Optimal Design of A Stepper-Driven Planar Linkage Using Entropy Methods

OPTIMAL DESIGN OF A STEPPER-DRIVEN PLANAR

LINKAGE USING ENTROPY METHODS
J. C. Musto1*
Mechanical Engineering Department,
Milwaukee School of Engineering,
1025 N. Broadway, Milwaukee, WI 53202-3109, USA
1

ABSTRACT
Stepper motors can be used to provide open-loop motion control in
mechanisms. Unlike servo controlled mechanisms, however, the rotational
drive input error cannot be resolved below the step error in the motor.
Therefore, there is a fixed level of rotational position error that must be
accepted in stepper driven mechanisms, and this rotational position error
will inevitably propagate to kinematic position error in the mechanism. In
this paper, the direct linearization method will be used to derive a model for
kinematic position error based on uncertainty in the rotational input angle
of a mechanism. Using this model, a method of constrained optimization
to design a mechanism to minimize the effect of uncertain input conditions
on kinematic position will be presented. The method is based on entropy
minimization techniques that have been applied in a variety of robotic
system applications. The method will be demonstrated in a case study,
and will be shown to optimize the positioning reliability of a mechanism
under input angle errors. The method will be shown to accurately predict
drive error propagation, through comparison to Monte Carlo simulation.
When coupled with entropy-based system reliability optimization methods,
optimal mechansims can be synthesized in response to various positioning
constraints.
KEYWORDS: Mechanisms; Optimization; Direct linearization; Entropy

1.0 INTRODUCTION
Stepper motors offer the ability to provide precise motion control
without the need for closed-loop control systems. As the motor will
index a fixed rotational increment with each pulse applied to the motor,
simple step-counting control methods can be used for position control.
However, the positioning accuracy of a stepper motor is a physical
limitation of the motor itself. Stepper motor accuracy is typically rated
as a percentage of the motor step size; this accuracy is a property of
* Corresponding author email: musto@msoe.edu

ISSN: 2180-1053

Vol. 6

No. 1

January - June 2014

Journal of Mechanical Engineering and Technology

motor construction, and cannot be increased through traditional closedloop control methods. For example, a typical stepper motor might be
rated at 1.8/step, with an accuracy of +/- 5%. Since this motor accuracy
is a function of motor design, it cannot be reduced through feedback
control methods.
Since stepper motor error is inherent in the design of the motor, kinematic
mechanisms driven by stepper motors are prone to positioning errors,
as the actuator error propagates through the mechanism. The effect of
kinematic errors on planar mechanisms has been studied by numerous
researchers, including many recent efforts. Pandey and Zhang
propose an entropy method for optimizing the reliability of robotic
manipulator positioning (Pandey & Zhang, 2012; Zhang & Pandey,
2013). This approach uses simulation trials to evaluate the position
error distribution of the manipulator under joint angle uncertainties.
Approaches based on strain-energy error minimization (Aviles, et al.,
2009) have been developed for optimal synthesis of planar linkages.
A method based on reliability optimization using covariance matrix
estimates has also been proposed (Huang & Zhang, 2010, 2012). Each
of these methods can be used to provide some measure of design under
uncertainty for a planar mechanism.
Methods based on direct linearization to estimate the effect of
parameter variation on mechanism position provide perhaps the most
promising approach to quantifying uncertainty in planar mechanisms
(Wittwer et al., 2004; Leishman & Chase, 2010). In this paper, these direct
linearization methods will be adapted for use in the optimal design
of stepper-driven planar mechanisms. These methods will be shown
to be consistent with other entropy-based approaches to kinematic
optimization (Musto, 2002).
2.0

