Mechanism and Machine Theory 93 (2015) 127–146

Contents lists available at ScienceDirect

Mechanism and Machine Theory

journal homepage: www.elsevier.com/locate/mechmt

Optimal dynamic balancing and shape synthesis of links in

planar mechanisms
Kailash Chaudhary ⁎, Himanshu Chaudhary
Department of Mechanical Engineering, Malaviya National Institute of Technology Jaipur, JLN Marg, Jaipur 302017, India

a r t i c l e i n f o a b s t r a c t

Article history: This paper presents a two stage optimization procedure for dynamic balancing of planar mecha-
Received 21 November 2014 nisms and finding optimum link shapes. In the first stage, the shaking force and shaking moment
Received in revised form 2 July 2015 are minimized by optimizing mass distribution of links using the equimomental system of point-
Accepted 8 July 2015
masses for each link. Then for the optimum inertial parameters of the balanced mechanism, the
Available online 25 July 2015
optimum links shapes are synthesized systematically using closed parametric curve such as
cubic B-spline in the second stage. The control points of cubic B-spline curve are taken as the
Keywords: design variables for link shape formation to minimize the percentage error in the resulting link
Dynamic balancing
inertia values. The constraints on design variables are defined for both symmetrical and non-
Equimomental system
symmetrical shapes in the optimization problem formulation. The proposed method of balancing
Link design
Optimization and shape synthesis can be applied to any planar single and multiloop mechanism with revolute
Planar mechanisms as well as prismatic joints. Its effectiveness is demonstrated by applying it to four-bar, five-bar, six-
bar and slider-crank mechanisms.
© 2015 Elsevier Ltd. All rights reserved.

1. Introduction

The resultant inertial forces and moments of the moving links of a mechanism are termed as shaking force and shaking moment
[1]. These forces and moments transmitted to the frame result in vibrations, wear and noise which adversely affect the dynamic per-
formance of the mechanism. Several methods to reduce the shaking force and shaking moment based on various principles are
discussed in [2–4]. The complete force balancing can be achieved by making the total mass center of moving links stationary either
using mass redistribution [5,6] or by adding counterweights [7]. Force balancing and trajectory tracking are achieved in a five-bar
real-time controllable mechanism using adjusting kinematics parameter approach [8]. However, the full force balance generally in-
creases the other dynamic performance characteristics like shaking moment, driving torque and bearing reactions in the joints [9].
Similarly, the complete moment balancing needs elimination of the total angular momentum of the moving links. The complete
elimination of the total angular momentum using link mass distribution and/or adding counterweights is not possible [10]. Therefore,
normally the shaking moment is reduced by adding inertia or disk counterweights [11–13], duplicate mechanism [14] and moment
balancing idler loops [15] along with the full force balancing. Alternatively, shaking force and moment can be minimized by properly
designing and making the moving links of bi-material [16]. The effects of dynamic balancing on the elastodynamic properties of the
mechanisms are investigated in [17–21]. In all these methods, the overall mass is increased and mechanism becomes more complex.
Instead of complete balance of shaking force and shaking moment, some methods are developed to minimize them simultaneously
through optimization. The conventional optimization technique is used to optimally balance the planar four-bar mechanism [22,23]
and to analyze the sensitivity of shaking force and shaking moment to the design variables [24]. The shaking moment is minimized in

⁎ Corresponding author.
E-mail addresses: k.chaudhary.mech@gmail.com (K. Chaudhary), hchaudhary.mech@mnit.ac.in (H. Chaudhary).

http://dx.doi.org/10.1016/j.mechmachtheory.2015.07.006
0094-114X/© 2015 Elsevier Ltd. All rights reserved.
128 K. Chaudhary, H. Chaudhary / Mechanism and Machine Theory 93 (2015) 127–146

five-bar manipulator through constrained nonlinear optimization problem in which shaking force elimination is presented as the
balancing constraints [25]. An optimization procedure can be formulated to determine link masses, their locations of mass centers
(CMs) and moment of inertias that provide minimum shaking force and shaking moment. Such as a convex optimization technique
[26] is used to determine counterweight parameters to balance the planar mechanisms. Therefore, the mechanism balancing problem
may be formulated as multi-objective optimization problem and solved using evolutionary optimization techniques like particle
swarm optimization (PSO) [27] and genetic algorithm (GA) [28,29] considering design constraints depending upon application.
After obtaining the optimized inertial properties of links of the mechanism, their shapes must be decided to carry loads. One simple
way to form link shape is to discretize initial assumed shape into small mass elements and locate them systematically along the link
length [30]. The shape synthesis can also be done based on maximum work done by taking volume of all links as constraints [31]. The
topology optimization based on parametric curves [32] and non-intersecting closed polygons for given geometric constraints [33] are
some other methods to form shape.
In some other applications, the shaft shape for rotating machinery is optimized using Evolutionary Structural Optimization (ESO)
method which gradually removes the ineffectively used material from the design domain [34,35]. The simple geometric shapes of
mechanism links are determined for interference free motion by identifying feasible material domain associated with link geometries
[36]. However, these methods used for finding link shapes have limitations as (1) they do not consider the dynamic balance for
mechanisms and (2) require a pre-defined shape or design domain to start with.
This paper presents a two stage optimization method to synthesize link shapes to minimize the shaking force and shaking moment
in the planar mechanisms. In the first stage, the optimization problem for balancing is formulated by modeling rigid links of mecha-
nism as dynamically equivalent system of point-masses, known as equimomental system [37–40]. This multi-objective optimization
problem is formulated as single objective using appropriate weighting factors and solved using genetic algorithm.
In the second stage, shapes of the mechanism links are found for the optimum inertial properties resulted in the first stage. The link
geometries are modeled as closed parametric curves, i.e., cubic B-spline curve. The difference between desired optimum inertia value
and resulting link inertia value is minimized taking the positions of the control points of curve boundary as the design variables. Note
that genetic algorithm doesn't require initial values of the design variables to solve an optimization problem. Therefore initial shape or
design domain for links is not required in this method. The desired mass and location of mass centers of the links are considered as the
constraints in this stage. In each iteration of optimization, the boundary domain defined by parametric curves is evaluated to obtain
mass and inertia of each link through Green's theorem [41]. Thus the full force balancing and partial dynamic balancing are achieved
by synthesizing the link shapes.
The paper is organized as follows. Section 2 presents dynamics and link shape for dynamic balancing of the planar mechanisms.
The optimization problem formulation is presented in Section 3. In Section 4, four numerical examples are solved using the proposed
method and results are discussed. Finally, conclusions are summarized in Section 5. The three Appendices A, B and C are provided to
derive the equations of motion for rigid link moving in a plane, the equimomental point-mass system and definition of cubic B-spline
curve with corresponding inertial properties calculations, respectively.

