Você está na página 1de 6

Precision Engineering 38 (2014) 915920

Contents lists available at ScienceDirect

Precision Engineering
journal homepage: www.elsevier.com/locate/precision

On the modeling of exure hinge mechanisms with nite beam


elements of variable cross section
R. Friedrich , R. Lammering, M. Rsner
Helmut-Schmidt-University/University of the German Federal Armed Forces, Institute of Mechanics, Holstenhofweg 85, 22043 Hamburg, Germany

a r t i c l e

i n f o

Article history:
Received 24 September 2013
Received in revised form 28 May 2014
Accepted 3 June 2014
Available online 11 June 2014
Keywords:
Flexure hinge
Compliant mechanism
Modeling
Amplication
Dynamics

a b s t r a c t
In this paper, a modeling technique for exure hinge mechanisms is studied. Beam elements of variable cross sections are deployed within a nite element procedure to model a circular exure hinge.
The resulting nite element model has very few degrees of freedom and is accurate in both static and
dynamic analysis. Furthermore the modeling approach is applied to an amplier mechanism. Comparing
the results of the proposed model with a 3D nite element reference model, high accuracy for a broad
spectrum of hinge parameters is reported while reducing the number of degrees of freedom immensely.
2014 Elsevier Inc. All rights reserved.

1. Introduction
A exure hinge is an innovative engineering solution for providing relative motion between two adjacent stiff members by the
elastic deformation of a exible connector. Classical pin joints are
meanwhile commonly substituted by exure hinges in applications
where friction, wear and backlash are objectionable features which
impact precision and controllability strongly. Furthermore monolithic structures incorporating exure hinges, commonly called
compliant mechanisms, do not need any lubrication, tend to be
lighter and cheaper and provide a smooth motion which qualies them for the employment in high-precision positioning stages,
medical devices and aeronautical applications. An even broader
range of application of exure hinges is restricted by their limited
rotation capability due to stress concentration.
A number of mathematical approaches have been developed
for modeling exure hinges that describe their static and dynamic
behavior as a function of the geometric parameters. Most commonly this is done by integration of the differential equations taking
the exure hinge geometry into account [13]. Inverse conformal mapping [4] and empirical approaches [5,6] have also been
proposed. All these approaches derive closed-form stiffness or
compliance equations for a single exure hinge. Thereof further
properties like the center of rotation and the mechanical stress
within the exure hinge are deduced.

The application of these closed-form equations to describe subsequent exure hinges within a compliant mechanism is feasible
yet laborious because of their complex kinematics [7,8]. Therefore
usually 2D plain stress or 3D nite element analysis is carried out
with partially an enormous number of degrees of freedom (DOF).
Because of their size, these models are impractical for optimization
or control purposes. Zettl et al. [9] addressed the need for smaller
and highly accurate models by developing the so-called equivalent
beam methodology (EBM). The exure hinge is represented by a set
of ctitious beams whose load-dependent parameters are determined preliminarily by a costly 3D nite element analysis. Lobontiu
and Garcia also used nite beam elements for modeling exure
hinges taking the variable cross section and shear deformation into
account [7,10].
In this paper, a higher order beam element for modeling planar exure hinges using Hermite shape functions is presented.
Beam elements of variable cross section are employed to model
the hinge geometry [11]. The simple benchmark problem shown in
Fig. 1 is applied to verify the modeling approach. Given the variable
thickness

for 0 x

t(x) = ts + 2r 2
Corresponding author. Tel.: +49 40 6541 2745; fax: +49 40 6541 2034.
E-mail address: robert.friedrich@hsu-hh.de (R. Friedrich).
http://dx.doi.org/10.1016/j.precisioneng.2014.06.001
0141-6359/ 2014 Elsevier Inc. All rights reserved.

x(2r x)

for
for

L
2

L

r
+r

L



L

xL

+r


(1)

916

R. Friedrich et al. / Precision Engineering 38 (2014) 915920

L
h
b

71 GPa
2770 kg m 1
0.10 m
0.01 m
0.01 m

Fig. 1. System and loading of the benchmark problem with a circular exure hinge.