DIRECT LINEARIZATION MODEL OF A STEPPERDRIVEN FOUR BAR LINKAGE WITH ERROR IN THE
DRIVE ANGLE

Researchers have demonstrated that the direct linearization method

(DLM) can be used to accurately model the effect of manufacturing
tolerances on link lengths in a kinematic linkage (Wittwer et al., 2004;
Leishman & Chase, 2010). In this section, the DLM will be modified
to generate a model for kinematic position error based on error in
the input angle. It is this input angle, which is inherent when stepper
motors are used to drive mechanisms, which will be considered in the
2

ISSN: 2180-1053

Vol. 6

No. 1

January - June 2014

BAR
BARLINKAGE
LINKAGEWITH
WITHERROR
ERRORIN
INTHE
THEDRIVE
DRIVEANGLE
ANGLE
Researchers
demonstrated
that
method
(DLM)
can
used
Researchers have
have
demonstrated
that
thedirect
directlinearization
linearization
method
(DLM)
canbe
be
used
Optimal
Design of
A the
Stepper-Driven
Planar Linkage
Using
Entropy
Methods
toto accurately
model
the
effect
of
manufacturing
tolerances
on
link
lengths
accurately model the effect of manufacturing tolerances on link lengths inin aa
kinematic
kinematiclinkage
linkage(Wittwer
(Wittweretetal.,
al.,2004;
2004;Leishman
Leishman&&Chase,
Chase,2010).
2010).InInthis
thissection,
section,the
the
DLM
will
be
modified
totogenerate
aamodel
for
kinematic
position
error
based
on
error
DLM
will
be
modified
generate
model
for
kinematic
position
error
based
on
error
remainder
of this paper.
ininthe
theinput
inputangle.
angle.ItItisisthis
thisinput
inputangle,
angle,which
whichisisinherent
inherentwhen
whenstepper
steppermotors
motorsare
areused
used
totodrive
mechanisms,
which
will
be
considered
in
the
remainder
of
this
paper.
drive mechanisms, which will be considered in the remainder of this paper.

Consider the four-bar linkage, shown in Figure 1.

Consider
Considerthe
thefour-bar
four-barlinkage,
linkage,shown
shownininFigure
Figure1.1.

Figure
four-bar
Figure
1. 1.The
linkage
Figure
1.The
Thefour-bar
four-barlinkage
linkage

The
equation
for
mechanism is:
Theloop
loop
equation
forthis
this
is:
The
loop
equation
formechanism
this mechanism
is:
++==00

(1)
(1)

This
Thisloop
loopequation
equationcan
canbe
beexpanded
expandedtotothe
thefollowing
followingscalar
scalarequations:
equations:

This loop equation can be expanded to the following scalar equations:

1 (2 , 3 , 4 ) = cos 2 + cos 3 cos 4 = 0

(2)
1 (2 , 3 , 4 ) = cos 2 + cos 3 cos 4 = 0
(2)
2 (2 , 3 , 4 ) = sin 2 + sin 3 sin 4 = 0
(3)
+ sin
cos33sin
cos
(2)
cos22 +
21(2 , 3 , 4 ) = sin
(3)
4 4=0 = 0
If the coupler point is considered to be the point that represents the desired position of
) =equations
the
mechanism,
describing
position
point can
also be
If the
coupler
point
bethe
represents
the desired
position
of
2 (
sin to
, 3 ,isthe
sin point
3 the
that
sin
4 = 0of this
(3)
2then
4considered
2+
If
coupler
point
considered
to be
pointof that
represents
written
in
scalar form:
thethe
mechanism,
then
the is
equations
describing
thethe
position
this point
can alsothe
be
written
in position
scalar
form:
If the coupler
point
isofconsidered
to be the point
represents
the desired
positionthe
of
desired
the mechanism,
thenthat
the
equations
describing
(
)
the
mechanism,
then
the
equations
describing
the
position
of
this
point
can
also
be
,

cos

cos(
+

)
(4)

