Mechanism and Machine Theory 45 (2010) 1924–1941

Contents lists available at ScienceDirect

Mechanism and Machine Theory

j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / m e c h m t

Compliant folded beam suspension mechanism control for rotational dwell

function generation using the state feedback linearization scheme
Ayşe Tekeş, Ümit Sönmez ⁎, Bilin Aksun Güvenç
Mechanical Engineering Department, MEKAR Laboratory, Istanbul Technical University, Gumussuyu, 34437, Istanbul, Turkey

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

Article history:
The control of a slider–crank mechanism driven by a compliant, folded beam mechanism is
Received 13 March 2009
Received in revised form 17 July 2010
investigated in this study. The compliant folded beam mechanism, also called the compliant
Accepted 25 July 2010 double arm mechanism, has application areas at both the micro and macro levels. The
compliant folded beam mechanism control in micro dimensions, driven by a force actuator is
investigated here. The kinematic synthesis of the mechanism has been studied using
Keywords:
Compliant mechanisms
mathematically exact nonlinear Elastica theory. The equivalent stiffness of the large
Compliant double parallel arm mechanisms deflecting fixed-free flexible beam, the single arm beam and the complete folded beam
Compliant folded beam mechanisms mechanism (double arm mechanism) are computed and, the complete mechanism stiffness is
Slider–crank mechanism represented by a polynomial function that is curve fitted to the nonlinear inextensible exact
Dynamics of compliant mechanisms beam theory solution. The dynamic response of the mechanism is obtained by solving the
Control of nonlinear springs nonlinear equation of motion numerically using Runge–Kutta methods. The emphasis of the
Trajectory control paper is placed on the subject of control of compliant mechanisms incorporating nonlinear stiffness
Function generation
(having large deflections) for function generation. The trajectory control of double cranks is
achieved by considering linearization by the state feedback since the compliant folded beam
mechanism (the double parallel arm mechanism) has a nonlinear stiffness. A PD controller is
then applied to the feedback linearized system. The controller coefficients are determined to
satisfy desired specifications on the output of the system.
© 2010 Elsevier Ltd. All rights reserved.

1. Introduction

The concept of compliant mechanisms started with the challenge to exploit the flexibility of linkages for beneficial purposes.
Instead of avoiding linkage flexibility problems of mechanisms, employing the flexibility itself to one's benefit leads to advantages
over rigid body mechanisms. A compliant mechanism can exploit the flexibility of its components for the creation of motion which
can be very difficult and expensive to generate by rigid body mechanisms.
A controller design for function generation of a parallel arm mechanism is presented in this paper. The mechanism employs
flexible members undergoing large deflections. Modeling of such mechanisms draws on a broad range of knowledge in several
areas including: compliant mechanisms, function generation, and control of nonlinear systems for the desired motion.
A compliant mechanism is called fully or partially compliant [1] depending on the existence of traditional links and joints. Fig. 1
shows a fully compliant five-bar mechanism designed using the traditional rigid links and the elastic joints. A fully compliant
mechanism shown in Fig. 1 does not require the assembly of its parts.
Compliant joints have certain disadvantages such as a decreased fatigue life and increased geometric stress concentration
factor [1,2]. The use of flexible members can make a mechanism comparatively lighter, thus enhancing its use in applications
requiring low weight. The design of a compliant mechanism [3] is more complicated than the design of a rigid mechanism as it

⁎ Corresponding author.
E-mail addresses: atekes@itu.edu.tr (A. Tekeş), usonitu@gmail.com (Ü. Sönmez), guvencb@itu.edu.tr (B.A. Güvenç).
URL: http://mekar.itu.edu.tr (Ü. Sönmez).

0094-114X/$ – see front matter © 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.mechmachtheory.2010.07.005
 A. Tekeş et al. / Mechanism and Machine Theory 45 (2010) 1924–1941 1925

Fig. 1. A fully compliant five-bar mechanism design (made from one piece).

requires large deflection analysis, consideration of the stored strain energy and the solution of transcendental loop closure
equations.
It is common in compliant mechanism design to explore the flexible member deflection to its outmost limit, thus requiring a
large deflection analysis. The large deflection analysis of compliant mechanisms might be achieved using one of the three methods
of; 1) using a Pseudo Rigid Body Model, 2) using a nonlinear FEM software package, and 3) using the Elastica theory. These
methods are summarized as follows:

• A compliant mechanism might be modeled with a Pseudo Rigid Body Model (PRBM), which is a rigid body equivalent of the flexible
linkage having equivalent torsional springs at its joints [4]. PRBM [5] makes the compliant mechanism analysis relatively easier.
Compliant mechanisms are usually designed for simple motion requirements and the trial and the error method are commonly used
to design novel compliant mechanisms. Once a desired designed is found, the structural properties of a compliant mechanism might
be checked and revised according to the required load and flexible beam maximum bending stress [3].
• Another efficient analysis tool of compliant mechanisms is to use nonlinear FEM. Numerical techniques for large deflection
analysis of structures are available in the literature. These techniques address geometric and material nonlinearities of
structures. Most of these techniques use Finite Element Methods (FEMs) of different kinds. There are mainly two kinds of
incremental nonlinear FEMs considering geometric nonlinearities, the total Lagrangian [6] and the updated Lagrangian [7,8]
methods. Material nonlinearities have also been considered in compliant mechanism design. Jung and Gea [9] presented
compliant mechanism design with nonlinear materials using topology optimization, for example. They used a general
displacement functional with a nonlinear material model in the topology optimization formulation to maximize the mechanical
advantage of a force inverter mechanism.
• Elastica theory is another commonly used compliant mechanism analysis tool. Elastica theory provides a closed form of the
solution. In other words, the functional solution relation between the applied load and the deflection cannot be expressed
explicitly in the form of y = y(x). The mathematical exact equations need to be solved first to express such explicit y = y(x)
relationship, then the exact solution might be represented by an accurate polynomial [10]. The only disadvantage of the Elastica
theory is that its application is restricted to simple geometries and simple loading conditions. Therefore compliant mechanism
designs that can be investigated using the Elastica theory consist of simple flexible members and simple loading conditions.

