Você está na página 1de 14

See

discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/261758078

A Load Independent Pseudo-Rigid-Body 3R


Model for Determining Large Deflection of
Beams in Compliant Mechanisms

Conference Paper January 2008


DOI: 10.1115/DETC2008-49041

CITATIONS READS

11 93

1 author:

Hai-Jun Su
The Ohio State University
124 PUBLICATIONS 1,132 CITATIONS

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Shape Morphing Arm Robotic (SMART) Manipulators for Simultaneous Safe Human-Robot Interaction
and High Performance in Manufacturing View project

DNA Origami Mechanisms View project

All content following this page was uploaded by Hai-Jun Su on 07 May 2014.

The user has requested enhancement of the downloaded file. All in-text references underlined in blue are added to the original document
and are linked to publications on ResearchGate, letting you access and read them immediately.
Proceedings of IDETC/CIE 2008
ASME 2008 International Design Engineering Technical Conferences &
Computers and Information in Engineering Conference
August 3-6, 2008, New York City, NY, USA

DETC2008-49041

A LOAD INDEPENDENT PSEUDO-RIGID-BODY 3R MODEL FOR DETERMINING


LARGE DEFLECTION OF BEAMS IN COMPLIANT MECHANISMS

Hai-Jun Su
Assistant Professor, ASME Member
Department of Mechanical Engineering
University of Maryland, Baltimore County
1000 Hilltop Circle Baltimore, Maryland 21250
Email: haijun@umbc.edu

ABSTRACT plicit kinematic and static constraint equations derived from the
Modeling flexible beams that undergo large deflection is one model. To demonstrate the use of the PRB 3R model, a compli-
of the key steps in analyzing and synthesizing compliant mecha- ant 4-bar linkage is studied and verified by a numerical example.
nisms. Geometric nonlinearities introduced by large deflections The result shows a maximum tip deflection error of 1.2% com-
often complicate the analysis of mechanism systems comprising pared with the FEA model.
such members. Several pseudo-rigid-body (PRB) or multi seg-
ment models in the literature have been proposed to approximate
the tip deflection and slope. However these models are either de- 1 INTRODUCTION
pendent on external loads or too complicated to analyze. They Compliant mechanisms [1] comprise of flexible links that
are neither appropriate for analyzing mechanisms in which loads undergo large deflection have many potential advantages such as
change significantly as they move, nor for synthesizing mecha- reduced part number, reduced wear, reduced backlash and error.
nisms where a parametric model is preferred. In this paper, a However the nonlinearity of large deflection beams often compli-
load independent PRB 3R model which comprises of four rigid cates analysis and design of compliant mechanisms and prevents
links joined by three revolute joints and three torsion springs is its wide applications. One major difficulty comes from the fact
proposed. The traditional PRB 1R models are first studied for that the tip load on the flexible beams of a compliant mechanism,
both small deflection beams and large deflection beams. These determined by the magnitude of force and moment as well as the
studies provide fundamental insights to the geometric nonlinear- force direction, can vary significantly as the mechanism moves.
ity of large deflection beams. Numerical integration is applied to The two extreme cases are when the load is force only (or domi-
compute tip deflections for various loads. A three-dimensional nant) and moment only (or dominant).
search routine has been developed to find the optimal set of char- Parametric approximation for the beam tip deflection is crit-
acteristic radius factors for the proposed PRB 3R model. De- ical for higher level design especially design synthesis. Sev-
tailed error analysis and comparison against the result by the eral pseudo-rigid-body (PRB) models have been developed to
numerical integration and the PRB 1R model are accomplished approximate tip deflection of flexible beams for various loads.
for different load modes. Howell and Midha proposed a PRB 1R model that is comprised
The benefits of the PRB 3R model include (a) high accuracy of two rigid links joined at a pin joint and a torsion spring along
for large deflection beams, (b) load independence which is crit- the beam. Here R represents a revolute or pin joint. They
ical for applications where loads vary significantly and (c) ex- have found that the position of the pin joint is determined by the

1 c 2008 by ASME
Copyright
so-called characteristic radius factor which equals 0.85 for y y
Fo
moment only load [2] and 0.735 for force only load [3]. And Mo
the equivalent spring stiffness K for these two extreme cases k=2EI/3l 2l/3 k=EI/2l l/2
bF x bM
are about 1.51 and 2.65 respectively. If the load is combined end x
l/3 EI, l l/2 EI, l
force and end moment, and K will be anywhere between the (a) (b)
two extreme cases. Furthermore the tip slope angle must be con-
Figure 1. Pseudo-rigid-body 1R model for a linear elastic beam sub-
strained below a certain bound, e.g. 77 for vertical force load.
jected to (a) a vertical end force and (b) a pure end moment
However this model is not appropriate for the applications where
the load varies significantly and application where the mecha-
nism undergoes exceptionally large deflection. 2 PSEUDO-RIGID-BODY 1R MODEL FOR SMALL DE-
FLECTION BEAMS
To approximate tip deflection of for initially straight can-
In this section, we study the small deflection beams and our
tilever beams subjected to combined end force and moment, Sax-
goal is to develop a PRB 1R model for a small combined end
aena and Kramer [5] modified the PRB 1R model by introducing
force and end moment. This analysis will help us gain insight
two linear springs to restrain the change of characteristic radius
to the highly nonlinear large deflection beams. When an initially
factor for different load modes. Lyon et al. [6] decomposed a
straight cantilever beam of length l and flexural rigidity EI is
flexible beam into two segments, each of which is approximated
subjected to a small end force F or a small end moment M , the
by a PRB 1R model. However the error introduced by these two
tip slope angle and tip deflection b/l are calculated as
models are exceptionally large. On the other extreme, Saggere
and Kota (2001 [8]) proposed a finite element model that evenly
decomposes a flexible beam into n 3 segments joined by n tor- F l 2 F l 2
= , bF /l = : end force, (1)
sion springs of stiffness EI/l. This model is completely load 2EI 3EI
independent and can accommodate a wide range of loads. How- M l M l
= , bM /l = : end moment. (2)
ever the approximation accuracy is determined by the number of EI 2EI
segments. The larger the number of segments n, the smaller the
approximation error. Unfortunately a large n can significantly This result can be found in mechanics of materials book such
complicate the kinematic and mechanic equations of compliant as [9]. For readers convenience, a short review of this result is
mechanisms. given in the Appendix.
Figure 1 shows a PRB 1R model for linear elastic beams
By compromising the PRB 1R model and the finite element subject (a) a small end force and (b) a small end moment. The
model aforementioned, this paper proposes a new PRB 3R model model is comprised of two rigid links connected by a pin joint
for initially straight cantilever beams subjected to a combined wit yh a torsion spring of stiffness k. The location of the torsion
force and moment. The model is comprised of three revolute spring is determined by the characteristic radius factor , com-
joints, each accompanied by a torsion spring. The kinematic and puted from Eqs. (1-2) as
mechanic parameters are load independent. Since the analytical
inverse and forward kinematics are readily available for 3R se- (
rial chains, the kinematic and mechanic equations are relatively b/l F = bF/l = 2/3 : end force
= = (3)
simple. M = bM/l = 1/2 : end moment