where L is the total length of the specimen and h is the thickness


of its straight sections. r and ts are the radius of the right circular
exure hinge and its minimum thickness, respectively. The cross
section and the second moments of inertia are

mass matrix Me and the element stiffness matrix Ke are derived


after discretization of the continuum into nite beam elements of
length l and applying the principle of virtual work:

b t(x)
A(x) = b t(x) and I(x) =
12

(2)

dN1u

dx

B=

NT A(x)N dx
0

given the constant width b. In this paper only circular exure hinges
are regarded, yet the thickness function t(x) can be any meaningful function to describe exure hinge shapes, e.g. conic sections
or llets. Classical 2-node beam elements (see Appendix A) with
3 DOF (axial displacement u, deection w and deection angle )
are employed to model straight sections of the benchmark. The
exure hinge region is modeled by a dedicated 3-node beam element approach. The resulting models have a signicantly reduced
number of DOF compared to 3D modeling approaches. The range
of validity of the proposed beam model is explored comparing it
to a 3D nite element reference model in static and dynamic analysis, where displacements, maximum stress and eigenfrequencies
are regarded. Later the proposed approach is applied to model a 2D
amplier mechanism.

Me = 

Ke = E

A(x)

(4)

BT 0

Iy (x)

Iy (x) B dx,

Iy (x)

Iy (x)

(5)

where matrix N contains the shape functions and B their derivatives, both formulated in the local element coordinate system x.
The classical 2-node beam element, which is applied to model the
straight segments of the benchmark problem (Fig. 1), is inappropriate for modeling the exure hinge region due to the complex
deformation of the variable cross section. Therefore, a 3-node beam
element (Fig. 2) is proposed. The N and B matrices for the proposed
element are

N=

N1w

N1

d N1w
2

N2w

N2

d N2w

dx

dx

N3w

N3

dN3u

d N3w

dx

ue = u1

The ordinary differential equations for the axial displacement


u and the bending displacement w of an EulerBernoulli beam are
given in [12]:
2

A(x)

+
t 2 x
2

A(x)

+
t 2
x2

EA(x)

u
x

EI y (x)

1

u2

w2

w
x2

2

u3

w3

3

u(x) = a0 + a1 x + a2 x

= qz (x, t)

where  and E are the mass density and Youngs modulus, respectively. The cross section area A(x) and the second moment of
inertia Iy (x) are functions of the beam thickness t(x) in the global
coordinate system. The well known expressions for the element

(7)

dx

T

(8)

(10)

It should be noted that (x) is obtained from the Bernoulli


hypothesis via (x) = dw/dx. The resulting Hermite shape functions are:
N1u = 1 3
N2u = 4

x
x
+2 2
l
l

(11)

x
x
4 2
l
l

N3u =

N2w = 16
N3w = 7

(12)

x
x
+2 2
l
l

N1w = 1 23

Fig. 2. The 3-node Bernoulli beam element, nodal degrees of freedom and local
coordinate system.


N3
2

(9)
3

w(x) = b0 + b1 x + b2 x + b3 x + b4 x + b5 x .
(3)

A 2nd order polynomial serves as ansatz for the axial displacements


u while the deection w is approximated by a 5th order polynomial:

= n(x, t)
2

w1

while the column matrix of the unknowns is


2. Modeling approach

dx

(6)

dx


N2
2

N3u

dx


N1
2

N2u

dN2u

dx

N1u

(13)

x
x
x
x
+ 66 3 68 4 + 24 5
l
l2
l
l

x
x
x
32 3 + 16 4
l2
l
l
2

(14)

(15)
5

x
x
x
x
34 3 + 52 4 24 5
l
l2
l
l

(16)

R. Friedrich et al. / Precision Engineering 38 (2014) 915920

917

directly computed from the unknown deformations u and the Bmatrix applying

(x) = E Ka Ba ua + Kb


N2

x
x
x
x
+ 13 2 12 3 + 4 4
l
l
l
l

N3 =

(17)

x
x
x
x
+ 32 2 40 3 + 16 4
= 8
l
l
l
l

(18)

x
x
x
x
+5 2 8 3 +4 4
l
l
l
l

(19)

The nite element procedure is implemented using MATLAB


[13]. Integrals for the computation of Me and Ke are evaluated
numerically using an adaptive Gauss-Kronrod quadrature with an
absolute error tolerance of 106 , which is readily implemented in
MATLAB (quadgk-function). Finally, the system mass matrix M and
the system stiffness matrix K are assembled so that the problems
formulated in (3) can be rewritten as a linear system of equations:
M u + K u = f.