In this investigation, the Elastica theory is used since the solution of a large deflection, fixed-free cantilever beam (which makes
one half of the compliant mechanism arm considered here) is available in literature [11]. Magnetic actuation is frequently used in
precision instruments. As an example, a pin actuator for a printer mechanism [12] is shown in Fig. 2. In this printing device, a
hammerhead attached to a flexible link is accelerated by a magnetic force. The hammerhead's kinetic energy is used to push ink
from a ribbon onto paper.
Dynamics of compliant mechanisms started to appear in literature since the early 90s. Snyder and Wilson [13] studied the
dynamics of a single limb compliant robot arm with a payload at the tip. They derived and simulated the nonlinear equations of
1926 A. Tekeş et al. / Mechanism and Machine Theory 45 (2010) 1924–1941

Fig. 2. A printer head driven by a magnetic force actuator, adapted from Hendriks [12].

motion for a range of system parameters, emphasizing the ramp pressure loading. Nahvi [14] studied static and dynamic behavior
of a compliant mechanism consisting of single flexible elements using the geometrically nonlinear FEM method. Lyon et al. [15]
used PRBM to predict the first modal frequencies of compliant mechanisms and compared them with experimental results. Boyle
et al. [16] derived the analytical equations of a compliant constant compression force mechanism using the Pseudo Rigid Body
Model and a lumped parameter representation of flexible elements. Yu et al. [17] developed the dynamic equations of general two
dimensional compliant mechanisms utilizing the PRBM method based on dynamic equivalence principles. Sönmez [18]
investigated the dynamic buckling response of an imperfect Elastica subject to various loading conditions.
The remainder of the paper is organized as follows. In Section 2, the general description of the device and the nonlinear theory
of a fixed-free, flexible beam are presented. This is followed by the load deflection representation (equivalent stiffness function) of
a fix free beam, of a flexible arm and, of the complete compliant mechanism. In Section 3, the feedback linearization application
and the subsequent trajectory control design for the lumped mass, the damper and the nonlinear spring model used for controller
design is presented. In Section 4, two function generation examples are presented. The possible application areas and the
concluding remarks are presented in Sections 5 and, 6 respectively.

2. Compliant folded beam (double arm) mechanism

In this section, a compliant folded beam (double arm) compliant mechanism design is investigated both kinematically and
dynamically. This mechanism consists of eight flexible arms, two rigid side arms, two rigid cranks, and a rigid coupler (a shuttle),
actuated by a comb drive. The load deflection behavior of the parallel arms is obtained by using the cantilever beam's large
deflection results. First, the nonlinear spring rate of this member (a fixed-free beam) is found using the inextensible Elastica
theory. Then, the spring rates of the arm beam and the complete mechanism are calculated. Finally, the calculated nonlinear spring
rate of the complete mechanism is used to obtain the equation of motion of the whole system.
The dynamic simulation of the compliant parallel arm mechanism is studied using numerical simulation procedures
considering the desired force actuation coming from a comb drive unit. The simulation results of the compliant double arm
mechanism have been obtained by including the translational inertia of the rigid coupler (the shuttle). The relation between the
compliant double arm mechanism and the crank angle is derived by its geometry alone since the crank is assumed to be massless
(its mass is quite small comparing to the shuttle). Load deflection characteristics of the flexible beams are represented by
polynomial curve fits (obtained from nonlinear inextensible exact beam theory) that are then used as the nonlinear stiffness
characteristics in the lumped parameter model.
An investigation of a small deflecting parallel arm mechanism (see Fig. 3) driven by a magnetic actuator for nano precision was
presented by Smith et al. [19]. A similar fixture of a folded beam mechanism in micro dimensions is investigated in this paper.
The compliant double arm mechanism considered in this paper has large deflecting arms as shown in Fig. 4. A macro level
prototype of the same compliant double arm mechanism manufactured for visualization purposes is displayed in Fig. 5. It should

Fig. 3. A small deflection compliant parallel arm mechanism driven by a magnetic force actuator for nano precision, adapted from Smith et al. [19].
 A. Tekeş et al. / Mechanism and Machine Theory 45 (2010) 1924–1941 1927

Fig. 4. Slider–crank mechanism driven by a compliant parallel double arm mechanism.

be noted that an arbitrary function generation can be achieved through proper trajectory control of this mechanism. This makes
the controlled compliant parallel arm mechanism a valuable tool for motion generation.

