Mechanism and Machine Theory 157 (2021) 104214

Contents lists available at ScienceDirect

Mechanism and Machine Theory

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

Research paper

Kinematic modeling and optimal design of a partially

compliant four-bar linkage using elliptic integral solution
Ke Xu, Haitao Liu∗, Wei Yue, Juliang Xiao, Yabin Ding, Guofeng Wang
Key Laboratory of Mechanism Theory and Equipment Design of Ministry of Education, Tianjin University, Tianjin 300354, China

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

Article history: This paper deals with the kinematic modeling and optimal design of compliant mecha-
Received 26 October 2020 nisms with one flexible joint designed from a rigid four-bar linkage. An equivalent mech-
Revised 23 November 2020
anism is proposed to describe the large-deflection motion of a compliant four-bar linkage.
Accepted 2 December 2020
In light of the elliptic integral solution, the kinematics and strain energy of the compliant
four-bar linkage are derived and verified by a numerical comparison study. Then, by taking
Keywords: the path deviation and strain energy as two objectives, the multi-objective optimization
Compliant mechanism problem is solved at a given initial configuration of the rigid mechanism, followed by a
Four-bar linkage discussion of the best initial configuration where the rigid joint is replaced by elastic flex-
Combined flexible beam ures. Both the finite element analysis and experimental verification are carried out. The
Flexible joint results show that the output motion of the optimized compliant mechanism matches the
Large deflection
desired path very well, which demonstrates the feasibility of designing compliant mecha-
nisms by this approach.
© 2020 Elsevier Ltd. All rights reserved.

1. Introduction

Compliant mechanisms (CMs) have at least one flexible link or joint that transfers motion and force through the deflec-
tion of flexible members. Hence, they provide many benefits, e.g., ease of manufacture and assembly, no backlash and light
weight, etc. [1-3]. Fully CMs are generally limited to small ranges of displacement. While partially CMs composed of both
flexible members and traditional rigid links and joints can overcome this limitation and have many applications, such as the
linear actuator [4], soft exoskeleton [5], and impact contact force generator [6]. However, these CMs typically exhibit large
deflections that give rise to nonlinearities in both geometry and material behavior, which makes the modeling, analysis, and
design of CMs more difficult. In addition, many methods have been explored to design rigid mechanisms [7-11]. Therefore,
how to design a CM simply from a rigid mechanism that can accomplish a specific task (e.g. desired path) has attracted the
interests of many researchers.
The easiest way to design CMs from rigid mechanisms is to use flexible members to directly replace the rigid links or
joints in the rigid mechanisms through the enumeration process [12]. Unfortunately, the obtained CMs often fail to repro-
duce the path of the original rigid mechanism because the influence of the deflection of the flexible members on the motion
of the CM is not fully considered. To replace the original rigid mechanism equivalently, rigid-body replacement combined
with a pseudo-rigid-body model (PRBM) has been presented [1,13-14]. The PRBM proposed by Howell and Midha used two
rigid links joined at a characteristic pivot with a torsion spring to predict the deflection of flexible beams [15]. Rigid-body

∗
Corresponding author.
E-mail address: liuht@tju.edu.cn (H. Liu).

https://doi.org/10.1016/j.mechmachtheory.2020.104214
0094-114X/© 2020 Elsevier Ltd. All rights reserved.
K. Xu, H. Liu, W. Yue et al. Mechanism and Machine Theory 157 (2021) 104214