2. Dynamics and link shape

2.1. Shaking force and shaking moment

Fig. 1 shows a six-link multiloop mechanism where the fixed link is detached from the moving links to show reactions. The shaking
force for such mechanism having n moving links is defined as the reaction of the vector sum of all the inertia forces whereas the
shaking moment is defined as the reaction of the resultant of the inertia moment and the moment of the inertia forces about a
fixed point. Thus, the shaking force and the shaking moment transmitted to the fixed link are given as [1]:
Xn

f sh ¼ − fi ; ð1Þ
i¼1

Xn 
 
nsh ¼ − ni ‐ri xf i ; ð2Þ
i¼1

where n⁎i and fi⁎ are the inertia moment and the 2-vector of inertia force in the plane of motion, respectively, acting about and at the
origin, Oi, of the ith link. Moreover, the 2-vector, ri, is the position of origin, Oi, from a fixed point which is the origin of fixed frame or
can be any other suitable point. As the shaking force and shaking moment have different units, they are normalized with respect to
any reference link of the mechanism as [42]:
 
2
f ¼ j f j= maω ; ð3Þ
 
2 2
n ¼ n= ma ω ; ð4Þ

where a and m are the length and mass of the reference link whereas ω is its angular velocity.
 K. Chaudhary, H. Chaudhary / Mechanism and Machine Theory 93 (2015) 127–146 129

Fig. 1. Definitions of various parameters for a multiloop mechanism.

2.2. Link shape synthesis

In this subsection, the link shape is approximated using parametric closed cubic B-spline curve as shown in Fig. 2. The curve inter-
polates or approximates a set of n + 1 control points, P0, P1,…, Pn [43,44] and defined in Eq. (5). Furthermore, the thickness of the link
is taken uniform normal to the plane of motion and can be different for different link in the mechanism considered.
Xn
PðuÞ ¼ Pi Ni;k ðuÞ; 0 ≤ u ≤ umax ð5Þ
i¼0

yi

Pi
Pn/2-1
P2
P1
Pn/2 P0, Pn
xn/2 Oi ai Oi+1 x0 xi
Pn-1
Pn/2+1
Pj

(a) Binary link

yi1
xi2 P31 P2n2
P3n3 y31 y2n2
ai3 ai2 P21
y3n3 Oi+2 y21
xi3
ai1 Oi+1 y1n1 xi1
y11 Oi
yi3
P1n1
P11
yi2

(b) Ternary link

Fig. 2. Closed cubic B-spline curve representing link shape and its control points.
130 K. Chaudhary, H. Chaudhary / Mechanism and Machine Theory 93 (2015) 127–146

where parameter k controls the degree of curve and B-spline blending functions Ni,k(u) are defined in Appendix C. The control points
form the vertices of the characteristic polygon of the B-spline curve as shown in Fig. 2. Note that the cubic B-spline curve is a composite
sequence of curve segments connected with C2 continuity which blends two curve segments with same curvature. The coordinates of
any point on the ith segment of the curve is given as:

α 1 xi‐1 þ α 2 xi þ α 3 xiþ1 þ α 4 xiþ2

xi ðuÞ ¼ ð6Þ
6

α 1 yi‐1 þ α 2 yi þ α 3 yiþ1 þ α 4 yiþ2

yi ðuÞ ¼ ð7Þ
6

where the terms α1, α 2, α 3 and α 4 are defined in Appendix C and (xi − 1, yi − 1), (xi, yi), etc. are the coordinates of points Pi − 1, Pi, etc.
respectively.
The mass and inertia of the link that is synthesized using closed cubic B-spline curve are calculated using Green's theorem [45]. As
defined in Appendix C, the area A, centroid (x; y) and area moment of inertia about centroidal axes [Ixx, Iyy, Izz] of the closed curve made
of n cubic B-spline segments are calculated as:

n Z
X ui
0
A¼ xi ðuÞy i ðuÞ du; ð8Þ
i¼1 ui‐1

Z
1 X n ui
2 0
x¼− yi ðuÞx i ðuÞ du; ð9Þ
2Ai i¼1 ui‐1

Z
1 X n ui
2 0
y¼ xi ðuÞy i ðuÞ du; ð10Þ
2Ai i¼1 ui‐1

Z
1X n ui
3 0
Ixx ¼ − yi ðuÞx i ðuÞ du; ð11Þ
3 i¼1 ui‐1

Z
1X n ui
3 0
Iyy ¼ xi ðuÞy i ðuÞ du; ð12Þ
3 i¼1 ui‐1

Izz ¼ Ixx þ Iyy : ð13Þ

The first derivatives xiV(u) and yiV(u) of xi(u) and yi(u) w.r.t. u, respectively, in Eqs. (8–12) are given by:

0 β1 xi‐1 þ β2 x i þ β3 xiþ1 þ β4 xiþ2

x i ðuÞ ¼ ; ð14Þ
6

0 β1 yi‐1 þ β2 y i þ β3 yiþ1 þ β4 yiþ2

y i ðuÞ ¼ ; ð15Þ
6

where

2 2
β1 ¼ ‐3u þ 6ui‐3i ; ð16Þ

2 2
β2 ¼ 9u þ 2uð3‐9iÞ‐3 þ 9i ‐6i; ð17Þ

2 2
β3 ¼ ‐9u þ 2uð‐6 þ 9iÞ−9i þ 12i; ð18Þ

2 2
β4 ¼ 3u þ 2uð3−3iÞ þ 3 þ 3i −6i: ð19Þ
 K. Chaudhary, H. Chaudhary / Mechanism and Machine Theory 93 (2015) 127–146 131

For geometric properties defined in Eqs. (8–13), the mass and mass moment of inertia of a link with shape represented by closed
curve are calculated as:

m ¼ Atρ; ð20Þ

I ¼ Izz tρ; ð21Þ

where t and ρ represent thickness and material density for the link, respectively.

3. Optimization problem formulation

3.1. First stage optimization—dynamic balancing

The optimization problem for minimizing the shaking force and shaking moment is formulated using the concept of
equimomental point-mass system in this section. Consider a multiloop mechanism having n moving links, i.e., i = 1,…,n as shown
in Fig. 1. Each link is systematically converted into a system of k equimomental point-masses and the point-mass parameters are
taken as the design variables. As each point mass needs three parameters to identify it, a 3 k-vector, xi, of design variables for ith
link is defined as:

T
xi ¼ ½ mi1 li1 θi1 mi2 li2 θi2 … mik lik θik  ð22Þ

where mij is jth point mass of ith link, and lij and θij are polar coordinates of it in the body fixed frame as shown in Fig. A1(b). Hence, the
3kn design vector, x, for the mechanism having n moving links is given by:
h iT
x ¼ xT1 T
x2 … xn
T
: ð23Þ

Considering the RMS values of the magnitude of shaking force, fsh,rms, and shaking moment, nsh,rms, defined in Eqs. (3–4), the
optimization problem is posed as weighted sum of the force and moment as:

Minimize Z ¼ w1 f sh;rms þ w2 nsh;rms ð24Þ

X X 2
Subject to mi; min ≤ mi j ≤mi; max ; I i; min ≤ mi j li j
j j
yi ¼ y j ð for symmetrical linkÞ ð25Þ
yi ≠y j ð for non−symmetrical linkÞ
for i ¼ 1; 2; …; n and j ¼ 1; 2; …; k

where w1 and w2 are the weighting factors used to assign weightage to shaking force and shaking moment, respectively, in the
objective function (Eq. 24). The constraints on design variables depend on the allowable minimum and maximum values of mass
and inertia of the links. The minimum mass, mi,min, of the ith link and its mass distribution can be decided by load bearing capability
of its material. Furthermore, the maximum mass, mi,max, can be taken into account according to what extent the shaking force and
shaking moment are eliminated. Similarly, the limits on parameters,lij, can be determined based on the limiting values of the moment
of inertia. The solution of this optimization problem finds the optimum values of design variables, i.e., the point-mass parameters, for
minimizing the objective function.