2.1. Large deflection analysis of a fixed-free cantilever beam subject to a vertical applied load

Classical formulation of large deflection analyses was initiated by Jacob Bernoulli who showed that beam curvature at any point
is proportional to the bending moment at that point. Euler [20] obtained several solutions of Elastica problems. When deflections
become large, beam curvature can no longer be approximated. The governing differential equations are expressed in terms of
nonlinear functions that are typical of flexible straight beams, generally leading to highly implicit relationships involving elliptic
integrals and functions.
The large deflection analysis of a fixed-free beam formulated by Bishop [11], Frisch-Fay [21] and Sathyamoorthy [22] is
summarized first. Let us assume a cantilever beam subject to a vertical applied load that deflects with the horizontal δx and vertical
δy tip displacements, as shown in Fig. 6. The notations used in [22] are adopted in this paper.
The Bernoulli–Euler moment curvature relationship may be written as,

−EIϕ;s = −P ðL−x−δx Þ ð1Þ

Where EI is the flexural rigidity of the beam, ø is the beam angle and ø,s = dø/ds is the beam's curvature. Taking the derivative of
the above equation and using dxds
= cosϕ and dyds
= sinϕ the following equation may be obtained:

2
−EIϕ;ss = −ðP = EI Þcosϕ = −α cosϕ ð2Þ

Fig. 5. Macro compliant double parallel arm mechanism prototype made for visualization purposes.
1928 A. Tekeş et al. / Mechanism and Machine Theory 45 (2010) 1924–1941

Fig. 6. Cantilever beam subject to an end load.

Multiplying the above relation with ϕ, s and using the boundary conditions of s = L, ϕ = ϕ1 and the curvature at the free end
being ϕ, s = 0 and integrating, one could obtain the following:
pffiffiffi 1=2
ϕ;s = 2αðsinðϕ1 Þ−sinϕÞ ð3Þ

The value of ϕ1 cannot be obtained directly from Eq. (3). Using the inextensibility of the neutral axis, the following relationship
may be written:
pffiffiffi −1 = 2
2Lα = ∫½sinðϕ1 Þ−sinϕ dϕ ð4Þ

Defining a new variable, θ, such that, ð1 + sinϕÞ = ð1 + sinϕ1 Þsin2 θ = 2k2 sin2 θ and rewriting Eq. (4) one may obtain the
following relation:
 2
−1 = 2
2
Lα = ∫ 1−k sin θ dθ ð5Þ

Using the elliptic integral definitions Eq. (5) might be written as:

Lα = F ðkÞ−F ðk; θ1 Þ: ð6Þ

pffiffiffi
where, sinθ1 = 1 = 2k F ðkÞ and F(k,θ1) are complete and incomplete elliptic integrals of first kind. In order to calculate the
vertical deflection δy considering the following relation,
  
w;ϕ ϕ;s = w;s = sinϕ ð7Þ

and integrating Eq. (3), the vertical deflection might be obtained as

 pffiffiffi  sinϕ
δy = ∫dw = 1 = 2α ∫ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi dϕ ð8Þ
ðsinϕ1 −sinϕÞ

The above Eq. (8) might be expressed in terms of θ as

 2

  2k2 sin θ−1
δy = L = ð1 = LαÞ∫ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi dθ: ð9Þ
2
1−k2 sin θ

The above relation may be rewritten using elliptic integrals as

 
δy = L = ð1 = LαÞT½F ðkÞ−F ðk; θ1 Þ−2EðkÞ + 2Eðk; θ1 Þ: ð10Þ

where E(k) and E(k,θ) are complete and incomplete elliptic integrals of the second kind. In order to calculate the horizontal tip
deflection δx: let us reconsider Eqs. (1) and (3) by taking x = ϕ = 0 and substituting into Eq. (1) to obtain
pffiffiffi 
1=2
ð1−δx Þ = L = 2 = αL ðsinϕ1 Þ ð11Þ
 A. Tekeş et al. / Mechanism and Machine Theory 45 (2010) 1924–1941 1929

Fig. 7. Normalized load deflection plots of a cantilever beam (large deflection theory).

These nonlinear equations express the large tip deflection of a cantilever beam subject to a vertical load in closed form. These
equations are solved numerically using the Newton–Raphson method and the normalized load PL2/EI and normalized deflection δx
and δy plots shown in Fig. 7 are obtained. Considering the linear theory, cantilever beam vertical displacement may be represented
by the well known formula; δy = PL3/(3EI) and the vertical tip displacement obtained using the linear theory [23] is also presented
in Fig. 7. It is seen that the linear and the nonlinear theory give the same results up to the normalized vertical displacement
of δy/L = 0.18.
In order to obtain the equivalent stiffness of the cantilever beam in terms of simple functions, a rational function is fit to the
exact solution. The following rational function is found to be suitable for the required high accuracy (corresponding correlation
function Г N 0.99999).

27050u
p= ð12Þ
u3 −13470u2 + 3321u + 8347

In order to find the deflected shapes for the fixed-free cantilever beam, the upper limit for the following integral should be
changed according to (x,y) point of the beam,

up
dθ
∫ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð13Þ
2
low 1−k2 sin θ

For the flexible beam tip deflection, the lower and the upper limits are represented by θlow and θup respectively. At the fixed end
the beam angle ϕ = 0 and at the free end ϕ = ϕ1 therefore,
 
−1 1
θlow = sin pffiffiffi ð14Þ
k 2
θupp = π = 2

The coordinates of the Elastica are given by

1
yElastica = T½F ðk; θÞ−F ðk; θlow Þ−2Eðk; θÞ þ 2Eðk; θlow Þ ð15Þ
α

xElastica = hT½cosðθlow Þ−cosðθÞ: ð16Þ

where h is given by, h = 2k / α. The deflected shapes of the fixed-free beam subjected to vertical end loading are shown in Fig. 8.
The shape of the large deflection flexible beam may be obtained either with Elastica theory or with nonlinear FEM analysis.
These shapes are helpful to determine the functional and the workspace requirements of the designed mechanism. The compliant
parallel arm mechanism functioning, subject to a quasi-statically increasing load is shown in Fig. 9. The rigid coupler in Fig. 9
translates parallel to the ground.