2
3
4
2
3
position of this point can also be written in scalar form:
written in scalar form:
(2 , 3 , 4 ) = cos 2 + cos( + 3 )
(4)
(2 , 3 , 4 ) = sin 2 + sin( + 3 )
(5)
cos22 +
+ sin(
cos(++33))
(4)
(2 , 3 , 4 ) = sin
(5)
where and represent the x and y positions of the coupler point .
(2 , 3the
, 4x)and
= ysin
2 + of
sin(
+ 3 ) point .
(5)
where and represent
positions
the coupler
In the positioning of the mechanism, the input angle 2 is uncertain; the error associated
where
angle
and isrepresent
the
x and
ythe
positions
ofthe
.the error
In
thethe
positioning
input
angle
coupler
with
of2 . the
Themechanism,
error
associated
with
in point
deviations
fromassociated
nominal
2 is uncertain;
2 results
where
andisP
the
x and
y positions
of deviations
the
coupler
point
P.
with the
.represent
The
associated
2 results
in
fromEquations
nominal
values
forPangle
4y2as
well;error
these
errors
arewith
defined
as 3 and
4 . Applying
x3 and
In
the
positioning
of
the
mechanism,
the
input
angle

is
uncertain;
the
error
associated
(2)
to (5)
the mechanism
with errors
uncertainty
in positioning
to the following
values
for to3 and
are defined
as 3 and leads
Equations
2
4 as well; these
4 . Applying
withthe
angle
is mechanism
errorwith
associated
withthe
2 positioning
results
deviations
from
nominal
modified
equations:
(2)
tothe
(5)
to the
uncertainty
in
leads toisthe
following
In
positioning
of the
mechanism,
inputinangle
uncertain;
2 . The
2
modified
equations:
values
for

and

as
well;
these
errors
are
defined
as

and

.
Applying
Equations
3
4
3
4
the error associated
with the angle is 2. The error
associated
with 2
) =mechanism
(2)1to((5)
with
the= following
, 4the
cos(2 +
cos(3 +in
3 )positioning
cos(4 leads
+ 4 )to

0
(6)
2 , 3to
2 ) +uncertainty
results
in
deviations
from
nominal
values
for

and

as
well;
these
3
4
modified
1 (2 , equations:
3 , 4 ) = cos(2 + 2 ) + cos(3 + 3 ) cos(
(6)
4 + 4 ) = 0
errors
as 3 and
. Applying Equations (2) to (5) to the
2 (2are
, 3 , defined
(7)
4 ) = sin(2 +
2 ) + 4 sin(3 + 3 ) sin(4 + 4 ) = 0
cos(22 +
+22)) +
+ sin(
cos(33++33))sin(
cos(
(6)
21(2 , 3 , 4 ) = sin(
(7)
4 )=0 = 0
4 4++
4)
(2 , 3 , 4 ) = cos(2 + 2 ) + cos( + 3 + 3 )
(8)
Vol.
6) +No.
1 )January
- June
2014
) = 2 cos(
2 (2 , 3,(
ISSN:
=3 ,2180-1053
+ 2 )2+
(7)3
+ 2sin(
+
cos(
+ sin(
(8)
42),
3
3
4)+ 4 ) = 0
4sin(
3 + 3
(2 , 3 , 4 ) = sin(2 + 2 ) + sin( + 3 + 3 )
(9)

(2 , 3 , 4 ) = sin 2 + sin( + 3 )
(5)
where
and represent
the x and ythe
positions
of the coupler
point .
In
the positioning
of the mechanism,
input angle
2 is uncertain; the error associated
with
2 . The the
error
associated
withofthe
in point
deviations
from nominal
2 results
and isrepresent
x and
y positions
coupler
.
wherethe angle
In the positioning
the mechanism, the input angle 2 is3uncertain;
the error Equations
associated
values
for 3 and of
and 4 . Applying
4 as well; these errors are defined as
withtothe
angle
is mechanism
2 . The error
associated
within2 positioning
results in deviations
fromfollowing
nominal
(2)
(5)
to
the
with
uncertainty
leads
to
the
In the positioning
of the mechanism, the
angle is uncertain;
thethe
errorfollowing
associated
mechanism
with
ininput
positioning
values forequations:
3 and
4 asuncertainty
well; these errors
are
defined 2as 3leads
and 4 .toApplying
Equations
modified
with the angle
is 2 . The error associated with 2 results in deviations from nominal
modified
equations:
(2) to (5) to
the mechanism
with uncertainty in positioning leads to the following
values
for