replacement takes a rigid mechanism that is capable of performing the desired path as a PRBM. Consequently, the joints
in the rigid mechanism are transformed into characteristic pivots. Then, one or more of the rigid links and/or joints are
replaced with equivalent compliant members (e.g. flexible cantilever beams). A CM designed by this method can reproduce
the desired path in a small deflection very well. Most of the existing studies have investigated this flexible cantilever beam
to design CMs by the concept of rigid-body replacement. For instance, due to the accurate prediction of small deflection
by the PRBM, CMs with microscale motion have been designed by replacing the revolute joint in a rigid mechanism with
small-length flexible beams [16-18]. Compliant mechanical amplifiers have been extensively explored based on parallel four-
bar and symmetric five-bar linkages for displacement amplification [19-21]. However, compared with the rigid mechanism,
the motion range of the CM designed by small-length flexible beams is still limited.
To increase the motion range of the CM, a long-length flexible beam that can produce large deflection subjected to end
loads has attracted the interest of researchers. In related work, a compliant four-bar linkage has been synthesized for gener-
ating 3-precision-point motion by replacing the rigid link in a rigid four-bar linkage with a long-length flexible beam [22].
Based on a PRBM, Mattson used a straight flexible cantilever beam to replace one rigid link in a parallelogram four-bar link-
age to design a compliant parallel mechanism with a large deflection [23]. The obtained CM was successfully used in a bi-
cycle derailleur and parallel-motion bicycle brakes. Taking long-length flexible beams as revolute joints in a serial kinematic
chain, Chen [24] designed fully compliant statically balanced mechanisms with large deflections. By replacing traditional
rigid joints with flexible joints composed of flexible beams, Yu [25,26] designed a CM satisfying the performance require-
ment of the original mechanism. Kuresangsai [27] used a straight flexible beam to replace the rigid joints in a Hoeken’s
linkage and designed flexure-jointed planar mechanisms by an optimization technique, but the large-deflection behavior of
flexible members was described by polynomial curvature functions. A monolithic compliant constant force mechanism has
also been designed from a compliant slider-crank mechanism, which can provide a constant force within a rather limited
deflection range [28]. Besides, Tanik [29] designed a fully compliant large stroke slider-crank mechanism with the aid of a
straight flexible cantilever beam. In this design, the prismatic joint is replaced with a compliant parallel-guiding mechanism.
Nevertheless, relatively large errors would be introduced in the analysis of the CM based on PRBM when large deflections
occur in the complaint members [30]. To avoid a non-trivial task of large deflection analysis, Megaro [31] developed a
computer-aided design tool for CM. In this approach, a designer starts with a rigid mechanism designed to accomplish the
desired task. Rigid joints rotating through a limited range of angles in the mechanism are then selected and automatically
replaced with flexible joints in the software. Finally, a partially CM composed of rigid links, initially curved beams, and
full-revolution joints can be synthesized by the proposed optimization method. This assistive tool allows users to generate
a diverse set of compliant versions of conventional mechanisms, as exemplified by the applications to Chebyshev Linkage,
Jansen’s Linkage, steering mechanism, etc. [31]. However, since it adopts the PRBM as a basis, the motion tracking accuracy
is still at an unsatisfactory level. Besides, it is also hard to reveal the influence of the parameters of flexible joints on the
motion of CM designed from the rigid mechanism. The difficulty lies in finding suitable compliant flexures and precisely
describing their large deflections.
Bearing this issue in mind and motivated by the work in [31], this paper takes a crank-rocker mechanism as an example
to carry out the design of a partially CM by replacing the hinge connecting the coupler and rocker with a flexible counter-
part. To preserve the output motion of the rigid mechanism as much as possible, a combined flexible beam (CFB) composed
of two straight and one circular-arc segments is employed as the parameterized flexible joint. Note that a parametric model
approximately predicting the large deflection of a CFB is of crucial importance for the synthesis of compliant mechanisms.
Compared with the PRBM and the chained beam constraint model (CBCM) [32], the elliptic integral solution is known as
an analytical method with higher precision for modeling flexible beams with constant curvatures [33-36]. Large deflection
can be directly evaluated by this approach instead of dividing the beams into serval segments. Therefore, as the core of this
study, the elliptic integral solution is used for precisely predicting the deflection and strain energy of the CFB.
The study aims to systematically investigate the effects of design variables on the output motion and provide fundamen-
tal guidelines for the design of other partially CMs having complex structures. The rest of this paper is organized as follows.
Section 2 presents a general method for modeling the deflections and strain energy of the CM. A numerical comparison
study is given to verify the effectiveness and accuracy of the method. In Section 3, the optimization is explored by mini-
mizing the tracking error and the strain energy in one full cycle of the mechanism’s motion. Finally, conclusions and future
studies are summarized in Section 4.

2. Kinematics and strain energy

This section introduces a general method for modeling the large-deflection motion and strain energy of the compliant
four-bar linkage with one flexible joint by means of elliptic integrals.

2.1. Structure description

A rigid crank-rocker four-bar linkage OABC is shown in Fig. 1. Link OA is the crank with an input angle ψ1 . ψ2 is the
angle between links AB and BC, and ψ3 is the rotation angle of the link BC. A point D on the link AB is used to measure the
output of the linkage. Then, a path of point D can be generated when the crank moves over a complete cycle. As shown in
Fig. 1, the rigid joint RB (denoted by capital letter B) is replaced with a CFB composed of two straight and one circular-arc

2
K. Xu, H. Liu, W. Yue et al. Mechanism and Machine Theory 157 (2021) 104214

Fig. 1. Compliant four-bar linkage and the its structural parameters.

Fig. 2. Parameters of original and deflected CFB.

segments at the initial configuration of the rigid mechanism when ψ1 = ψ10 . The other structural parameters are described
as follows. Point o is the position of one end point of the CFB on link BC, and point E is another end point of the CFB. β1
is the angle between rigid link oC and the first segment of the undeflected CFB. On the CFB, the lengths of two straight
segments are L1 and L3 , respectively, while the length of the circular-arc segment is L2 = Rβ2 . Here, R and β2 denote the
radius and central angle. The cross-section of the CFB is a rectangle. Its length and width are d and w, respectively. The
output point D in the compliant four-bar linkage will then produce a path due to the deflection of the CFB driven by the
crank OA.

2.2. Kinematic modeling

A global coordinate system O − XY is established at point O (see Fig. 1). A local coordinate system o − xy is established
at the point o. The x axis is along the first segment of the undeflected CFB. To obtain the output path of point D in the CM,
the force-displacement relationship of the CFB needs to be established.

3
K. Xu, H. Liu, W. Yue et al. Mechanism and Machine Theory 157 (2021) 104214

Fig. 3. Equivalent mechanism of compliant four-bar linkage.

As shown in Fig. 2, a b and θ are the tip positions and slope angle of the deflected CFB measured in o − xy. FO and MO
are the applied force and moment at the end point E, and the angle ϕO defines the direction of FO measured in o − xy.
The deflections of the CFB can be calculated by the prescribed moment and external force at the tip point E. However, it
is difficult to determine the values of MO , FO , and ϕO due to the coupling between the force and displacement in large-
deflection behavior [30]. To avoid this problem, an equivalent mechanism of the compliant four-bar linkage is presented as
shown in Fig. 3. It is assumed that there is an actuated prismatic joint between points A and C. The link BC is considered as
the frame of the equivalent mechanism. The bending behavior of the CFB driven by the crank OA is equivalent to that driven
by this virtual prismatic joint. In addition, the rigid link AE will be taken as the fourth segment of the CFB. Then, we can
directly predict the path of point A rather than that of point E. Two purposes of this treatment are given as follows. First,
the direction of the external force FO applied to the point A is coincident with the motion direction of the prismatic joint.
Second, letting point A be subjected to a pure force FO can avoid the difficulty of calculating the resultant moment MO .
In this manner, the kinematic modeling of the compliant four-bar linkage can be transformed into solving the deflection
of the CFB with four segments subject to a pure force FO . The deflection of this special CFB can be considered as the super-
position of deflections of four segments. The deflection solutions by elliptic integrals and the path of point D will be derived
as follows.