2.2. The equivalent stiffness of the folded beam mechanism

In this section, the equivalent stiffness of the compliant folded beam (the double parallel arm) mechanism is acquired by using
a rational function fit to the related exact solution of the cantilever beam. First, the load deflection curve is obtained for the fixed-
1930 A. Tekeş et al. / Mechanism and Machine Theory 45 (2010) 1924–1941

Fig. 8. Large deflection shapes of the fixed-free cantilever beam subjected to a vertical end load.

free beam by using the normalized load deflection curve of the cantilever beam presented in the previous section. Then, the arm
beam and the complete double parallel arm mechanism load-deflection relation are obtained by starting from the fixed-free beam
results. Fig. 10 shows the upper half of the suspension can be used to help in visualizing the relations concerning the fixed-free
beam, the whole arm beam and the complete compliant folded beam mechanism.
Micro machined springs are often folded or bent around [24] a radius to improve the performance of the overall device.
Consequently, the spring constant estimates require combinations of individual segment values. The load-deflection relation of the
compliant double parallel arm is computed by using the load-deflection plot of fix free beam and relating the necessary physical
parameters of the cantilever beam to the complete double arm mechanism. The displacement and the load relationship of the
compliant double arm mechanism are related to the normalized displacement and the normalized load of the fixed-free beam as

DisplacementTotal = δy × L × 4
EI ð17Þ
LoadTotal = × loadnorm × 8
L2

The load deflection relations of the fixed-free beam, the arm beam, the complete mechanism are shown in Fig. 11 for Case I.

2.3. Lumped parameter modeling of the slider–crank mechanism driven by a compliant parallel double arm mechanism

The compliant parallel double arm mechanism has a single D.O.F. and it can be modeled with a classical lumped mass-spring
system (see Fig. 12). The lumped system stiffness is represented by its Elastica equivalent function kNF(x).
The equation of the lumped mass–spring–damper system is
::
m x + c ẋ + kNF ðxÞ = F ðt Þ ð18Þ

Fig. 9. The large deflection trajectories of the parallel arm mechanism subject to an increasing load.
 A. Tekeş et al. / Mechanism and Machine Theory 45 (2010) 1924–1941 1931

Fig. 10. Upper half of the compliant suspension.

The aim of the system is to transform the linear motion to the rotational motion. The linear motion is achieved by actuating the
parallel double arm mechanism and the rotational motion is achieved by attaching a slider–crank mechanism to the compliant
folded beam mechanism's shuttle (Fig. 13). Since the coupler link and the crank mass are small as compared to the inertia of the
shuttle, they might be neglected.
The relation between the crank angle and the slider position is
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2
x = rcosðθÞ + LcosðφÞ = rcosðθÞ + L2 −r2 sin θ−x0 ð19Þ

In order to compute the rotation crank angle at the four quadrants of the unit circle; the sine, and the cosine relations are both
necessary. Crank angle is obtained by MATLAB function ‘atan2’ by using the required relations:

r 2 + x2 −l2
cos ðθÞ = ð20Þ
2xr
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
!ffi
u
u r
2
+ x
2
−l
2 2
sinðθÞ = tr 2 −
2xr

3. Mechanism control

The control of a system for the arbitrary initial conditions and the arbitrary desired dynamic behavior requires the controller to
force the system to exhibit a dynamic response that is as close as possible to the desired one. Tracking of a desired trajectory is one

Fig. 11. The load deflection characteristics of the fixed-free beam, the arm beam, and the complete mechanism.
1932 A. Tekeş et al. / Mechanism and Machine Theory 45 (2010) 1924–1941

Fig. 12. Single D.O.F. mass, damper and spring system.

of the most basic control objectives. The equation of motion of compliant mechanisms is usually nonlinear due to nonlinear system
stiffness. This makes trajectory tracking with nonlinear controllers quite significant for nonlinear complaint mechanism control
[25].
The compliant folded beam mechanism might be used as an indexing or a dwell mechanism if its rigid shuttle can be positioned
at the desired locations at specified times and kept there for the specified durations. In this investigation, only the desired motion
control is studied. It should be noted that the dwell motion is one of the many possible function generations that can be achieved
with the trajectory control of this mechanism. This is also demonstrated by choosing several different functions at the non-dwell
part of the motion. The magnetically actuated parallel arm mechanism with a control law is well suited for use as a function
generator. If a desired velocity is achieved at a specified output location, the rigid coupler might be used as a punch mechanism.
This idea is explored [26] concerning a compliant slider–crank mechanism with a buckling beam.
Even though local PID controllers are well suited for position control, they might cause an overshoot of the response. Overshoot
is an undesired behavior in some applications. Simultaneous local control loops result in overall delays. Trajectory tracking is
required in high-speed applications [27]. In practice, the system dynamics performance depends on several different variables. If a
variable under question is measurable, it can be used to characterize the system response or to tune the controller parameters [28].
The mechanical system model concerning the equation of motion is nonlinear due to the nonlinear behavior of the elastic
members modeled by higher order polynomials. In order to design linear controllers, a linear model of the system has to be
obtained. A linear model is obtained by linearizing the nonlinear governing equations of motion about the operating point. This
method will work best when the generated functions have the same amplitude as that of the operating point. An alternative
method is to linearize the actual system behavior by feedback cancellation of the nonlinearities in feedback linearization. The
feedback linearization method [29] has superiority to the operating point linearization of the model, because it linearizes the
system for the whole range of input variables instead of working with a linear model of a nonlinear system that is only accurate in
the vicinity of the operating point.

