A new approach towards the synthesis of six-bar double dwell mechanisms

Mohan Jagannath and Sandipan Bandyopadhyay

Abstract This paper presents a new formulation for the synthesis of planar six-bar linkages with only rotary joints showing two dwells in each cycle of the input crank. The formulation combines the concepts of instantaneous kinematics and optimisation to produce a simple, one-step synthesis method. The method is illustrated with a numerical example, where it is seen to produce better results in terms of accuracy and computational eciency with respect to reported works.

1 Introduction
The study of the dwell mechanisms has a long history [1, 2]. While dwells can be easily achieved by means of a cam mechanism, the search for a suitable linkage with only rotary pairs continues to attract researchers in the mechanisms community. Early solutions to this problem was based upon the concept of coupler curve synthesis, where in one would rst synthesise a four-bar mechanism having approximately1 circular arc(s) in its coupler curve. Then, the coupler point would be linked to a xed base through an RR dyad to generate a six-bar mechanism. Of this dyad, the rst link would have a length equal to the radius of curvature of the circular portion, and the other link would show dwell when the coupler point traversed the circular portion. For multiple dwells there is the additional constraint that all the corresponding
Mohan Jagannath Indian Institute of Technology [email protected] Madras, Chennai 600036, India, e-mail:

Sandipan Bandyopadhyay Indian Institute of Technology Madras, Chennai 600036, India, e-mail: [email protected]
1 It is well known that the coupler curve of a planar four-bar mechanism with all rotary joints cannot have an exactly circular arc of nite arc length (see, e.g. [3]).

Mohan Jagannath and Sandipan Bandyopadhyay

the circular arcs on the coupler curve should have the same radius of curvature. Finding out a suitable coupler curve within the atlas by Hrones and Nelson [4] has been a tedious but straight-forward approach popular among the early practitioners and researchers. However, more recently, the use of optimisation techniques has been more common for solving this problem as well as many others in the domain of mechanism synthesis [5, 6]. In general, the optimisation approaches to this problem has three stages: specication of the coupler curve with the desired properties, the subsequent identication of a four-bar mechanism whose coupler curve matches the prescribed one, and extension of the four-bar to an appropriate six-bar [5]. This approach has the advantage of being simple and intuitive. However, one can easily identify several shortcomings of it as explained below. In most of the cases, the entire coupler curve is prescribed, while only some parts of it are of actual interest. The formulation is over-constrained in this sense. Dwell is essentially an instantaneous phenomena having to do with the relative motion of the output with respect to the input, and as such it does not depend on the position and orientation of the entire mechanism with respect to some global frame, or the scale of the mechanism. However, since the coupler curve has to be specied in a xed global frame, the formulation needs to carry the overhead of as many as four non-contributing variables in the design process (see section 2 for details). The decomposition of the synthesis process in the stages mentioned above reduces the eectivity of some of the design variables. For instance, the link-lengths of the RR dyad and the location of third xed pivot are of secondary importance since they cannot oset by any means the deviation from circularity of the coupler curve generated by the four-bar. The coordination of the dwell with the crank motion needs to be specied additionally. In this paper, we adopt a direct approach which addresses the above issues. By denition, a link in a mechanism is said to dwell when its velocity coecient with respect to the input is zero (see, e.g. [7]). Therefore, we propose a formulation for the synthesis of the six-bar in its entirety where the velocity coecient of the output link is approximately zero over two prescribed intervals of the input cranks motion. The resulting optimisation problem is seen to yield results which seem to fare better than those reported in recent papers, such as [5]. The approach is simple, easily implementable, and requires much less computational eort than other reported works. The paper is organised as follows: Section 2 deals with the mathematical formulation of the present problem. Section 3 illustrates the same by means of a numerical example adopted from [5] and presents a discussion of the results. Section 4 summarises the conclusions of the paper.

A new approach towards the synthesis of six-bar double dwell mechanisms

2 Formulation
In this section, we describe the formulation of the optimisation problem very briey. The mathematical development and the condition of dwell follows [7].

2.1 Kinematics
Fig. 1 shows a planar six-bar mechanism. The input link is l1 and the output

Y l4 a l2 5 4 2 3 l0 o2 X l5 o3 (o , o ) 3x 3y

l1 1 o1

l3

Fig. 1 Schematic of the planar six-bar double dwell mechanism

is l5 , which is to show dwell for two specied intervals of the motion of link 1. To identify the condition for dwell, we obtain the velocity coecient of the output with respect to the input symbolically, and set it to zero. The required velocity coecient is obtained in the following steps: Form the loop-closure equations in the form (1 , ) = 0, where = (2 , 3 , 4 , 5 )T (see Fig. 1 and the Appendix). Dierentiate the loop-closure equation with respect to time to obtain the following: =0 1 + 1 (1)