(20)

In a static analysis, the rst term is neglected and the system is


solved for the unknowns u. The eigenfrequencies of the system
are obtained solving the characteristic equation
(K 2 M) = 0.

(22)

where the superscripts a and b refer to the axial and bending


portion, respectively. Ka and Kb are the axial and bending stress concentration factors, respectively [14]. The errors in the displacement
w(x = l) and the maximum bending stress  x (x = L/2) in comparison to the reference model are depicted in Fig. 4. Applying the end
loads Fx and My leads to very similar results, hence only the results
for Fz are shown.

Fig. 3. ANSYS reference model with 240,000 DOF.

N1 = x 6

t(x) b b
B u
2

(21)

3. Results
3.1. 3D nite element reference model
For reference the benchmark problem stated in Fig. 1 is modeled
in ANSYS using 20-node SOLID186 elements with a total of 80,000
nodes. The mesh in the exure hinge region is strongly rened. Fig. 3
shows the employed mesh. The applied unit loads are distributed
over the side area of the model to avoid concentrated forces at single
nodes.
3.2. Static analysis
The number of elements within the exure hinge region nelFH
necessary to properly describe the static behavior is analyzed rst.
With nelFH = 1 the total number of DOF is 15, with nelFH = 3 it is 27
while the total number of DOF is 39 for nelFH = 5. The error of the
beam model is computed with respect to the 3D reference model.
In the static analysis, force Fz is applied to the benchmark problem and the resulting bending deformation of the load application
point is compared with the reference model. The deection error of
the beam model is shown in Fig. 4(a) depending on dimensionless
slenderness ratio = ts /r. A broad spectrum from = 0.02 (slender
exure hinge) to = 0.5 (relatively thick exure hinge) is analyzed.
The blue data points are the results obtained when one element is applied to model the exure hinge (nelFH = 1). Red (nelFH = 3)
and green (nelFH = 5) data points show the results for a rened
discretization. The mechanical stress within the exure hinge is

3.3. Dynamic analysis


The eigenvalue problem stated in equation (21) is solved to
extract the rst axial and the rst three bending eigenfrequencies.
In Fig. 5 the errors of the computed eigenfrequencies in comparison to the reference model are depicted with the according
eigenmodes. Modeling the exure hinge region with one element
(nelFH = 1) appears to be insufcient as the results for the static analysis suggest. This is even more pronounced in the dynamic analysis.
Therefore, the respective data points are excluded from Fig. 5.
3.4. Discussion
It is clearly demonstrated that the static and dynamic behavior of the 3D reference model (240,000 DOF) can be replicated
by the much smaller beam model (1539 DOF) using the proposed approach. The study of the benchmark problem shows that
the discretization of the exure hinge geometry with only one
beam element of variable cross section (nelFH = 1) leads to an unacceptable discrepancy compared to the reference model. Regarding
Fig. 4(a) and (b) one instantly recognizes a pronounced error for
very slender exure hinges of < 0.1, where the area of signicant deformation lies at the very center of exure hinge. This area
becomes smaller the more slender the exure hinge is, eventually becoming smaller than the employed discretization, resulting
in the observed error. Since the maximum stress  x is obtained
from the displacements by derivation, the magnitude of error is
larger compared to the displacement error. The deection error of
58% for thicker exure hinges ( > 0.4) arises due to the increasing inuence of shear on the deection which is not covered by the
Bernoulli-model.
Following conclusions can be formulated regarding the range of
validity of the proposed approach:
For exure hinges of a slenderness ratio of > 0.08, the displacement w(x = l) can be predicted with errors below 8% using three
elements to model the exure hinge. The use of ve elements
extends the range of validity to > 0.04.
 max (x = L/2) can be predicted by the (nelFH = 5)-model for > 0.08
with errors of less than 5%, while the (nelFH = 3)-model fails for
< 0.2.
In the dynamic analysis, where the rst axial and the rst three
bending modes are under consideration, the (nelFH = 5)-model
yields reliable results for the entire spectrum of regarded , while
(nelFH = 3)-model begins to deviate for < 0.1.
4. Modeling example
The simple amplier mechanism in Fig. 6, which is described in
detail in [15], is chosen as example to demonstrate the simplicity
and effectiveness of the proposed modeling method. The mechanism consists of four straight segments ((1), (2), (4) and (6)) and two
U-shaped circular exure hinges ((3) and (5)) of the same radius r

918

R. Friedrich et al. / Precision Engineering 38 (2014) 915920

10
0
-10

10
0
-10

-25

-25

-50

-50

-75

-75

-100

0.1

0.2

0.3

0.4

0.5

(a) Error: deection w (x = L ) due to F z

-100

0.1

0.2

0.3

(b) Error: maximum stress

0.4

0.5

x (x = L / 2 ) due to F z

Fig. 4. Error of the beam model in a static analysis.

