Professional Documents
Culture Documents
A Local and Analytical Curvature-Smooth Method With
A Local and Analytical Curvature-Smooth Method With
KEYWORDS: Smooth curvature, Local transition methods, Feedrate scheduling, Linear toolpath, Machining efficiency
Linear toolpath, which is always in forms of a large amount G01 codes, is most popular in practical Computer Numerical Control
(CNC) systems. The linear toolpath discontinuity will bring the unavoidable fluctuations on velocity and acceleration, leading to
unsatisfied machining efficiency and quality. To improve the G01 machining, corners are normally smoothened by various smooth
curves and in this paper, a novel two-step method including path-smoothing and feedrate scheduling is proposed. In path-smoothing,
a B-spline curve is adopted to blend sharp corners locally and the smooth curvature along transition toolpath is obtained. With
constrained chord error, the analytical curvature extreme is proved to occur at the B-spline midpoint and optimal curvature extreme
is searched to enhance machining efficiency. In feedrate scheduling, based on the extreme points, a look-ahead methodology is utilized
to determine the reachable feedrate at these critical points and a jerk-continuous scheduling scheme is developed afterwards. To verify
the proposed method, tracking and machining experiments are conducted to demonstrate the efficiency, contour performance as well
as machining quality compared to point-to-point and curvature-continuous methods.
Manuscript received: August 22, 2017 / Revised: July 21, 2018 / Accepted: August 3, 2018
3
1. Introduction toolpath and obtained better geometric continuity locally. Zhao et al.
presented a B-spline transition scheme to generate curvature-continuous
Computer-aided manufacturing (CAM) systems always transform toolpath and the transition curve was analyzed as two symmetrical
4
complicated sculptured surfaces into a series of linear G01 segments in Bezier curves without optimization. On this basis, Du et al. optimized
1,2
practice and send them to CNC systems directly. Those G01 codes the toolpath with curvature variation energy (CVE) to achieve higher
are always traveled by point-to-point method, leading to much lower efficiency. Moreover, to obtain smoother toolpath, two symmetric
5
machining efficiency. Otherwise, when higher corner feedrate is chosen, Bezier curves were applied by Fan et al. to smoothen the G01 toolpath
kinematical constraints including acceleration and jerk would be violated, with iteration method, and this method became complex due to the two
resulting in poor surface finish. In order to improve the efficiency as quartic Bezier curves. In order to blend corners with one whole spline
6
well as the quality of G01 toolpath machining, most researchers have curve, Sencer et al. gained curvature-continuous toolpath with a single
proposed transition methods to raise the toolpath smoothness. Since quintic Bezier curve. However, the curvature was not changed
these methods usually replace corners with various smooth trajectory monotonically; in this case, the feedrate at transition curve was constant,
locally, it is high-efficient and transition-error-constrained, making it leading to some efficiency loss. For five-axis machine tools, Sneha et
3-9 7
most popular in CNC areas. For the local methods, path-smoothing al. employed one B-spline curve to generate curvature-smooth toolpath.
and feedrate scheduling are two necessary procedures and in the However, the transition B-spline toolpath was not optimized in
following, the related researches about the two aspects are both provided. machining efficiency and a time-consuming optimal scheduling method
8
In the stage of path-smoothing, most researches have applied various only can be conducted offline. Zhang et al. utilized double cubic
curve, such as Circle, Bezier and B-spline ones, to smoothen the linear NURBS curve to smoothen five-axis linear toolpath and a curvature-
continuous toolpath was obtained. The features of the toolpath were not
elaborated and it was not optimized neither. In recent studies, Zhang et
9
al. proposed the novel one-step methods. He applied a quadratic time
function to obtain the constant acceleration at corners and a minimum
time transition algorithm was proposed. Nevertheless, discontinuous
acceleration at junction points would lead to low machining quality,
which is improper for high performance machining.