For a non-singular conguration (i.e. det( ) = 0), express the joint rates of the non-actuated links in terms of the actuation rate 1 :

Mohan Jagannath and Sandipan Bandyopadhyay

1 = K v 1 , 1

Kv =

(2)

where K v = (2 /1 , 3 /1 , 4 /1 , 5 /1 )T is the vector of velocity coecients. We are interested in the last element of K v , corresponding to the link 5, which is found as: Kv5 = l1 (a sin(1 + 3 ) sin( + 2 4 ) l2 sin(2 + 3 ) sin(1 4 )) l2 l5 sin(2 + 3 ) sin(4 + 5 ) (3)

2.2 Formulation of the optimisation problem

2.2.1 Design variables The variables that actively inuence this problem are the parameters dening the link geometry, i.e. l0 , l1 , l2 , l3 , l4 , l5 , a, , and the location of the third xed pivot, given by (O3x , O3y ). Note that the global location of the linkage has been omitted by setting the location of the pivot O1 at the origin. Likewise, the orientation of the base has been omitted by aligning it to the X axis. This process eliminates three variables from the formulation without any loss of generality. Further, since the scale of the mechanism does not inuence the relative motions between the links, we eliminate one more variable by setting l1 = 1 and scaling all other linear dimensions accordingly. Therefore, the nal set of design variables is: x = (l0 , l2 , l3 , l4 , l5 , a, O3x , O3y , )T .

2.2.2 Objective function The expression for Kv5 holds only instantaneously. To set it to approximately zero over nite spans of 1 , we evaluate Kv5 over a number of points2 in each span (denoted by N1 and N2 ) respectively), and set the mean of the squared deviation to zero. Thus the objective function to be minimised is:
2

For the sake of simplicity, we have taken the points to be uniformly distributed. However, any other suitable distribution can also be used.

A new approach towards the synthesis of six-bar double dwell mechanisms

1 E= N1

N1 2 Kv (1i ) 5 i=1

1 + N2

N2 2 Kv (1j )
5

(4)

j=1

It may be noted that Kv5 is not available symbolically as a function of 1 as depicted in equation (4). It can, nevertheless, be calculated numerically by solving rst the non-actuated variables, i , from the loop-closure equations and then using equation (3). As shown in Fig. 2, there are four sets of values
Branch 1 2 0 2 0 2 4 6 2 0 2 0 2 4 6 Branch 2

Branch 3 1 0 1 2 3 4 0 2 4 6 1 0 1 2 3 4 0

Branch 4

Fig. 2 Four possible congurations of the planar six-bar mechanism

of possible for any given 1 , and therefore, the designer needs to choose the appropriate conguration at this point. This gives the designer the exibility to incorporate the over-all packaging considerations into the synthesis process, and obviates the possibility of branch error as well.

2.2.3 Design constraints The design constraints arise from several considerations: Positive link lengths: l0 , l2 , l3 , l4 , l5 , a > 0. Full cycle mobility of link 1: This requires that the four-bar consisting of links 0,1,2,3 is a Grashof mechanism with link 1 as the smallest in the same: 2(max(l0 , 1, l2 , l3 ) + min(l0 , 1, l2 , l3 )) < l0 + 1 + l2 + l3 , and l0 , l2 , l3 > 1. Packaging considerations: To ensure a compact envelope of all the moving parts and from manufacturing considerations, it is required that no link in

Mohan Jagannath and Sandipan Bandyopadhyay

the mechanism is very long and no other is very short. This is enforced in this case via the following constraint: max(l0 , 1, l2 , l3 , l4 , l5 , a)/min(l0 , 1, l2 , l3 , l4 , l5 , a) where the numerical value of can be chosen as appropriate. We use = 5 in this paper. Location of the third xed pivot, O3 : The designer has the exibility of suggesting a desirable location of the third pivot. In this case, however, that can be done directly through the specication of the lower and upper bounds of the variables O3x and O3y .

3 Illustrative example
In this section, we demonstrate the above formulation with a numerical example. For the sake of comparison, we adopt the same problem denition as in a recent contribution [5]. The task is to obtain dwells during 1 [15 , 15 ] and 1 [160 , 220 ]. We have implemented the above formulation in computing language and also performed the optimisation using the built-in non-linear constrained minimisation routine . Apart from the specication of the objective and constraint functions, the upper and lower bounds on the variables and their initial (guessed) values are required by this routine. The values used by us are shown in Table 1. We use N1 = 100 and N2 = 400 in equation (4) for
Table 1 Bounds on the design variables and the initial guesses Variable Lower bound Upper bound Initial guess l0 1 5 2 l2 1 5 2 l3 1 5 2 l4 0 5 5 l5 0 5 3 a O3x 0 l0 5 (l0 + 2) 3 4 O3y 2 60 1 40 1 45

computing the objective function, and select the solution branch three, as shown in Fig. 2, as the preferred branch for its compactness. For these set of inputs, the optimisation routine is seen to converge in 23 iterations to the results shown in Table 2. Fig. 3 shows the variation of the output angle as a function of the input. The dwell portions have been marked, and it can be seen clearly that the output curve is nearly at in those regions, indicating a