2.2.1. Constraint equations

As shown in Fig. 4, the ith segment in this special CFB is fixed at and moves along with the free end of the (i − 1)th
segment. A local frame oi − xi yi is established at the tip of the (i − 1)th segment with the xi axis being initially tangential
to the ith segment. In Fig. 4, (ai , bi ) and (ai , bi )are the tip positions of the ith deflected segment measured in oi − xi yi and
o − xy, respectively. While αi and θi are the corresponding deflection angles measured in those two frames. Especially for
the first segment, we have

a0 = b0 = 0, θ0 = 0 (1)
The direction ϕO of the external force FO as shown in Fig. 3 can be calculated as

ψ4 + β1 + π , A→C
ϕO = (2)
ψ4 + β1 , C→A

where A → C denotes that the direction of FO is pointing from A to C and vice versa. ψ4 is the angle between rigid link oC
and the virtual prismatic joint. According to the conditions of static equilibrium, the forces and moments applied on the ith
segment measured in oi − xi yi satisfy

Fi = FO , ϕi = ϕO − θi−1 , i = 1, . . . , 4 (3)

Mi−1 = Fi ai cos ϕi − Fi bi sin ϕi + Mi , M4 = 0 (4)

where ϕi is the direction of Fi measured in oi − xi yi . Geometric constraint needs to be satisfied in the motion process of
compliant linkage

4
K. Xu, H. Liu, W. Yue et al. Mechanism and Machine Theory 157 (2021) 104214

Fig. 4. Variables used for the deflection analysis of the ith segment moving along with the free end of the (i-1)th segment subject to applied force Fi and
moment Mi. Dashed line: the initial position of the ith segment; solid line: the deflected segment.

 T  T  T
loC cos β1 sin β1 + lAC cos ψ4 sin ψ4 = a4 b4
2
lAC 2
= lOA 2
+ lOC − 2lOA lOC cos ψ1 (5)
where loC , lOC , and lOA are the lengths of rigid links oC , OC , and OA, respectively. In what follows, the method for the evalu-
ation of (ai , bi ) and θi of the ith segment will be investigated.

2.2.2. Deflection of the ith segment of the CFB

When i = 4, this segment is a rigid beam. The end of this beam is only subjected to force FO with angle ϕO . The displace-
ment of point A can then be expressed in o4 − x4 y4
  
a4 = lAE , b4 = 0, α4 = ψ5 = arctan y0A − y0E / x0A − x0E − β2 (6)
where ψ5 is the angle between the third undeflected segment and the rigid link AE with length lAE , and this angle is
constant in the CFB bending process; (x0A , y0A ) and (x0E , y0E ) are the coordinates of points A and E when the CFB is undeflected.
When i < 4, according to the Bernoulli Euler equation, the relationship between the curvature of the ith segment at a
given point P (as shown in Fig. 4) and the moment applied on the segment expressed in oi − xi yi can be formulated as
   
dαP dαP0 Fi sin ϕi’ a’i − xP − Fi cos ϕi’ b’i − yP + Mi
KP − KP0 = − = (7)
ds ds EI
where KP (KP0 ) and αP (αP0 ) are the curvature and slope angle of the deflected (undeflected) beam at point P, respectively;
dαP /ds is the rate of the change of angular deflection along the deflected beam length; (xP , yP ) is the coordinates of point
P at the deflected segment measured in oi − xi yi ; E is the Young’s modulus of the material; I is the area moment of inertia
of the cross section. Differentiating Eq. (7) with respect to the arc length s gives
 
d dαP dαP Fi  
= sin αP − ϕi’ (8)
dαP ds ds EI

where dxP /ds = cos αP and dyP /ds = sin αP are used. Integrating Eq. (8) with respect to αP results in
 2  2
1 dαP −Fi   Fi   1 0 Mi
= cos αP − ϕi’ + cos αi − ϕi’ + K + (9)
2 ds EI EI 2 i EI

Note that at the tip of the ith segment, αP = αi and dαi /ds = Ki0 + Mi /(EI ), where Ki0 is the initial curvature at the tip.
Solving for dαP /ds yields

dαP 2Fi   
= λi − cos αP − ϕi’ (10)
ds EI

5
K. Xu, H. Liu, W. Yue et al. Mechanism and Machine Theory 157 (2021) 104214

2
where λi = cos(αi − ϕi’ ) + EI/(2Fi ) · (Ki0 + Mi /(EI )) . Integrating Eq. (10) with respect to s, for s between 0 and Li , results in

2Fi 1 αi dαP
ti = =   (11)
EI Li 0
λi − cos αP − ϕi’
Similarly, the tip deflection in terms of ai and bi can be computed by
   T
a’i 1 αi cosαP dαP αi sinαP dαP
= 0
λi −cos (αP −ϕi’ )
0
λi −cos (αP −ϕi’ ) (12)
b’i ti
To avoid dividing the beams into serval segments (such as the chain algorithm, beam constraint model) and achieve high
modeling precision directly, three integrals in Eqs. (11) and (12) are solved utilizing the elliptic integral solution [33-36]. The
results are given as follows.
When |λi | > 1,
√
ti = 2k i f i / L i (13)
  √  
a’i 2ki cos ϕi’ − sin ϕi’ λi fi − (λi + 1 )ei
= (14)
b’i ti sin ϕi’ cos ϕi’ gi
where   
fi = F (φi , ki ) − F (φi−1 , ki ), ei = E (φi , ki ) − E (φi−1 , ki ), gi = λi + 1( λi − cos(αi − ϕi’ ) − λi − cos ϕi’ )
1+cos(αi −ϕi’ ) 1+cos ϕi’
sin2 φi = 2 , sin2 φi−1 = 2 , k2i = 2/(λi + 1 )
When |λi | ≤ 1,
√
ti = 2 f i /Li (15)
  √ 
a’i 1 cos ϕi’ − sin ϕi’ 2 ( f i − 2ei )
= (16)
b’i ti sin ϕi’ cos ϕi’ 2 gi
where  
fi = F (κi , ui ) − F (κi−1 , ui ), ei = E (κi , ui ) − E (κi−1 , ui ), gi = λi − cos(αi − ϕi’ ) − λi − cos ϕi’
1+cos(αi −ϕi’ ) 1+cos ϕi’
sin2 κi = λi +1 , sin2 κi−1 = λi +1 , u2i = (λi + 1 )/2

