Professional Documents
Culture Documents
discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/265052777
CITATIONS READS
16 308
2 authors:
Some of the authors of this publication are also working on these related projects:
Tool path generation and accuracy improvement of five-axis CNC machining View project
All content following this page was uploaded by Jixiang Yang on 23 January 2015.
The user has requested enhancement of the downloaded file. All in-text references underlined in blue are added to the original document
and are linked to publications on ResearchGate, letting you access and read them immediately.
International Journal of Machine Tools & Manufacture 88 (2015) 923
art ic l e i nf o a b s t r a c t
Article history: Nonlinear and conguration-dependent ve-axis kinematics make contouring errors difcult to estimate
Received 5 June 2014 and control in real time. This paper proposes a generalized method for the on-line estimation and
Received in revised form control of ve-axis contouring errors. First, a generalized Jacobian function is derived based on screw
11 August 2014
theory in order to synchronize the motions of linear and rotary drives. The contouring error components
Accepted 14 August 2014
contributed by all active drives are estimated through interpolated position commands and the
Available online 24 August 2014
generalized Jacobian function. The estimated axis components of contouring errors are fed back to the
Keywords: position commands of each closed loop servo drive with a proportional gain. The proposed contouring
Five-axis error estimation and control methods are general, and applicable to arbitrary ve-axis tool paths and any
Control
kinematically admissible ve-axis machine tools. The proposed algorithms are veried experimentally
Contouring errors
on a ve-axis machine controlled by a modular research CNC system built in-house. The contouring
Screw theory
CNC errors are shown to be reduced by half with the proposed method, which is simple to implement in
existing CNC systems.
& 2014 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.ijmachtools.2014.08.004
0890-6955/& 2014 Elsevier Ltd. All rights reserved.
10 J. Yang, Y. Altintas / International Journal of Machine Tools & Manufacture 88 (2015) 923
i.e., the controller of one drive will take no action when the where the contouring errors calculation is not affected by the
contouring accuracy is deteriorated by the imperfect tracking curvature of the desired contour. Sun and Tong [18] presented
performance of other drives. Koren [9] estimated contouring errors a synchronization control approach to minimize multi-axis con-
as linear relationships with the drives' tracking errors, and touring errors. They dened the synchronization as the differential
compensated the contouring control signals to the velocity loop errors between each axis and its adjacent axes, and a decentralized
of each drive with the cross-coupled controller scheme (CCC). tracking controller was designed to minimize both tracking errors
Further enhancements of the CCC mainly focused on developing and synchronization errors. Tang et al. [19] and El Khalick et al.
an accurate contouring estimation algorithm [1013], or designing [20] proposed a contour control methodology which combined
new compensation structures [7,14,15] to improve the contouring model predictive control and adaptive feed rate regulation for the
accuracy. Koren and Lo [10] designed a variable gain CCC as an bi-axis CNC system. Altintas and Khoshdarregi [21] presented a
extension of [9] to improve the estimation accuracy of contouring vibration-avoidance and contouring error pre-compensation algo-
errors for nonlinear trajectories. Yeh and Hsu [11] approximately rithm for feed drives. The excitation of the structural modes was
estimated the contouring error vectors for the CCC by projecting avoided by applying input shaping lters on the reference axis
the tracking error vectors on the normal direction of the trajectory commands. Tracking errors were estimated from the closed loop
at the reference points. Erkorkmaz and Altintas [12] estimated the transfer function of drives, and were used to predict the contour-
contouring errors along an arbitrary two-axis tool path by line- ing errors which were mapped to each axis for pre-compensation.
