Professional Documents
Culture Documents
net/publication/260133433
CITATIONS READS
2 262
5 authors, including:
Bernard Anselmetti
Ecole normale supérieure de Cachan
63 PUBLICATIONS 770 CITATIONS
SEE PROFILE
All content following this page was uploaded by Nabil Anwer on 22 May 2015.
Abstract—Tolerance analysis of a planar 5R parallel uncertainty. Joint clearance may produce uncertain pose
mechanism with link dimension deviation and joint errors of the end effector, the unconstrained motions of
clearance are performed under the hypothesis of small the end effector are bounded by real dimensions which
displacements and clearance link. The influences of are random variables. The effects of joint clearance are
different link dimension deviations are derived by the highly non-repeatable and can not be rectified with any
linearization of kinematic equations, and the influences kind of calibration. However, joint clearance is necessary
of clearance link at active/passive joints are transformed for the assembility and mobility of over-constrained
to a mechanism reachability problem and derived parallel mechanisms.
individually. The geometric behaviors of link dimension Vukobratović et al. [2] investigated the influences of
deviations and joint clearances are studied and the robot kinematic parameter deviations on the gripper
compared by a numerical example. All these above pose errors and determined the specific manufacturing
provide the basis of the accuracy design theory for new tolerances which cause the maximal effector deviations
parallel mechanisms. to ensure a desired accuracy portrait. Chase et al. [3]
Keywords: parallel mechanism, tolerance analysis, link dimension presented the direct linearization method for tolerance
deviation, joint clearance
analysis of 2D and 3D mechanical assemblies, using a
vector-loop-based assembly tolerance models to describe
I. Introduction
the relationships between the resultant assembly
Parallel mechanisms are now widely used in real-life dimension and manufactured component dimension. Ting
applications such as precise positioning devices, motion et al. [4] presented an application of invariant link
simulator, ultra-fast pick-and-place robot, machine tools rotatability laws to identify the worst position and
and haptic devices [1]. The demand for high precision orientation errors due to joint clearance of planar linkage
tasks for parallel mechanism is continuously growing and containing revolute joints. Venazi et al. [5] applied the
the design methodology granting the prescribed accuracy principle of virtual work to study the kinematic
is becoming a critical issue. influences of clearance-affected pairs on the position and
The performance quality of an articulated mechanism orientations of the links of spatial mechanisms. Based on
is usually referred to accuracy and repeatability. The screw theory, Frisoli et al. [6] proposed a method to
accuracy is defined as a measure of the robot capability compute the effect of clearances on the motion accuracy
to attain commanded position or orientation with respect and the best/worst performance of a spatial manipulator
to the fixed absolute reference coordinate frame. The under a generic pose error function.
repeatability is defined as the capability of the robot to Because of the inheritance of the variations in
return the end effector to a previously reached and manufacturing and assembly processes, the dimension
memorized pose. The performance deviation is usually deviations and the performance variations of the
caused by manufacturing and assembly errors, which is mechanisms are unavoidable. How the link dimension
accumulated by two kind of small displacements. First deviation and joint clearance in parallel mechanisms
the relative deviation between different positioning contribute to the output deviations and what analysis
features that belong to a part which is always called link technique could be used to evaluate the performance are
dimension deviation, then the displacement due to the important issues. At the conceptual design stage of a
clearance between the two surfaces of the two parts parallel mechanism, we should take account of both
in contact. Link dimension deviation leads to invariable factors and assess its effects on the accuracy and
position and orientation deviation, its components are assembility of parallel mechanism.
random variables and can be well calibrated and they The paper will treat the problem of tolerance analysis
will remain the same during operation and cause no of a planar 5R parallel mechanism, taking into account
the effects of all link dimension deviation and joint
*
yangjx@bjut.edu.cn clearance. Because the magnitude of link dimension
†
jywang@tsinghua.edu.cn
‡
anwer@lurpa.ens-cachan.fr
deviation and joint clearance is limited and its value is
‡
anselm@lurpa.ens-cachan.fr reasonably of a lower order of magnitude than the
1
13th World Congress in Mechanism and Machine Science, Guanajuato, México, 19-25 June, 2011 A11_306
nominal mechanism dimensions. Under the hypothesis of There are four solutions for the inverse kinematic
small displacements, two important assumptions can be problem of the 5R parallel mechanism. For the
drawn: (1) the small displacements at the end effector, configuration shown in Fig.4, if the position of output
due to link dimension deviations on different parts and point C is known, the inputs to reach the position can be
joint clearances in different joints, are independent of obtained as
each other; (2) the contributions of link dimension ⎛ − b + b 2 − 4a c ⎞
deviations and joint clearances, calculated by given θ 1 = 2 tan −1 ⎜ 1 1 1 1 ⎟ (6)
⎜ 2a1 ⎟
tolerances and according to the kinematic equations, are ⎝ ⎠
composed as a linear summation to determine the ⎛ − b − b − 4a c ⎞
2
resulting motion of the end effector. θ 4 = 2 tan −1 ⎜ 2 2 2 2 ⎟ (7)
⎜ 2a 2 ⎟
⎝ ⎠
II. Description of Mechanism where
As shown in Fig.1, the planar 5R parallel mechanism a1 = x 2 + y 2 + l12 − l22 + 2xl1
is such a typical mechanism with the minimal degrees of b1 = −4 yl1
freedom in the field, which can be used for positioning a
c1 = x 2 + y 2 + l12 − l22 − 2xl1
point freely on a region of a plane that is known as the
workspace. It consists of five bars that are connected end a 2 = ( x − r5 ) 2 + y 2 + l 42 − l32 + 2( x − l5 )l 4
to end by five revolute joints, two of which that are b2 = −4 yl 4
connected to the base are actuated. Such a mechanism c 2 = ( x − l 5 ) 2 + y 2 + l 42 − l32 − 2( x − l5 )l 4
has attracted many researchers, who have investigated its
position analysis, workspace, singularity, assembly
III. Tolerance Analysis of Link Dimension Deviation
modes, performance atlases and kinematic design [7].