Fig. 5. Errors in the computed eigenfrequencies and respective eigenmodes

and minimum thickness ts . The amplier mechanism has a double symmetry consequently only one quarter of the mechanism
is modeled. It is actuated by the input force Fin . The ratio of output
motion uout to input motion uin denes its amplication a = uout /uin .
Five 3-node beam elements of variable cross section are
employed for each exure hinge, while the straight sections are
modeled with classical 2-node beam elements (see Appendix A).

The problem formulated in Eq. (3) in one dimension has to be


extended to account for the two-dimensional nature of the amplier mechanism. The element mass matrix Me derived in Eq. (4) is
transformed using

Mne2D = (rn ) Mne1D rn ,


T

with

cos i

rn = sin i

sin i

cos i

0.

1
(23)

Table 1
Dimensions and modeling of the amplier mechanism.

Fig. 6. Model of the amplier mechanism.

Segment i

Element type

li [m]

hi [m]

i [ ]

(1)
(2)
(3)
(4)
(5)
(6)

2-node beam
2-node beam
3-node beam
2-node beam
3-node beam
2-node beam

0.02
0.01
2r
0.085
2r
0.015

0.02
0.02
Circular
0.02
Circular
0.02

90
0
0
4
0
0

R. Friedrich et al. / Precision Engineering 38 (2014) 915920

919

Table 2
Modeling results of the amplier mechanism
ts [mm]

r [mm]

= ts /r

 max [MPa]

1 [Hz]

Beam

3D FEM

Beam

3D FEM

Beam

3D FEM

1.0

3
4
5

0.333
0.250
0.200

14.7
15.1
15.4

14.5
14.9
15.3

1.32
1.30
1.28

1.36
1.34
1.33

80.6
75.9
73.0

79.1
75.6
73.3

0.5

3
4
5

0.170
0.125
0.100

15.2
15.5
15.9

15.1
15.5
15.8

4.94
4.82
4.71

5.10
5.06
5.04

33.6
31.8
30.6

34.1
32.5
31.5

0.25

3
4
5

0.083
0.063
0.050

15.3
15.6
16.0
16.0

15.3
15.6
15.9

1.82
1.73
1.63
1.84

1.97
1.97
1.96

14.1
13.4
13.0
12.9

14.6
13.8
13.4

Eq. (23) is given in nodal form, where n is the node number. The
same transformation is performed for the element stiffness matrix
Ke . All geometry and modeling parameters are summarized in
Table 1 where li is the length of the segment i and hi its height.
i is alignment angle with respect to the global coordinate system.
The mechanism modeled in ANSYS by 3D elements has about
120,000 DOF, which makes it impractical e.g. for optimization
purposes. In contrast, the amplier mechanism modeled by the
proposed approach (75 DOF) reduces the number of DOF by over
thousand. In terms of the computational time on a regular quadcore PC, the ANSYS-model takes 100 seconds to compute, while the
MATLAB-model needs only 0.8 s.
The 3D model again serves as a reference to evaluate the results
obtained with the beam model. Amplication a, maximum bending stress  max and the rst eigenfrequency 1 are regarded at
different hinge geometries. For the static analysis, an input force
of Fin = 1 N is assumed. Aluminum material properties of E = 71 GPa
and  = 2770 kg/m3 are applied. The width b of the mechanism is
set to 0.015 m.
In Table 2 the results of both the proposed beam model and the
3D model are compared for different exure hinge geometries. The
proposed beam modeling approach is able to replicate the results
of the 3D nite element simulation with high accuracy despite its
minimal number of DOF.
The beam model computes the amplication a with an error
of 1% or less. Also the maximum stress values  max are in good
agreement with the predicted results with errors below 8%. For very
slender exure hinges with < 0.08 the deviation of the computed
maximum stress from results predicted by the 3D model is more
pronounced, which conrms the conclusions from Section 3.4. This
can be eliminated by employing seven 3-node beam elements of
variable cross section for modeling the exure hinge, which is done
for the results in the last row of Table 2. The rst eigenfrequency
can be reproduced accurately with errors below 4% for all regarded
slenderness ratios .
5. Conclusion and outlook
In this paper, a modeling approach for exure hinge mechanisms
using beam elements is presented. Although in this study only circular exure hinges are regarded, the modeling approach can be
easily applied for other exure hinge shapes. Resulting models have
by about a factor of 1000 reduced number of DOF compared to commonly employed 3D modeling techniques for exure hinges and
compute models 50100 times faster. It is shown that the exure
hinge has to be modeled with at least three of the proposed elements to properly describe the static and dynamic system behavior.
The range of validity is shown to be limited by the slenderness ratio
of the exure hinge. Applying the modeling approach to an amplier mechanism it is demonstrated that the amplication can be