3.2. Second stage optimization—shape formation for balanced mechanism

In this subsection, an optimization problem is formulated to find the link shapes corresponding to the optimized inertial
parameters of the balanced mechanisms. To formulate the optimization problem, the Cartesian coordinates of control points of
cubic B-spline curve can be taken as design variables as shown in Fig. 2. In this paper, the genetic algorithm is used as the solver
for the optimization problem which doesn't require initial values of the design variables to start with and hence no initial shape is
required.
Link length between Oi to Oi + 1 is divided into equal parts. Hence, the x-coordinates of the control points lying between Oi and Oi + 1
are fixed according to the link length. Now, the y-coordinates are taken as the design variables. Furthermore, the extension of link beyond
Oi and Oi + 1 is controlled by P0, P1, Pn − 1 at one end and Pn/2 − 1, Pn/2, Pn/2 + 1 at other end. Hence, x coordinate of P0, y coordinates of P1
and Pn − 1 are chosen as the design variables and same is done at left end. Finally, the design vector for binary link is defined as:

h iT
X ¼ x0 y1 … yn=2−1 xn=2 yn=2þ1 … yn−1 : ð26aÞ
132 K. Chaudhary, H. Chaudhary / Mechanism and Machine Theory 93 (2015) 127–146

The conditions for symmetrical and non-symmetrical shapes are imposed by yi = yj and yi ≠ yj, respectively. In addition to
manufacturing benefits, the symmetrical shapes have zero products of inertia. For a ith ternary link Oi Oi + 1 Oi + 2 shown in
Fig. 2(b), the link length, ai, can be defined as summation of the distances between joints, i.e., ai = ai1 + ai2 + ai3. The number of
control points between two joints can be decided according to the distance between them.
If n1, n2, and n3 are number of control points for lengths ai1, ai2, and ai3, respectively, then total number of control points is sum of
n1, n2, and n3. At each joint, two points coincide and their y-coordinate can be determined by considering the local coordinate frame in
the link as shown in Fig. 2(b). The design vector in this case can be defined as:
h iT
X ¼ y11 … y1n1 y21 … y2n2 y31 … y3n3 : ð26bÞ

Note that for the link having three or more joints, the shapes can be synthesized by selecting y-coordinates for each segment of the
length between joints. The inertial properties of resulting shapes are constrained by the optimal properties. These constraints ensure
that the links with optimum shapes have the same inertial properties as that of the dynamically balanced mechanism links. The
objective function is formulated to minimize the percentage error in resulting links inertia values as:

ðIi −Ii Þ
Minimize Z ¼ x 100 ð27Þ
Ioi

  
Subject to mi ¼ mi ; xi ¼ xi ; yi ¼ yi for i ¼ 1; 2; …; n: ð28Þ

Here parameters with superscript ‘*’ represent optimum parameters obtained in the first stage optimization for dynamically balanced
mechanism and subscript ‘i’ is used for ith link of mechanism. The flow chart shown in Fig. 3 illustrates the proposed optimization
method.

4. Applications, results and discussion

After formulating the optimization problem, it can be solved by using either conventional or evolutionary optimization algo-
rithms. The conventional algorithms use gradient information of objective function with respect to design variables. These
methods converge on the optimum solution near to the initial guess point and thus produce local optimum solution [46,47]. Al-
ternatively, the genetic algorithm is evolutionary search and optimization algorithm based on the mechanics of natural genetics
and natural selection [48,49]. This algorithm evaluates only the objective function and genetic operators—selection, crossover
and mutation are used for exploring the design space. One can specify the bounds and nonlinear constraints for the variables
in this algorithm.
In this section, the effectiveness of proposed optimization method is shown by applying it to four planar mechanism problems. The
balancing problems can be framed as single objective or multi-objective optimization problems to simultaneously minimize shaking
force and shaking moment. As shaking force and shaking moment are of different units, these quantities need to be dimensionless for
adding them in a single objective function. For this, the mechanism parameters are made dimensionless with respect to the
parameters of the driving link of the mechanism. To keep the dimension of the problem moderate, here each link is represented by
three equimomental point-masses. Further the dimension of the problem can be reduced by assigning five parameters for each
link as:

θi1 ¼ 0; θi2 ¼ 2π=3; θi3 ¼ 4π=3 and li2 ¼ li3 ¼ li1 : ð29Þ

Out of nine variables, mij, lij, θij, for j = 1, 2, 3, for ith link, the other four point-mass parameters, mi1, mi2, mi3 and li1 are brought into
the optimization scheme as the design variables. The conversion of ith rigid link into a set of three equimomental point-masses is
discussed in Appendix B. To solve these problems using evolutionary optimization algorithms, i.e., the genetic algorithm, “ga” function
in Genetic Algorithm and Direct Search Toolbox of MATLAB with default values of genetic operators are used.
Considering mi,min = 0.25moi , mi,max = 5moi and Ii,min = 0.25Ioi for ith link, the optimization problem as explained in Eqs. (24–25) is
solved using genetic algorithm. The superscript ‘o’ represents parameters for original mechanism. Then, to find link shapes, the thick-
ness of links is taken as 10% of the driving link length and the link material is chosen as the mild steel (density = 7850 kg/m3) for
deciding the density and maximum permissible stress. The stress at the weakest section in each link is calculated for the maximum
joint force occurred in the second stage of the optimization algorithm. Dynamic stresses due to variable loading and fatigue failure
are not considered in this paper. Moreover, the von mises stresses for the peak load is considered to determine minimum cross-
section of each link. The inertial properties of links are calculated using Eqs. (20–21).

4.1. Planar four-bar mechanism

A numerical problem of planar four-bar mechanism [11,27] as shown in Fig. 4 is solved using the method proposed in this paper.
The link length, mass and other geometric parameters of the unbalanced mechanism are given in Table 1. For this example, the driving
link, i.e., link #1, rotates with a constant speed of 100 rad/s.
 K. Chaudhary, H. Chaudhary / Mechanism and Machine Theory 93 (2015) 127–146 133

Dynamic Balancing

Design variables:
Point mass parameters of
mechanism links

Objective function and

constraints:
Minimize Z w1 f sh,rms w2 nsh,rms

First stage optimization

Subject to mi , min mij mi , max
j

I i , min mijlij2
j

for i = 1, 2,…,n and j = 1, 2,…,k

Solution using “ga”

function in MATLAB

Optimized values of shaking force and

shaking moment and corresponding
design variables

Optimized mechanism parameters

Shape Optimization

Design variables:
Coordinates of control points of cubic
B-spline curve

Objective function and constraints:

Second stage optimization

( I i* I )
Minimize Z x100
I i*
* * *
Subject to mi mi ; xi xi ; yi yi
for i = 1, 2,…,n

Solution using “ga” function in

MATLAB

Optimized values of control

points for cubic B-spline curve

Optimum link shapes for

balanced mechanism

Fig. 3. Two stage optimization scheme to balance mechanism and shape synthesis.
134 K. Chaudhary, H. Chaudhary / Mechanism and Machine Theory 93 (2015) 127–146

'
2
C2 r2 #2 C3
m3
Y d2 m2 #3 d3
3
2 a2 a3
3
4
C1 #1 2 n03
m1 f03
1
a1
d1 -f03 - n03
e
n01 n1
1 X
O4
f01 a0
-f01 #0
- n01

O1

Fig. 4. Four-bar mechanism detached from its frame.

