Proceedings of the RAAD03, 12th International Workshop on Robotics in Alpe-Adria-Danube Region Cassino, May 7-10, 2003 MECHANICAL EFFICIENCY

ANALYSIS OF A CARDAN JOINT WITH MANUFACTURING TOLERANCES

E. Pennestr` , L. Vita Dipartimento di Ingegneria Meccanica Universit` a degli Studi di Roma Tor Vergata via del Politecnico, 1, 00133 Roma, Italy

Introduction
Cardan joints are common devices for transmitting the motion between misaligned intersecting axes. Their capability of easy mounting, of resisting high loads and commercial availability makes them an attractive solution, as a substitute of spherical pairs, in parallel robots. Although their structure has been known for centuries, only recently a complete dynamic analysis has been presented in a series of papers authored by F. Freudenstein and his coworkers [1, 2, 3, 4]. In the mentioned references friction is not included. Likely the rst scientic contribution on the mechanical efciency of Cardan joints is due to Morecki [5]. This model, based on a simplied static analysis and including the losses in the yoke bearings only, has been veried with a different analytical approach and rened by including also the losses in the xed bearings. The results have been plotted in a design chart which allow to compute the efciency as a function of the angle between input-output shaft axes [8]. The modeling of manufacturing errors in Cardan joints is introduced by considering a kinematically equivalent RCCC mechanism. In this paper a kinematic and static analysis of the RCCC mechanism by means of the dual numbers algebra is carried out rst. Then, the effects of friction are included. For this purpose, the following hypotheses are adopted: - Coulomb friction; - absence of stiction; - negligible inertia forces; - absence of backlash in the kinematic pairs; 1

- rigid bodies. As a main contribution, it is herein presented a model for computing the mechanical efciency of this mechanism. To the best of the authors knowledge, in the scientic literature is not addressed the problem of computing the mechanical efciency of Cardan joints in presence of manufacturing errors.

zi

i Link i

zi+1

Link i-1 -si

i xi ai

y i+1 xi+1

Figure 1: Denavit-Hartenberg parameters

Nomenclature
- ai : minimum distance between zi and zi+1 axes; - D: distance between the bearings of the cross; - di : diameter of shaft i; - f : friction coefcient; - Fix ,Fiy ,Fiz joint forces cartesian components at the ith joint;

- Fix ,Fiy ,Fiz joint forces at the ith joint; - xi yi zi : moving cartesian system attached to the ith body, as in the Denavit-Hartenberg convention; - si : relative linear displacement of the links measured according to the Denavit-Hartenberg convention (see Figure 1); - i angle between zi and zi+1 axes; - dual unity 2 = 0 - mechanical efciency of the Cardan jojnt; - i : angular velocity of the ith body, measured in the cartesian system o xiyi zi ; - i : relative angular displacement of the links measured according to the Denavit-Hartenberg convention (see Figure 1); - f : frictional torque at the ith revolute joint; - the denote dual quantities; - Dots denote differentiation w.r.t. time.
(i)

where [I ] is the identity matrix. The above equation can be rewritten in the form
2 3 1 2 1 4

= AT

AT

(5)

Carrying out the matrix products and equating the elements on the same rows and columns one obtains [4]: D sin 4 + E cos 4 + F = 0 , where D = s1 s3 s1 , E = s3 c1 s4 + s1 c4 c1 , , (7) (8) (9) (6)

F = c2 + c3 c1 c4 s1 s4 c1

and

Kinematic analysis of the RCCC linkage

Let us denote with i = i + si , i = i + ai , (1) (2)

s2 = c2 = s3 =

s1 c3 s4 + s3 c4 c4 + s3 c1 s4 s2 c1 c2 c3 c4 + s3 s4 c4 , s1 s2 s1 s1 c4 + c4 c1 s4 + c1 s4 s4

, (10) (11)

, (12) s2 the dual numbers which dene, respectively, the relative position between adjacent links and the geometry of the s1 s4 c1 + c2 c3 c1 c4 . (13) c3 = ith link. s2 s3 With reference to Figure 1, the transform matrix from coordinate system oi+1 xi+1 yi+1 zi+1 to oi xi yi zi , Thus, the dual angles 2 , 3 and 4 are computed as folin terms of such numbers is given by1 2: lows ci ci si si si i = si ci ci si ci . (3) A i+1 D D2 + E 2 F 2 0 si ci , (14) 4 = 2 tan1 F E The closure condition for the RCCC mechanism shown 2 = ATAN2 sin 2 , cos 2 , (15) in Figure 2 is expressed by the matrix product
1 2 3 1

A
1 c = cos 2 ATAN2

= [I ] ,

(4)

3 = ATAN2 sin 3 , cos 3

(16)

and s = sin functions with dual numbers as arguments can be computed by means of the procedure presented in the Appendix.