After path-smoothing, feedrate scheduling is essential in motion
planning with smooth acceleration/deceleration (ACC/DEC, AD)
10,11
feedrate profiles, and researchers have developed various approaches,
mainly focus on jerk-limited and jerk-continuous ones. Polynomial jerk-
11
limited profiles are implemented by Sun et al. and a look-ahead
method was presented. Jahanpour et al. proposed a new adaptive 12
Fig. 1 Description of curvature-smooth transition curve
NURBS interpolation method based on time-optimal S-shaped quintic
13
feedrate planning. Fan et al. designed a 15-phase AD profile with
continuous jerk. The superiority of jerk-continuous feedrate profile was 6
discussed over jerk-limited ones. Moreover, to obtain smoother AD r (u ) = ∑ Ai (u )Pi , u ∈ [0,1] (3)
14 i =0
profiles, Huang et al. utilized sine series to obtain optimal trigonometric
feedrate profile to improve the machining efficiency with heavier As shown in Eq. (A.1), B-spline curve functions can be divided into
computational load. Besides, a time-optimal bidirectional scanning two sections by u = 0.5 and considering the symmetry, we analyze half
15
algorithm was studied by Dong et al. and it is generally utilized in of it in this section. Since geometric continuity (Gn continuity) is
3-5
motion planning process. parameterized in terms of arc-length and independent on the toolpath
In this paper, a path-smoothing method is presented and a parameterization, it is more suitable to represent the inherent smoothness
corresponding jerk-continuous feedrate scheduling scheme is provided. of toolpath than parametric continuity (Cn continuity). Therefore, the
A smoother toolpath than most previous literatures is generated to obtain geometric continuity is analyzed below. Furthermore, the B-spline
3
higher machining efficiency as well as better machining quality. The transition methods are designed to generate a G continuous toolpath,
reminder of this paper is organized as follows. In section 2, a B-spline which also means curvature-smooth one.
curve transition method is proposed to generate a curvature-smooth
toolpath with analytical and optimal curvature extreme. Thereafter, 2.1 Curvature-smooth transition methods
section 3 provides a novel jerk-continuous feedrate scheduling scheme, In order to generate a curvature-smooth toolpath, a single B-spline
which is utilized in a look-ahead methodology with bidirectional scanning with seven control points is adopted between two adjacent linear
algorithms. In section 4, the tracking and machining experiments are segments. From Fig. 1, the control points P , P , P and P , P , P lie 0 1 2 4 5 6
conducted and compared to previous works. Finally, the conclusions on the two adjacent lines respectively, ensuring the G continuity at P 1
0
are given in section 5. and P . Furthermore, e , e are the unit orthogonal vectors of XY
6 x y
coordinate system, and transition angle θ is the half angle between two
neighbor liner segments.
2. B-Spline Curve Transition Methods Considering the symmetry of transition curve, half of the B-spline
is discussed to evaluate the whole toolpath continuity. For convenience,
In this paper, quintic B-spline is implemented to blend the sharp we denote d :d :d = 1:k :k and λi = d0 secθ 〈 e1, ei〉 , where i = x, y, z
0 1 2 1 2
16
corners. The transition curve is parametrically defined as and e = PtranPstr / ||PstrPtran|| is the unit direction vector of line PstrPtran.
1
i
Compared to the transition point P , the control points on line PstrPtran
trans
r (u ) =
j
∑
= −
i p
N j, p
(u )P , u ∈ [0,1]
j
(1) yields:
P3 = Ptrans ,
Where p = 5, Pj, j = 0, …, 6 are the seven control points and the
knot vector is [0 0 0 0 0 0 0.5 1 1 1 1 1 1]. The ith basic functions of
P2 = Ptrans +
i
∑ λe
=x , y ,z
i i
quintic B-spline curve are denoted as Nj,p(u), which is shown as follows: P1 = Ptrans + (1 + k ) ∑ λ ei 1 i
(4)
= x , y ,z
⎧⎪1, u ≤ u ≤ u +
i
N (u ) = ⎨ P0 = Ptrans + (1 + k + k ) ∑ λ ei
j j 1
⎪⎩0, else
j ,0 1 2 i
i =x , y ,z
(2)
u −u u −u
N (u ) = N − (u ) + + +
j
N j p 1 j
−1 (u )
For convenience, the first and second derivatives of curve are
u + −u u + + −u +
j, p j, p 1 j, p
j p j j p 1 j 1
denoted as:
Additionally, the transition curve can be simplified to the following
dr d 2r
equations, where Ai(u), i = 0, …, 6 are 5th order polynomial deduced ru = , ruu = 2 (5)
du du
from de Boor-Cox algorithm recursively. The detailed forms of Ai(u),
i = 0, …, 6 are all given in Appendix A. As analyzed above, the continuity at junction point P is G . If the 0
1
INTERNATIONAL JOURNAL OF PRECISION ENGINEERING AND MANUFACTURING Vol. 19, No. 10 OCTOBER 2018 / 1531
second derivative about parameter ruu becomes zero at junction point these extreme points. The feedrate extreme is given as:
P , G continuity would be ensured and the following equations need
2
2 2δ A J
0
(15)
8ε 8ε m ref[6] 21
d0 = < (9) κ max =
k1 + 5 k1 + 5 40ε tan 2 θ
If the transition point locates on the coordinate origin, the corner Therefore, the proposed curvature-smooth transition method has
error ε can be separated from the curve function r. The curve yields: potential to obtain higher machining efficiency than previous methods.