where F (φi , ki ) and E (φi , ki ) are the incomplete elliptic integrals of the first and second kind with an amplitude φi and a
modulus of ki , respectively. Considering that the deflection of the CFB is caused by a pure force, the values of φi and κi
are unique, and each segment has no inflection point during deflection [33-36]. Furthermore, measured in o − xy, the tip
position and angle of the ith segment (i = 1, . . . , 4) can be obtained by
     
ai cos θi−1 − sin θi−1 a’i ai−1
= + (17)
bi sin θi−1 cos θi−1 b’i bi−1

θi = αi + θi−1 (18)

2.2.3. Position of the output point D

Eqs. (3)-(4), (6), (14), and (16)-(18) constitute deformation equations of the CFB including five unknown parameters FO ,
ϕO , α1 , α2 , and α3 . Eqs. (5), (13), and (15) provide five constraints. Therefore, the deflections of point A (i.e., a4 , b4 and θ4 )
and point E (i.e., a3 , b3 and θ3 ) can be evaluated by numerically solving these deformation equations. Then, the position of
output point D measured in o − xy can be calculated
 T  T  T
xD yD = a3 b3 +lDE cos(θ3 + ψ6 ) sin(θ3 + ψ6 ) (19)
where ψ6 = arctan (y0D − y0E )/(x0D − x0E ) − β2 is the angle between the third undeflected segment and link DE with length
lDE , as shown in Fig. 3; (x0D , y0D ) is the coordinates of point D when the CFB is undeflected. By means of homogeneous
transformation, the position of point D measured in O − XY can be expressed as
     
XD cos γ3 − sin γ3 −xD cos γ2
= +lOo (20)
YD sin γ3 cos γ3 yD sin γ2
l2 + l2 − l2 l 2 + l 2 − loC
2
γ3 = ϕO − π − nγ1 , γ1 = arccos OC AC OA , γ2 = − arccos OC Oo
2lOC lAC  2lOC lOo
 1 , ψ ∈ [ 0 , π ]
oC OC cos (ψ2 − β1 ), n =
1
lOo = loC
2 + l 2 − 2l l
OC −1, ψ1 ∈ (π , 2π ]
Finally, given the structural parameters of the compliant four-bar linkage and the material properties of the CFB, the
deflected configurations of the CM driven by crank OA can be predicted.

6
K. Xu, H. Liu, W. Yue et al. Mechanism and Machine Theory 157 (2021) 104214

Table 1
Parameters of the compliant four-bar linkage.

lOA lOC lAB lBC loB β1 L1 R β2 L3 d w E

20mm 60mm 96mm 60mm 16mm 5° 12mm 9mm 135° 16mm 6mm 1mm 2.734GPa

2.3. Strain energy modeling

Strain energy is an important factor used to evaluate the performance of a CM. Therefore, it is necessary to establish
its mathematical model. According to Hook’s law, the stress σ of a material at a distance ω from a neutral layer can be
written as σ = E ω/ρ , where ρ is the curvature of the neutral layer. The strain energy stored in the microsegment ds can be
obtained as
E
  M2
dU = ω2 dA ds = i ds (21)
2ρ 2
A 2EI
where A is the area of the cross-section. Integrating Eq. (21), the strain energy of the ith segment of the CFB can be ex-
pressed as
Li Mi2 EI αi dαP
Ui = ds = dαP (22)
0 2EI 2 αi0 ds

where αi (αi0 ) are the tip slope angles of the ith deflected (undeflected) segment. Substituting Eq. (10) into Eq. (22) leads to
 
EIti αi   αi0  
Ui = λi − cos αP − ϕ i dαP − λi − cos αP − ϕ i dαP (23)
2 0 0

By means of the elliptic integral solution, the strain energy of the ith segment can be evaluated

2EI · fi · ei /Li − Ui 1 < |λi |
Ui = (24)
EI · fi ( (λi − 1 ) fi +2ei )/Li − Ui 0 < |λi | ≤ 1

where the values of fi and ei are identical to those in Eqs. (13)-(16). Since α10 = α30 = 0, we have U1 = U3 = 0. According
to α20 = β2 and Eqs. (7) and (22), U2 can be expressed as

U2 = (EI/R )2 L2 /(2EI ) = EIβ2 /(2R ) (25)

Therefore, the strain energy stored in the CFB can be derived as


3
U= Ui (26)
i=1

2.4. Verification

Numerical calculation (NC) and finite element analysis (FEA) simulation are employed to verify the correctness of the
kinematic model and strain energy model of the compliant four-bar linkage with prescribed structural parameters listed in
Table 1. In the FEA model (see Fig. 5(a)), the CFB has meshed into 2233 elements. The type of element is quadrilateral. The
material is UV-curable resin with a Young’s modulus of 2.734Gpa. Linkages OA, AE and oC are considered as rigid bodies.
The connection between two rigid linkages is a frictionless revolute joint. The CFB is bonded to the surface of rigid part. For
the transient structural analysis, the large deflection is activated in the solver.
For the comparison study, the initial configuration of the CM is defined as the configuration where three revolute joints
RA , RO and RC are collinear, as shown in Fig. 5(a), i.e., ψ10 = π . The midpoint of link AB is selected as the output point D.
Comparisons of the output path and strain energy are shown in Figs. 5(b) and 5(c), respectively. According to the results, the
maximal path deviation and energy error are 0.0175mm and 0.0022J, respectively, which illustrates that the derived models
are capable of accurately predicting the motion and strain energy of the compliant four-bar linkage.