.
2 z2 x2 s4 y2 s2

Fx2 =

s2 c2 s1 s2 s2

Fz1 Fz2 + Fz3 s2 s2 (22)

s1 c2 + s2 c2 c1 s1 s2 s2

. .
4 z4 y 4

x4 a1 a4 a2 a3 s3

Fy2 =

Fk1 c1 Fz2 s1 s1

(23)

.
y 3 3

x1 z1

z3 x3

Fx3 =

s2 s1 s2 s2

Fz1 Fz2 + c2 s2 s2 Fz3 (24)

.
1

s1 c2 c2 + s2 c1 s1 s2 s2

Figure 2: The RCCC kinematically equivalent linkage Fy3 =

Static analysis of the RCCC linkage

Let us denote with the dual numbers Fx j = Fx j + Mx j , Fy j = Fy j + My j , Fz j = Fz j + Mz j , (17) (18) (19) s2 s2 Fx4 = c3 s1 s2 Fz1

Fz2 c2 Fz3 s2 s2

(25)

c3 c2 s1 c2 + c3 s2 c1 s3 s1 s2 s1 s2 s2 s3 s2 c2 c3 c2 Fz3

Fz2 (26)

for j = 1, 2, 3, 4, the joint forces. Imposing the static equilibrium of the links one obtains3 : Fx1 = + + c1 s2 c2 s2 s2 c1 s1 s1 s2 s2 s1 s2 s2 c1 s2 s2 Fz3 Fz1 Fz2 (20) Fz1 Fz2 (21)

Fy4 = +

s2 c3 s3 s1 s2 s2

Fz1

(27) Fz2

s2 s1 s2 c1 c2 c1 s2 c1 c2 s1

c3 s3 c2 s1 c2 + c3 s3 s2 c1 + c3 c3 s1 s2 s1 s2 s2 c3 s3 c2 + c3 c3 s2 c2 s3 s2 s2 s2 s2 s3 s3 s1 s2 Fz1 Fz3

(28)

Fy1 = +

c1 s2 c1 s2 + s1 s2 c2 s1 s2 s2 s1 s2 s2

Fz4 = +

s1 s2 c2 c1 + s1 s1 c2 + s2 s2 c1 s1 s2 s2

c2 s3 s1 c2 s3 + s2 c3 s1 s3 + s3 c1 s2 s3 s1 s2 s2 c2 s3 s3 + s2 c3 c2 s3 + s2 s2 c3 s2 s2 Fz3

Fz2

Fz3

(29)

3 The authors found that many of the algebraic expressions for the static force analysis of the RCCC linkage presented in [4] have some misprints. For this reason, all the equations have been deduced again and reported (hopefully) correct in this paper.

The modeling of friction in the kinematic pairs of a Cardan joint

In the ideal Cardan joint, the resultant of the reaction force at the kinematic pairs is zero and there is not any displacement along the joint axes. This is not necessarily true in presence of manufacturing errors. In revolute joints s i = 0, thus the effects of friction on the reaction force component Fiz are herein neglected. However, these can be taken into account when detailed informations on the geometry of the revolute joint are available.

Revolute pairs
The frictional forces at the ith revolute joint arise from two sources: - reaction forces Fxi , Fyi and Fzi ; - reaction moments Mxi and Myi The resistant action about the z axis, arising from the reaction moment Mxi and Myi , can be modeled [6, 7, 8] according to the scheme presented in Figure 3. In particular, for our purposes, the torque Mix is substituted by two parallel and opposite forces F acting normally to the revolute joint axis. Because of the presence of friction, these forces generate the frictional torque xi f = f di Mxi , Li (30)

Figure 3: Modeling of friction in the revolute pairs It is clear that the presence of friction alter the equilibrium of the links. However, because of the high efciency values this effect is not herein considered.

Cylindric pairs
In the frictionless cylindric pair s i = 0, Mzi = 0 and Fzi = 0. When friction is considered the values of Mzi and Fzi must be computed. However, the computation of Mzi is carried out with the same model adopted for the revolute pair. Thus, the following substitutions are required
2 + M2 Mxi yi

where di is the diameter of the journal bearing, Li the distance between bearing supports4 and f the friction coefcient. Similarly, the torque Myi generate the frictional torque di yi Myi . (31) f = f Li Therefore the friction reaction torque component along the z axis is computed by letting i Mzi = sign where f = f
4 For

Fzi = sign (s i ) f i Mzi = sign

(i)

2 + F2 + 2 Fxi yi

Li

(34) (35)

f +

f di 2

2 + F2 Fxi yi

1 Numerical example

f +

(i)

f di 2

2 + F2 Fxi yi

(i)

di Li

2 + M2 . Mxi yi

The average efciency m of a Cardan joint, according to the model described in [8] can be obtained using the (32) chart of Figure 5, where a = 2 f di /(Li ) is an adimensional parameter introduced by A. Morecki [5]. In the mentioned model, the Cardan joint has no manufacturing tolerances and the energy losses are computed in(33) cluding all the four revolute joints.

a single support bearing, Li represents the length of the bearing.