Six different cases are investigated to balance this mechanism:

In case (1), in order to make full force balancing, the weighting factors, w1 and w2, are taken as 1 and 0, respectively, in Eq. (24).
Link #1 and link #3 are considered for mass redistribution keeping link #2 intact.
In case (2), the simultaneous minimization of force and moment is considered taking the weighting factors, w1 and w2, as 0.5. In
this case, the point-masses of all three links are chosen to find their optimum mass distribution.
Particularly, four-bar mechanism may also be balanced by converting coupler into physical pendulum as did in [11,27]. To
investigate the effect of this conversion, the cases (1) and (2) are repeated after making link #2 as the physical pendulum and reported
as cases (3) and (4), respectively. In Case (5), the mechanism balancing is achieved without changing the original link masses while
balancing using non-symmetric shapes is presented as Case (6).
The comparison of original RMS values and peak values of shaking force and shaking moment with those of optimum values is
provided in Table 2. Figs. 5 and 6 show the variations of the shaking force, shaking moment and driving torque over the complete

Table 1
Original parameters of four-bar mechanism.

Link Length Mass Moment of inertia di θi

i ai (m) mi (kg) Ici,zz (kg-m2) (m) (°)

1 0.1 0.392 0.0014 0.05 0

2 0.4 1.570 0.0841 0.20 0
3 0.3 1.177 0.0356 0.15 0
0 0.3 ∑moi = 3.139 –

Table 2
The values of dynamic quantities in the original and optimized mechanisms.

Case Type of balancing Shaking force Shaking moment Driving torque

RMS Peak RMS Peak RMS Peak

Original mechanism 5.9604 11.8837 10.7250 12.3939 3.0588 5.0480

Case 1: w1 = 1, w2 = 0; Force balance 0.0087 0.3395 16.1133 16.3722 5.4359 9.4000
Keeping link #2 intact (−99.85%) (−97.14%) (+50.24%) (+32.10%) (+77.71%) (+86.21%)
Case 2: w1 = 0.5, w2 = 0.5; Force and moment balance 2.9620 4.4830 3.4484 5.1770 0.8769 1.6990
All links are considered (−50.31%) (−62.27%) (−67.85%) (−58.23%) (−71.33%) (−66.34%)
Case 3: w1 = 1, w2 = 0; Force balance 0.0688 0.1861 27.1472 30.6055 8.7761 15.7899
Link #2 is made as physical pendulum (−99.85%) (−98.43%) (+153.12%) (+146.94%) (+186.91%) (+212.79%)
and then kept intact
Case 4: w1 = 0.5, w2 = 0.5; Force and moment balance 4.0330 6.3998 5.9502 8.5520 1.5747 3.0076
Link #2 is made as physical pendulum (−37.34%) (−46.15%) (−44.52%) (−30.99%) (−48.52%) (−40.42%)
and then all links are considered
Case 5: w1 = 0.5, w2 = 0.5; Force and moment balance 4.3761 7.2856 5.5974 6.7374 1.5819 2.7111
Keeping link masses unchanged (−27.66%) (−38.69%) (−48.12%) (−45.64%) (−48.49%) (−46.29%)
Case 6: w1 = 0.5, w2 = 0.5; Force and moment balance 4.0172 7.4941 6.3747 7.4163 1.7958 2.9809
Non-symmetrical shapes (−32.95%) (−36.94%) (−40.59%) (−40.16%) (−41.30%) (−40.95%)
 K. Chaudhary, H. Chaudhary / Mechanism and Machine Theory 93 (2015) 127–146 135

12 12
Original value Original value

Normalised Shaking Force

Normalised Shaking Force

10 Case 1 10 Case 3
Case 2 Case 4
8 8

6 6

4 4

2 2

0 0
0 2 4 6 0 2 4 6
Time (sec) Time (sec)
12
Normalised Shaking Force Original value
10 Case 5
Case 6
8

0
0 2 4 6
Time (sec)

(a) Shaking force

40 40
Normalised Shaking Moment

Normalised Shaking Moment

20 20

0 0

-20 -20

-40 Original value -40 Original value

Case 1 Case 3
-60 Case 2 -60 Case 4
0 2 4 6 0 2 4 6
Time (sec) Time (sec)

40
Normalised Shaking Moment

20

-20

-40 Original value

Case 5
-60 Case 6
0 2 4 6
Time (sec)
(b) Shaking moment
Fig. 5. Variations of shaking force and shaking moment without converting link #2 into physical pendulum: Case (1), (2), (5) and (6); with link #2 as physical
pendulum: Case (3) and (4).
136 K. Chaudhary, H. Chaudhary / Mechanism and Machine Theory 93 (2015) 127–146

20 20

Normalised Driving Torque

Normalised Driving Torque

10 10

0 0

-10 -10

Original value Original value

-20 -20
Case 1 Case 3
Case 2 Case 4
-30 -30
0 2 4 6 0 2 4 6
Time (sec) Time (sec)

Normalised Driving Torque 20

10

-10

Original value
-20
Case 5
Case 6
-30
0 2 4 6
Time (sec)

Fig. 6. Variations of driving torque without converting link #2 into physical pendulum: Case (1), (2), (5) and (6); with link #2 as physical pendulum: Case (3) and (4).

crank cycle. The optimized link parameters are then found by using the equimomental conditions explained in Appendix B and given
in Table 3.
Next, the optimization problem for link shape formation presented in Eqs. (27–28) is solved and the resulting link shapes are
shown in Fig. 7.

4.1.1. Force balancing

The effectiveness of the optimization methodology proposed in Section 3 is compared with the analytical full force balancing
conditions [50] which are as follow:
a1 0 a
m1 d1 ¼ m2 r 2 and θ1 ¼ θ 2 ; m3 d3 ¼ m2 d2 3 ; and θ3 ¼ θ2 þ π ð30Þ
a2 a2
where all the parameters are defined in Fig. 4.

Table 3
Link parameters of the balanced four-bar mechanism.

Case Link Mass Total mass (kg) di θi Moment of inertia

i mi (kg) (m) (°) Ici,zz (kg-m2)

Case 1 1 1.857 7.505 0.0453 180 0.0131

2 1.570 0.2000 0 0.0841
3 4.078 0.0540 180 0.1112
Case 2 1 0.245 1.964 0.0222 0 0.0005
2 1.044 0.0594 0 0.0299
3 0.675 0.0702 0 0.0115
Case 3 1 2.436 10.777 0.0463 180 0.0208
2 2.310 0.2000 0 0.1577
3 6.031 0.0566 180 0.1781
Case 4 1 0.355 2.877 0.0180 0 0.0008
2 1.458 0.0824 0 0.0545
3 1.064 0.0541 0 0.0196
Case 5 1 0.390 3.138 0.0134 0 0.0007
2 1.570 0.0962 0 0.0468
3 1.178 0.0601 0 0.0192
Case 6 1 0.289 2.439 0.0314 6.72 0.0008
2 1.253 0.1417 358.15 0.0488
3 0.897 0.1087 1.4229 0.0225
 K. Chaudhary, H. Chaudhary / Mechanism and Machine Theory 93 (2015) 127–146 137

(a) Case 1 (b) Case 2

(c) Case 3 (d) Case 4

(e) Case 5 (f) Case 6

- - - - initial mechanism optimized mechanism

denotes joint denotes mass center

Fig. 7. Original and optimized link shapes corresponding to different cases for four-bar mechanism [all figures drawn on scale].