3. Optimal design

With the aid of the models derived in Section 2, this section deals with the parameter optimization of the CFB by
minimizing the path deviation and the strain energy of the CM. Given an initial configuration of the rigid four-bar linkage,
the Pareto front of the multi-objective optimization problem is obtained first. Then, the optimal results are determined
according to the ideal point method [37]. Finally, a discussion of initial configurations is given, followed by the FEA and
experimental study.

7
K. Xu, H. Liu, W. Yue et al. Mechanism and Machine Theory 157 (2021) 104214

Fig. 5. Initial configuration of the compliant four-bar linkage (a) and comparisons of the output path (b) and strain energy (c).

3.1. Objectives and constraints

The design problem of a compliant four-bar linkage with one flexible joint is a dimension-based design of β1 , L1 , R, β2 ,
L3 , and the position of point o, which achieves the minimal path deviation. Besides, the strain energy will be produced
during the bending of the CFB, which should be kept as small as possible to reduce the energy required for generating
motion. The parameters of the original four-bar linkage are listed in Table 1. The Young’s modulus of the material is given
at first. Note that the parameters of the cross-section of the CFB, i.e. d and w, are related to stress and fatigue life [1].
To achieve high lateral stability and stiffness along the direction of the rotation axis, these two parameters are prescribed
according to the scale of the four-bar mechanism. As a consequence, the design variable vector can be expressed as
 T
x = loB β1 L1 R β2 L3 (27)
Given a value of ψ10 , i.e. an initial configuration of the rigid linkage, the deviation between the desired and actual paths
can be estimated by

N  2  2
e= xid − xiD + yid − yiD (28)
i=1

where (xid , yid ) and (xiD , yiD ) are coordinates of the ith (i = 1, . . . , N) desired and actual points for the same input angle;
N is the number of discrete points on the path. Note that a large number of discrete points will definitely increase the
computational burden. To determine N, the path deviation and strain energy of the CM with different parameters of the
CFB (see Table 2) are further investigated at a given initial configuration ψ10 = −π /2. Fig. 6(a) shows the desired and actual
paths of three examples obtained using the derived models, from which it can be found that path deviation is increasing

8
K. Xu, H. Liu, W. Yue et al. Mechanism and Machine Theory 157 (2021) 104214

Table 2
Three groups of the design variables x.

No. loB (mm) β1 (°) L1 (mm) R(mm) β2 (°) L3 (mm)

1 15 5 10 5 125 5
2 20 10 15 10 130 10
3 25 15 20 15 135 15

Table 3
Maximal values of path deviation and strain energy.

No. Path deviation (mm) Strain energy (J)

NC FEA Error NC FEA Error

1 3.279 3.262 0.017 0.123 0.1206 0.0024

2 6.175 6.163 0.012 0.061 0.0589 0.0021
3 9.367 9.356 0.011 0.034 0.0324 0.0016

Fig. 6. Path deviation (a) and strain energy (b) of the CM with different parameters of the CFB.

with the increase of the deflection of the CFB. Referring back to Fig. 1 and Fig. 3, the largest deflection of the CFB occurs
when three points A, O and C are collinear, i.e. ψ1 = 0 or ψ1 = π . For these three examples, the largest deflections are all
achieved at ψ1 = 0. The path deviations evaluated by NC and FEA are listed in Table 3. Hence, it is reasonable to reduce
the computational burden of Eq. (28) by letting N = 1 and just minimizing the largest path deviation at ψ1 = 0 or ψ1 = π .
Therefore, the objective of path deviation can be defined as
 
f1 (x ) = max eψ1 =0 , eψ1 =π (29)
Analogously, the maximal value of strain energy is also achieved when the CFB takes the largest deflection (see Fig. 6(b)),
resulting in the second objective
 
f2 (x ) = max Uψ1 =0 , Uψ1 =π (30)
Consequently, the optimization design problem can be formulated as
 
minimize F (x ) = f 1 (x ) f 2 (x )
 T
over x = loB β1 L1 R β2 L3
subjectto c1 : max (M )w/(2I ) < [σ ] (31)
c2 : L1 + L2 + Rβ2 < (lAB + lBC )/3
c3 : L1 + L2 + Rβ2 +lAE + loC ≥ lOA + lOC
c4 : x l ≤ x ≤ x u
where max(M ) is the maximal moment applied on the CFB; [σ ] is the allowable stress of the material; xl and xu are the
lower and upper bounds of x, respectively. The constraint c1 avoids material failure. During the bending of a flexible beam,
material failure will be closely correlated to the maximum stress, which is a function of the deflection and the moment of
inertia of the beam and can be calculated by max(M )w/(2I ) [1,13]. max(M ) can be evaluated from Eq. (4). The constraint
c2 gives a constrain on the total length of the CFB, which is shorter than one-third of the total lengths of links AB and BC,
in order to achieve a compact design. Besides, it should be noted that if the sum of lengths of the crank OA and the fixed

9
K. Xu, H. Liu, W. Yue et al. Mechanism and Machine Theory 157 (2021) 104214

Table 4
Lower and upper bounds of the design variables x.

loB (mm) β1 (°) L1 (mm) R(mm) β2 (°) L3 (mm)

xl 10 0 5 5 120 1
xu 30 20 25 25 140 20

Table 5
Algorithm parameters of NSGA-II.