segment approximation of the original contour. Later, Huo et al. However, two important factors limit the direct application of
[13] proposed a similar approach as presented in [12] by estimat- past contouring error minimization approaches on a ve-axis
ing the contouring errors from the actual and reference positions machine tool. First, the ve-axis contouring accuracy is affected
of the machine. In order to simultaneously improve tracking and not only by the tool tip position, but also by the tool orientation,
contouring performances, Su and Cheng [14] designed a position which is illustrated in Fig. 1(a). Furthermore, even though the tool
error compensator combined with the cross-coupled controller. tip positions may be tracked with high accuracy, the tool orienta-
Shih et al. [15] and Yang and Li [7] developed new CCC structures tion discrepancy may still cause large over-cut or under-cut, as
for bi-axis contouring paths. The CCC structures presented in shown in Fig. 1(b). Consequently, unlike two- or three-axis
[15,7] carried out the compensation at the position loop rather machines, tool tip position and tool orientation should be con-
than at the velocity loop. trolled simultaneously to minimize contouring errors in ve-axis
There have also been alternative methods to CCC systems, machines [1]. The second factor is that the ve-axis contouring
which transfer the contour error to a new coordinate frame, or operations have non-linear, structure-dependent kinematic rela-
estimate the contouring error and decouple it as axis tracking tionships with the drives' motions [22]. Lo [23] designed a coupled
errors for feedforward compensation. Chiu and Tomizuka [16] PID controller to minimize contouring and tracking errors in the
formulated the contouring control problem in the local task workpiece coordinate frame, and then transferred the control
coordinate frame which moved along the trajectory. The controller signals into the machine coordinate system by using the inverse-
was designed to minimize the tangential and normal components Jacobian matrix. Sencer et al. [24] provided a ve-axis contour
of contouring errors. Hu et al. [17] proposed a global task error estimation method for tool tip position spline and tool
coordinate frame-based contouring controller for bi-axial systems, orientation spline separately. The tool tip position contour error
Fig. 1. (a) Five-axis contour error illustration; (b) cutting error caused by tool orientation discrepancy.
J. Yang, Y. Altintas / International Journal of Machine Tools & Manufacture 88 (2015) 923 11
T
O Oi , O j , Ok PN
Tool
O 1
PN-1 z
T
c Lac, z
P Px , Py , Pz
Workpiece LTya ,z
P1 P2
Table
P0
a
Zw x
Yw Zm Ym
y
Xw Xm
T
P Px , Py , Pz T
M x, y , z , a, c
T
O Oi , O j , Ok
a b
Fig. 4. Kinematics structure of the experimental machine: (a) kinematics structure model; (b) kinematics chains.
2 3
as follows [30]: 1 0 0 0
2 3 2 3 60 1 0 0 7
P x Oi 0 0 6 7
g bt 0 6 7
6P O 7 60 07 40 0 1 LTa;z 5
6 y j 7 ^ ^ ^ ^ ^ 6 7
6 7 g bw 0 1 Ue c c e a a e x x e y y ez z Ug bt 06 7 0 0 0 1
4 P z Ok 5 40 15
1 0 1 0
3 In order to synchronize the movements of translational drives
with and rotary drives that affect contouring accuracy, the generalized
" # " # Jacobian function for ve-axis machines is derived by differentiat-
vx T
vy ing the forward kinematics function given in Eq. (2). According to
x ; vx 1 0 0 ; y ;
031 031 the screw theory [31], the instantaneous spatial velocity of the
" #
vz cutting tool frame as seen from the workpiece frame is:
vy 0 1 0 T ; z ; vz 0 0 1 T ;
031 s
" # V^ wt g_ wt g wt 1 4
a qa T T
a ; a 1 0 0 ; qa 0 0 0 ;
a
" # The vector V swt corresponds to the instantaneous spatial velo-
c qc T city twist of the cutting tool frame related to the workpiece frame.
c ; c 0 0 1 ;
c g_ wt is the instantaneous time variance of the transfer matrix from
2 3 the workpiece frame to the tool frame. The spatial velocity can be
1 0 0 0
60 evaluated from the chain rule as:
6 1 0 0 7 7
qc 0 0 0 T ; g bw 0 6 7;
40 0 1 Lac;z 5 s 5 g wt _ 5 g wt
V^ wt i g wt 1 Ug wt 1 _ i 5
0 0 0 1 i1 i i1 i
J. Yang, Y. Altintas / International Journal of Machine Tools & Manufacture 88 (2015) 923 13
It must be noted that the tool tip position in the workpiece ^ ^ ^ ^ ^ r pt
Jpy g bw 0 1 e c c e a a e x x ^ y e y y ez z g bt 0
frame is 1
P 1 ^ c c ^ a a ^ x x ^ y y ^ ^ z z
r pt
P ; Jpz g bw 0 e e e e z e g bt 0
1 1
1 ^ c c ^ ^ a a ^ x x ^ y y ^ z z
r pt
where P P x ; P y ; P z T . The tool tip velocity related to the work- Jpa g bw 0 e a e e e e g bt 0
1
piece coordinate frame is calculated as
" # ^ ^ ^ ^ ^ r
Jpc g bw 0 1 ^ c e c c e a a e x x e y y ez z g bt 0
pt
P_ 5 g wt 14
P_
s
V^ wt P g wt 1 P _ i Jp _ 6 1
0 i1 i
where r pt 0 0 0 T is the cutting tool position vector relative
where to the cutter coordinate frame.