The rest of paper is organized as follows. Section 2 stud-


ies a PRB 1R model for small deflection beams subjected to a And the stiffness of the torsion spring is k = M/ = EI/l.
combined end force and moment. In Section 3, large deflection When a combination of end force F and moment M is ap-
equations of elastic beams are solved by numerical integration plied on the beam, the tip slope angle and tip deflection can be
in order to generate the complete data set for various end loads calculated by
which will be used to develop PRB models. Section 4 studies
the PRB 1R model for an elastic beam undergoing large deflec- = + , (4)
tion. In Section 5, a PRB 3R model for large deflection beams is b/l = F + M . (5)
proposed and analytical kinematics and static equations based on
this model are derived. Section 6 studies an optimized load inde-
pendent PRB 3R model and evaluates it by detailed error analysis And the characteristic radius factor is computed as
for various loads. And lastly in Section 7, a numerical example is
provided to demonstrate the benefits of the model and the result b/l F + M F + 2M
= = = , (6)
is compared with a FEA model. + 1 + 2

2 c 2008 by ASME
Copyright
where the parameter is called the load ratio, written as 3 LARGE DEFLECTION EQUATIONS FOR CAN-
TILEVER BEAMS WITH COMBINED END FORCES
M AND MOMENT
= = : load ratio. (7) Figure 3 shows an initially straight cantilever beam sub-
2 F l
jected to a combined end force F and moment M . The direc-
And the spring stiffness k is still proportional to , calculated as tion of the end force is denoted by angle . Parameters , a and
k = EI/l. A large corresponds to a load dominated by end b are tip slope angle and the horizontal and vertical tip deflec-
moment. A small represents a load dominated by end force. tion respectively. In this article, we assume that there is not sign
Figure 2 shows the plot of vs. [0, +). It can be seen that change in the curvature of the beam. Hence the deflection angle
starts at F = 2/3 when force is dominant, drops dramatically of the beam is either monotonically increasing or monotonically
between [0, 2], and gradually approaches to M = 1/2 when decreasing. Because of symmetry, here we only study the former
> 20. case and assume [0, ] and M 0.
Since the large deflection beam equations have been stud-
ied extensively in the literature [3, 7], only the final result is pre-
sented here. For readers convenience, a sketch of the derivation is
force dominates provided in the Appendix. By large deflection beam theory, the
0.65 beam tip deflection is determined by the following three equa-
tions,

Z
0.60 1 0 d
= p , (8)
2 0 [cos( ) cos( )] +
Z 0
1 cos d
0.55 a/l = p , (9)
2 0 [cos( ) cos( )] +
moment dominates Z 0
1 sin d
b/l = p , (10)
20 40 60 80 100 2 0 [cos( ) cos( )] +

Figure 2. The characteristic radius factor for a linear elastic beam sub-
where a/l, b/l are normalized tip deflection and is the load
jected combined force and moment vs. linear load ratio
index, defined as

Please note that this model is exact for both force and tip 2
= = : load index. (11)
deformation calculation. More importantly it gives us an insight 4 2
to the nonlinearity of large deflection beams. In what follows,
we will study the large deflection beams. Note and are the tip slope angle for small deflection
beams defined in (1) and (2) which are called force index and
moment index respectively. See [5] for similar definitions.
Fo o And the term is the load ratio defined in (7). A large rep-

a resents a moment dominant load. As one can see that the load
Mo index (for large deflections) is not only dependent on the load
ratio but also on moment index . This high nonlinearity intro-
y duces difficulties in large deflection analysis.
b
(x, y)
3.1 Numerical Integration of Deflection Equations
Large deflection equations (8-10) consist of three load pa-
x EI, l rameters , , and three deformation parameters , a/l, b/l.
Given any three parameters, one can find the other three by solv-
Figure 3. Large deflection of a cantilever beam subjected to a combined ing these equations. However it is not convenient to directly
end force and moment solve (8-10) for deformation parameters with known load pa-
rameters since the tip slope is one of the integration limits.
A common way to overcome this difficult is to solve (8) for

3 c 2008 by ASME
Copyright
k=0. k=0.1
with a given and then use (9-10) to determine the tip deflec- 1.0 =99

1.0
tion. Howell and Midha (1995) [3] provided an elliptic integrals =135
0.8 0.8
solution and Saxena and Kramer (1998) [5] solved these equa- 0.6 0.6
tions with numerical integration. In this article, the numerical in- 0.4 =9