In the following, the transition toolpath is optimized to obtain the
r = ε rˆ (10) minimum time. From Eqs. (13) and (14), to improve machining
efficiency, the curvature extreme at critical points should be minimized
where r̂ is a function of transition parameters except corner error ε. by searching the only uncertain parameter ε along the whole transition
For the curvature, since corner error can also be separated from ru and toolpath.
ruu, the curvature can be rewritten as follows: As shown in Fig. 1, the transition length can be calculated as follows:
ru × ruu rˆu × rˆuu κ̂
κ= 3
= 3
= (11) Ltran = (1 + k1 + k2 )ε secθ = 1.45ε secθ (16)
ru ε r$ u ε
Therefore, corner error ε can simply scale the transition curve when Obviously, the sum of the two transition lengths should be no longer
5
taking the transition point as the coordinate origin. Since the geometric than the linear length and satisfy the following conditions:
features are unrelated with position of transition point, the conclusion +1
Litran + Litran < Li , i = 2...N − 1
also applies to the situation that the transition point is not located on
L1tran < L1 (17)
coordinate origin.
Theorem 1. When k = k /2, k ∈ [0.3, 1], the curvature extreme
LNtran−1 < LN −1
2 1 1
occurs at u = 0.5, and it can be calculated analytically by the following With the constraints before, the optimization is provided as:
equation.
N −1
min ∑ κ i
4(k1 + 1)(k1 + 5)
κ max = (12) i =1
5(k1 + 3) 2 ε tan 2 θ ⎧ε i ≤ ε m
⎪ i i +1 (18)
The proofs of Theorem 1 are given in Appendix B. ⎪ Ltran + Ltran < Li , i = 2...N − 1
s.t. ⎨ 1
⎪ Ltran < L1
2.2 Transition toolpath optimization ⎪ LN −1 < L
⎩ tran N −1
Since the curvature extreme happens at the center of the transition
curve, feedrate at these points should be constrained to satisfy the In the offline process, linear programming approaches are always
acceleration and jerk limitations. Besides, the acceleration and jerk applied to find the optimal solutions. However, those optimal approaches
3,18
constraints are supposed to be no larger than the tangential ones at would consume much time and be difficult to realize in real-time
1532 / OCTOBER 2018 INTERNATIONAL JOURNAL OF PRECISION ENGINEERING AND MANUFACTURING Vol. 19, No. 10
⎧ t ⎧ t2
⎪4 J T ⎪2 J
⎪ ⎪ T
⎪ (t − 12 T ) ⎪ 2t 2 T
j = ⎨−4 J , a = ⎨− J ( − 2t + )
⎪ T ⎪ T 4
⎪ t −T ⎪ 2t 2
⎪4 J ( T ) ⎪J ( − 4t + 2T )
⎩ ⎩ T
(19) Fig. 3 An ACC period with continuous jerk profile
⎧ 2 t3
⎪v0 + J
⎪ 3 T
⎪ 2t 3 2 tT T 2 J
v = ⎨v0 − J ( −t + − ) feedrate determined by jerk and length vT is achieved by the following
⎪ 3T 4 48
equation.
⎪ 2t 3 13T 2
⎪v0 + J ( − 2t 2 + 2tT − )
⎩ 3T 24 1 )2
(vTJ + v0 ) 2 (vTJ − v0 ) = Li J m (22)
2
where T is the AD period. From Eq. (19), the velocity at the end of AD
period yields: Eq. (22) can be solved by Newton-Raphson methods efficiently in
real-time CNC systems. Moreover, as observed in Fig. 3, the maximal
1
vT = v(T ) = v0 + J mT 2 (20) acceleration occurs at the midpoint of ACC period and the maximum
8 A
is am = JmT/4 < Am. Thus, the feedrate determined by acceleration vT
Then the AD period time is represented as T = 8 (vT – v0 )/Jm . also needs to qualify the following equation simultaneously.