These conditions are satisfied for the complete force balanced mechanism achieved for cases (1) and (3). Thus the shaking force in
a mechanism can be completely eliminated by finding optimum link shapes (Fig. 7(a) and (c)). Note that this is achieved by
redistributing masses of link #1 and #3. Conventionally this is done by adding counterweights to link #1 and #3 which is not useful
from the mechanical design point of view. However, the total mass is increased to 7.505 kg and 10.777 kg from mass of 3.139 kg of
unbalanced mechanism, respectively in both approaches. As shown in Table 2, the RMS values of driving torque and shaking moment
138 K. Chaudhary, H. Chaudhary / Mechanism and Machine Theory 93 (2015) 127–146

C2 3 C3
#2 #3
d2 m2 m3
Y d3
2 a2 a3 3
2
4
#1
m1
m4 #4
C1 a1
1 C4 4

d1 a4
d4
e
n01 n e
1 1
X n04 n4
5
f01 f04
-f01
-f04 -n04
-n01 #0
a0
O1 O5

Fig. 8. Planar five-bar mechanism detached from its frame.

are increased by 77.71% and 50.24%, respectively for the complete force balancing (Case 1). This confirms the results found in inves-
tigations carried out by Lowen et al. [9] which shows that both the driving torque and shaking moment are increased in a force bal-
anced mechanism. In case (3), the force balancing is achieved after converting link #2 into the physical pendulum. It requires more
mass to be added for balancing in comparison of case (1) as given in Table 3. Moreover, the driving torque and the shaking moment
increments over original are 186.91% and 153.12%, respectively. Hence, the conversion of coupler into physical pendulum has adverse
effect on the driving torque and shaking moment.

4.1.2. Combined force and moment balancing

The method for shaking force and shaking moment balancing presented in [11,27] uses physical pendulum, force and inertia coun-
terweights. Alternatively, here reductions in the shaking force and shaking moment are achieved by redistributing masses optimally
as shown in Fig. 7(b), (d), (e) and (f). Treating link #2 as the physical pendulum does not improve the results as shown in Table 2 and
Fig. 5. In Case (5), the link masses are kept unchanged by using the equality constraint of the same mass in the optimization algorithm.
This results in the reductions of about 28%, 48% and 49% in driving torque, shaking force and shaking moment, respectively. The
mechanism balancing is achieved using the non-symmetric links as Case (6) which reduces these properties by 33%, 40% and 41%.
In this case, the condition of non-symmetry is used for the design variables, i.e., the control points as defined in Eq. (26a,26b,). The
results clearly show that the case (2) is the best solution for the problem considered as the reductions of about 50%, 68% and 71%
in RMS values of shaking force, shaking moment and driving torque, respectively, are achieved in this case.
Hence, complete shaking force balancing is achieved numerically by redistribution of link masses. Also, conversion of coupler into
physical pendulum will not improve balancing. These are key contribution of the paper among others. Furthermore, in case (2), the
peak values of shaking force, shaking moment and driving torque are reduced by about 62%, 58% and 66%, respectively.

4.2. Planar five-bar mechanism

The method proposed in this paper is used to solve the balancing problem of a planar five-bar mechanism, Fig. 8, reported in [51].
The parameters of original mechanism as reported in [51] are given in Table 4 where the input motions of links 1 and 4 are considered
as the cycloidal motion and the harmonic motion, respectively. In [51], a method is presented for only shaking force balancing of the

Table 4
Parameters of original and balanced planar five-bar mechanism.

Link Length Standard mechanism Balanced mechanism

i ai
Mass Moment of inertia di θi Mass Moment of inertia di θi
(m)
mi ICi,zz (kg-m2) (m) (°) mi Ici,zz (kg-m2) (m) (°)
(kg) (kg)

1 0.02 0.03 1.00e−6 0.01 0 0.015 2.93e−6 0.0017 180

2 0.10 0.15 1.25e−4 0.05 0 0.052 1.07e−4 0.0146 0
3 0.10 0.15 1.25e−4 0.05 0 0.038 8.72e−5 0.0312 0
4 0.04 0.06 8.00e−6 0.02 0 0.021 8.05e−6 0.0061 0
 K. Chaudhary, H. Chaudhary / Mechanism and Machine Theory 93 (2015) 127–146 139

4
x 10
8000 Original value 5

Normalised Shaking Moment

Normalised Shaking Force
Optimized value
6000
0

4000
-5
2000
Original value
-10 Optimized value
0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Time (sec) Time (sec)

(a) Shaking force (b) Shaking moment

Fig. 9. Variations of shaking force and shaking moment for complete cycle for planar five-bar mechanism.

mechanism using natural orthogonal complement dynamic modeling and is solved by conventional optimization method. It uses non-
linear constraint optimization in which the center of mass parameters of moving links are chosen as the design variables. However,
the resulting effect on shaking moment and driving torque was not considered. For the same numerical problem, both shaking
force and shaking moment are simultaneously minimized in this paper using proposed methodology and the genetic algorithm.
Fig. 9 shows the variations of the shaking force and shaking moment over the complete crank cycle. The optimized link parameters
are then found by using the equimomental conditions and given in Table 4.
The reductions of 80.21% and 84.75% were found in the RMS values of shaking force and shaking moment, respectively whereas
reductions in peaks are 89% for both quantities. The corresponding shapes of mechanism links are shown in Fig. 10.

4.3. Planar Stephenson six-bar mechanism

The analytical conditions are presented in [50] for only shaking force balancing of planar six-bar mechanism in which counter-
weight methodology is used. In another method [52], a convex optimization technique is used in determining the optimal shape,
position and mass of the counterweights. The method proposed in this paper is used to solve the balancing problem of a planar
Stephenson six-bar mechanism as reported in [52] shown in Fig. 1 for which parameters of original mechanism are given in
Table 5. For the constant angular velocity of 2π rad/s for link #3, both the shaking force and the shaking moment are minimized by

Symbols: and denote joints

and mass centers

C3

4
C2

2
C4
1 5

- - - - - - initial mechanism optimized mechanism

Fig. 10. Original and optimized link shapes of five-bar mechanism [figure drawn on scale].
140 K. Chaudhary, H. Chaudhary / Mechanism and Machine Theory 93 (2015) 127–146

Table 5
Parameters of original and balanced planar Stephenson six-bar mechanism.

Link Length Standard mechanism Balanced mechanism

i ai
Mass Moment of inertia di θi Mass Moment of inertia di θi
(m)
mi Ici,zz (kg-m2) (m) (°) mi Ici,zz (kg-m2) (m) (°)
(kg) (kg)

1 0.0559 0.060 4.98e−5 0.0286 3 0.031 3.72e−5 0.0249 0

2 0.1206 0.082 3.27e−4 0.0630 0 0.060 2.00e−4 0.0409 0
3 0.0032 0.075 7.27e−7 0.0031 5 0.019 2.27e−6 0.0057 180
4 0.1397 0.173 1.21e−3 0.0836 19 0.058 2.88e−4 0.0566 0
5 0.0444 0.039 1.53e−5 0.0197 0 0.018 8.12e−6 0.0566 0

redistributing the link masses as against the addition of counterweights as suggested in [50,52]. Also, the optimum shape of
mechanism links is obtained for the balanced mechanism using the proposed method.
The variations of shaking force and shaking moment over the complete crank cycle are shown in Fig. 11. The optimized link
parameters for the balanced mechanism are given in Table 5.
The reductions of 72.37% and 77.07% were found in the values of shaking force and shaking moment, respectively whereas
reductions in peaks are 69% and 81%. The corresponding shapes of balanced mechanism links are shown in Fig. 12.