0.4
tegration approach is used to compute tip deflection for various =171

0.2 0.2
combination of end force and moment. This is done by using
Mathematica software. -0.5 0.0 0.5 1.0 -0.5 0.0 0.5 1.0

Before computing the integration, the limit of slope angle 1.0


k=1.
1.0
k=1.5

=171
max for given load parameters must be determined. By requir-
0.8 0.8
ing the term under the square root of equation (8) being positive
0.6 0.6
for [0, ], we obtain
0.4 0.4
=9
0.2 0.2

+ cos1 (1 ) : 02 -0.5 0.0 0.5 1.0 -0.5 0.0 0.5 1.0
max = (12)
, : 2 k=2. k=2.5
1.0 1.0
=171 =171
0.8 0.8
See [7] for similar results. 0.6 0.6

=9
Figure 4 shows the deflected tip loci computed via numerical 0.4 =9 0.4

integration for various load indices [0, 50] and various force 0.2 0.2
direction angles [9 , 171 ]. Note for simplicity, we set the -0.5 0.0 0.5 1.0 -0.5 0.0 0.5 1.0
upper bound of to be , that is max = min[, + cos1 (1
k=5. k=50.
)]. All loci start with a circular arc when is small and diverge 1.0 1.0
=171
as the load increases. When the end force dominates the load( < 0.8 0.8
0.1), the tip locus is straighten out as the load increases. And 0.6 0.6
the smaller the angle , the earlier this occurs. When the end 0.4
=9

0.4
force and end moment are comparable (1 < < 2), the tip loci
0.2 0.2
diverge significantly for different angles . The smaller the angle
, the more the loci bend inwards. This is buckling phenomenon -0.5 0.0 0.5 1.0 -0.5 0.0 0.5 1.0
caused by axial force dominant loads. As increases, the effect
Figure 4. Trajectory of deflected tip point arranged in terms of load in-
of on the tip locus decreases. When the moment dominates
dex [0, 50]. The maximum slope angle at the tip is determined by
the load ( > 5), the tip locus is almost identical to that of pure
moment case. max = min[, + cos1 (1 )]
Essentially Figure 4 provides a comprehensive data set of
actual tip deflection which will be used to analyze and compare where max is the material yield strength and t is the beam thick-
PRB models. In what follows, we will first discuss the tradi- ness. Substituting the definitions (1) and (2) yields
tional PRB 1R model and then propose a new PRB 3R model,
which is load independent and can accurately approximate large Et a b
deflections of a cantilever beam. max [ + 2( sin cos )], (14)
2l l l

or
3.2 Limit of Tip Slope Angle by Maximum Stress
l max a b
Equation (12) gives the upper bound of the end slope angle. 2 + 2 sin cos = Mnet , (15)
t E l l
In practical designs, the slope angle is often limited by the max-
imum stress of the beam which is at the fixed end for a beam
with monotonic curvatures. To prevent the beam from failure, where the term l/t (ratio of length to thickness) defines the beam
the following condition for rectangular section must be satisfied slenderness and max /E (ratio of yield strength to Youngs mod-
ulus) defines the material flexibility. The larger these two quan-
tities, the larger slope angle the beam can be bent to. The right
tMmax t side of (15) Mnet is the normalized net moment applied to the
max = [M + F (a sin b cos )], (13) fixed end of the beam.
2I 2I

4 c 2008 by ASME
Copyright
Figure 5 shows that max for a flexible beam under moment routine that maximizes the angle for a prescribed allowable
load is determined by various materials and beam slenderness error bound. An alternative way to compute and pseudo-rigid-
l/t. The values of max /E for various materials are from p.30 body angle for each tip point (a/l, b/l) is via kinematics of 1R
of reference [1]. It can be seen that a polypropylene beam with chain, calculated as
l/t = 100 can sustain 5 rad or 286 tip bending angle without
failure. And steel beams with l/t = 500 can only be bent to
(1 a/l)2 + (b/l)2
an angle of 50 . This figure can help us identify material and = (16)
2(1 a/l)
slenderness needed for a specific application.
1 b/l
= tan (17)
a/l + 1
0.8
b/l
o= o=/2
It has been shown [5] that is predominantly dependent on
o=2.5
0.6
o=2
Polypropylene Titanium (Ti-13)
the load ratio and only slightly correlated to force directional
l/t=50 l/t=100 o=1.4 angle . For each load ratio , we average the values of over
Aluminum [0, ] and plot the average value as dots in Fig. 7. It is remarkable
l/t=100
o=4 0.4 to see that this curve is very similar to the linear deflection case
Polyethylene
o=0.87 shown in Fig. 2. Hence it is reasonable to approximate versus
l/t=100
o=3/2 Steel
l/t=500
load index by the following function
0.2
o=5
Polypropylene
l/t=100 F + 2M (c )
o=0 (18)
-0.2 0.0 0.2 0.4 0.6 0.8 1.0
a/l 1 + 2c

Figure 5. Limit of tip slope max under pure moment load determined
where F = 0.842, M = 0.735 are the optimal characteristic ra-
by maximum stress
dius factors for end force and end moment respectively. And
c = 4.694 can be calculated by one parameter linear regression.
Comparing with (6), one can immediately see that c = 1 for lin-
ear elastic (small) deflections. As far as the author is aware, c
Fo has not been previously discussed in the literature. Let us name
variable c as nonlinearity index.
y k=k(EI/l) o
o =c
l Mo

0.84 approximation
x
0.82
k EI, l
(1)l 0.80


Figure 6. The pseudo-rigid-body 1R model with parameter 0.78
actual
[0.735, 0.85] depending on the load 0.76