3.1. The theory of linearization by state feedback

The theory of linearization by state feedback is summarized in this section. A nonlinear system can be represented by the
following set of equations,

ẋ = f ðxÞ + g ðxÞu
ð21Þ
y = hðxÞ

Where f(x) and g(x) are smooth vector fields on ℜnand h(x) is a smooth nonlinear function, x = ½x1 ; x2 ; …xN T is the state vector,
u is the system input and y is the output of the system. The nonlinear system can be linearized for all operating points by
introducing the new input variable v, which makes the system exactly linear from the new input v to the output y. This is different
from approximate Jacobian linearization.

Fig. 13. Slider–crank mechanism.

 A. Tekeş et al. / Mechanism and Machine Theory 45 (2010) 1924–1941 1933

Differentiating the output y in Eq. (21) above with respect to time results in

∂hðxÞ ∂hðxÞ ∂hðxÞ

ẏ = ẋ = f ðxÞ + gðxÞu
∂x ∂x ∂x ð22Þ
= Lf hðxÞ + Lg hðxÞu

where Lf hðxÞ is the Lie derivative of h with respect to (wrt) f and Lg hðxÞ is the Lie derivative of h wrt g.
For a given nonlinear system,
• If Lg hðxÞ N δ1 N 0, then the state feedback control law defined as

1 h i
u= −Lf hðxÞ + v ð23Þ
Lg hðxÞ

where δ1 is a positive real number and v is the new input expressed by ˙y = v yields a first order linear system from new input v
to output y.
• If Lg hðxÞ≡0; ∀x∈U, then by taking one more derivative of Eq. (22) results in
:: 2
y = Lf hðxÞ + Lg Lf hðxÞu ð24Þ

• If Lg Lf hðxÞ N δ2 N 0, where δ2 is positive real number, the control law

1 h i
2
u= −Lf hðxÞ + v ð25Þ
Lg Lf hðxÞ

results in the second order linear system ÿ = v between the new input v and the output y.
In order to generalize the rule; if Lg Lif hðxÞ≡0 where i = 0, 1, …, n and n is the smallest integer for which Lg Ln−1
f N δn , where δn is a
positive real number, the control law is derived as

1 h i
n
u= −Lf hðxÞ + v ð26Þ
Lg Lfn−1 hðxÞ

and results in an nth order linear system from the input v to the output y [29].

3.2. Linearization by state feedback

The
 displacement  and the velocity of the shuttle part of the mechanism are selected as the state variables as:
x1 ðpositionÞ
x= ; y = ½x1 ðpositionÞ:
x2 ðvelocityÞ
The nonlinear functions required to represent the nonlinear system are
2 3 2 3
x2 0
f ðxÞ = 4 kNF ðx1 Þ 5; g ðxÞ = 4 1 5; hðxÞ = x1 ð27Þ
−
m m

The system is forced to behave like a linear system by state feedback linearization. The derivative of the output function is taken
until the input–output relation is obtained. The first and the second Lie derivatives are

∂hðxÞ ∂hðxÞ
ẏ = f ðxÞ + g ðxÞ = Lf hðxÞ + Lg hðxÞu
∂x ∂x ð28Þ
⇒ ẏ = x2

The first derivative of the output does not include the input u. Therefore, the second derivative is written as
::
y = L2f hðxÞ + Lg Lf hðxÞu
::
v= y
  ð29Þ
1 h :: i 1 kNF ðxÞ
2
u= y−Lf hðxÞ = +v
Lg Lf hðxÞ 1=m m

where, v is the synthetic input or the control input.

1934 A. Tekeş et al. / Mechanism and Machine Theory 45 (2010) 1924–1941

The trajectory control of double cranks is studied considering the linearization by state feedback since the compliant double
arm mechanism has a nonlinear stiffness. After the feedback linearization, a PD controller is used as the main controller. The
coefficients of the PD controller are found by using the desired specifications expected from the output of the system. The control
rule is developed for the system to generate a desired function of dwell motion. The PD controller is written as

v = Kp ðxdes −xÞ + KD ð ẋdes − ẋÞ = Kp eðt Þ + KD ėðt Þ ð30Þ

where Kp and Kd are the proportional and the derivative gains of the PD controller, respectively.

3.3. Setting up the system control law

Linear control techniques might be used after this stage noting that linearized double integral relationship trajectory controller
design is relatively simple. The error function may be defined as;

e = yr ðt Þ−yðt Þ
ė = ẏr ðt Þ− ẏðt Þ ð31Þ
:: :: ::
e = yr ðt Þ− yðt Þ

The input function could be written as,

::
u = y−α1 ð ẏr − ẏÞ−α2 ðyr −yÞ ð32Þ

The characteristic equation of the closed loop system trajectory error becomes
::
e + α1 ė + α2 e = 0 ð33Þ

The characteristics equation gives exponentially stable response if the roots of the equation are on the left side of the complex plane.
Therefore, if the initial conditions are assumed to be zero, the error e(t) converges to zero with increasing time. The coefficients α1
and α2 in Eq. (33) are chosen to result in characteristic equation roots that provide desired trajectory following. The MATLAB
Simulink model of the compliant arm and its control unit are given in Appendix A.
The first step in the simulation of controlled dynamic response of the compliant mechanism considered here is to
perform feedback linearization by using Eq. (24). The input of the nonlinear lumped mass–spring–damper system is the
applied force and the output is the displacement. After feedback linearization, this changes and the new input becomes the
desired trajectory (generated function) and the output is the displacement. The controller parameters are determined for
obtaining satisfactory trajectory following. The results are then multiplied by constant gains to convert the displacement
from m to cm and mm.