4.4. Planar slider-crank mechanism

A numerical problem of planar slider-crank mechanism is solved using the proposed method for which a cam mechanism with
counterweight was used in [53] to simultaneously reduce the shaking force and shaking moment. Table 6 shows the inertial
parameters for original and balanced mechanisms. The variations of the shaking force and shaking moment over the complete
crank cycle are shown in Fig. 13.

20 20
Original value
Normalised Shaking Moment
Normalised Shaking Force

Optimized value
15 10

10 0

5 -10
Original value
Optimized value
0 -20
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Time (sec) Time (sec)

Fig. 11. Variations of shaking force and shaking moment for complete cycle for Stephenson six-bar mechanism.

7
4 2 C4

C2
C5

65
C1 3
C3

1 Symbols: and denote joints

and mass centers
- - - - - - initial mechanism optimized mechanism

Fig. 12. Original and optimized link shapes of Stephenson six-bar mechanism [figure drawn on scale].
 K. Chaudhary, H. Chaudhary / Mechanism and Machine Theory 93 (2015) 127–146 141

Table 6
Parameters of original and balanced planar slider-crank mechanism.

Link Length Standard mechanism Balanced mechanism

i ai (m)
Mass mi Moment of inertia di θi Mass mi Moment of inertia di θi
(kg) ICi,zz (kg-m2) (m) (°) (kg) ICi,zz (kg-m2) (m) (°)

1 0.292 2 0.03 0.146 0 3.782 0.0494 0.0027 180

2 0.427 3 0.14 0.214 0 1.555 0.0285 0.1633 0
3 – 4 – 0 0 4.000 – – –

The reductions of 48.45% and 44.14% were found in the RMS values of normalized shaking force and shaking moment, respectively
and corresponding shapes of mechanism links are shown in Fig. 14.
One of the advantages of the proposed method is that the links of the balanced mechanism are of the uniform thickness while the
force and inertia counterweights added to the original mechanism in traditional methods [11,27,50] are of large thickness and radius
compared to the original link parameters. Also, it doesn't require any pre-defined shapes or design domain to start with as suggested
in [27,52]. The percentage error of resulting inertia values were found within ±5%. Also, the resulting stresses for all the links of bal-
anced mechanism can be calculated at the weakest sections under external loads.

5. Conclusions

A two stage optimization method for optimum dynamic balancing and synthesis of link shapes for planar mechanisms with rev-
olute as well as prismatic joints is proposed in this paper. It is demonstrated that the conversion of the rigid links into equimomental
system of point-masses is a very effective tool in the balancing problem. The optimal mass distribution of links masses by taking point-
mass parameters as the design variables reduce the inertial force and moment transmitted to the frame significantly. In the case of
four-bar mechanism, the best solution reduces the RMS values of driving torque, shaking force and shaking moment by about 71%,
53% and 68%, respectively, in the case (2). The method is quite general and equally applicable for all single or multiloop mechanisms
where the analytical solutions are not available. The proposed method also demonstrates genetic algorithm as a solver in mechanism

3.5
Original value
Optimized value
Normalised Shaking Moment

3 0.5
Normalised Shaking Force

2.5

2 0

1.5

1 -0.5

0.5 Original value

-1 Optimized value
0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Time (sec) Time (sec)

Fig. 13. Variations of shaking force and shaking moment for complete crank cycle for slider-crank mechanism.

2
C2

1
C1

Symbols: and denote joints

and mass centers

- - - - - - initial mechanism optimized mechanism

Fig. 14. Original and optimized link shapes for planar slider-crank mechanism [figure drawn on scale].
142 K. Chaudhary, H. Chaudhary / Mechanism and Machine Theory 93 (2015) 127–146

balancing. In addition, the optimized values of link mass and inertia are effectively converted into physically possible symmetrical and
non-symmetrical shapes of links using closed B-spline curves. The novelty of the methodology is that it combines the dynamics and
design solution for mechanisms. It can also be extended logically for spatial mechanisms.

Nomenclature
ai vector from Oi to Oi + 1 of ith link
ai magnitude of ai, link length
A area of region defining link shape
Ci the mass center of ith link
di position vector of the mass center Ci of ith link, from origin Oi
D boundary of region defining link shape
fi 2-D vector of resultant force at Oi acting on ith link
Ii moment of inertia about Oi of ith link
lij distance of jth point mass from origin Oi of ith link
mi mass of ith link
mij mass of jth point mass of ith link
ni resultant moment about Oi acting on ith link
Oi origin of ith link where local coordinate frame Oi Xi Yi fixed to the link
Vi 2-D linear velocity vector of Oi of ith link whose vix and viy are components
  
vi 2-D linear acceleration vector of Oi of ith link whose vix and viy are components
w1, w2 weighting factors for multiple objective functions
αi angular position of ith link
θi angular position of mass center Ci of ith link in the local frame
θij angular position of point mass mij in the local frame
ωi angular velocity of ith link

ωi angular acceleration of ith link

Note that vectors are shown as bold font.

Appendix A

Equations of motion of rigid link moving in a plane

Consider an ith rigid link having motion in XY plane for which a local frame, Xi Yi, is fixed at Oi as shown in Fig. A1(a). The Newton–
Euler (NE) equations of motion for the link in the fixed inertial frame, OXY, are written as [1]:
  
ωi Ii −vix mi di sinðθi þ α i Þ þ viy mi di cosðθi þ α i Þ ¼ ni ; ðA:1Þ

  2
mi vix −ωi mi di sinðθi þ α i Þ−ωi mi di cosðθi þ α i Þ ¼ f ix ; ðA:2Þ

  2
mi viy þ ωi mi di cosðθi þ α i Þ−ωi mi di sinðθi þ α i Þ ¼ f iy ; ðA:3Þ

Oi+1
Oi+1
mi1 ai
Ci ai
Yi Xi
Yi mi li1
i Xi lij mij
di i1
i i2
Oi
Oi ij
Y Y mi2 li2

X X
O O

Fig. A1. (a) The ith rigid link moving in XY plane, and (b) its equimomental system of point-masses.
 K. Chaudhary, H. Chaudhary / Mechanism and Machine Theory 93 (2015) 127–146 143

where, ωi is the angular velocity about the axis perpendicular to plane of motion and vix and viy are the components of linear velocity of
  
origin Oi. Moreover, ωi, vix and viy are the time derivatives of ωi,vix and viy, respectively. Also, ni is resultant moment about Oi, and fix and
fiy are the components of the resultant force at Oi, respectively. Furthermore, mi and Ii are the mass of ith link and its mass moment of
inertia about Oi.