Table 2 Optimal values of the design variables Variable l0 l2 l3 l4 l5 a O3x O3y Optimal value 1.9161 1.9052 2.0145 4.9809 2.5580 2.9916 4.0000 1.4199 43.9287

A new approach towards the synthesis of six-bar double dwell mechanisms

good approximation to dwell. However, the plot of Kv5 vs. 1 gives a better

66 64 62 60 58 56 54 100

Dwell period (15 to 15 )

(degrees)

Dwell period (160 to 220 )

100 200 (degrees)

300

400

Fig. 3 Variation of the output angle 5 vs. the input crank angle 1

quantitative description of the performance of the mechanism. In order to benchmark our results, in Fig. 4 we plot Kv5 computed using the results obtained by us along with the one obtained using the best among the results produced in [5]. The plots show that the present method produces more ac0.03

0.02

0.01
Dwell period (15 to 15 )
0 0 5

Kv

0
Dwell period (160 to 220 )
0 0

0.01

0.02
Calculated from [5] Obtained results

0.03 50

50

100 1 (degrees)

150

200

250

Fig. 4 Variation of the velocity coecient Kv5 vs. 1 in the dwell regions

Mohan Jagannath and Sandipan Bandyopadhyay

Table 3 Comparison of the obtained results with those in reference [5] Dwell period Deviations max(|Kv5 (1 )|) Total variation of 5 max(|Kv5 (1 )|) Calculated from [5] Present work 0.0295 0.1754 0.0211 0.5215 53310 0.0141 0.0441 0.0061 0.0846 241

15 to 15 160 to 220

Total variation of 5 Total number of function evaluations

curate dwells over the same intervals. A comparison of accuracy as well as computational eorts involved is given in Table 3. It may be noted here that in our formulation, the values of 5 at which the dwells occur have not been specied. We believe that keeping this open imparts a certain exibility to the optimisation process, allowing it to obtain better results. In case two specic values are required in an application, they can be easily obtained at the cost of the additional complexity of incorporating two mating gears, one driven by link 5 and the other driving the output link, set at a desired angular oset with link 5. Optionally, it is also possible to modify the formulation to incorporate these constraints directly.

4 Conclusion
This paper presents a simple formulation for the optimal synthesis of planar double dwell mechanisms based on six-bar linkages with only rotary joints. The formulation is based entirely on the instantaneous kinematics of the mechanism, as opposed to an intermediate coupler-curve synthesis approach used commonly in literature. The formulation is illustrated with a numerical example which yields better results than a recent contribution to the relevant literature, and requires much lesser number of function evaluations as well. The novel idea of using the velocity coecients directly in synthesis can easily be used in synthesising a host of other mechanisms.

A new approach towards the synthesis of six-bar double dwell mechanisms

Appendix
Loop-closure equations of the six-bar mechanism (see Fig. 1): l1 cos 1 + l2 cos 2 + l3 cos 3 l0 = 0 l1 sin 1 + l2 sin 2 l3 sin 3 = 0 l1 cos 1 + a cos( + 2 ) + l4 cos 4 + l5 cos 5 O3x = 0 l1 sin 1 + a sin( + 2 ) + l4 sin 4 l5 sin 5 O3y = 0

References
1. E. Sandgren, Design of single- and multiple-dwell six-link mechanisms through design optimization, Mechanism and Machine Theory, vol. 20, no. 6, pp. 483490, 1985. 2. W. C. Barris, S. Kota, D. R. Riley, and A. G. Erdman, Mechanism synthesis using the workstation environment, Computer Graphics and Applications, IEEE, vol. 8, pp. 39 50, March 1988. 3. A. K. Mallik, A. Ghosh, and G. Dittrich, Kinematic Analysis and Synthesis of Mechanisms. CRC-Press, 1994. 4. J. Hrones and G. Nelson, Analysis of the Four Bar Linkage. MIT Press and Wiley, 1951. 5. Shiakolas, Koladiya, and Kebrle, On the optimum synthesis of six-bar linkages using dierential evolution and the geometric centroid of precision positions technique, Mechanism and Machine Theory, vol. 40, pp. 319335, September 2005. 6. N. Nariman-Zadeh, M. Felezi, A. Jamali, and M. Ganji, Pareto optimal synthesis of four-bar mechanisms for path generation, Mechanism and Machine Theory, vol. 44, pp. 180191, January 2009. 7. S. Bandyopadhyay and A. Ghosal, Analysis of conguration space singularities of closed-loop mechanisms and parallel manipulators, Mechanism and Machine Theory, vol. 39, pp. 519544, May 2004.