4. The function generation of the double parallel arm mechanism

In this section, the dimensions of an example mechanism are presented first. Then two function generation examples are
studied. After that trajectory control results are presented.

4.1. Mechanism dimensions

If the mechanism needs to be moved in the large deflection range of flexible beams or the generated function needs to
be modified for different tasks, then different linearized stiffness functions should be considered for each different
generated function. This approach requires a look up table with a stack of linearized stiffnesses. Therefore, a more direct
approach is needed. Linearization by the state feedback method is suitable to control the system with a nonlinear
stiffness. The controlled force may be generated easily by a comb drive in micro dimensions and by a magnetic drive in
macro dimensions.
The following dimensions of the compliant parallel arm mechanism are chosen for an example problem as shown in Table 1,
and they are similar to the dimensions presented in Compliant Mechanisms Book [5], chapter 3.3.10. The cross sectional area of the
flexible beam is chosen from commonly used MEMS beam manufacturing methods (one layer 2 μm, two layers 3.5 μm and three
layers 4.5 μm). The rigid shuttle mass is chosen in accordance with mechanism flexible arms. Slider–crank mechanism dimensions
(r, L) are designed with respect to the maximum displacement of the flexible arms of the double arm mechanism.

4.2. Dwell motion as a desired trajectory

A dwell motion is defined as the mechanism output task. A dwell is defined as zero output motion for nonzero input motion or
force. A dwell motion can be described by four parameters: rise time, dwell time, return time and lift as shown in Fig. 14. In
 A. Tekeş et al. / Mechanism and Machine Theory 45 (2010) 1924–1941 1935

Table 1
Parallel arm mechanism's dimensions and its structural properties.

Mechanism part Dimension and properties

Flexible arm length 250 μm

Rigid coupler (shuttle) mass 4.025 × 10− 12 kg
Beam cross section In plane thickness × Out of plane width(3 μm–1 μm) × 3.5 μm (Case I–Case II)
I = bh3/12 7.1458 × 10− 24 m4
E (elasticity modulus) 1.4 ⁎ 1011 N/m2 (poly-silicon)
r (length of the crank) Case I: 40 × 10− 6 m = 40 μm
Case II: 100 × 10− 6 m = 100 μm

compliant dwell mechanisms, these characteristics depend on the force and displacement interaction between the input link and
the output link [30] rather than displacement interaction as in regular rigid body mechanisms.
The example mechanism with its parameters defined in Table 1 might deflect ±70 μm and ±190 μm in its elastic range for
Case 1 and Case 2, respectively. The maximum bending stresses for these cases are 864 MPa and 830 MPa, corresponding to slider
displacements of ±70 μm and ±190 μm, respectively.
Two cases are presented. The first case includes dwell periods with different rise functions including a triangle and a sinus
function to demonstrate that the method proposed in this paper is suitable for any arbitrary function generation. The second case
includes different rises in the nonlinear range showing that, the method proposed in this paper is capable of handling nonlinear
stiffness functions in the large deflection range.

4.2.1. Case I
A dwell rotation with the same rise and different functions are obtained first. Figs. 15–17 show the crank's desired trajectory
and the model output, corresponding slider's desired trajectory and its model output and the controlled mechanism trajectory
following error. This control example could have been achieved with a local linearization at the operating point (the dwell rise)
and then using Laplace transform techniques.
The maximum slider deflection is 37.72 μm and, the corresponding cantilever beam deflection is 9.42 μm, the normalized
cantilever beam vertical deflection is 0.07 which is in the linear range of beam deflection. The maximum crank angle error is 0.5°
(very small) and goes to zero at the waiting periods.

4.2.2. Case II
A triple rotational dwell profile with different rise amplitudes is considered for this example (see Fig. 18). The desired crank
angle profile is used to form the desired slider positions and the corresponding slider positions are shown in Fig. 19.
It is difficult to obtain the above trajectory using a linearized model because the operating points are different in each dwell. A
stack of the linearized models might be used for each dwell case but this requires defining the linearized sets of stiffness

Fig. 14. Dwell motion characteristics.

1936 A. Tekeş et al. / Mechanism and Machine Theory 45 (2010) 1924–1941

Fig. 15. Desired triple dwell motion with different profiles of the crank for Case I.

Fig. 16. Corresponding desired triple dwell motion profile of the shuttle for Case I.

parameters for each case, and makes the model more complicated. The results are obtained by considering the input trajectory in
Fig. 18, and the following results are presented in Figs. 19–20.
The maximum slider deflection is 179.5 μm and the corresponding normalized cantilever beam vertical deflection is 0.356,
which is in the nonlinear range. The maximum sudden change of the crank angle error is −2° and goes to zero at waiting periods.