Eq. (4) and (5) can be linearized using a first order
C (x, y) Taylor’s series expansion, which yields
l2 l3 [ A ][ Δc] = [B][Δl] + [C][ Δθ] (8)
where
B D
[ Δc] = [ Δx Δy ]
T
θ1
l5 θ4 ⎡ x − l1 cos θ1 y − l1 sin θ1 ⎤
[ A] = ⎢ x − l y − l4 sin θ 4 ⎥⎦
4 cos θ 4 − l5
A x E ⎣
Fig. 1. The planar 5R parallel mechanism ⎡ x cos θ1 + y sin θ1 − l1 l2 0 0 0 ⎤
[B] = ⎢ 0 0 l3 x cos θ 4 + y sin θ 4 − l4 − l5 cos θ 4 x − l4 cos θ 4 − l5 ⎥⎦
⎣
A kinematic model of the planar 5R parallel
mechanism ABCDE is developed as shown in Fig. 1. A ⎡ yl cos θ1 − xl1 sin θ1 0 ⎤
[C] = ⎢ 1 ⎥
fixed global reference system A-xy is located at point A ⎣ 0 ⎦
with the y-axis normal to AE and the x-axis directed If the input angular displacement is assumed to be
along AE. The dimension of each link is denoted as li generated exactly, (i.e. Δθ1=Δθ4=0), the position errors of
(i=1~5). The position of the output point C on link BC in output point C can be calculated approximately using
the reference system A-xy can be described by the [ Δc] = [ A ] [B][Δl] = [S][Δl]
−1
(9)
position vector c0, and there is
where [S] = [ A ]−1 [B] ∈ R 2×5 is the sensitivity matrix,
c0 = ( x y )T (1)
describing how deviations in the link dimensions affect
In the reference frame A-xy, the position vectors of the position errors.
point B and D can be written as The worst-case position errors can be found using the
b 0 = (l1 cos θ1 l1 sin θ1 ) T (2) following equations
d 0 = (l4 cos θ 4 + l5 l4 sin θ 4 ) T
(3) ⎧ 5
where θ1 and θ4 are two actuated joint variables measured ⎪Δx = ∑ S1i ⋅ t i (10)
⎪ i =1
counterclockwise from the positive x-axis. The inverse ⎨ 5
⎪ Δy =
kinematic problem then can be solved by following ⎪⎩ ∑
i =1
S 2i ⋅ t i
constraint equations
c 0 − b 0 = l2 (4) where ti is the tolerance associated with the link
dimension li.
c 0 − d 0 = l3 (5)
2
13th World Congress in Mechanism and Machine Science, Guanajuato, México, 19-25 June, 2011 A11_306
⎪ i =1 (11)
⎨
⎪ 5
⎪Δy = ∑ (S 2i ⋅ t i )
2
⎩ i =1 A′
In order to verify Eq. (10) and (11) for use in
determining position errors of planar 5R parallel
Fig. 3. Clearance link at active joint
mechanism, it can be compared with vertex analysis
method [8] and Monte Carlo method [9]. The vertex If only taking account of the clearance link at active
analysis method uses all combinations of the extreme joint A, as shown in Fig. 3, the angle between link AE
tolerance of link dimension. The Monte Carlo method and A′B remains θ1 when crank AA ′ rotates around the
uses repeated random sampling on link dimension point A, the position vectors of point C can be written as
deviation to evaluate its effect on position errors.