Dimension Population size Distribution index Directional crossover probability Crossover probability Generation

6 40 20 0.5 0.9 200

Fig. 7. Results of multi-objective optimization. (a) Pareto front; (b) paths of three Pareto-optimal solutions.

Table 6
Three Pareto-optimal solutions.

No. Variables Objectives

loB (mm) β1 (°) L1 (mm) R(mm) β2 (°) L3 (mm) f 1 (mm) f 2 (J)

I 16.1999 19.5722 16.8544 8.4273 125.0881 1.0466 7.881e-3 0.0751

II 10.0392 12.4623 20.2058 9.4889 123.2196 11.3322 8.0484 0.0367
III 15.9835 19.7007 20.4507 12.3347 121.3545 1.0488 2.7348 0.0486

link OC is larger than that of the rest parts, the compliant four-bar linkage will reach a dead point (i.e., the deflected CFB
turns to a straight beam in this case). Therefore, a Grashof-like condition c3 should be satisfied to ensure the full rotation
of the crank.
Moreover, it can be seen from Fig. 6 that the larger the strain energy, the smaller the path deviation, which indicates
that two objectives are in conflict with each other. Hence, there is no unique global solution to Eq. (31) but a set of so-
lutions. To solve the multi-objective optimization problem, we employ an effective algorithm NSGA-II [38] implementing
in MATLAB for searching the Pareto front. A brief description of the algorithm is given as follows. First, the initial popu-
lation with a certain scale is generated randomly. After non-dominated sorting, the first-generation offspring population is
obtained through three basic operations of the genetic algorithm, i.e. selection, crossover, and mutation. Secondly, starting
from the second generation, the parent population and the offspring population are merged to carry out fast non-dominated
sequencing. At the same time, the crowding degree of individuals in each non-dominated layer is calculated. According to
the non-dominated relationship and the crowding degree of individuals, the appropriate individuals are selected to form a
new parent population. Finally, a new generation population is generated by the basic operations of the genetic algorithm.
Iteratively, the optimization problem can be solved until the end condition of the program is satisfied.

3.2. Results and discussion

Given [σ ] = 70 MPa, the bounds of x (see Table 4), the parameters of NSGA-II referring to [39] (see Table 5), and the
initial configuration ψ10 = π , the optimization results (i.e. Pareto front) are obtained as shown in Fig. 7. In order to achieve
a unique solution on the Pareto front, a decision-making strategy known as the ideal point method is adopted. More de-
tails could be found in [38]. The final optimal results (Solution III) together with two extreme points on the Pareto front

10
K. Xu, H. Liu, W. Yue et al. Mechanism and Machine Theory 157 (2021) 104214

Fig. 8. Scatter matrix for the design variables and two objectives of the Pareto front.

Fig. 9. Design variables as functions of two objectives.

(Solutions I and II) are listed in Table 6. Solution I has the smallest path deviation but the largest strain energy. Solution
II achieves the minimal strain energy but results in the largest path deviation. Solution III is a compromise between two
objectives.
Furthermore, the interdependency between design variables and two objectives of the Pareto front is revealed by means
of the scatter matrix, as shown in Fig. 8. The diagonal elements represent the probability density charts of each variable. The
upper and lower triangular elements represent the correlation scatter distribution diagram and the correlation coefficient
between the two variables, respectively. Fig. 9 illustrates the design variables as functions of objectives. Form Fig. 8 and
Fig. 9, the following conclusions can be drawn:

• f1 and f2 are heavily dependent on each other as their correlation coefficient is −0.9331, which is consistent with the
result shown in Fig. 6.

11
K. Xu, H. Liu, W. Yue et al. Mechanism and Machine Theory 157 (2021) 104214

0
Fig. 10. Variations of two objectives versus lAC .

Fig. 11. Results of the FEA of the CM at ψ1 = 80.4◦ , 180◦ , 279.6◦ , and 360◦ .

• loB and L1 are key factors affecting f1 and f2 . f1 is more dependent on β1 and L3 than f2 . As shown in Fig. 9, β1 and L3
converge to constants in the ranges of 0mm < f1 ≤ 4.2mm and 0.041J ≤ f2 ≤ 0.0751J.
• Both f1 and f2 are less dependent on R and β2 (i.e., two parameters of the circular-arc segment in the CFB). It demon-
strates that the circular-arc segment is not the main cause of large deflections for the Pareto-optimal solutions. However,
R and β2 are heavily dependent on each other as their correlation coefficient is −0.9367.

12
K. Xu, H. Liu, W. Yue et al. Mechanism and Machine Theory 157 (2021) 104214

Fig. 12. Experimental setup (a), four configurations of the CM driven by the crank (b)~(e), and path deviation (f).

Table 7
Optimized parameters of the CM given ψ10 = 80.4◦ .

loB (mm) β1 (°) L1 (mm) R(mm) β2 (°) L3 (mm)

11.5480 16.6277 20.8561 8.4211 139.7144 1.0548

It should be noted that since the optimal design is carried out at a given initial configuration, it is essential to make
a comparison of the optimization results by taking different values of ψ10 . To this end, we use the initial length of lAC
(see Fig. 3), denoted by lAC 0 , instead of ψ 0 for the optimizations. According to the parameters of the rigid linkage given in
1
Table 1, the range of lAC can be evaluated, which is [40mm, 80mm]. Then, taking different values of lAC 0 in this range at a
0
step of 5 mm, the variations of the objectives of different optimization results versus lAC are depicted in Fig. 10. It can be
found that the minimal values of the two objectives are achieved around lAC 0 = 60mm, and the corresponding initial angle is

ψ1 = 80.4 or ψ1 = 279.6 . Since the two initial angles are related by symmetry with respect to the x-axis, the optimized
0 ◦ 0 ◦

parameters of two CMs have the same values as listed in Table 7.