g wt g
Jp g wt 1 P; ; wt g wt 1 P ; _ _ 1 ; _ 2 ; _ 3 ; _ 4 ; _ 5 T 7 2.2. On-line estimation of contouring errors
1 5
where
Xw
^ ^ ^ ^ ^ r pt
Jpx g bw 0 1 e c c e a a ^ x e x x e y y ez z g bt 0
1 Fig. 5. Illustration of the contouring error estimation method.
14 J. Yang, Y. Altintas / International Journal of Machine Tools & Manufacture 88 (2015) 923
!
The drive motions of the machine can be transferred to the ratio hr 1 from Pa to vector P r P r 1 is:
workpiece frame through the Jacobian function as follows,
P r P a T P r 1 P r r a T JTp;r Jp;r r 1 r
hr 1 T
P r 1 P a Jp;r 1 r 1 a ; P r P r 1 Jp;r 1 r r 1 17 P r 1 P r P r 1 P r r 1 r T JTp;r Jp;r r 1 r
By substituting Eq. (17) into Eq. (16), the ratio hr is evaluated in 22
the machine frame as:
If hr 1 is still larger than 1, the iterative calculation is continued
r 1 a T JTp;r 1 Jp;r 1 r r 1 until one 0 r hr k r 1, then the drive commands corresponding
hr 18
r r 1 T JTp;r 1 Jp;r 1 r r 1 to the contour error pose P c ; Oc are calculated as
c 1 hr k r k 1 hr k r k 23
The tracking errors contributed by each drive on the total
contouring error are estimated for the following ve cases as
follows. Case 4 (Fig. 6(d)): If hr o0, while hr 1 4 1, then the nearest
point is taken as P r 1 . The drive commands corresponding to
Case 1 (Fig. 6(a)): If 0 r hr r 1, the locally nearest point from the contour error pose P c ; Oc are calculated as:
actual position Pa to the reference path lies between P r 1 and c r 1 24
Pr. The relative drive commands c c;x ; c;y ; c;z ; c;a ; c;c T
corresponding to the contour error pose P c ; Oc are expressed
Case 5 (Fig. 6(e)): If hr 4 1 and hr 1 o0, the nearest point is
as:
assumed as Pr. The drive commands corresponding to the
c 1 hr r 1 hr r 19 contour error pose P c ; Oc are calculated as
c r 25
Case 2 (Fig. 6(b)): If hr o0, the locally nearest point from Pa to
the referenced trajectory lies before P r 1 , and the perpendi-
!
cular ratio hr 1 from Pa to vector P r 2 P r 1 is
Hence, the drive commands corresponding to the contour error
P r 2 P a T P r 1 P r 2 pose are obtained for each case. Consequently, the contribution of
hr 1
P r 1 P r 2 T P r 1 P r 2 the contour error to each axis x ; y ; z ; a ; c T can be evaluated
r 2 a T JTp;r 2 Jp;r 2 r 1 r 2 easily as:
20
r 1 r 2 T JTp;r 2 Jp;r 2 r 1 r 2 c a c;x ; c;y ; c;z ; c;a ; c;c T a;x ; a;y ; a;z ; a;a ; a;c T 26
The maximum iteration time depends on the magnitude of the
If hr 1 is still smaller than 0, the iterative calculation is tracking errors along the trajectory in terms of the machines'
continued until 0 rhr k r 1, and then the drive commands sampling intervals along the arc-length, but it is constrained by
corresponding to the contour error pose P c ; Oc are calculated the computational power of the CNC's CPU. Suppose that the
as maximum limitation of the iterative time is M, if the contour error
cannot be determined after M interpolation samples which means
c 1 hr k r k 1 hr k r k 21
hr M o 1 or hr M 4 1, then the drive commands corresponding to
the contour error pose P c ; Oc are calculated as c r M or
Case 3 (Fig. 6(c)): If hr 4 1, the locally nearest point from Pa to c r M . Although the estimated contour error is not exactly
the referenced trajectory lies after Pr, and the perpendicular equal to the actual one, it is smaller than the contour error
hr 0 Pr Pr Pr hr 1 Pr
Pr 1 0 hr 1 Pr Pr
1 Pr 1
1
2
Hr
Pa Pa
Pa
Pr 2 Pr 1
Pr 1 Pr Pr Pr 1
hr 0 hr 1 hr 0
1 1 hr 1
Pa Pa
Fig. 6. Different contouring error estimation conditions.