10
20 30 40
50

Figure 7. Actual and approximated characteristic radius factor vs. load


4 PSEUDO-RIGID-BODY 1R MODEL FOR LARGE DE- index for beams under large deflections. The actual values of are
FLECTION BEAMS obtained by averaging over [0, ]
Figure 6 shows the classical PRB 1R model for large de-
flection beams which is comprised of two rigid links connected
by a pin joint with a torsion spring. The location of the torsion The approximated values of is plotted as the continuous
spring is determined by the characteristic radius factor . And line in Fig. 7. The Pearsons correlation coefficient of the actual
the tip point is determined by a PRB angle . To determine value and approximated values is 0.98. They represent errors less
and , Howell and Midha [3] provided a one dimensional search than 0.016 or 2%. Comparing with the result in [5], function (18)

5 c 2008 by ASME
Copyright
is simpler and has a smaller approximation error. Furthermore where c1 = cos(1 ), s1 = sin(1 ), c12 = cos(1 + 2 ),s12 =
function (18) is infinity order continuous for > 0, which is im- sin(1 + 2 ),c123 = cos(1 + 2 + 3 ), s123 = sin(1 + 2 +
portant in applying numerical methods for design synthesis [12] 3 ).
and optimization. On the other hand, given tip point Q and slope , the PRB
Similarly the spring stiffness k and the ratio c of the tip angles i are computed by analytical inverse kinematics of a 3R
slope angle to the PRB angle can be also calculated for each serial chain. A sketch of the solution process is shown below.
set of , a/l, b/l. It should be pointed out that model parameters For convenience, we denote the vector from P1 to P3 by P13 =
, k , c are dependent on the load parameters and . This (px , py )T .
is not convenient for analyzing compliant mechanisms that are
subjected to a wide range of loads. Here we do not further discuss
px = Qx 3 cos 0 , (21)
this model, but instead focus on a new 3R model in the following
sections. py = Qy 3 sin , (22)
!
1
p2x + p2y 21 22
2 = cos , (23)
Fo 21 2
o =1+2+3 1 = Atan2(sin 1 , cos 1 ), (24)
Mo 3 = 1 2 , (25)
T T
3l 3 Q=(Qx ,Qy ) =( a/l,b/l)
where
y k3 P3
2l px (1 + 2 cos 2 ) + py 2 sin 2
cos 1 = ,
k2 2 21 + 22 + 21 2 cos 2
1l P2
k1 py (1 + 2 cos 2 ) px 2 sin 2
1 sin 1 = .
O 21 + 22 + 21 2 cos 2
0l x EI, l
P1
See Tsai (1999) [11] for detailed derivation of inverse kinemat-
Figure 8. A pseudo-rigid-body 3R model for cantilever beam subjected ics. Note equation (23) yields two solutions , + for elbow-down
to combined end force and moment solution and - for elbow-up solution. For beams with mono-
tonic curvature, we choose either the elbow-down solution or the
elbow-up solution for all .

5 A PSEUDO-RIGID-BODY 3R MODEL 5.2 Statics Equations


Figure 8 shows a PRB 3R serial chain for modeling a flex- Given the external load on the tip of the 3R chain, the torques
ible beam subjected to a combined end force and moment. The at the three pin joins are given by
PRB model is comprised of four rigid links joined by three pin
joints and torsion springs. Let us denote the deflection angles and
1 Fx l
spring stiffness by 1 , 2 , 3 and k1 , k2 , k3 respectively. And the 2 = [J T ] Fy l (26)
ratio of the four link length to the length of the beam are charac-
3 M
teristic radius factors, denoted by 0 , 1 , 2 , 3 satisfying

where we have redefined Fx = F cos , Fy = F sin . And matrix


0 + 1 + 2 + 3 = 1. (19)
[J T ] is transpose of the non-dimensional Jacobian of the 3R chain
obtained by differentiating kinematic equation (20), written as
5.1 Forward and Inverse Kinematics Equations
The tip point Q = (Qx , Qy )T = (a/l, b/l)T and slope angle
of the 3R chain are calculated via the forward kinematics 1 s1 2 s12 3 s123 1 c1 + 2 c12 + 3 c123 1
[J T ] = 2 s12 3 s123 2 c12 + 3 c123 1 . (27)
3 s123 3 c123 1
Qx = 0 + 1 c1 + 2 c12 + 3 c123
Qy = 1 s1 + 2 s12 + 3 s123 (20)
If the spring torques are proportional to the PRB angles, the
= 1 + 2 + 3

6 c 2008 by ASME
Copyright
spring stiffness can be calculated by normalizing (26) To determine the maximum slope angle max , we sub-
stitute the beam slenderness l/t = 100 and material flexibility

k1 1 k1 max /E = 30 into equation (15). The search routine found the
k2 2 = k2 (EI/l) (28) optimal characteristic radius factors

k3 3 k3
0 = 0.1, 1 = 0.35, 2 = 0.40, 3 = 0.15. (30)
where

k1 2 cos /1
k2 = [J T ] 2 sin /2 . (29)
linear fit of 1
k3 /3 actual 3
0 i=(EI/l)0
5.3 Computation of Spring Stiffness 4 actual 2
For a given set of characteristic radius factors 0 , 1 , 2 , 3 ,
the spring stiffness k1 , k2 , k3 are determined in two steps: 3 actual 1 linear fit of 2,3
1. substitute pre-computed tip point a/l, b/l and slope (plot-
2
ted in Fig. 4) into the inverse kinematics equations (21-25)
to compute the PRB angles 1 , 2 , 3 and
1
2. substitute the angles 1 , 2 , 3 and load parameters , ,
into (29) to compute the stiffness coefficients k1 , k2 , k3 .
0.5 1.0 1.5
i