In order to verify the feasibility of the proposed method, the numerical simulation and experimental study are car-
ried out. The CFB is designed using the parameters given in Table 7. Fig. 11 shows the results of the FEA of the CM at
ψ1 = 80.4◦ , 180◦ , 279.6◦ , and 360◦ . The evaluated maximal path deviation and strain energy of the mechanism are 0.795mm

13
K. Xu, H. Liu, W. Yue et al. Mechanism and Machine Theory 157 (2021) 104214

Fig. 13. A partially CM designed from Jansen’s Linkage. (a) Jansen’s Linkage and (b) partially CM.

and 0.0228J, respectively. While the maximal errors of the predicted path and strain energy between the NC and FEA
are 0.0102mm and 0.0013J, respectively. Besides, a prototype of the CM is also fabricated using the parameters given in
Table 1 and Table 7 (see Fig. 12(a)). Fig. 12(b)–(e) illustrate four positions of the mechanism, in which the black curve is the
path of the original rigid four-bar linkage. The path derivation is approximately measured using the graph paper and plotted
in Fig. 12(f). Ignoring the manufacturing error and the uncertainty of measurement, the result shows that the maximal path
deviation is less than 1 mm, which verifies the feasibility of the optimal design method.

4 Conclusions

An investigation of the optimal design of a CM obtained by replacing a conventional joint in a rigid four-bar linkage with
a CFB is presented in this paper. An equivalent mechanism is presented to describe the motion of the compliant four-bar
linkage. Drawing on the elliptic integral solution, the deflections and strain energy of the compliant four-bar linkage are
obtained based on the equivalent mechanism. The effectiveness and accuracy of the modeling method are verified by a
numerical comparison study between the derived models and the FEA software. Besides, the maximal path derivation and
strain energy of the CM are taken as two objectives to be minimized for the optimal design of the structural parameters. The
Pareto front has been obtained under a prescribed initial configuration. The interdependency between design variables and
the objectives on the Pareto front has been analyzed using a scatter matrix. Finally, the best initial configuration to design
the compliant four-bar linkage is discussed. The corresponding maximal path deviation and strain energy stored in the CFB
of the optimal solution are 0.795mm and 0.0228J, respectively, which verifies the feasibility of the proposed method.
Further research will be devoted to extending our method to design CM from rigid linkages with complex topological
structures, such as the Jansen’s Linkage (see Fig. 13), and exploring other methods for precisely modeling initially curved
beams for a design. Different curvatures of the flexible joint and shapes of the cross-section are also worthy of investigation.
Meanwhile, the influence of external force on the kinematic and dynamic performance of the partially compliant mechanism
is an interesting topic for future work.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have
appeared to influence the work reported in this paper.

14
K. Xu, H. Liu, W. Yue et al. Mechanism and Machine Theory 157 (2021) 104214

Acknowledgment

This work was partially supported by the National Natural Science Foundation of China (91948301, 51775376, and
51675369) and Tianjin Science and Technology Program (17JCZDJC40100).

Supplementary materials

Supplementary material associated with this article can be found, in the online version, at doi:10.1016/j.mechmachtheory.
2020.104214.