4.3. Stress calculation for bending failure

The flexible bending beam stress analysis is presented in this section. The flexible beam's curvature changes as the flexible
beam deflects. Considering only bending stress assuming the beam middle axis is inextensible (Bernoulli Beam), maximum
bending moment occurs at the fixed end. The primary bending stress in case of nonlinear deflection is expressed by

M 6P ðL−δx Þ
σb = y= ð34Þ
I bh2

Fig. 17. The rotation error between the reference crank angle and the model for Case I.
 A. Tekeş et al. / Mechanism and Machine Theory 45 (2010) 1924–1941 1937

Fig. 18. Desired triple dwell motion profile of the crank (Case II).

Fig. 19. The shuttle position of the mechanism when controlled by feedback linearization (Case II).

Where P and δx may be found from the nonlinear load deflection plot corresponding to δy. Considering the linear theory, the tip
deflection of the fixed-free beam is

PL3
δy = ð35Þ
3EI

Therefore, the stress formula reduces to

3Ehδy
σb = ð36Þ
2L2

Fig. 20. The rotation error (Case II).

1938 A. Tekeş et al. / Mechanism and Machine Theory 45 (2010) 1924–1941

Fig. 21. Part of the experimental valve control fixture presented by Nitu et al. [31].

Using the equations above, the maximum bending stress of the rectangular cross section area may be calculated where b and h are
cross sectional dimensions, out of plane width and in plane thickness respectively.
The maximum bending stresses for Case I and Case II are calculated as 462.33 MPa and 778.05 MPa, respectively, and they are
both below the ultimate tensile strength value of Sut = 930 MPa for polycrystalline silicon (polysilicon). Maximum normalized
displacement of the fixed-free beam for Case I is 0.07 and for Case II is 0.36. The first one is in the linear range and the second one is
in the nonlinear range.

5. Possible application areas

The controlled double parallel arm mechanism might find application in different areas. For example, a double parallel arm
mechanism might be used to open and close a valve, controlling the fuel intake of a micro or a macro combustion engine.
Electromechanically driven cam mechanisms have appeared in literature. Nitu et al. [31] have presented the development of a
fuel distribution system for thermal engines, electrically actuated and electronically controlled (see Fig. 21), aimed to replace the
classical cam-follower mechanism. The valve was supported by two oppositely placed linear springs to ensure its return to the
original configuration as shown in Fig. 22. This work consisted of design, modeling and simulation of an electromagnetic actuator,
able to drive a valve with a certain dynamic performance, within the desired stroke and development of an experimental system
including electronic control and corresponding software.
The double parallel arm mechanism might be an alternative to the traditional cam controlled mechanical system with valves.
Instead of using linear springs and their attachments described above, one-piece compliant shaft and valve carrier mechanism
might be used. A similar mechanism in 3D suspension was designed and built for micro-scribing processes using principles of
compliant mechanisms [32]. The compliant end-effector design is a significant improvement over existing scribing alternatives
and can produce smooth and uniform scribed lines that exhibit less material chipping. The proposed mechanism shown in Fig. 23
can be actuated with a linear solenoid drive. Linear solenoids either convert electrical energy into mechanical power via a plunger
with an axial stroke in a push or pull action. Electrical current is supplied to a coil housed in a frame. The resulting magnetic field
draws the plunger from its un-powered, extended position to a seated position against a backstop or pole piece. Some of the linear
solenoids are especially designed (clappers or flappers are hinged solenoids) for low-force, long-life applications which are

Fig. 22. Adapted from Nitu et al. [31].

 A. Tekeş et al. / Mechanism and Machine Theory 45 (2010) 1924–1941 1939

Fig. 23. The proposed valve carrier mechanism of a combustion-engine fuel-inlet. The compliant mechanism consists of three parallel double arm mechanisms as a
suspension.

suitable to actuate the proposed parallel arm valve suspension mechanism. The proposed system could open and close the valve at
the desired time and the desired positions (function generation). Since the double parallel arm shaft axis do not change in time,
the proposed valve carrier mechanism can be actuated with a solenoid driven actuator having a control unit.
The double parallel arm mechanism also finds extensive applications in micro dimensions as a suspension system for the comb
drives, the translational motion of the shuttle suspended above the ground can be converted to a rotational motion with the
addition of a coupler and a crank pair as shown in Fig. 24. The force control of the comb drive would be used to obtain desired dwell
output with different rise angles and periods, which might find application in several areas including opening and closing of
several flow channels, connecting different electrical circuits and use as a mechanical password key for security purposes.

6. Conclusion

In this paper, the control of a parallel arm double mechanism for a desired dwell trajectory is presented. The rotational dwell
motions are defined as the compliant mechanism motion task and the corresponding desired displacement, desired velocity and
desired acceleration functions are used with a feedback linearization system to achieve the desired trajectory. The feedback

Fig. 24. Compliant MEMS design driven by a comb drive for function generation.
1940 A. Tekeş et al. / Mechanism and Machine Theory 45 (2010) 1924–1941

linearization system is found suitable for the compliant mechanism control due to its nonlinear stiffness function. Several
examples, designs are presented and introduced where the proposed method may find application areas in both macro and micro
dimensions.

Acknowledgment

This project is sponsored by Turkish Scientific and Technological Research (TÜBİTAK) Center Career Award 105M191 of the
corresponding author.