Meanwhile, the length of one period is calculated by the period time T
2 Am 2
and it is expressed as follows: vTA = + v0 (23)
Jm
1
Li = v0T + J mT 3 (21) With the feedrate constraints in Eqs. (13), (22) and (23), a
16
bidirectional scanning algorithm6,17 has been implemented to determine
Besides, the arc-length between two neighbor critical points can the reachable corner feedrate. Thereafter, the proposed AD profile is
also be estimated by Simpson’s rules and the estimated length Li would applied to connect them. The main differences compared to the methods
be utilized in the following. Considering Eqs. (19) and (21), the corner in literatures is that the jerk-continuous AD profile is adopted in the
INTERNATIONAL JOURNAL OF PRECISION ENGINEERING AND MANUFACTURING Vol. 19, No. 10 OCTOBER 2018 / 1533
)
Type 1 L i ≥ Lr (1 )
Te = 8 ( vF – ve )/Jm
v m = vF
Calculated Ts by Eq. (27)
2 8 ( vs – v e )
Te = Ts + --------------------
-
Jm
)
Type 2 Lr2 < Li < Lr1
Tc = 0
1 2
vm = vs + --- Jm Ts
8
Ts = 8 ( ve – vs )/Jm
)
Li ≤ Lr2 Te = 0
Type 3
Fig. 4 Four types jerk-continuous AD profiles. (a) Type 1, (b) Type 2, vs ≤ ve Tc = 0
(c) Type 3, (d) Type 4 v m = ve
Te = 8 ( vs – ve )/Jm
)
proposed algorithms and the smooth toolpath is mixed by linear and B- Li ≤ Lr2 Ts = 0
Type 4
spline segments. Since the procedures have been addressed in Refs. 20, vs > ve Tc = 0
21 clearly, these procedures will not be elaborated here. v m = vs
After determining the reachable start and end velocity vs, ve in the
bidirectional scanning algorithms, the AD parameters to determine are
acceleration, constant feedrate, deceleration period as well as maximum
velocity Ts, Tc, Te, vm. All kinds of AD profiles are given in Fig. 4, and
the type-determinant procedures are described as follows.
(1) Reference length Lr1 represents the length accelerating from vs
to vF as well as decelerating form vF to ve. If Li ≥ Lr1, the AD profile
is shown in Fig. 4(a) and it corresponds to type 1. In this case, ACC,
CV (constant velocity) and DEC profile all exist.
2(vF − vs ) 2(vF − ve )
Lr1 = (vs + vF ) + (ve + vF ) (24)
J J
(2) Similarly, reference length Lr2 means the length accelerating or Fig. 5 Layout of the tracking experimental platform
decelerating form vs to ve. Thus, only one of the ACC or DEC profile
exists when Li ≥ Lr2. The AD profile is shown in Figs. 4 (c) or (d),
which corresponds to type 3 or 4. )
J m (vs − ve )Ts 4 − 4 J m LiiTs 3 + 8(vs − ve ) 2 Ts 2
2 vs − ve ) ) 64 (27)
Lr = (vs + ve ) (25) − 64 Li vsTs + 32 Li 2 + (ve − vs )(ve + vs ) 2 = 0
2
J J
(3) If Lr2 < Li < Lr1, this case is type 2. In Fig. 4(b), the maximum In the end, the AD parameters of all types are listed in Table 1.
feedrate vm is smaller than command feedrate and the CV profile does In the stage of interpolation, a predictor-corrector interpolation
not exist while ACC and DEC profile both occur. In this type, the AD method22 is employed to reduce feedrate fluctuations between linear
block length Li can be calculated based on Eq. (21): segments and spline segments whereas the typical second-order Taylor
interpolation is applied to other positions. Hence, the interpolation
1 1
Li = vsTs + J mTs 3 + veTe + J mTe 3 (26) positions can be computed in CNC systems and transferred to driver
16 16
systems.
As observed in Eq. (20), the relations of the deceleration and
2
acceleration time can be calculated as Te = Ts + 8 ( vs – ve )/Jm .