References

[1] L.L Howell, Compliant Mechanisms[M], John Wiley & Sons, New York, 2001.
[2] A.E. Albanesi, V.D. Fachinotti, M.A Pucheta, A review on design methods for compliant mechanisms[J], Mecnica Comput. (3) (2010) 59–72.
[3] G.B. Hao, J.J. Yu, H.Y Li, A brief review on nonlinear modeling methods and applications of compliant mechanisms[J], Front. Mech. Eng. 11 (2016)
119–128.
[4] A. Tekes, H. Lin, K. McFall, Design, analysis, experimentation, and control of a partially compliant bistable mechanism[J], J. Dyn. Syst. Meas. Contr. 142
(1) (2020) 011008.
[5] T. Bützer, O. Lambercy, J. Arata, et al., Fully wearable actuated soft exoskeleton for grasping assistance in everyday activities[J], Soft Robot. (2020).
[6] B. Demirel, M.T. Emirler, Ü. Sönmez, et al., Semicompliant force generator mechanism design for a required impact and contact forces[J], J. Mech.
Robot. 2 (4) (2010) 045001.
[7] P. Zhao, X. Ge, B. Zi, et al., Planar linkage synthesis for mixed exact and approximated motion realization via kinematic mapping[J], J. Mech. Robot. 8
(2016) 051004.
[8] 1111 S. Bai, J. Angeles, Coupler-curve synthesis of four-bar linkages via a novel formulation[J], Mech. Mach. Theory 94 (2015) 177–187.
[9] S. Hadizadeh Kafash, A. Nahvi, Optimal synthesis of four-bar motion generator linkages using circular proximity function[J], Mech. Mach. Theory 115
(2017) 18–34.
[10] J. Sun, P. Wang, W. Liu, et al., Non-integer-period motion generation of a planar four-bar mechanism using wavelet series[J], Mech. Mach. Theory 121
(2018) 28–41.
[11] R. Singh, H. Chaudhary, A.K Singh, A novel gait-based synthesis procedure for the design of 4-bar exoskeleton with natural trajectories[J], J. Orthop.
Transl. 12 (2018) 6–15.
[12] L.L. Howell, A. Midha, The topological synthesis of compliant mechanisms[J], Mech. Mach. Theory 31 (2) (1996) 185–199.
[13] L.L. Howell, S.P. Magleby, B.M Olsen, Handbook of Compliant Mechanisms[M], John Wiley & Sons, 2013.
[14] L.L. Howell, Compliant mechanisms[C], in: 21st Century Kinematics, Springer, London, 2013, pp. 189–216.
[15] L.L. Howell, A. Midha, Parametric deflection approximations for end-loaded, large-deflection beams in compliant mechanisms[J], J. Mech. Des. 117 (1)
(1995) 156–165.
[16] Y. Tian, B. Shirinzadeh, D. Zhang, Design and dynamics of a 3-DOF flexure-based parallel mechanism for micro/nano manipulation[J], Microelectron.
Eng. 87 (2) (2010) 230–241.
[17] Q. Meng, Y. Li, J. Xu, A novel analytical model for flexure-based proportion compliant mechanisms[J], Precis. Eng. 38 (3) (2014) 449–457.
[18] R. Wang, H. Wu, H. Wang, et al., Design and stiffness modeling of a four-degree-of-freedom nanopositioning stage based on six-branched-chain com-
pliant parallel mechanisms[J], Rev. Sci. Instrum. 91 (6) (2020) 065002.
[19] P.R. Ouyang, W.J. Zhang, M.M Gupta, A new compliant mechanical amplifier based on a symmetric five-bar topology[J], J. Mech. Des. 130 (10) (2008)
104501.
[20] H. Yu, C. Zhang, B. Yang, et al., The design and kinetostatic modeling of 3PPR planar compliant parallel mechanism based on compliance matrix
method[J], Rev. Sci. Instrum. 90 (4) (2019) 045102.
[21] X.D. Chen, Z.L. Deng, S.Y. Hu, et al., Design of a compliant mechanism based four-stage amplification piezoelectric-driven asymmetric microgripper[J],
Micromachines (Basel) 11 (1) (2020) 25.
[22] L.L. Howell, A. Midha, A loop-closure theory for the analysis and synthesis of compliant mechanisms[J], ASME J. Mech. Des. 118 (1) (1996) 121–125.
[23] C.A. Mattson, L.L. Howell, S.P. Magleby, Development of commercially viable compliant mechanisms using the pseudo-rigid-body model: case studies
of parallel mechanisms[J], J. Intell. Mater. Syst. Struct. 15 (3) (2004) 195–202.
[24] G. Chen, S. Zhang, Fully-compliant statically-balanced mechanisms without prestressing assembly: concepts and case studies[J], Mech. Sci. 2 (2011)
169–174.
[25] Y. Yu, Z. Cui, X. Zhao, et al., Design and experiment of parallel robot with compliant joints[J], Trans. Chin. Soc. Agric. Mach. 44 (7) (2013) 274–278.
[26] Y. Li, Y. Yu, Feasibility analysis of compliant joints instead of traditional joints in parallel robot[J], Trans. Chin. Soc. Agric. Mach. 47 (4) (2016) 343–348.
[27] P. Kuresangsai, M.O.T Cole, Kinematic modeling and design optimization of flexure-jointed planar mechanisms using polynomial bases for flexure
curvature[J], Mech. Mach. Theory 132 (2019) 80–97.
[28] P. Bilancia, G. Berselli, Design and testing of a monolithic compliant constant force mechanism[J], Smart Mater. Struct. 29 (4) (2020) 044001.
[29] C.M. Tanik, E. Tanik, Y. Yazicioglu, et al., On the analysis and design of a fully compliant large stroke slider-crank (rocker) mechanism[J], Mech. Sci. 11
(1) (2020) 29–38.
[30] J Su H, A pseudorigid-body 3R model for determining large deflection of cantilever beams subject to tip loads[J], ASME J. Mech. Robot. 1 (2) (2009)
021008.
[31] V. Megaro, J. Zehnder, M. Bcher, et al., A computational design tool for compliant mechanisms[J], ACM Trans. Graph. 36 (4) (2017) 82.
[32] G.M. Chen, F.L. Ma, G.B. Hao, et al., Modeling large deflections of initially curved beams in compliant mechanisms using chained beam-constraint–
model[J], J. Mech. Robot. 11 (2019) 011002.
[33] G.L. Holst, G.H. Teichert, B.D Jensen, Modeling and experiments of buckling modes and deflection of fixed-guided beams in compliant mechanisms[J],
J. Mech. Des. 133 (5) (2011) 051002.
[34] Y.J Song, An elliptic integral solution to the multiple inflections large deflection beams in compliant mechanisms[J], Adv. Mater. Res. (2014)
349–352.
[35] B. Todd, B.D. Jensen, S.M. Schultz, et al., Design and testing of a thin-flexure bistable mechanism suitable for stamping from metal sheets[J], J. Mech.
Des. 132 (7) (2010) 071011.
[36] A.M. Zhang, G.M Chen, A comprehensive elliptic integral solution to the large deflection problems of thin beams in compliant mechanisms[J], J. Mech.
Robot. 5 (2) (2013) 021006.
[37] F. Bre, V.D Fachinotti, A computational multi-objective optimization method to improve energy efficiency and thermal comfort in dwellings[J], Energy
Build. 154 (2017) 283–294.
[38] K. Deb, A. Pratap, S. Agarwal, et al., A fast and elitist multiobjective genetic algorithm: NSGA-II[J], IEEE Trans. Evol. Comput. 6 (2) (2002) 182–197.
[39] G.L Wu, Optimal structural design of a Biglide parallel drill grinder[J], Int. J. Adv. Manuf. Technol. 90 (2017) 2979–2990.

15