Xi−Yi×dqi
(6)
Yi+1=Xi×dqi+Yi
2.2. Typical circular interpolation method

The relation between the length of circular arc and the Then the amount of feed at point i+1 can be achi-
central angle can be described as follows (see Fig. 1). eved, i.e.
|p1p2|⫽dq⫻R
where dq is the value of central angle between op1 and
(1)
再 dXi+1=Xi+1−Xi=−dqiYi
dYi+1=Yi+1−Yi=dqiXi
(7)

op2. R is the radius. Then, the second-order interpolation

may be simplified into linear interpolation for central
angle (dq). Supposing the center of the circle locates at
origin of coordinate. P1(x0,y0) and P2(xe,ye) represent the 2.3. Proposed elliptic interpolation method
reference positions at starting point and ending point,
A standard ellipsis can be described by the follow-
ing equations,

再 X=acosq
Y=bsinq
(8)

where 2a, 2b represent the lengths of major axis and

minor axis respectively. q is a parameter. Differential
equation of length (dL) can be expressed as follows:

dL⫽冑a2sin2q+b2cos2qdq (9)

Fig. 1. Circular arc and central angle.

 W. Hu / International Journal of Machine Tools & Manufacture 42 (2002) 473–478 475

Because the length of the curve segment is very small,

we can get an approximate equation, i.e.
dL⬇冑a2sin2q+b2cos2qdq (10)

Supposing feedrate (f) is equal to dL, and substituting

the parameters of point i into Eq. (10) yields:

冪冉 冊冉 冊 2
a 2 b
dqi⬇fi/ Yi) + Xi (11)
b a
Fig. 2. Machining sketch drawing of the example elliptic gear.

According to Eq. (8), we can obtain Eq. (12)

(Xi+1,Yi+1)⫽(acosqi+1,bsinqi+1)⫽(acos(qi (12)
⫹dqi),bsin(qi⫹dqi))

Using the same method as the circular interpolation,

we can obtain Eq. (13) that describes the point i+1,

冦
a
Xi+1=Xi− Yidqi
b
(13)
b
Yi+1=Yi+ Xidqi
a

Fig. 3. The contour error caused by the interpolation methods for the
Then, the amount of feed at point i+1 can be example elliptic curve.
obtained, i.e.
interpolation method [2]. To evaluate the circular and

冦
a elliptic interpolation methods, an example of elliptic gear
dXi+1=Xi+1−Xi=− Yidqi
b is done on a PC-Pentium-166MHz and the sampling per-
(14) iod is T=0.01 s. The evaluation equation is demonstrated
b
dYi+1=Yi+1−Yi= Xidqi in the following.
a
The contour curve of example gear that is described
by ellipse is shown as follows:
These are the recursive equations of the elliptic interp- F(x,y)⫽x2⫹4y2⫺80y⫽0
olation.
The starting point of the path is [0,0] and the ending
point is [0,20]. The feedrate is 50 mm/s. Corresponding
to the two aspects needed to evaluate, the machining
3. Evaluation sketch drawing and the contour errors caused by the cir-
cular and elliptic interpolation methods are shown in
The above elliptic interpolation algorithm is claimed Figs. 2 and 3, respectively. And the performance of the
to be more suitable for the design and manufacture of interpolation methods are summarized in Table 1 and
elliptic gears than the exiting circular interpolation Fig. 4.
method. According to the computing and comparing
methods of Lo [2], the author draws some important con-
Table 1
clusions. In this section, the elliptic and circular interp- Performance of the interpolation methods
olation methods are evaluated and the conclusions (or
the remarks) are given in detail. The evaluation includes Interpolation method Circular Elliptic
two aspects: (1) the contour error (or the interpolation
accuracy) produced by the above two methods, and (2) Maximum contour error(µm) 17 1.3
Maximum feedrate error(%) 0.3 0.3
the amount of the computational time needed to generate Computation time in a sample period 23 19
the reference points at each sampling period. The two (µm/T)
aspects are regarded as important indices to assess an
476 W. Hu / International Journal of Machine Tools & Manufacture 42 (2002) 473–478