5.4 Three-Dimensional Search for Optimal Character- Figure 9. Linear regression of spring stiffness k1 , k2 , k3 for end mo-
istic Radius Factors ment loads. The range of tip slope angle is [0, 3/2]
The PRB 3R model consists of four kinematic parameters
0 , 1 , 2 , 3 and three mechanic parameters k1 , k2 , k3 . The goal
is to find the optimal model that fits all load modes within an
acceptable error range. To achieve this goal, we evenly divide
the range of i [0.05, 0.5](i = 0, . . . , 3) with step size 0.05. For 5.5 Evaluation of the PRB 3R Model with Moment
any set of i , the spring stiffness are computed for two extreme Load
loads, pure moment and pure vertical force. The optimal set For the moment load ( = ), the actual values of the PRB
of i is the one which has the minimum difference between the angles i are plotted as dots in Fig. 9 in which the range of
spring stiffness for end moment load and end force load. Because tip slope is [0, 3/2]. A linear regression process is used to
these two load modes represent the extreme cases, the difference compute the approximated values of spring stiffness k1 , k2 , k3
of spring stiffness for all other load modes will be guaranteed as
smaller. Note this is a three dimensional search procedure since
the sum of i equals one. The details of the procedure is listed k1 = 3.51933, k2 = 2.78518, k3 = 2.79756. (31)
below
1. Pick a set of i [0.05, 0.5](i = 0, . . . , 3). The linear fit curves are plotted as continuous lines in Fig. 9.
2. Compute spring stiffness for end moment load ( = ) us- The error of tip locus of the PRB 3R model is plotted in Fig.
ing the procedure in Section 5.3 and name them as km j ( j = 10. The maximum error of the tip deflection is 2.3% of beam
1, 2, 3) length for [0, 3/2]. And the relative error of the tip slope
3. Compute spring stiffness for vertical end force load ( = angle is given by
0, = /2) and name spring stiffness as k f j ( j = 1, 2, 3)
4. Repeat step 2 and 3 for each set of i [0.05, 0.5]
5. The optimal set of characteristic radius factors is the one in |1/k1 + 1/k2 + 1/k3 1| = 0.064% (32)
which km j is closest to k f j , that is
As a comparison, the PRB 1R model give an error of 31.8% for
( ! )
3 tip locus and an error of 25.8% for tip slope angle at = 3/2.
i | min (km j k f j )2 , i [0.05, 0.5] When using PRB 1R model, the slope angle must be limited be-
j=1 low 100 . Figure 11 shows the tip loci of the PRB 1R model

7 c 2008 by ASME
Copyright
tip error b/l
0.30 1.0 PRB 3R Model
PRB 1R model (=0.735)
Actual locus
0.25
0.20 0.8
0.15 PRB 1R model
0.10 PRB 3R model 0.6 (=0.85)
0.05

0 0.4
1 2 3 4

Figure 10. Comparison of tip error of PRB 1R model with the PRB 3R 0.2
model for beams under moment load =/2

0.0 a/l
Numerical Integration 0.0 0.2 0.4 0.6 0.8 1.0
b/l PRB 3R Model
0.8 o=37/50 PRB 1R Model (=0.735) Figure 12. Comparison of tip locus of the PRB 1R model and the PRB
PRB 1R Model
(=0.735) PRB 3R Chain Tip 3R Model for beams under force load.
0.6
Deflected beam
6 THE LOAD INDEPENDENT PRB 3R MODEL
0.4
The goal of this section is to find a PRB 3R model that is
3R chain
independent on the external load. For the optimal set of i given
Deflected beam o=9/50 in (30), we plot the spring stiffness k1 , k2 , k3 vs. [0, ]
3R chain
0.2 3R chain with = 0 in Fig. 13(a). It can be seen that the spring stiffness is
o=3/2 only slightly correlated to force direction angle . Hence the ef-
Deflected beam fect of can be neglected in finding the desired load independent
a/l model.
-0.2 0.0 0.2 0.4 0.6 0.8 1.0
Figure 13(b) shows the plot of the spring stiffness vs.
Figure 11. Comparison of tip locus of the PRB 1R model and the PRB [0, 25] with fixing = /2. The spring stiffness k1 drops from
3R Model for beams under moment load. 3.716 at = 0 to 3.445 at = 5 and then remains constant. Simi-
larly k2 increases from 2.871 at = 0 to 3.031 at = 5 and then
remains constant. However k3 suffers a sharp increase from
and PRB 3R model. It can be seen that the PRB 1R model de-
2.264 at = 0 to 2.600 at = 0.1 and then flattens out.
viates off the actual curve when > 120 , while the PRB 3R
model is still accurate. The shape of the deflected beam and cor-
responding configuration of the PRB 3R chain at three key po- =0 =/2
sitions = 9/50, = 37/50, = 3/2 are also plotted in k1 3.6
3.6
3.4 k1
Fig. 11. 3.4
3.2 3.2
3.0
2.8 k2 3.0 k2
2.6 2.8
5.6 Evaluation of the PRB 3R Model with Force Load 2.4 k3
2.6 k3
Similarly the spring stiffness are computed for force only 0.0 0.5 1.0 1.5 2.0 2.5 3.0
5 10 15 20 25
load ( = 0) as (a) (b)

Figure 13. (a) Plot of stiffness ki vs. the force directional angle
k1 = 3.71591, k2 = 2.87128, k3 = 2.26417, (33) [0, ] for = 0 and (b) Plot of stiffness ki vs. the load index [0, 25]
for = /2
where the characteristic radius factors in (30) are used and the
maximum tip slope angle is = by (12). Figure 12 shows
the tip loci of PRB 1R model and the PRB 3R model for beams The spring stiffness for the final load independent PRB 3R
under a vertical force load ( = /2). Once again, the PRB 3R model is computed by averaging over [0, 25]. And the kine-
model has a smaller error of 1.2% than the error of 3.6% for PRB matic and mechanic parameters of the load independent PRB 3R
1R model at = /2. model are summarized in Table 1.