Appendix B

Equimomental system of point-masses

A rigid link can be modeled as a dynamically equivalent set of point-masses in order to identify the design variables for the formu-
lation of optimization problem. This dynamically equivalent system of point-masses is referred to equimomental systems [37]. The
rigid link and the system of point-masses will be dynamically equivalent (equimomental) if they have same mass, same center of
mass and same inertia with respect to the same coordinate frame [39]. Hence, a set of rigidly connected n point-masses, mij, located
at (lij, θij), must satisfy the following conditions:

X
mi j ¼ mi ; ðB:1Þ
j

X  
mi j li j cos θi j þ α i ¼ mi di cosðθi þ α i Þ; ðB:2Þ
j

X  
mi j li j sin θi j þ α i ¼ mi di sinðθi þ α i Þ; ðB:3Þ
j

X 2
mi j li j ¼ Ii : ðB:4Þ
j

The first subscript i denotes link number and the second subscript j represents the point-mass belong to the link (Fig. A1(b)). Using
equimomental conditions, B.1 to B.4, the Newton–Euler equations of motion, Eqs. (A.1–A.3), are rewritten as:

X X   X  
 2  
ωi mi j li j ‐vix mi j li j sin θi j þ α i þ viy mi j li j cos θi j þ α i ¼ ni ; ðB:5Þ
j j j

X X   X  
  2
mi j vix ‐ωi mi j li j sin θi j þ α i −ωi mi j li j cos θi j þ α i ¼ f ix ; ðB:6Þ
j j j

X X   X  
  2
mi j viy þ ωi mi j li j cos θi j þ α i −ωi mi j li j sin θi j þ α i ¼ f iy : ðB:7Þ
j j j

Appendix C

B-spline curve

For a B-spline curve of degree (k − 1), the B-spline function Ni,k(u) is computed iteratively [43,44]:

Ni;k‐1 ðuÞ   Niþ1;k‐1 ðuÞ

Ni;k ðuÞ ¼ ðu‐ui Þ þ uiþk ‐u ðC:1Þ
uiþk‐1 ‐ui uiþk ‐uiþ1

where


1; ui ≤ u ≤ uiþ1
Ni;1 ¼ : ðC:2Þ
0; otherwise
144 K. Chaudhary, H. Chaudhary / Mechanism and Machine Theory 93 (2015) 127–146

In Eq. (C.2), Ni,1 is a unit step function and ui are known as parametric knots or knot values. These values form a sequence of
nondecreasing integers called the knot vector. The parametric equation of ith curve segment of a cubic B-spline curve having control
points Pi − 1, Pi, Pi + 1 and Pi + 2 for u ∈ [ui ‐ 1, ui] is given as:

α 1 P i‐1 þ α 2 P i þ α 3 P iþ1 þ α 4 P iþ2

P i ðuÞ ¼ ðC:3Þ
6

where

3 2 2 3
α 1 ¼ ‐u þ 3u i‐3ui þ i ; ðC:4Þ

 
3 2 2 3 2
α 2 ¼ 3u þ u ð3‐9iÞ þ u ‐3 þ 9i ‐6i −3i þ 3i þ 3i þ 1; ðC:5Þ

 
3 2 2 3 2
α 3 ¼ ‐3u þ u ð‐6 þ 9iÞ þ u −9i þ 12i þ 3i −6i þ 4; ðC:6Þ

 
3 2 2 3 2
α 4 ¼ u þ u ð3−3iÞ þ u 3 þ 3i −6i −i þ 3i −3i þ 4: ðC:7Þ

Green's theorem

For two functions P(x, y) and Q(x, y) over a closed region D in the plane with boundary ∂D, Green's theorem states:
 
∂Q ∂P
∬ − dxdy ¼ ∮ ∂D ðPdx þ QdyÞ: ðC:8Þ
D ∂x ∂y

The area of closed region D is calculated as:

A ¼ ∬ dxdy: ðC:9Þ
D

This area is calculated using Green's theorem by taking P(x, y) = 0 and Q(x, y) = x that gives:
 
∂Q ∂P
A¼∬ − dxdy; ðC:10Þ
D ∂x ∂y

A ¼ ∮ ∂D xdy: ðC:11Þ

For a plane curve specified parametrically as (x(u), y(u)) for u ∈ [u0, u1], Eq. (C.11) becomes:
Z u1
0
A¼ xy du: ðC:12Þ
u0

Similarly, the moment about x-axis and y-axis of plane are computed as:using P = ‐ y2/2 and Q = 0
Z u1
1 2 1 2 0
Mx ¼ ∬ ydxdy ¼ − ∮ y dx¼− y x du ðC:13Þ
2 2 u0

using P = 0 and Q = x2/2

Z u1
1 2 1 2 0
My ¼ ∬ xdxdy ¼ ∮ x dy ¼ x y du: ðC:14Þ
2 2 u0

The geometric centroid ðx; yÞ of plane curve is given by x ¼ My =A and y ¼ Mx =A. Finally, the area moments of inertia can be
computed as:using P = − y3/3 and Q = 0
Z u1
2 1 3 1 3 0
Ixx ¼ ∬ y dxdy ¼ − ∮ y dx ¼ − y x du ðC:15Þ
3 3 u0
 K. Chaudhary, H. Chaudhary / Mechanism and Machine Theory 93 (2015) 127–146 145

using P = 0 and Q = x3/3

Z u1
2 1 3 1 3 0
Iyy ¼ ∬ x dxdy ¼ ∮ x dy ¼ x y du: ðC:16Þ
3 3 u0

Appendix D. Supplementary data

Supplementary data to this article can be found online at http://dx.doi.org/10.1016/j.mechmachtheory.2015.07.006.