J. Yang, Y. Altintas / International Journal of Machine Tools & Manufacture 88 (2015) 923 15
happens at the reference commands r . More importantly, the contouring errors estimated from the Virtual CNC with the
corresponding tool pose after compensation is still guaranteed on proposed method are compared against the actual contouring
the desired trajectory. The maximum iteration time is constrained errors evaluated off-line by calculating the minimum distance
to 15 iterations which are completed within 1 ms servo control between the actual and the desired paths. The rst trajectory is a
intervals during the experiments reported in the paper. spatial helix, and the second trajectory is the fan prole used in
One explicit advantage of the proposed contouring estimation [28]. Both trajectories are generated by using the C3 continuous
model is that it is directly performed in the machine frame as non-singular tool path generation algorithm proposed in [30]. The
opposed to the workpiece frame reported by the previous feed rate optimization method with drive constraints [32] is used
research, which requires both forward kinematics and inverse to generate reference position commands to drives with a max-
kinematics calculations at each estimation cycle. The requirement imum tangential feed rate of 100 mm/s. The drives are controlled
of the inverse kinematics calculation lmits the generality of the with a lead compensation, closed loop controller presented in
contour error estimation and control model in the sense that it Section 3 without contouring error control.
greatly depends on the machine topology, i.e., the estimation and The rst trajectory and its interpolated drive commands are
control method developed for one machine topology is difcult to shown in Fig. 8. The difference between the estimated and actual
be adapted to another one. However, the proposed method avoids contouring errors is negligibly small as shown in Fig. 9. The fan-
the non-generality problem by proposing the generalized Jacobian shaped trajectory and its reference drive commands are shown in
function to transfer all the calculations to the workpiece frame, Fig. 10, and its estimated contouring errors are also very close to
where only the reference drive commands and their actual the real values as shown in Fig. 11. The simulation results show
positions are needed during the calculation. A detailed comparison that the estimated tool tip contouring and tool orientation errors
with the workpiece frame method presented in [23] and the are within 0.1 mm and 4 mrad of the actual values, respectively. The
proposed approach is illustrated in Fig. 7, where P aw ; Oaw and discrepancy is within an acceptable range for the real-time
P rw ; Orw are the actual and reference position and orientation contouring error prediction and compensation in ve-axis CNC
vectors in the workpiece frame, while pw ; ow are the position machines.
and orientation contour errors in the workpiece frame. The In order to further analyze the two trajectories, the curvatures
comparison shows that the proposed model is signicantly sim- of their tool tip splines are evaluated as shown in Fig. 12. It can be
plied and shortened with the proposed method. Furthermore, seen that the curvature for the rst trajectory is low and constant.
the Jacobian function required in [23] is not always reversible, and However, the curvature for the second trajectory is high and
thus the calculation may face singular points. However, the variable, and the proposed contouring error estimation method
proposed model will not face the singular problem since no works well for both curvatures. The reason for the robust perfor-
inverse calculation is needed. mance of the proposed contour error estimation method is that
the reference drive commands are generated with constrained
2.3. Verication of the contour error estimation with simulations chord errors, hence the linear, discrete segments are small enough
to represent the splined path. Since the method uses only the
A series of simulations have been performed to contour two reference and actual position values at discrete control intervals, it
sample trajectories in the Virtual CNC platform [33,34]. The does not require the evaluation of the contouring errors in the
a ,a Position Contour c, y
Forward Paw , Oaw Inverse
a ,c
Error pw , ow Contour c, z
Kinematics Jacobian
r,x
Prw , Orw Estimation Position
c, a
r,y Function Function
Reference Model c,c
r,z
Position
r ,a
r ,c
Machine Machine
Coordinate Coordinate
frame frame
a,x
a, y
a,z Actual c, x
a ,a Position Contour c, y
a ,c Error Contour c, z
r,x Estimation Position
c,a
r, y
Reference Model
r,z c,c
Position
r ,a
r ,c
Fig. 7. Contouring error estimation structures: (a) traditional methods [23], (b) proposed method.
16 J. Yang, Y. Altintas / International Journal of Machine Tools & Manufacture 88 (2015) 923
Z: mm 15
Ok
10 0
5
10 -1
0 10 1
Y:
mm 0 0
-10 1
-10 mm Oj 0
X: -1 -1
Oi
Position: rad
200 0
-1
0
-2
0.5 1 1.5 2 0 0.5 1 1.5 2
Time: s Time: s
Fig. 8. The helix trajectory and its interpolated motion commands.