8 c 2008 by ASME
Copyright
Table 1. Kinematic and mechanic parameters for the optimized PRB 3R
, , by solving the static equations (29),
model
2. compute the tip point a/l, b/l and tip slope angle with
0 = 0.1, 1 = 0.35, 2 = 0.40, 3 = 0.15 PRB angles computed from step 1 by forward kinematics
k1 = 3.51 EIl , k2 = 2.99 EIl , k3 = 2.58 EIl (20) and
3. compute error of the tip point and slope angle with the nu-
merical integration solution plotted in Fig. 4.
Numerical Integration PRB 3R Model
Figure 14 shows the actual and approximated tip loci for dif-
1.0 b/l 1.0
=0 =0.1 ferent load indices with fixed force angle = /2. The max-
0.8 0.8
imum error for the tip point is about 2.2% of the beam length.
And Fig. 15 shows the actual and approximated tip loci for dif-
0.6 0.6 ferent force angles with fixed load index = 0. The maximum
error of tip is about 8.5% when = 175 which corresponds to
0.4 0.4 the maximum tip slope angle max = 153 .
The most difficult load mode to approximate is when = 2
0.2 (error/l)max=1.9% 0.2 (error/l)max=1.6% and is small (Fig. 4). This corresponds to a load in which
moment is comparable with an axially dominant force. For this
0.0 0.0
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0 case, tip deflection angles must be constrained below a certain
a/l value in order to maintain a certain error bound.
1.0 1.0
=1 =2
0.8 0.8
Numerical Integration PRB 3R Model
0.6 0.6
b/l =30
o
=60
o
1.0 1.0
0.4 0.4 0.8 0.8
0.6 0.6
0.2 (error/l)max=0.9% 0.2 (error/l)max=1.1%
0.4 0.4
(error/l)max=0.35% (error/l)max=0.97%
0.0 0.0
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0 0.2 0.2
a/l
1.0 1.0 -0.4-0.2 0.0 0.2 0.4 0.6 0.8 1.0 -0.4-0.2 0.0 0.2 0.4 0.6 0.8 1.0
=5 =50 o o
0.8 0.8
1.0 =120 1.0 =135
0.8 0.8
0.6 0.6
0.6 0.6

0.4 0.4 0.4 0.4


0.2 (error/l)max=3.8% 0.2 (error/l)max=4.9%
0.2 0.2
(error/l)max=1.2% (error/l)max=2.2% -0.4-0.2 0.0 0.2 0.4 0.6 0.8 1.0 -0.4-0.2 0.0 0.2 0.4 0.6 0.8 1.0
0.0 0.0 o o
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0 1.0 =150 1.0 =175
0.8 0.8
Figure 14. Approximation error of the PRB 3R model for various load
indices with = /2. Dots represent the deflection loci computed via 0.6 0.6
numerical integration and the solid lines represent the approximated tip 0.4 0.4
loci predicted the load independent PRB 3R model. 0.2 0.2
(error/l)max=6.2% (error/l)max=8.5%
-0.4-0.2 0.0 0.2 0.4 0.6 0.8 1.0 -0.4-0.2 0.0 0.2 0.4 0.6 0.8 1.0
6.1 Comparison of the Load Independent PRB 3R
Model with the Result of Numerical Integration Figure 15. Approximation error of the PRB 3R model for various force
The error analysis of the load independent PRB 3R model is direction angle with fixed load index = 0. Dots represent the actual
carried out via the following three steps: deflection loci computed via numerical integration and the solid lines rep-
resent the approximated tip loci predicted by the load independent PRB
1. compute the PRB angles 1 , 2 , 3 for a given external load 3R model.

9 c 2008 by ASME
Copyright
A
B
A Au
r
M
y Av F Q
M F
O
v
Q
1 x 2 3
u Figure 17. The internal force and moment at the connection point Q.
O
P1 P2 P3 l Q
the beam OQ at the point Q as F = (Fx , Fy )T and M respectively.
Figure 16. A compliant four-bar linkage is comprised of a fixed-fixed flex- See Fig. 17. They can be obtained by solving equation (26)
ible beam OQ and and two rigid links OA and AQ connected by two pin
joints A and B.
Fx l k1 1
Fy l = [J T ]1 k2 2 , (35)

M k3 3

6.2 Kinetostatic Analysis of a Compliant Four-Bar where inverse of the matrix (27) can be derived analytically and
with the PRB 3R Model written as
To demonstrate the use of the proposed PRB 3R model in an-
alyzing compliant mechanisms, here we study a compliant four-
2 c12 1 c1 2 c12 1 c1
bar linkage shown in Fig. 16. The crank link OA is hinged to the
[J T ]1 = 2 s12 1 s1 2 s12 1 s1 /(1 2 s2 ).
ground at the pint joint B and to the coupler link AQ at the pin
2 3 s3 3 (2 s3 + 1 s23 ) 1 (2 s2 + 3 s23 )
joint A respectively. Links OA and AQ are considered rigid. The
(36)
follower link OQ is initially straight and considered flexible with
The static equilibrium equation is obtained by vanishing the
one end fixed on the ground and the other end rigidly connected
resultant moment of the reaction force F0 and M 0 with respect to
to the coupler link AQ at the point Q.
the pin joint A, written as
The coordinate systems are defined as the following. The
origin of the global coordinate system xy is chosen at the fixed
end of the flexible beam with the x axis along the beam. For con- AQ F0 + M 0 k = 0, (37)
venience, a moving coordinate system uv is placed on the coupler
link at the connection point Q with the u axis along the slop of the where the symbol represents the vector cross product and
beam tip. Let Q = (Qx , Qy )T , B = (Bx , By )T and A = (Ax , Ay )T AQ = (AQx , AQy )T is the vector from the point A pointing to the
denote the coordinates of the points Q, B, A respectively. The point Q, calculated from (34) as
length and position of the crank link are denoted by r and the an-
gle respectively. The position of the pin joint A in the moving
frame system are denoted (Au , Av )T . AQx Qx Ax c123 s123 Au
= = (38)
The kinematic constraint equations are obtained by applying AQy Qy Ay s123 c123 Av
the loop closure at the pin joint A, written as
By using the fact that F0 = F and M 0 = M, Eq. (37) can
be re-written in scalar form as
Qx c123 s123 Au Bx cos
+ = +r , (34)
Qy s123 c123 Av By cos
AQx Fy AQy Fx + M = 0. (39)
And the coordinates of the beam tip Q can be written in terms of
three pseudo rigid angles via the forward kinematics (20). Substituting (35) and (38) into (39) and combing with the
By disconnecting the follower link and the coupler link at kinematic constraint equations (34) yield a system of three equa-
the point Q, we define the internal force and moment applied to tions in three unknowns 1 , 2 , 3 . Assuming the crank angle