Ax2⫹Bxy⫹Cy2⫹Dx⫹Ey⫹F⫽0 (15)

Through translation and traverse, Eq. (15) can be

transformed into standard circle, ellipse, hyperbola
or parabola as shown in Eqs. (2, 8, 16) and (24). By
using the above interpolation algorithms, the general
conic section can be solved in manufacturing to meet
different requirements of parts with various second-
order curves.
(2) As can be seen in Eqs. (7, 14, 23) and (29), all the
interpolation points are located on the curve. So
there is no error at each point of interpolation. But
Fig. 4. Performance of the computation methods.
there is an approximating error caused by chord
replacing arc. The error is related to the radius of
As can be seen in Figs. 2–4 and Table 1, the elliptic curvature and central angle. It increases as the radius
interpolation method matches the different part of the of curvature decreases or the feedrate increases. By
gear using different segments with corresponding radius using low feedrate, we can improve the accuracy of
of curvature. While the circular interpolation method the curve with small radius of curvature (Fig. 5).
approximates the elliptic gear by changing the curve (3) In this paper, in order for the algorithms to proceed,
equation. As can be seen in Table 1 and Fig. 4, because the variable dqi must be calculated first (Eqs. (3, 11,
the circular interpolation method has to calculate the 20) and (26)); then, the recursive equations run. The
radius of curvature repeatedly, it requires more multipli- recursive equations for different second-order interp-
cations (or divisions) than the elliptic interpolation olation are different, but algorithm flows are similar,
method. Consequently, the computational time required which can be shown in Fig. 5.
for the elliptic interpolation method (19 µm/T) is less
than that for the circular interpolation method (23 µm/T).
Remarks: As can be seen in Table 1 and Fig. 4, since
the feedrate errors caused by the two methods are very
small, both can be neglected (Table 1 shows df/f=0.3%).
The major and important problem is how to control the
contour error. The elliptic interpolation method provides
an accurate and efficient way to solve the problem.
Because the elliptic interpolation method results in a
much smaller contour error but requires less compu-
tational time than the circular interpolation method.
Moreover, Figs. 2–4 and Table 1 show that most of the
important indices of the elliptic interpolation method are
superior to the circular interpolation method. In fact, this
paper focuses on decreasing the contour error for the
elliptic gears. The proposed elliptic interpolation method
is the result of the development of the ideas.

4. Conclusions

The above interpolation method can be extend to

hyperbolic and parabolic interpolation to meet the
requirements of different second-order interpolation,
such as machining saddle (hyperbola), the mould of
focus lamp (parabola), etc. The equations of hyperbolic
and parabolic interpolation algorithms are given in
Appendix A [10].

(1) A general conic section can be expressed in an

implicit form: Fig. 5. Flow chart of the proposed interpolation.
 W. Hu / International Journal of Machine Tools & Manufacture 42 (2002) 473–478 477

Appendix A A.2. Parabolic interpolation algorithm

A.1. Hyperbolic Interpolation Algorithm A standard parabola can be described by the follow-
ing equations:

再
A standard hyperbola can be described by the follow-
x=2ptan2q
ing equation: q苸(⫺p/2,p/2) (24)

再
y=2ptanq
y=asecq
(16) and
y=tanq
q⫽arctan[y/(2p)] q苸(⫺p/2,p/2) (25)
where 2a and 2b are the lengths of real axis and imagin-
ary axis respectively. q is the centrifugal angle.
Because hyperbola has two branches: left branch and Using the same method as hyperbolic interpolation,
right branch and the two branches are symmetrical in we can obtain
terms of Y-axis, we only study the right branch.
Differential equation of length (dL) can be expressed dqi⫽fi/[(2⫹xi/p)(p2⫹y2i )1/2] (26)
as follows:
dL⫽冑a2sec2qtan2q+b2sec4qdq
According to Eq. (24), we can obtain Eq. (27):
(17)