and
4 as well; these errors are defined as 3 and 4 . Applying Equations
3
modified
equations:
(
)
,

cos(3 +in
3 )positioning
cos(4 leads
+ 4 )to

0
(6)

cos(
1
2
3
4
2
2 ) +uncertainty
(2) to (5) to the mechanism with
the= following

Journal of Mechanical Engineering and Technology

modified
equations:
(22, ,
33, ,44)) =
= sin(
cos(22++22))++ sin(
cos(
+ 3 ) sin(
cos(
21(
++4)4 )=0 = 0
3 3+ 3 )
44
, 3 , 4 )) = cos(2 + 2 ) + cos(3 + 3 ) cos(4 + 4 ) = 0
1 (
2 (22 ,
3(
, 24, =
sin( + 2 )2 +
2sin(
+ 3 ) +

3sin(
+
) + 3 cos(
+ 34)+ 4 ) = 0
3 , 4 ) = 2 cos(
2 (2 , 3 ,(
4 ) = sin(2 + 2 ) + sin(3 + 3 ) sin(4 + 4 ) = 0
cos(++
+ 3 )
= sin(
cos(22++22))++ sin(
(22,,33,,44)) =
3 3+ 3 )

(6)
(7)
(6)
(7)
(8)

(7)
(8)
(9)

(
, , ) = cos(2 + 2 ) + cos( + 3 + 3 )
(8)
(22 , 33 , 44 ) =identities,
sin(2 +and
2 ) +
sin( +by
3 +
3 )
(9)
Applying trigonometric
linearizing
invoking
the small angle
approximation,
yields:
Applying (
trigonometric
identities,
and
linearizing
by
invoking
the
(9)
2 , 3 , 4 ) = sin(2 + 2 ) + sin( + 3 + 3 )
Applying trigonometric
identities, and linearizing by invoking the small angle
small
angle approximation,
yields:
approximation,
1 (2trigonometric
, 3 ,yields:
4 ) = (cos
sin 2linearizing
) + (cos by
3 sin 3 ) the
small angle
2 2 and
3 invoking
Applying
identities,

(cos

sin

=
0
(10)
4
4
4
approximation,
yields:
1 (2 , 3 , 4 ) = (cos 2 2 sin 2 ) + (cos 3 3 sin 3 )
2 +4 sin
4 )
= 0
(10)
(cos 4
2 cos
2 (
2 , 3 , 4 ) = (sin
2 ) + (sin 3 + 3 cos 3 )
1 (
2 , 3 , 4 ) = (cos 2 2 sin 2 ) + (cos 3 3 sin 3 )

(sin

cos

)
=
0
(11)
4
4
4
(10)
(cos 4 4 sin 4 ) = 0
2 (2 , 3 , 4 ) = (sin
2 + 2 cos 2 ) + (sin 3 + 3 cos 3 )
(sin
+2
cos
=cos
0 (cos 3 3 sin 3 )
(11)
4 )
3 ,, 4 ), =)=(cos
sin
2 )+
(
2
22,(
(sin
42+
42 cos
2 3 4
2 ) + (sin 3 + 3 cos 3 )
sin (sin
(sin