Appendix A

References

[1] S. Kota, J. Joo, Z. Li, S.M. Rodgers, J. Sniegowski, Design of compliant mechanisms: applications to MEMS, Analog Integr. Circ. Signal Process. 29 (2001) 7–15.
[2] N. Lobontiu, Compliant Mechanisms: Design of Flexure Hinges, CRC Press, Boca Raton, 2003 p. 5.
[3] Ü. Sönmez, C. Tutum, Compliant bistable mechanism design incorporating Elastica buckling beam theory and pseudo-rigid-body model, ASME J. Mech. Des.
130 (2008).
[4] L.L. Howell, A. Midha, A loop closure theory for the analysis and synthesis of compliant mechanisms, ASME J. Mech. Des. 118 (1996) 121–125.
[5] L.L. Howell, Compliant Mechanisms, First ed., John Wiley, New York, 2001 pp. 1–20.
[6] K.J. Bathe, Finite Element Procedures in Engineering Analysis, Prentice-Hall, Englewood Cliffs, New Jersey, 1982.
[7] K.J. Bathe, E. Ramm, E.L. Wilson, Finite element formulation for large deformation dynamic analysis, Int. J. Numer. Methods Eng. 9 (1975) 353–386.
[8] T.Y. Yang, S. Saigal, A simple element for static and dynamic response of beams with material and geometric nonlinearities, Int. J. Numer. Methods Eng. 20
(1984) 851–867.
[9] D. Jung, H.C. Gea, Compliant mechanism design with non-linear materials using topology optimization, Int. J. Mech. Mater. Des. 1 (2004) 157–171.
[10] Ü. Sönmez, Compliant MEMS crash sensor designs: the preliminary simulation results, IEEE Intell. Veh. Symp. (2007) 303–308.
[11] K.E. Bishopp, D.C. Drucker, Large deflection of cantilever beams, Q. Appl. Math. 3 (1945) 272–275.
[12] F. Hendriks, Bounce and chaotic motion in print hammers, IBM J. Res. Dev. 27 (1983) 273–280.
[13] J.F. Wilson, J.M. Snyder, The Elastica with end load flip over, ASME J. Appl. Mech. 55 (1988) 845–848.
[14] H. Nahvi, Static and Dynamic Analysis of Compliant Mechanisms Containing Highly Flexible Members, Ph.D. thesis, Purdue University, West Lafayette, IN,
1991.
[15] S.M. Lyon, P.A. Erickson, M.S. Evans, L.L. Howell, Prediction of the first modal frequency of compliant mechanisms using the pseudo-rigid body model, ASME J.
Mech. Des. 121 (1999) 309–312.
[16] C. Boyle, L.L. Howell, S.P. Magleby, M.S. Evans, Dynamic modeling of compliant constant-force compression mechanisms, Mech. Mach. Theory 38 (2003)
1469–1487.
[17] Y.Q. Yu, L.L. Howell, C. Lusk, Y. Yue, M.G. He, Dynamic modeling of compliant mechanisms based on the pseudo-rigid-body model, ASME J. Mech. Des. 127
(2005) 760–765.
[18] Ü. Sönmez, Buckling dynamics of imperfect Elastica subject to various loading conditions, Int. J. Mech. Mater. Des. 3 (2006) 347–360.
[19] S.T. Smith, D.G. Chetwynd, S. Harb, A simple two-axis ultraprecision actuator, Rev. Sci. Instrum. 65 (1994) 910–917.
[20] L. Euler, Methodus inveniendi lineas curvas (Appendix 1, De Curvis Elasticis), in: W.A. Oldfather, C.A. Ellis, D.M. Brown (Eds.), Lausanne and Geneva, English
Translation, 20, 1993, pp. 72–160, Isis.
[21] R. Frisch-Fay, Flexible Bars, Butterworth & Co. (Publishers) Limited, London, 1962.
 A. Tekeş et al. / Mechanism and Machine Theory 45 (2010) 1924–1941 1941

[22] M. Sathyamoorthy, Nonlinear Analysis of Structures, CRC Press, Boca Raton, 1998 pp. 2–10.
[23] R.C. Hibbeler, Statics and Mechanics of Materials, 6th edition, Prentice Hall, New Jersey, 1993.
[24] Gad-el-Hak Mohammed, The MEMS Handbook, CRC Press, Boca Raton Florida, 2001.
[25] L.T. Grujic, On the tracking problem for nonlinear systems, in: S.G. Tzafestas (Ed.), Applied Control, Marcel Dekker, New York, 1993, pp. 325–343.
[26] B. Demirel, M.T. Emirler, A. Yörükoğlu, N. Koca, Ü. Sönmez, Compliant impact generator for required impact and contact force, IMECE-2008-68796, ASME
International Mechanical Engineering, Congress & Exposition Boston, Mech. Syst. Control 11 (2008) 373–379.
[27] H. Asada, J.J.E. Slotine, Robot Analysis and Control, John Wiley, New York, 1986 p. 139.
[28] D.K. Anand, R.B. Zmood, Introduction to Control Systems, 3rd ed.Butterworth, Heinemann, 1995 pp. 660–661.
[29] J.J. Slotine, W. Lee, Applied Nonlinear Control, Prentice Hall, 1990.
[30] Ü. Sönmez, Synthesis methodology of a compliant exact long dwell mechanism using elastica theory, Int. J. Mech. Mater. Des. 3 (2006) 73–90.
[31] C. Nitu, B. Gramescu, S. Nitu, Application of electromagnetic actuators to a variable distribution system for automobile engines, J. Mater. Process. Technol. 161
(2005) 253–257.
[32] B.R. Cannon, T.D. Lillian, S.P. Magleby, L.L. Howell, M.R. Linford, A compliant end-effector for microscribing, Precis. Eng. 29 (2005) 86–94.