1.000

=2
0.975

1 = 2 = 3 = 90 , 4 = 150 , Li =50, di =40, (i = 2, 3, 4), f =0.05.

=4
0.950

=6

0.925 0.900 0.875 0.850 0.825

=8

Conclusions
It has been proposed a formulation for computing the mechanical efciency of a Cardan joint with manufacturing tolerances or mounting errors. The model foresee an increase of energy losses due to presence of axes offsets. These losses can be explained because of the reciprocating relative motion between adjacent links along the shaft axes. In the ideal Cardan joint such motion 0.24 is absent. Current investigations carried out by this research unit include experimental tests and and theoretical of inertia effects and operating parameters .

=10

=1
=
=
24

20

=1

30

0.800 0.775 0.750 0.02

0.04

0.06

0.08

0.10

0.12

0.14

0.16

0.18

0.20

0.22

a=2 f d/(L)

Figure 4: Average efciency in a Cardan joint without manufacturing tolerances [8]

References
[1] Fischer, I., Freudenstein, F., Internal Force and Moment Transmission in a Cardan Joint with Manufacturing Tolerances, ASME Journal of Mechanisms, Transmissions and Automation in Design, vol.106, December 1984, pp.301-311. [2] Chen, C.K., Freudenstein, F., Dynamic Analysis of a Universal Joint with Manufacturing Tolerances, ASME Journal of Mechanisms, Transmissions and Automation in Design, vol.108, December 1986, pp.524-532.
0 1 2

0.070

With manufacturing tolerances No manufacturing tolerances

Efficiency loss (1-)

0.065

0.060

0.055

0.050

0.045

1 (rad)

[3] Freudenstein, F., Macey, J.P., The Inertia Torques of the Hooke Joint, Proc. Of the 21st Biennial ASME Mechanisms Conference, Chicago, September 16-19, 1990, DE-Vol.24, pp.407-413. [4] Fischer, I., Dual-Number Methods in Kinematics, Statics and Dynamics, CRC Press, Boca Raton, FL, 1999. [5] Duditza, F., Transmissions par Cardan, Editions Eyrolles, Paris, 1971. [6] Shih, C.W., Shih, M. Y., Haug, E.J., Dynamics of Mechanical Systems with Coulomb Friction, Stiction, Impact and Constraint Deletion - III, Mechanism and Machine Theory, vol.21, 1986, pp.417425. [7] Dhanaraj, C., Sharan, A.M., Efcient Modeling of Rigid Link Body Dynamic Problems with Friction, 5

Figure 5: Instantaneous efciency loss in a Cardan joint with and without manufacturing tolerances The present analysis compares the instantaneous efciency loss (1 ) for two cases: No manufacturing tolerances (i.e. a j = 0, ( j = 1, 2, 3, 4). Presence of manufacturing tolerances, estimated in the following values of axes offsets a1 = a2 = a3 = a4 = 0.5. Angular errors on j could be included as well. The remaining geometric parameters are as follows:

Mechanism and Machine Theory, vol.30, 1995, if ((abs(x1)<eps).AND.(y1>0.)) then C%pr=pi/2.;C%du=0. pp.749-764. return [8] Biancolini, M.E, Brutti, C., Pennestr` , E., Valen- end if tini, P.P., Dynamic, Mechanical Efciency and Fa- if ((abs(x1)<(eps)).AND.(y1<0.)) then tigue Analysis of the Double Cardan Homokinetic C%pr=3*pi/2.;C%du=0. Joint, to appear in International Journal of Vehicle return Design. end if [9] Fasse, D.,E., Some Applications of Screw Theory to Lumped Parameter Modeling of Visco-Elastic ! Quadrant n.1 Coupled Rigid Bodies, Proceedings of the Sympo- if ((x1>0.).AND.(y1>0.)) then sium Commemorating the Legacy Works and Life C%pr=atan(x);C%du=y/(1+x**2) of Sir Robert Stawell Ball, University of Cam- return end if bridge, Trinity College, July, 9-11, 2000. ! Quadrant n.2 and 3 if (x1<0.) then C%pr=atan(x)+pi;C%du=y/(1+x**2) Fortran procedure for computing the ATAN2 function return with dual numbers as arguments. This procedure re- end if quires those in the module dualnumbers imple- ! Quadrant n.4 mented by D.E. Fasse [9]. if ((x1>0).AND.(y1<0)) then C%pr=atan(x)+2*pi;C%du=y/(1+x**2) function dual_atan2(A,B) result(C) return type(dual),intent(in) :: A,B end if type(dual):: C

Appendix

type(dual)::A1,B1 real*8 :: eps,x1,y1,x,y eps=1.d-10 A1=A;B1=B x=A1%pr/B1%pr y=(A1%du*B1%pr-A1%pr*B1%du)/& (B1%pr*B1%pr) x1=B1%pr;y1=A1%pr

end function dual_atan2

Acknowledgement
The authors gratefully acknowledge the nancial support of the Ministry of Public Instruction through a COFIN grant.