cos

)
(12)
3 cos
3
3
(11)
4 +
4
4) = 0
2 sin
(2 , 3 , 4 ) = (cos 2
2 ) + cos (cos 3 3 sin 3 )
sin (sin 3 cos
+
(12)
3 2cos
) +3)sin (cos 3 3 sin 3 ) +
(
2 , 3 , 4 ) = (sin 2 + 2
(2 , 3 , 4 ) = (cos 2 2 sin 2 ) + cos (cos 3 3 sin 3 )
cos(sin
(sin3 +
+3cos
cos3))
(13)
sin
(12)
(2 , 3 , 4 ) = (sin 2 + 32 cos32 ) + 3 sin (cos 3 3 sin 3 ) +
cosinEquations
(sin 3 +(10)
3 cos
) yields:
(13)
Isolatingthe
error terms only
and3(11)
(2 , 3 , 4 ) = (sin 2 + 2 cos 2 ) + sin (cos 3 3 sin 3 ) +
cos 3 )(11) yields:
(13)
cosin Equations
(sin 3 + (10)
Isolating the error
terms
only
sin 4 3
sin 33 and
sin
0
2

(14)
Isolating the error
in Equations
(10)
(11)
terms
costerms
2 2inonly
cos
3
cos
4 and
0 yields:
4
Isolating the error
Equations
and
(11)
sin (10)
sin
4 yields:
3
sinonly
2
0
3

+
=
(14)
cos 2 2
cos 3 cos 4 4
0
sin 3 sin 4 3
sin 2
0

+
=
(14)
cos 2 2
cos 3 cos 4 4
0

which can be rewritten:

which can be rewritten:

which can be rewritten:
or:
or:
or:

0
2 + 3 =
0
3
4
2 + =
4
0

(15)
(15)

3 = 1 2
(16)
34 = 1 2
(16)
4
This relates the values of the dependent error terms 3 and 4 in terms of the
This
relates error
the values
of the
dependent
errorposition
terms
terms stepper
of the
independent
2 , which
corresponds
to the
error
theindriving
3 andof4
independent
error
, whichof
corresponds
to the position
error of3the
driving
This
the2values
the dependent
error terms
and
4 in stepper
terms
motor.relates
motor.
of the independent error 2, which corresponds to the position error of
Similarly
with stepper
Equationsmotor.
(13) and (14), the error terms can be isolated as:
the
driving
Similarly with Equations (13) and (14), the error terms can be isolated as:

cos sin 3 sin cos 3 0 3

sin
= sin 2 2 + cos
(17)
cos
sin cos 3 0 34
cos 22
sin sin 33 +

=
2 +

(17)
cos 2
sin sin 3 + cos cos 3 0 4

the position errors

coupler
point, -or:
4where and
ISSN:
Vol. 6of the
No.
1 January
June 2014
are2180-1053
where and are the position errors of the coupler point, or:

which can be rewritten:

4
3 = 1 2
(16)

0
or:
+ 43 3 =
1 terms 3 and 4 in terms of (15)
2dependent
This relates the values of the
error
the
4 = 0
(16)
2
Optimal Design of A Stepper-Driven
Planar
Linkage Using Entropy Methods
4 to the position
error
oftheindriving
stepper
independent
2 , which
This relates error
the values
of corresponds
the dependent
error
terms

and
terms
of
the
3
4

motor.
1position
2
(16)
3 to= the
or:
independent error 2 , which corresponds
error of the driving stepper
4
This relates the values of the dependent
error terms 3 and 4 in terms of the
motor.
Similarly
with
Equations
(13)
and
(14),
the
error
terms
can
be isolated
,
which
corresponds
theterms
position
error
of the
stepper
independent
error