Afterwards, substituting it into Eq. (26), a fourth order polynomial can 4. Simulations and Experiments
be obtained and the acceleration period Ts would be determined by
solving the following equation. In this section, two different examples are utilized to compare the
1534 / OCTOBER 2018 INTERNATIONAL JOURNAL OF PRECISION ENGINEERING AND MANUFACTURING Vol. 19, No. 10
Fig. 6 Maple leaf curve: (a) Toolpath smoothing, (b) Detailed transition results at corners
Fig. 7 The kinematic profiles: (a) Feedrate profile, (b) Acceleration profile, (c) Jerk profile
Fig. 9 Layout of the machining platform: (a) CNC system and machine, (b) Machining shape profile
Fig. 10 The feedrate profiles of two methods: (a) Feedrate profiles, (b) Detailed section of feedrate profiles
sampling period of this platform is set as 0.2 ms while the interpolation corresponding cycle time and contour error of all methods are
period is set as 1 ms. The position sensors work with a resolution of summarized in Table 2.
156.25 nm. Compared to P2P method, proposed curvature-smooth method
A curvedly maple leaf curve is utilized in this part to evaluate the could improve the machining efficacy obviously, and the improvement
efficiency and contour performance. In this section, the corner error indicates nearly 30.32%. As observed, P2P method can obtain the best
constraint is set as ε = 0.1 mm and the kinematic constraints are vm = contour performance with sacrificing machining efficiency, which is not
60 mm/s, Am = 2500 mm/s2, Jm = 180000 mm/s3. Within the given suitable for high speed machining. Compared to curvature-continuous
geometric tolerance, the toolpath after smoothing is shown in Fig. 6(a), method, since smaller curvature at corners can be obtained as analyzed
and the detailed results around several corners are illustrated in Fig. in section 2.2, proposed method can obtain higher machining efficiency,
6(b). In order to indicate the efficiency and contour performance more and the efficiency is improved about 12.35%. In addition, benefiting
accurately, the maple leaf curve in Fig. 6 is operated repeatedly for four from smoothness of the toolpath, the contour error of proposed method
times. also reduced in both mean and maximum contour error from Table 2.
Calculated by the interpolation points, Fig. 7 presents the kinematic
profiles of three methods, P2P, curvature-continuous method with jerk- 4.2 Machining experiments
limited planning and proposed method. It can be seen that feedrate, In order to verify the actual machining efficiency and quality, the
acceleration and jerk are all limited as we set. Furthermore, from the shape shown in Fig. 9(b) is implemented on a CNC machine by the
details that marked with rectangles in Fig. 7(b) and (c), continuous jerk proposed curvature-smooth method and previous curvature-continuous
and smooth acceleration are both guaranteed, which is consistent with method.
the analysis in section 4. The shape machined is composed of 1729 linear points and the
Moreover, Fig. 8 shows the practically measured contour errors for experimental parameters are set as follows: spindle speed S = 10000 r/
the toolpath interpolated by three methods. Specifically, the min, program feedrate F = 80 mm/s, acceleration limitation Am = 1000
1536 / OCTOBER 2018 INTERNATIONAL JOURNAL OF PRECISION ENGINEERING AND MANUFACTURING Vol. 19, No. 10
Fig. 11 The machining results: (a) Three-dimension white light interference surface topography instrument, (b) Machining shape and partial
enlarged view
mm/s2, jerk limitation Jm = 250000 mm/s2, and corner error constraint and B-spline toolpath with G3 continuity and a bidirectional scanning
ε = 0.05 mm. algorithm is provided in the feedrate scheduling scheme. Finally, the
Fig. 10 shows the feedrate profiles of two methods, curvature- proposed method is validated by tracking and machining experiments
continuous and the proposed method. Compared to curvature- respectively. It is also compared to P2P and curvature-continuous
continuous method, proposed curvature-smooth method could improve methods in terms of efficiency, contour performance and machining
the machining efficiency about 14.9%, which is mainly determined by quality at the same time. The experiments benchmarks indicate that the
the smaller curvature extreme values. The actual machining results are proposed method can improve the machining efficiency obviously with
illustrated in Fig. 11, and the portion marked in yellow dots are enlarged better contour performance and surface texture.
and measured by a three-dimension white light interference surface
topography instrument (ZYGONexView) with a resolution of 0.1 nm,
which is also shown in Fig. 11(a). ACKNOWLEDGEMENT
For proposed method, the 3D average surface roughness (Sa) is
reduced from 1.499 µm to 1.190 µm. Meanwhile, the root mean square This work was partially supported by the National Key Technology
height (Sq) and the maximum height (Sz) of the proposed methods are Support Program of China (No. 2015BAI0B16). The research was also
both improved also. The improvement in machining quality is due to supported by Guangdong ELESY Electric Co. Ltd. The authors are
the smoother transition and feedrate profile. In this case, the proposed grateful to Prof. Chuxiong Hu from Tsinghua University for the technical
method can obtain better machining quality and higher machining assistance in section 4.1.
efficiency at the same time.