⎡ r cos α1 + l1 cos θ1 + l2 cos β1 ⎤ (12)
Different random number distributions (normal, uniform, c1 = ⎢ 1 ⎥
trapezoidal, triangular, et al.) can be defined for the ⎣ r1 sin α1 + l1 sin θ1 + l2 sin β1 ⎦
toleranced parameters. where α 1 is the input angle of crank AA′ , β1 is the
output angle of link BC which can be solved by
IV. Tolerance Analysis of Joint Clearance following constraint equation
The joint clearance can be idealized as the difference c1 − d 0 = l3 (13)
of the diameters of the pin and hole a joint. Assuming Substituting Eq. (3) and (13) into Eq. (14) yields
constant contact between the pin and hole of a joint, a ⎛ − e + e 2 − 4d f ⎞
joint clearance may be modeled as a small link (called β1 = 2 tan −1 ⎜ 1 1 1 1 ⎟ (14)
clearance link) with the length ri (i=1~5) equal to one ⎜ 2d 1 ⎟
⎝ ⎠
half of the joint clearance. Thus, a 2-DOF five-bar where
linkage would become an equivalent 7-DOF ten-bar
linkage (Fig. 2). ⎧(l1 cos θ1 + r1 cos α1 − l4 cos θ 4 − l5 ) 2 + ⎫
⎪ ⎪
d1 = ⎨(l1 sin θ1 + r1 sin α1 − l4 sin θ 4 ) 2 + l22 − l32 ⎬
⎪−2(l cos θ + r cos α − l cos θ − l )l ⎪
C′ ⎩ 1 1 1 1 4 4 5 2⎭
3
13th World Congress in Mechanism and Machine Science, Guanajuato, México, 19-25 June, 2011 A11_306
C′
V. Numerical Example
For example, if l1=l4=1.2(m), l2=l3=1.0(m), l5=1.6(m), Fig. 6. Deviation sensitivities of link dimensions
to generate a circular trajectory, the coordinates of the
point C can be defined by the following parametric i min(S1i) max(S1i) min(S2i) max(S2i)
equations 1 -0.0407 0.5468 -0.3304 0.6288
⎧ x = 0.8 + 0.3 cos φ (20) 2 0.4346 0.7046 0.7097 4.3227
⎨ φ ∈ [0,2π ]
⎩ y = 1.6 + 0.3 sin φ 3 -0.7046 -0.4346 0.7097 4.3227
The history of actuated joint variables θ1 and θ4 4 -0.5468 0.0407 -0.3304 0.6288
corresponding to the parameter φ are given in Fig. 5. 5 0.4249 0.5751 -4.2937 -0.5054
TABLE 1. Values of deviation sensitivities
4
13th World Congress in Mechanism and Machine Science, Guanajuato, México, 19-25 June, 2011 A11_306
References
[1] J.-P. Merlet. Les robots parallèles. Hermès Science, Paris, 1997.
[2] M. Vukobratović, B. Borovac. Accuracy of the robot positioning
and orientation assessed via its manufacturing tolerances.
Mechanism and Machine Theory, 30(1): 11-32, January 1995.
[3] K. W. Chase, J. Gao, S. P. Magleby, C. D. Sorensen. Including
Fig. 7. Position deviation intervals due to joint clearances geometric feature variations in tolerance analysis of mechanical
assemblies. IIE Transactions, 28(10): 795-807, 1996.
[4] K.-L. Ting, J. Zhu, D. Watkins. The effects of joint clearance on
i min(xi1) max(xi1) min(xi2) max(xi2) position and orientation deviation of linkages and manipulators.
1 -7.0459 -4.3455 4.3456 7.0460 Mechanism and Machine Theory, 35(3): 391-401, March 2001.
[5] S. Venazi, V. Parenti-Castelli. A new technique for clearance
2 -7.0459 -4.3455 4.3456 7.0460 influence analysis in spatial mechanisms. ASME Journal of
3 -7.0460 -4.3456 4.3455 7.0459 Mechanical Design, 127(3): 446-455, May 2005.
4 -7.0459 -4.3456 4.3455 7.0459 [6] A. Frisoli, M. Solazzi, M. Bergamasco. A new method for the
estimation of position accuracy in parallel manipulators with joint
5 -7.0459 -4.3456 4.3455 7.0459 clearances by screw theory. In IEEE Int. Conf. on Robotics and
i min(yi1) max(yi1) min(yi2) max(yi2) Automation, pages 837-844, Pasadena, USA, May 19-23, 2008.
1 -43.234 -7.0972 7.0971 43.218 [7] X.-J. Liu, J. S. Wang, G. Pritschow. Kinematics, singularity and
workspace of planar 5R symmetrical parallel mechanisms.
2 -43.234 -7.0972 7.0971 43.218 Mechanism and Machine Theory, 41(2): 145-169, February 2006.
3 -43.234 -7.0972 7.0971 43.218 [8] W. L. Cleghorn, R. G. Fenton, J.-F. Fu. Optimum tolerancing of
4 -43.234 -7.0972 7.0971 43.218 planar mechanisms based on an error sensitivity analysis. ASME
5 -43.234 -7.0972 7.0971 43.218 Journal of Mechanical Design, 115: 306-313, June 1993.
[9] J. Stuppy, H. Meerkamm. Tolerance analysis of a crank mechanism
by taking into account different kinds of deviation. In 11th CIRP
TABLE 2. Minimal and maximal position deviations (μm) Int. Conf. on Computer Aided Tolerancing, Annecy, France, March
26-27, 2009.