When the length of the curve segment is very small,

再 xi+1=2ptan2(qi+dqi)
yi+1=2ptan(qi+dqi)
(27)

we can get an approximate equation, i.e.

dL⬇冑a2sec2qtan2q+b2sec4qdq
Using the Taylor equation, (xi+1,yi+1) can be solved
(18)
as follows:
xi+1=xi+2yi[1+xi/(2p)]dqi+2p{[1+xi/(2p)]1/2
Supposing feedrate (f) is equal to dL, we can obtain
the following equation: 冦 +(xi/p)[1+xi/(2p)]}(dqi)2
yi+1=yi+(2p+xi)dqi+yi[1+xi/(2p)](dqi) 2
(28)

dq⫽f/冑a2sec2qtan2q+b2sec4q (19)

Substituting the parameters of point i into Eq. (19)

yields The amount of feed at point i+1 can be gotten, i.e.
dxi+1=2yi[1+xi/(2p)]dqi+2p{[1+xi/(2p)]1/2

冦
dqi⫽fi/[(xiyi/b)2⫹(bx2i /a2)]1/2 (20)
+(xi/p)[1+xi/(2p)]}(dqi)2 (29)
According to Eq. (16), we can obtain Eq. (21) dyi+1=(2p+xi)dqi+yi[1+xi/(2p)](dqi) 2

再xi+1=asec(qi+dqi)
yi+1=btan(qi+dqi)
(21)
References

Using the Taylor equation, (xi+1,yi+1) can be solved [1] D. Hearn, M.P. Baker, Computer Graphics, Prentice-Hall Inter-
as follows: national, Inc., Beijing, 1997.

再
[2] C.C. Lo, Feedback interpolators for CNC machine tools, Trans.
xi+1=xi+(xiyi/b)dqi+(1/2)(xiy2i /b2+x3i /a2)(dqi)2 of ASME, Journal of Manufacturing Science And Engineering
(22) 119 (11) (1997) 587–592.
yi+1=yi+(b/a2)x2i dqi+(x2i /a2)yi(dqi)2 [3] A. Abramovich, T.R. Crawford, An interpolation algorithm for
control applications on microprocessors, Trans. of IEEE, Journal
of Industrial Applications of Microprocessors 3 (1978) 195–202.
The amount of feed at point i+1 can be obtained, [4] Q. Qiu, K. Cheng, Y. Li, Optimal circular arc interpolation for

再
NC tool path generation in curve contour manufacturing, Com-
dxi+1=xi+1−xi=(xiyi/b)dqi+(1/2)(xiy2i /b2+x3i /a2)(dqi)2 puter-Aided Design 29 (11) (1997) 751–759.
dyi+1=yi+1−yi=(b/a2)x2i dqi+(x2i /a2)yi(dqi)2 [5] M. Shpitalni, Y. Koren, C.C. Lo, Real-time curve interpolators,
Computer-Aided Design 26 (11) (1994) 832–838.
(23) [6] A. Shima, T. Sasaki, T. Ohtsuki, Y. Wakinotani, 64-bit RISC-
based Series 15 NURBS interpolation, FANUC Technical
Review 9 (1999) 23–28.
478 W. Hu / International Journal of Machine Tools & Manufacture 42 (2002) 473–478

[7] R.T. Farouki, Y.F. Tsai, G.F. Yuan, Contour machining of free- Huazhong University of Science and Technology Publishing
from surfaces with real-time PH curve CNC interpolators, Com- House, Wuhan, 1992.
puter Aided Geometric Design 16 (1999) 61–76. [10] B. Liu, Y.J. Xiao, Z.L. Zhang, S.H. Huang, Hyperbolic/parabolic
[8] Translated from Russian by P.S. Ivanov, Numerical control of interpolation algorithms based on central angle division, Journal
machine tools and robots, Mir Publishers, Moscow, 1994. of Huazhong University of Science and Technology 25 (2) (1997)
[9] X.G. Liao, Q.Q. Zhu, Numerical control of machine tools, 82–84.