REFERENCES
5. Conclusions
1. Zhao, X., Zhao, H., Li, X., and Ding, H., “Path Smoothing for Five-
In this paper, a novel smoothing and feedrate scheduling method is Axis Machine Tools Using Dual Quaternion Approximation with
proposed for discontinuous linear segments. Compared to other already Dominant Points,” International Journal of Precision Engineering
existing method, the proposed path-smoothing method improves the G0 and Manufacturing, Vol. 18, No. 5, pp. 711-720, 2017.
toolpath into curvature-smooth one with a whole quintic B-spline curve.
2. Ye, P., Shi, C., Yang, K., and Lv, Q., “Interpolation of Continuous
The curvature monotonicity is proved by a new method and it is proved
Micro Line Segment Trajectories Based on Look-Ahead Algorithm
that the analytical curvature extreme occurs at the midpoint of B-spline
in High-Speed Machining,” The International Journal of Advanced
curve exactly. Moreover, B-spline transition curve and its control points
Manufacturing Technology, Vol. 37, Nos. 9-10, pp. 881-897, 2008.
are optimally designed to obtain the smaller curvature extreme than
other methods, which can reduce cycling time in return. Thereafter, a 3. Zhao, H., Zhu, L., and Ding, H., “A Real-Time Look-Ahead
corresponding jerk-continuous feedrate profile is designed for the linear Interpolation Methodology with Curvature-Continuous B-Spline
INTERNATIONAL JOURNAL OF PRECISION ENGINEERING AND MANUFACTURING Vol. 19, No. 10 OCTOBER 2018 / 1537
Transition Scheme for CNC Machining of Short Line Segments,” 2359-2371, 2017.
International Journal of Machine Tools and Manufacture, Vol. 65,
15. Dong, J. and Stori, J., “A Generalized Time-Optimal Bidirectional
pp. 88-98, 2013.
Scan Algorithm for Constrained Feed-Rate Optimization,” Journal
4. Du, X., Huang, J., and Zhu, L.-M., “An Analytical Transition of Dynamic Systems, Measurement, and Control, Vol. 128, No. 2,
Algorithm for Real-Time CNC Machining of Linear Tool Path,” pp. 379-390, 2006.
Procedia CIRP, Vol. 56, pp. 344-348, 2016.
16. Moore, P. and Molloy, D., “Efficient Energy Evaluations for Active
5. Fan, W., Lee, C.-H., and Chen, J.-H., “A Realtime Curvature- B-Spline/NURBS Surfaces,” Computer-Aided Design, Vol. 47, pp.
Smooth Interpolation Scheme and Motion Planning for CNC 12-31, 2014.
Machining of Short Line Segments,” International Journal of
17. Bi, Q., Wang, Y., Zhu, L., and Ding, H., “A Practical Continuous-
Machine Tools and Manufacture, Vol. 96, pp. 27-46, 2015.
Curvature Bezier Transition Algorithm for High-Speed Machining
6. Sencer, B., Ishizaki, K., and Shamoto, E., “A Curvature Optimal of Linear Tool Path,” Intelligent Robotics and Applications, pp. 465-
Sharp Corner Smoothing Algorithm for High-Speed Feed Motion 476, 2011.
Generation of NC Systems Along Linear Tool Paths,” The
18. Lai, J.-Y., Lin, K.-Y., Tseng, S.-J., and Ueng, W.-D., “On the
International Journal of Advanced Manufacturing Technology, Vol.
Development of a Parametric Interpolator with Confined Chord
76, Nos. 9-12, pp. 1977-1992, 2015.
Error, Feedrate, Acceleration and Jerk,” The International Journal of
7. Tulsyan, S. and Altintas, Y., “Local Toolpath Smoothing for Five- Advanced Manufacturing Technology, Vol. 37, Nos. 1-2, pp. 104-
Axis Machine Tools,” International Journal of Machine Tools and 121, 2008.
Manufacture, Vol. 96, pp. 15-26, 2015.