10 c 2008 by ASME
Copyright
is controlled, we can then apply numerical solvers such as is plotted as blue lines in Fig. 18. To verify the results, a finite
Newton-Raphson based iteration method to obtain solutions of element analysis (FEA) model is built with ABAQUS software.
1 , 2 , 3 . Note it is possible to transfer the resulted equations The crank link and coupler link are defined as discrete rigid ele-
into a polynomial form by using the approximation techniques ment and the flexible follower link OQ is meshed into 20 planar
proposed by Su and McCarthy (2005) [13] and analytically solve beam elements. The linkage based on the FEA model is plotted
the polynomial system to obtain solutions. However this is out of as red lines Fig. 18.
the scope of this paper and shall be an interesting research topic It can be seen that the deformation of the PRB 3R model is
in the future. very close to the FEA model. To find the accuracy of the PRB
3R model, we calculate the normalized tip error relative to the
FEA model and plot it in Fig 19. It can be seen that a maximum
b/l 1.0 A 1.0
tip error is about 1.2% of the beam length for [0, 2], which
0.8
=/4 0.8
=/2 is acceptable for conceptual design of compliant mechanisms.
0.6
B 0.6

0.4 0.4 Tip Error/l


Q
0.2 0.2

O a/l 0.010
-0.4 -0.2 0.2 0.4 0.6 0.8 1.0 -0.4 -0.2 0.2 0.4 0.6 0.8 1.0
0.008
-0.2 -0.2

1.0 1.0
=3/4 = 0.006
0.8 0.8
0.004
0.6 0.6
0.002
0.4 3R model 0.4

0.2 0.2

1 2 3 4 5 6
FEA
-0.4 -0.2 0.2 0.4 0.6 0.8 1.0 -0.4 -0.2 0.2 0.4 0.6 0.8 1.0 Figure 19. Tip error of the optimized 3R model compared with the FEA
-0.2 -0.2 model for [0, 2]
1.0 1.0

0.8
=3/2 0.8
=7/4
0.6 0.6

0.4 0.4 8 CONCLUSION


This paper proposes a pseudo-rigid-body 3R model that is
0.2 0.2
independent on external loads for analyzing large deflection of
-0.4 -0.2 0.2 0.4 0.6 0.8 1.0 -0.4 -0.2 0.2 0.4 0.6 0.8 1.0 cantilever beams subjected to combined end force and moment.
-0.2 -0.2 The model is a planar serial chain with three revolute joints, each
of which is superimposed by a torsion spring. The deflection
Figure 18. Plots of the deflection of a partially compliant 4-bar mecha- of flexible beams depends on three load parameters: force in-
nism computed by the optimized PRB 3R model and by the FEA model dex, force direction angle and load index. Numerical integration
with the input crank [0, 2]. is used to compute the actual beam deflection (tip point and tip
slope) for various load modes. These three dimensional data sets
are then input to a three dimensional search routine in order to
find the optimal kinematic parameters (characteristic radius fac-
tors) and mechanic parameters (spring stiffness). The result is a
7 A NUMERICAL EXAMPLE serial 3R model that is independent on external loads. The error
Consider the partially
compliant4-bar shown in Fig. 16 analysis for various load modes shows that this model is accu-
and let A u /l = 1/ 2, A v /l = 1/ 2, r/l = 1 2/2, B = rate and can accommodate a wide range of load indices, force
(0, 1/ 2). By driving the crank angle [0, 2], the motion direction angles and slope angles.
of the flexible beam and the coupler link is determined by simul- Lastly a compliant mechanism example is used to demon-
taneously solving the kinematic constraint equations (34) and the strate how to use the proposed PRB 3R in analyzing compliant
static constraint equation (39). The linkage at six key positions mechanisms. A numerical example is provided and compared