60
1
40
0.5
20
0 0
0 0.5 1 1.5 2 0 0.5 1 1.5 2
Time [s] Time [s]
Discrepancy: Discrepancy:
Actual value - Estimation value Actual value - Estimation value
0.02 3
Contour error [microradian]
Contour error [micrometer]
0
2
-0.02
-0.04 1
-0.06
0
-0.08
-0.1 -1
0 0.5 1 1.5 2 0 0.5 1 1.5 2
Time [s] Time [s]
Fig. 9. The estimated contouring errors and their discrepancy with the actual values for the helix trajectory.
J. Yang, Y. Altintas / International Journal of Machine Tools & Manufacture 88 (2015) 923 17
1
Z: mm 2
-2
Ok
0 0
40 -1
-20 20
Y: 0
mm -40 mm -1 0
-20 X: -1 Oi
0
Oj 1 1
Position: rad
Position: mm
200 1
0
0
0 1 2 3 0 1 2
Time: s Time: s
Fig. 10. The fan-shaped trajectory and its interpolated motion commands.
workpiece or frenet frames as proposed in previous articles The axis error i is scaled with a controller gain m and added to
[23,24,26]. Furthermore, the proposed estimation method pro- the position command generated by the trajectory prole and path
vides not only the contour error values, but also their axis interpolator r;i .
components x ; y ; z ; a ; c , that can be directly used to compen- The actual output of the closed loop system with the added
sate contouring errors without computationally costly inverse and contouring error compensator becomes:
forward kinematic transformations at each control interval. In
summary, the proposed contouring error estimation method is m 1Gci Gpi mGci Gpi
a;i r;i ; i x; y; z; a; c
robust to curvature variance, and is trajectory type independent 1 m 1Gci Gpi r;i 1 m 1Gci Gpi c;i
and machine structure independent. 29
1 Gci Gpi 1
3.1. Structure of the contour error controller cl;i ; i x; y; z; a; c
1 m 1Gci Gpi i 1 mGci Gpi =1 Gci Gpi i
60 2
40
1
20
0 0
0 1 2 3 0 1 2 3
Time [s] Time [s]
Discrepancy: Discrepancy:
Actual value - Estimation value Actual value - Estimation value
0.05 4
2
0
-0.05
-2
-0.1 -4
0 1 2 3 0 1 2 3
Time [s] Time [s]
Fig. 11. The estimated contouring errors and their discrepancy with the actual values for the fan-shaped trajectory.
0.25
The discrete transfer function of the lead compensation con-
0.2 troller Gc z is given as:
0.15 h1 z h0
Gc z 33
z l0
0.1
Lead compensation controllers [37] are tuned to have the
0.05
desired crossover frequency (c 150) and phase margin
0 (m 601). The resulting lead compensating controllers for all
0 0.5 1 1.5 2 2.5 3
drives are given in Table 2. The zero order hold equivalent of the
Time [s]
drive's open loop transfer function is evaluated at 1 ms control
Fig. 12. Curvature analysis of the two trajectories used in the simulation. intervals as:
K a K t Rg 1 Gps q z q0
3.2. Controller design Gp s 2 -Gp z
Z 2 1 34
Js Bs 1z 1 s z p1 z p0
a,x
r ,x
Gc x Gp x
x
m
c, x
r, y a, y
Gc y Gp y
y
m
c, y a,z
r,z
Gc z Gp z
z
m
c, z
c, x
Actual
Closed c, y Contour Position
c, z Contour Error
r,x
loop Position Estimation
c, a Reference r,y
c,c
Model Position r,z
r ,a
r ,c
r ,a a ,a
Gca Gpa
a
m
c,a a ,c
r ,c
Gcc G pc
c
m
c,c
Fig. 13. The structure of the proposed ve-axis contouring error controller.