Fig. A.1. The classical 2-node Bernoulli beam element, nodal degrees of freedom
and local coordinate system.

predicted with an error of 1%, the maximum mechanical stress with


an error of maximal 8% and the rst eigenfrequency with an error
of maximal 4%.
Acknowledgement
The nancial support of the Deutsche Forschungsgemeinschaft
(DFG) Priority Program 1476 is gratefully acknowledged.
Appendix A. The classical 2-node beam element
Fig. A.1

1/3

Me = lA

1/6

13/35

11/210 l

9/70

1/105 l2

13/420 l

1/3

0
13/35

Ke =

EA/l

EA/l

12EI/l3

6EI/l2

12EI/l3

4EI/l

6EI/l2

EA/l

0
12EI/l3

sym.

ue = u1

w1

13/420l

1/140l2

11/210l
0

1/105l2
(24)

sym.

2EI/l
(25)

2
6EI/l
6EI/l2

4EI/l
1

u2

w2

2

T

(26)

Appendix B. Supplementary data


Supplementary data associated with this article can be found,
in the online version, at http://dx.doi.org/10.1016/j.precisioneng.
2014.06.001.

920

R. Friedrich et al. / Precision Engineering 38 (2014) 915920

References
[1] Paros JM, Weisbord L. How to design exure hinges. Mach Des
1965;37(27):1516.
[2] Lobontiu N, Paine JSN, Garcia E, Goldfarb M. Design of symmetric conic-section
exure hinges based on closed-form compliance equations. Mech Mach Theory
2002;37(5):47798.
[3] Dirksen F, Lammering R. On mechanical properties of planar exure hinges of
compliant mechanisms. Mech Sci 2011;2(1):10917.
[4] Tseytlin YM. Notch exure hinges: an effective theory. Rev Sci Instrum
2002;73(9):33638.
[5] Schotborgh WO, Kokkeler FG, Tragter H, van Houten FJ. Dimensionless design
graphs for exure elements and a comparison between three exure elements.
Precision Eng 2005;29(1):417.
[6] Yong YK, Lu T-F, Handley DC. Review of circular exure hinge design equations
and derivation of empirical formulations. Precision Eng 2008;32(2):6370.
[7] Lobontiu N. Compliant mechanisms: design of exure hinges. Boca Raton: CRC
Press; 2003.

[8] Jouaneh M, Yang R. Modeling of exure-hinge type lever mechanisms. Precision


Eng 2003;27(4):40718.
[9] Zettl B, Szyszkowski W, Zhang W. Accurate low DOF modeling of a planar
compliant mechanism with exure hinges: the equivalent beam methodology.
Precision Eng 2005;29(2):23745.
[10] Lobontiu N, Garcia E. Circular-hinge line element for nite element analysis of
compliant mechanisms. J Mech Des 2005;127(4):76673.
[11] Friedrich R, Lammering R. An improved modeling approach for circular exure
hinges with application to geometry optimization. PAMM 2013;13(1):956.
[12] Rao SS. Vibration of continuous systems. Hoboken: Wiley; 2007.
[13] Kwon YW, Bang H. The nite element method using MATLAB. 2nd edition Boca
Raton: CRC Press; 2000.
[14] Pilkey WD, Pilkey DF. Petersons stress concentration factors. 3rd edition Hoboken: Wiley; 2008.
[15] Lobontiu N, Garcia E. Analytical model of displacement amplication and stiffness optimization for a class of exure-based compliant mechanisms. Comput
Struct 2003;81(32):2797810.

Você também pode gostar