19. Barre, P.-J., Bearee, R., Borne, P., and Dumetz, E., “Influence of a
8. Zhang, L., Zhang, K., Yan, Y., Zhang, L., and Yan, K. Z. Y., “Local Jerk Controlled Movement Law on the Vibratory Behaviour of
Corner Smoothing Transition Algorithm Based on Double Cubic High-Dynamics Systems,” Journal of Intelligent and Robotic
NURBS for Five-Axis Linear Tool Path,” Strojniski Vestnik-Journal Systems, Vol. 42, No. 3, pp. 275-293, 2005.
of Mechanical Engineering, Vol. 62, No. 11, pp. 647-656, 2016.
20. Lin, M.-T., Tsai, M.-S., and Yau, H.-T., “Development of a
9. Zhang, Q., Gao, X.-S., Li, H.-B., and Zhao, M.-Y., “Minimum Time Dynamics-Based NURBS Interpolator with Real-Time Look-Ahead
Corner Transition Algorithm with Confined Feedrate and Axial Algorithm,” International Journal of Machine Tools and
Acceleration for NC Machining Along Linear Tool Path,” The Manufacture, Vol. 47, No. 15, pp. 2246-2262, 2007.
International Journal of Advanced Manufacturing Technology, Vol.
21. Lee, A.-C., Lin, M.-T., Pan, Y.-R., and Lin, W.-Y., “The Feedrate
89, Nos. 1-4, pp. 941-956, 2017.
Scheduling of NURBS Interpolator for CNC Machine Tools,”
10. Emami, M. M. and Arezoo, B., “A Look-Ahead Command Computer-Aided Design, Vol. 43, No. 6, pp. 612-628, 2011.
Generator with Control Over Trajectory and Chord Error for
22. Tsai, M.-C. and Cheng, C.-W., “A Real-Time Predictor-Corrector
NURBS Curve with Unknown Arc Length,” Computer-Aided
Interpolator for CNC Machining,” Journal of Manufacturing Science
Design, Vol. 42, No. 7, pp. 625-632, 2010.
and Engineering, Vol. 125, No. 3, pp. 449-460, 2003.
11. Sun, S., Lin, H., Zheng, L., Yu, J., and Hu, Y., “A Real-Time and
23. Hu, C., Wang, Z., Zhu, Y., Zhang, M., and Liu, H., “Performance-
Look-Ahead Interpolation Methodology with Dynamic B-Spline
Oriented Precision LARC Tracking Motion Control of a Magnetically
Transition Scheme for CNC Machining of Short Line Segments,”
Levitated Planar Motor with Comparative Experiments,” IEEE
The International Journal of Advanced Manufacturing Technology,
Transactions on Industrial Electronics, Vol. 63, No. 9, pp. 5763-
Vol. 84, Nos. 5-8, pp. 1359-1370, 2016.
5773, 2016.
12. Jahanpour, J. and Alizadeh, M. R., “A Novel Acc-Jerk-Limited
NURBS Interpolation Enhanced with an Optimized S-Shaped
Quintic Feedrate Scheduling Scheme,” The International Journal of
Advanced Manufacturing Technology, Vol. 77, Nos. 9-12, pp. 1889-
1905, 2015.
13. Fan, W., Gao, X.-S., Yan, W., and Yuan, C.-M., “Interpolation of
Parametric CNC Machining Path under Confined Jounce,” The
International Journal of Advanced Manufacturing Technology, Vol.
62, Nos. 5-8, pp. 719-739, 2012.
2
(k , u )1
Peiqing Ye
Professor in the Department of Mechanical
f (u ) = 400 tan θ ( A k + A k + A )
1 1 1
2
2 1 3
Engineering, Tsinghua University. His
(B.1)
A (u ) = (6u − 6u + u )
1
6 5 3
research interest is advanced manufacturing
equipment and automation control, servo
A (u ) = −(24u − 36u + 12u + 5u − 3u )
2
6 5 4 3 2
control.
A (u ) = 6(3u − 5u + 2u )
3
6 5 4
E-mail: yepq@tsinghua.edu.cn
When u∈[0, 0.5], obviously (A1, A2, A3) > 0, hence f1(u) > 0 is
satisfied and the derivative about parameter u is calculated as follows:
1 f (u ) f (u ) − 1.5 f (u ) f (u )
' '
κu = 1 2 2 1
F (k ) = 12 p k + 6 p k + 2 p
1
''
1 4
2
1 3
2
1 2
(B.4)