References

[1] H. Chaudhary, S.K. Saha, Dynamics and Balancing of Multibody Systems, Springer-Verlag, Berlin, 2009.
[2] G.G. Lowen, F.R. Tepper, R.S. Berkof, Balancing of linkages—an update, Mech. Mach. Theory 18 (3) (1983) 213–220.
[3] V.H. Arakelian, M.R. Smith, Shaking force and shaking moment balancing of mechanisms: a historical review with new examples, ASME J. Mech. Des. 127 (2005)
334–339.
[4] V.V. Wijk, J.L. Herder, B. Demeulenaere, Comparison of various dynamic balancing principles regarding additional mass and additional inertia, ASME J. Mech.
Robot. 1 (2009) (041006-1-9).
[5] F.R. Tepper, G.G. Lowen, General theorems concerning full force balancing of planar linkages by internal mass redistribution, ASME J. Eng. Ind. Trans. 94 (3)
(1972) 789–796.
[6] I.S. Kochev, A new general method for full force balancing of planar linkages, Mech. Mach. Theory 23 (6) (1988) 475–480.
[7] M.J. Walker, K. Oldham, A general theory of force balancing using counterweights, Mech. Mach. Theory 13 (1978) 175–185.
[8] P.R. Ouyang, Q. Li, W.J. Zhang, Integrated design of robotic mechanisms for force balancing and trajectory tracking, Mechatronics 13 (2003) 887–905.
[9] G.G. Lowen, F.R. Tepper, R.S. Berkof, The quantitative influence of complete force balancing on the forces and moments of certain families of four-bar linkages,
Mech. Mach. Theory 9 (1974) 299–323.
[10] I.S. Kochev, General theory of complete shaking moment balancing of planar linkages: a critical review, Mech. Mach. Theory 35 (2000) 1501–1514.
[11] R.S. Berkof, Complete force and moment balancing of inline four-bar linkage, Mech. Mach. Theory 8 (1973) 397–410.
[12] G. Feng, Complete shaking force and shaking moment balancing of 26 types of four-, five- and six-bar linkages with prismatic pairs, Mech. Mach. Theory 25 (2)
(1990) 183–192.
[13] I. Esat, H. Bahai, A theory of complete force and moment balancing of planar linkage mechanisms, Mech. Mach. Theory 34 (1999) 903–922.
[14] V.H. Arakelian, M.R. Smith, Complete shaking force and shaking moment balancing of linkages, Mech. Mach. Theory 34 (1999) 1141–1153.
[15] C. Bagci, Complete shaking force and shaking moment balancing of link mechanisms using balancing idler loops, ASME J. Mech. Des. 104 (1982)
482–493.
[16] R.C. Soong, K.S. Hsu, A design combining kinematic and dynamic balancing considerations with bi-material links for four-bar linkages, J. Inf. Optim. Sci. 28 (4)
(2007) 663–686.
[17] E. Raghu, A. Balasubramonian, Experimental study on the elastodynamic behavior of the unbalanced and the counterweighted four bar mechanisms, J. Mech. Des.
112 (3) (1990) 271–277.
[18] F. Xi, R. Sinatra, Effect of dynamic balancing on fourbar linkage vibrations, Mech. Mach. Theory 32 (6) (1997) 715–728.
[19] Y.Q. Yu, B. Jiang, Analytical and experimental study on the dynamic balancing of flexible mechanisms, Mech. Mach. Theory 42 (5) (2007) 626–635.
[20] A. Martini, M. Troncossi, A. Rivola, Elastodynamic effects of mass-balancing: experimental investigation of a four-bar linkage, Adv. Mech. Eng. (2013) 1–10, http://
dx.doi.org/10.1155/2013/949457.
[21] A. Martini, M. Troncossi, M. Carricato, A. Rivola, Elastodynamic behavior of balanced closed-loop mechanisms: numerical analysis of a four-bar linkage, Meccanica
49 (3) (2014) 601–614.
[22] T.W. Lee, C. Cheng, Optimum balancing of combined shaking force, shaking moment, and torque fluctuations in high speed linkages, ASME J. Mech. Transm.
Autom. Des. 106 (2) (1984) 242–251.
[23] H. Chaudhary, S.K. Saha, Balancing of four-bar linkages using maximum recursive dynamic algorithm, Mech. Mach. Theory 42 (2) (2007) 216–232.
[24] Z. Li, Sensitivity and robustness of mechanism balancing, Mech. Mach. Theory 33 (7) (1998) 1045–1054.
[25] G. Alici, B. Shirinzadeh, Optimum dynamic balancing of planar parallel manipulators based on sensitivity analysis, Mech. Mach. Theory 41 (2006)
1520–1535.
[26] B. Demeulenaere, M. Verschuure, J. Swevers, J.D. Schutter, A general and numerically efficient framework to design sector-type and cylindrical counterweights for
balancing of planar linkages, ASME J. Mech. Des. 132 (2010) (pp. 011002(1-10)).
[27] M.R. Farmani, A. Jaamialahmadi, M. Babaie, Multiobjective optimization for force and moment balance of a four bar linkage using evolutionary algorithms, J.
Mech. Sci. Technol. 25 (12) (2011) 2971–2977.
[28] S. Erkaya, Investigation of balancing problem for a planar mechanism using genetic algorithm, J. Mech. Sci. Technol. 27 (7) (2013) 2153–2160.
[29] K. Chaudhary, H. Chaudhary, Dynamic balancing of planar mechanisms using genetic algorithm, J. Mech. Sci. Technol. 28 (10) (2014) 4213–4220.
[30] B. Feng, N. Morita, T. Torii, A new optimization method for dynamic design of planar linkage with clearances at joints, ASME J. Mech. Des. 124 (2002)
68–73.
[31] H. Azegami, L. Zhou, K. Umemura, N. Kondo, Shape optimization for a link mechanism, Struct. Multidiscip. Optim. 48 (1) (2013) 115–125.
[32] D. Xu, G.K. Ananthasuresh, Freeform skeletal shape optimization of compliant mechanisms, ASME J. Mech. Des. 125 (2003) 253–261.
[33] G. Yoganand, D. Sen, “Link Geometry Synthesis for Prescribed Inertia”, Proc. 15th National Conference on Machines and Mechanisms, Nov 30 - Dec 02, IIT Madras,
India, 2011.
[34] Y.M. Xie, G.P. Steven, Evolutionary structural optimization for dynamic problems, Comput. Struct. 58 (6) (1996) 1067–1073.
[35] Y. Kim, A. Tan, B. Yang, W. Kim, B. Choi, Y. An, Optimum shape design of rotating shaft by ESO method, J. Mech. Sci. Technol. 21 (2007) 1039–1047.
[36] D. Sen, S. Chowdhury, S.R. Pandey, Geometric design of interference-free planar linkages, Mech. Mach. Theory 39 (2004) 737–759.
[37] E.J. Routh, A Treatise on The Dynamics of A System of Rigid Bodies, Elementary Part I, Dover Publication Inc., New York, 1905.
[38] A.A. Sherwood, B.A. Hockey, The optimization of mass distribution in mechanisms using dynamically similar systems, J. Mech. 4 (1968) 243–260.
[39] N.C. Huang, Equimomental system of rigidly connected equal particles, J. Guid. Control. Dyn. 16 (6) (1983) 1194–1196.
[40] H. Chaudhary, S.K. Saha, Balancing of shaking forces and shaking moments for planar mechanisms using the equimomental systems, Mech. Mach. Theory 43
(2008) 310–334.
[41] J.J. Crisco, R.D. McGovern, Effective calculation of mass moments of inertia for segmented homogenous three-dimensional objects, J. Biomech. 31 (1998)
97–101.
[42] F.L. Conte, G.R. George, R.W. Mayne, J.P. Sadler, Optimum mechanism design combining kinematic and dynamic-force considerations, ASME J. Eng. Ind. 95 (2)
(1975) 662–670.
[43] I. Zeid, R. Sivasubramanian, CAD/CAM – Theory and Practice, Tata McGraw-Hill, New Delhi, India, 2009.
[44] M.E. Mortenson, Geometric Modeling, McGraw Hill Education (India) Private Limited, New Delhi, India, 2006.
[45] S. Brlek, G. Labelle, A. Lacasse, The discrete Green theorem and some applications in discrete geometry, Theor. Comput. Sci. 346 (2005) 220–225.
[46] R.T. Marler, J.S. Arora, Survey of multi-objective optimization methods for engineering, Struct. Multidiscip. Optim. 26 (6) (2004) 369–395.
[47] J. Mariappan, S. Krishnamurty, A generalised exact gradient method for mechanism synthesis, Mech. Mach. Theory 31 (4) (1996) 413–421.
146 K. Chaudhary, H. Chaudhary / Mechanism and Machine Theory 93 (2015) 127–146

[48] K. Deb, Optimization for Engineering Design—Algorithms and Examples, PHI Learning Private Limited, New Delhi, 2010.
[49] Y. Gao, L. Shi, P. Yao, “Study on Multi-Objective Genetic Algorithm”, Proc. 3rd World Congress on Intelligent Control and Automation, June 28–July 2, Hefei, P R China,
2000.
[50] R.S. Berkof, G.G. Lowen, A new method for completely force balancing simple linkages, ASME J. Eng. Ind. 91 (1) (1969) 21–26.
[51] D. Ilia, R. Sinatra, A novel formulation of the dynamic balancing of five-bar linkages with application to link optimization, Multibody Sys. Dyn. 21 (2009) 193–211.
[52] M. Verschuure, B. Demeulenaere, J. Swevers, J.D. Schutter, “On The Benefits of Partial Shaking Force Balance in Six-Bar Linkages”, Proceeding of 12th IFToMM World
Congress, June 18–21, Besancon, 2007.
[53] V. Arakelian, S. Briot, Simultaneous inertia force/moment balancing and torque compensation of slider-crank mechanisms, Mech. Res. Commun. 37 (2010)
265–269.