Feedforward Friction
compensation Actual friction
Physical plant
1 z 1 T
Computer
T
R Tz K K u
T
e u T D A T 1
Lead controller K K R
u ZOH Js B
Fig. 14. Block-diagram of the single drive using lead compensation controller.
which is denoted as f z a3 z3 a2 z2 a1 z1 a0 . According to the The proposed ve-axis contouring error estimation and control
Jury criterion [38,39], the necessary and sufcient criterion for the methods are experimentally veried on a ve-axis experimental
20 J. Yang, Y. Altintas / International Journal of Machine Tools & Manufacture 88 (2015) 923
Workstation PC
-Decode NC Program
-Generate Trajectory
Encoder Feedback
Without compensation
150 Compensation gain m+1=2.5
100
50
0
0 0.5 1 1.5 2 2.5 3
Time [s]
Without compensation
4
Compensation gain m+1=2.5
3
0
0 0.5 1 1.5 2 2.5 3
Time [s]
Fig. 16. Contour error experimental results.
J. Yang, Y. Altintas / International Journal of Machine Tools & Manufacture 88 (2015) 923 21
G G
m
G G
Contour m
error
pre-
Actual
Compensation Closed Contour Position
[8] Contour Error
loop Position Estimation Reference
Model Position
G G
m
G G
Contour error m
pre-compensation [8]
Fig. 17. Combination of the proposed on-line contour error control and the pre-compensation algorithm reported in [8].
Without compensation
150
With pre-compensation
Combine pre-compensation with on-line control
100
50
0
0 0.5 1 1.5 2 2.5 3
Time [s]
4 Without compensation
With pre-compensation
3
Combine pre-compensation with on-line control
2
0
0 0.5 1 1.5 2 2.5 3
Time [s]
Fig. 18. Simulation results of tool tip position and tool orientation contour error under different controllers.
22 J. Yang, Y. Altintas / International Journal of Machine Tools & Manufacture 88 (2015) 923
Because the proposed contour error control is an on-line exponential evaluation of a rotary twist is
method which uses only the commanded and actual positions of " #
^ e^ i i I 33 e^ i i i vi
the drives, it can be integrated to look-ahead-based pre- ei i e^ i i I 33
^ i sin i
013 1
compensation algorithms such as those presented in [8,40]. These
algorithms predicted the contouring errors ahead of the axis ^ i 2 1 cos i
A:2
control loops, and decoupled them into three linear and two
rotary components. The error commands were pre-compensated
The operator 4 ' is used to map i A 6 and i A 3 to ^ i A se3
before they were sent to the servo drives for execution. The
and ^ i A so3, while the operator 3 ' performs the opposite
combination of the proposed on-line contour error control and
function as follows:
the pre-compensation algorithm reported in [8] is taken as an " #
example whose block diagram is shown in Fig. 17, where comp;i are vi ^ i vi
i ) ^ i ;
the pre-compensated drive commands of drive i(i x; y; z; a; c). i 0 0
More details of the pre-compensation algorithm can be found in 2 3
0 i;3 i;2
[8]. The simulations indicated that the contouring error can be 6
^ i 4 i;3
i i;1 ; i;2 ; i;3 T ) 0 i;1 7
5 A:3
further minimized by implementing both pre- and on-line com-
i;2 i;1 0
pensation methods as shown in Fig. 18, provided that the plant
parameters are identied accurately. If there are uncertainties in
the drive parameters, pre-compensation of contouring errors The opposite mapping is evaluated as:
suffers because the axis errors are estimated from the transfer
function of the drives. However, the proposed method uses the ^ i 3
^ i 3 i
i ; A:4
measured axis positions, and hence it is not affected by the
uncertainties of the drive parameters. Similarly, the mapping is carried out for a translational joint as
follows:
" #
vi ^ I 33 i vi
5. Conclusions i ; ei i A:5
031 0 1
Five-axis machine tools are widely used because of their ability
where vi A 3 is the unit vector pointing to the positive direction
to contour the complex, sculptured surfaces commonly found in
of the linear axis.
die, mold, aerospace and implant applications. The productivity of
The vectors i , qi and vi are all specied relative to the
such machines is hindered by their low accuracy in contouring
foundation frame, which is located at the intersection of two
complex, splined tool paths while simultaneously positioning ve
rotary axes when the motions of all drives are zero.