11 c 2008 by ASME
Copyright
with the result of a FEA model. The result shows a maximum Bistable Compliant Four-Bar Mechanisms Using Polyno-
tip error of 1.2% when the optimized PRB 3R is used. This mial Homotopy, ASME Journal of Mechanical Design,
model is an important contribution to the field and is very suitable 129(10):1094-1098.
for kinematic/static synthesis leading to the conceptual design of [13] Su, H.-J., and McCarthy, J. M., 2005, A Polynomial Ho-
compliant mechanisms. motopy Formulation of the Inverse Static Analysis of Planar
Compliant Mechanisms,, ASME Journal of Mechnaical De-
sign, in press.
ACKNOWLEDGMENT
The author wish to express his thanks to Mr. Kun He and
Mr. Fenqi Luo from Department of Mechanical Engineering, APPENDIX: DEFLECTION EQUATIONS OF FLEXIBLE
University of Maryland, Baltimore County for their help with BEAMS
using ABAQUS software. The deflection of a flexible beam is governed by Bernoulli-
Euler equation which states that the bending moment at any point
on the beam is proportional to its curvature, written as
REFERENCES
[1] Howell, L. L., 2001, Compliant Mechanisms, Wiley- M d d 2 y/dx2
Interscience. = = , (40)
EI ds [1 + (dy/dx)2 ]3/2
[2] Howell, L.L., 1991, The Design and Analysis of Large-
Deflection Members in Compliant Mechanisms, M.S. The-
where l and EI be the length and flexural rigidity of the beam
sis, Purdue University, West Lafayette, Indiana.
respectively, and s are the deflection angle and the arc length
[3] Howell, L. L., and Midha A., 1995 Parametric Deflection
along the beam. And M is the internal moment at any point (x, y)
Approximations for End-Loaded, Large-Deflection Beams in
along the beam.
Compliant Mechanisms, ASME Journal of Mechnaical De-
sign, 117(1):156-165.
[4] Howell, L. L., Midha A., and Norton, T. W., 1996, Evalua- Small Deflection Beam Equations
tion of Equivalent Spring Stiffness for Use in a Pseudo-Rigid- Figure 20 shows an initially straight cantilever beam sub-
Body Model of Large-Deflection Compliant Mechanisms, jected to (a) small vertical force F and (b) small moment M .
ASME Journal of Mechnaical Design, 118(1):126-131. Under the small deflection assumption [10], the square of the
[5] Saxena, A., and Kramer, S. N., 1998 A Simple and Accu- slop (dy/dx)2 can be assumed to be small in the denominator of
rate Method for Determining Large Deflections in Compliant Eq. (40). As a result, Bernoulli-Euler equation can be simplified
Mechanisms Subjected to End Forces and Moments, ASME as
Journal of Mechnaical Design, 120(3):392-400.
[6] Lyon, S. M., Howell, L. L., Roach, G. M., 2000, Modeling M d2y
= 2. (41)
Fixed-Fixed Flexible Segments via the Pseudo-Rigid-Body EI dx
Model, Proceedings of the ASME Dynamics and Control Di-
vision, at the 2000 ASME International Mechanical Engineer-
ing Congress and Exposition, Orlando, FL, Nov. 5-10, DSC-
Vol. 69-2, pp. 883-990. Fo
Mo
y y
[7] Saxena, A., 1997, A New Pseudo-Rigid-Body Model for
Flexible Members in Compliant Mechanisms, Masters The-
bF bM
sis, University of Toledo, Toledo, OH, USA. x x
[8] Saggere, L. and Kota, S., 2001, Synthesis of Planar, Com- EI, l EI, l
(a) (b)
pliant Four-Bar Mechanisms for Compliant-Segment Mo-
tion Generation, ASME Journal of Mechnaical Design, Figure 20. The force index and moment index are the tip slope angle
123(4):535-541. of a cantilever beam subjected to (a) a small vertical force F and (b) a
[9] Beer, F.P., Johnston, E.R., DeWolf, J.T., 2006, Mechanics of small moment M respectively.
Materials, 4th edition, McGraw-Hill, New York.
[10] Roark, R.J. and Young, W.C., 1982, Formulas for Stress
and Strain, McGraw-Hill, New York. Substituting the beam slope
[11] Tsai, L. W., 1999, Robot Analysis, The Mechanics of Serial
and Parallel Manipulators, Wiley-Interscience. dy
[12] Su, H. J., and McCarthy, J. M., 2007, Synthesis of = (42)
dx

12 c 2008 by ASME
Copyright
View publication stats

and moment for each load case into (41) to obtain Note a + sign is used when taking the square root due to the
load assumption aforementioned. If there is a inflection point in
F (l x) d the beam, the curvature could be positive, zero or negative. For
= : end force (43) the negative curvature case, a - sign should be used instead.
EI dx
M d By using the terms defined in (48-49), Eq. (52) can be re-
= : end moment (44) written as
EI dx
q
d 1
Integrating the above equations with the following boundary con- = 4[cos( ) cos( )] + 2 . (53)
ditions, ds l

Separating variables in (53) and integrating yields


|x=0 = 0, y|x=0 = 0 (45)
Z 0
1 d
yields = p , (54)
2 0 [cos( ) cos( )] +

F x F x2
= (2l x), y = (3l x) : end force (46) where is called the load index, defined by
2EI 6EI
M M 2
= x, y = x : end moment (47) 2
EI 2EI = = : load index (55)
4 2
Evaluating Eqs. (46-47) at x = l gives the tip slope angle and
tip deflection b/l for end force and end moment independently, Similarly the normalized tip deflection a/l, b/l can be computed
by
F l 2 F l 2 Z 0
= , bF /l = : end force, (48) 1 cos d
2EI 3EI a/l = p , (56)
M l M l 2 0 [cos( ) cos( )] +
= , bM /l = : end moment. (49) Z 0
EI 2EI 1 sin d
b/l = p . (57)
2 0 [cos( ) cos( )] +
Large Deflection Beam Equations
When a beam undergoes a large deflection, the exact
Bernoulli-Euler equation (40) must be used. By geometry (Fig. Equations (54) and (56-56) relate the deformation parame-
3), the internal moment at any point (x, y) along the beam can be ters , a/l, b/l and the load parameters , , . Solving them
calculated as with specified three parameters yields the other three parameters.

M = F [(a x) sin (b y) cos ] + M . (50)

Substitute (50) and differentiate (40) with respect to s to obtain

d2 F
= sin( ), (51)
ds2 EI

where dx/ds = cos and dy/ds = sin are used. Integrating (51)
with boundary conditions d/ds = M /(EI) at the tip =
yields

s
d 1 2F l 2 M l 2
= [cos( ) cos( )] + (52)
ds l EI EI

13 c 2008 by ASME
Copyright

Você também pode gostar