drives at high feed speeds. The contouring errors are reduced only
In Eq. (2), g bw 0 and g bt 0 represent the rigid body transfer
if low tangential feeds are used, which leads to lower axis
matrix of the workpiece coordinate and the tool coordinate frames
velocities, hence reducing the productivity of the machines. The
relative to the foundation frame when the motions of all drives are
previously presented contouring error control methods require
zero, i.e.,
both forward and inverse kinematics transformations between the " # " #
machine and workpiece coordinate system. They are computa- I33 wo I33 t o
g bw 0 ; g bt 0 A:6
tionally demanding, and prone to numerical round-off errors in 013 1 013 1
real time applications. This paper presents an on-line contouring
error prediction and compensation method which uses only the where wo wox ; woy ; woz T and t o t ox ; t oy ; t oz T are the offset
commanded and measured positions of the drives, and does not vectors between the workpiece coordinate frame and the tool
require any kinematic transformations. The proposed method can coordinate frame relative to the foundation frame when the
be easily applied in existing CNC systems, and is proven to reduce motions of all drives are zero.
contouring errors by more than 50% with a simple, proportional
error feedback controller. References
[1] L. Tang, R.G. Landers, Multiaxis contour control-the state of the art, IEEE Trans.
Control Syst. Technol. 21 (6) (2013) 19972010.
Acknowledgments [2] F. Huo, A.N. Poo, Precision contouring control of machine tools, Int. J. Adv.
Manuf. Technol. 64 (14) (2013) 319333.
[3] R. Ramesh, M.A. Mannan, A.N. Poo, Tracking and contour error control in CNC
This research is supported by the NSERC Pratt & Whitney
servo systems, Int. J. Mach. Tools Manuf. 45 (3) (2005) 301326.
Canada Research Chair and the National Natural Science Founda- [4] M. Tomizuka, Zero phase error tracking algorithm for digital-control, J. Dyn.
tion of China under Grant No. 51005086. Syst. Meas. Control-Trans. ASME 109 (1) (1987) 6568.
[5] Y. Altintas, K. Erkorkmaz, W.H. Zhu, Sliding mode controller design for high
speed feed drives, CIRP Ann. 2000: Manuf. Technol. (2000) 265270.
[6] K. Erkorkmaz, Y. Altintas, High speed CNC system design. Part III: high speed
Appendix. The Generalized Forward Kinematics Model tracking and contouring control of feed drives, Int. J. Mach. Tools Manuf. 41
(11) (2001) 16371658.
[7] J.Z. Yang, Z.X. Li, A novel contour error estimation for position loop-based
Details of the generalized forward kinematics model given in cross-coupled control, IEEEASME Trans. Mechatron. 16 (4) (2011) 643655.
Eqs. (1) and (2) are given as follows. More details about the [8] K. Zhang, A. Yuen, Y. Altintas, Pre-compensation of contour errors in ve-axis
CNC machine tools, Int. J. Mach. Tools Manuf. 74 (2013) 111.
derivation process can be found in [30].
[9] Y. Koren, Cross-coupled biaxial computer control for manufacturing systems,
For a revolute joint, the twist i A 6 has the form: J. Dyn. Syst. Meas. Control Trans. ASME 102 (4) (1980) 265272.
" # [10] Y. Koren, C.C. Lo, Variable-gain cross-coupling controller for contouring, CIRP
vi Ann.Manuf. Technol. 40 (1) (1991) 371374.
i ; vi i qi A:1
i [11] S.S. Yeh, P.L. Hsu, Estimation of the contouring error vector for the cross-
coupled control design, IEEEASME Trans. Mechatron. 7 (1) (2002) 4451.
[12] K. Erkorkmaz, Y. Altintas, High speed contouring control algorithm for CNC
where i A 3 is the unit vector in the positive direction of the machine tools, in Proceedings of the ASME Dynamic Systems and Control
rotary axis, and qi A 3 is any point on the rotary axis. The Division, 1998. pp. 463469.
J. Yang, Y. Altintas / International Journal of Machine Tools & Manufacture 88 (2015) 923 23
[13] F. Huo, A.N. Poo, Improving contouring accuracy by using generalized cross- [27] M.T. Lin, S.K. Wu, Modeling and improvement of dynamic contour errors for
coupled control, Int. J. Mach. Tools Manuf. 63 (2012) 4957. ve-axis machine tools under synchronous measuring paths, Int. J. Mach.
[14] K.H. Su, M.Y. Cheng, Contouring accuracy improvement using cross-coupled Tools Manuf. 72 (2013) 5872.
control and position error compensator, Int. J. Mach. Tools Manuf. 48 (1213) [28] R.V. Fleisig, A.D. Spence, A constant feed and reduced angular acceleration
(2008) 14441453. interpolation algorithm for multi-axis machining, Comput.-Aided Des. 33 (1)
[15] Y.T. Shih, C.S. Chen, A.C. Lee, A novel cross-coupling control design for bi-axis (2001) 115.
motion, Int. J. Mach. Tools Manuf. 42 (14) (2002) 15391548. [29] A. Yuen, K. Zhang, Y. Altintas, Smooth trajectory generation for ve-axis
[16] G. Chiu, M. Tomizuka, Contouring control of machine tool feed drive systems: machine tools, Int. J. Mach. Tools Manuf. 71 (2013) 1119.
a task coordinate frame approach, IEEE Trans. Control Syst. Technol. 9 (1) [30] J. Yang, Y. Altintas, Generalized kinematics of ve-axis serial machines with
(2001) 130139. non-singular tool path generation, Int. J. Mach. Tools Manuf. 75 (0) (2013)
[17] C.X. Hu, B. Yao, Q.F. Wang, Global task coordinate frame-based contouring 119132.
control of linear-motor-driven biaxial systems with accurate parameter [31] R.M. Murray, Z. Li, S.S. Sastry, A Mathematical Introduction to Robotic
estimations, IEEE Trans. Ind. Electron. 58 (11) (2011) 51955205. Manipulation, CRC Press, Florida, 1994.
[18] D. Sun, M.C. Tong, A synchronization approach for the minimization of [32] B. Sencer, Y. Altintas, E. Croft, Feed optimization for ve-axis CNC machine
contouring errors of CNC machine tools, IEEE Trans. Autom. Sci. Eng. 6 (4)
tools with drive constraints, Int. J. Mach. Tools Manuf. 48 (78) (2008)
(2009) 720729.
733745.
[19] L. Tang, R.G. Landers, Predictive contour control with adaptive feed rate, IEEE
[33] C.H. Yeung, Y. Altintas, K. Erkorkmaz, Virtual CNC system. Part i. System
ASME Trans. Mechatron. 17 (4) (2012) 669679.
architecture, Int. J. Mach. Tools Manuf. 46 (10) (2006) 11071123.
[20] A. El Khalick, N. Uchiyama, Discrete-time model predictive contouring control
[34] K. Erkorkmaz, C.H. Yeung, Y. Altintas, Virtual CNC system. Part ii. High speed
for biaxial feed drive systems and experimental verication, Mechatronics 21
(6) (2011) 918926. contouring application, Int. J. Mach. Tools Manuf. 46 (10) (2006) 11241138.
[21] Y. Altintas, M.R. Khoshdarregi, Contour error control of CNC machine tools [35] Y. Altintas, Manufacturing automation: metal cutting mechanics, machine tool
with vibration avoidance, CIRP Ann.Manuf. Technol. 61 (1) (2012) 335338. vibrations, and CNC design, 2nd Ed., Cambridge University Press, 2012.
[22] T. Moriwaki, Multi-functional machine tool, CIRP AnnalsManufacturing [36] K. Erkorkmaz, Y. Altintas, High speed CNC system design. Part II: modeling and
Technology 57 (2) (2008) 736749. identication of feed drives, Int. J. Mach. Tools Manuf. 41 (10) (2001)
[23] C.C. Lo, A tool-path control scheme for ve-axis machine tools, Int. J. Mach. 14871509.
Tools Manuf. 42 (1) (2002) 7988. [37] R.C. Dorf, R.H. Bishop, Modern control systems, 12th Ed., Pearson, Upper
[24] B. Sencer, Y. Altintas, E. Croft, Modeling and control of contouring errors for Saddle River, New Jersey, 2010.
ve-axis machine tools-part i: modeling, J. Manuf. Sci. Eng.Trans. ASME 131 [38] E.I. Jury, Simplied stability criterion for linear discrete systems, Proc. Inst.
(2009) 0310063. Radio Eng. 50 (6) (1962) 14931500.
[25] B. Sencer, Y. Altintas, Modeling and control of contouring errors for ve-axis [39] S.M. Shinners, Advanced modern control system theory and design, John
machine tools-part ii: precision contour controller design, J. Manuf. Sci. Eng. Wiley & Sons, New York, 1998.
Trans. ASME 131 (2009) 0310073. [40] M.R. Khoshdarregi, S. Tappe, Y.Y. Altintas, Integrated ve-axis trajectory
[26] A. El Khalick M, N. Uchiyama, Estimation of tool orientation contour errors for shaping and contour error compensation for high-speed CNC machine tools
ve-axis machining, Robot. Comput.-Integr. Manuf. (0) . (IEEE/ASME Trans.), Mechatronics (99) (2014) 113 (pp).