Similarly with Equations

(13) and (14), the
error
can be
isolated
as:driving
2
3 to
This

and

in
terms
of (16)
the
= error
1 terms
motor.relates the values of the dependent
as:
3
4
4 error terms2 can be isolated as:
Similarly with Equations (13) and (14), the
independent
the position
error
of
the
driving
stepper
error
costo sin

sin

cos

0
2 , which
sin 2 corresponds

3
3
with
Equations
=
and
the error terms
3
(17)
2 +(14),
motor.
Similarly
(13)
be cos
isolated
as:

cos

3 +
can
cos
0

the
This
relates
values
error
terms
and
of the
sin
3 cos
33 4 in
0 terms
sin
cos sin
sin
43

sinof
22 the dependent
3
=

2 +
(17)
independent
which
corresponds
theposition
error
stepper
2 and
the
sinto
sin
+ cosbe
cosof
3 the
0 driving

error
2 ,cos
Similarly
with
(13)
(14),
error
terms

point,
sin isolated
cos
3 as:
0 43
cos
the
sincoupler
33 can

sin
position
Equations
2
where

and

are
the
errors
of
or:

motor.
=
2 +

(17)
and
sin errors
sin 3 +of the
cosare
2 the
cos coupler
cos 3 0point,

where
P
position
and
are sin
ofthe
coupler
point,
or: 0 4 or:
where P
x
ytheposition errors

sin

cos

cos
sin

3
3
2
3 be isolated
Similarlywith
(13)
and
the error
terms
as: 3
Equations
=


(17)
2 +(14),
2coupler
can
cos
(18)
sinof=
the
sin
+
cos
3 0 4

and
arecos
the2position errors
point,
or:
where
3+

43
cos 3 0
(18)

cos=
sin2 +3
sin
sin 2
4

= are(16)
Equation

(17)
3
and
the position
of the
coupler point,
or:
where

2 + errors
Substituting
Equation
into
(18)
yields:
3

sin

sin

cos

cos

cos

2
3
3
4
= 2 +
(18)
4
Substituting Equation (16) into Equation (18) yields: 1
3)

=the
( coupler
+
(19)
where and
errors
of=

point,
2or:
(18)
are the position
Substituting
Equation
(16) into
Equation
2 +(18)
4yields:
Substituting Equation (16) into Equation (18) yields:
1
= ( + ) 2
(19)
which relates the position error of the coupler point to the
rotational position error of the

3
Substituting
Equation
(16) intomatrix
Equation
(18)
yields:
1
1)
, which
(18)
=

(+

)
(19)
2++
describeserror
how
driving
stepper

2
4 rotational
which relates
the motor.
positionThe
error of the term
coupler(
point
to the
position
of the
the
rotational
error inmotor.
the drive
propagates
to position
1error in known as the sensitivity
driving stepper
Theangle
matrix
term
(
+

1) , which describes how the

)

=described
(
+
which relates
thethe
position
of the
coupler
point
to in
theFigure
rotational
position
errormatrix
of(19)
the
matrix
().error
For
typical
four-bar
linkage
this
sensitivity
Substituting
Equation
(16)error
into
Equation
(18)toyields:
rotational
in the
the
drive
angle
propagates
position
error 2in 1,
known
as the
sensitivity
which
relates
error
of the
the
rotational
driving stepper
motor.position
The matrix
term
( +coupler
1 ) ,point
which to
describes
how the
is:
matrixrelates
(). For
the
typicalerror
four-bar
linkage
in
Figure
1, this
sensitivity
matrix
-1 the
which
thein
position
of propagates
the
couplerdescribed
point toThe
the
rotational
position
of
rotational error
thethe
drive
angle
to
as (C+DB
theerror
sensitivity
position
error
of
driving
stepper
matrix
term
A),
1error
=motor.
(position
+ 1
)
2in known
(19)
is:
) in
driving
stepper
motor.
The
matrix
term
(
+

)
,
which
describes
how
the
sin(+

sin(

)
matrix
().
For
the
typical
four-bar
linkage
described
Figure
1,
this
sensitivity
matrix
2
4
3
which describes how the
rotational
error
in
the
drive
angle
propagates
sin 2 +
sin(3 error
)
rotational
error in the drive angle propagates
to
position
in
known
as
the
sensitivity
4
is:
sin(+
sin(
2
3 ) (S).position
=
which
relates the
position
error
ofsin
thethe
point
to4 ) the
rotational
error
of(20)
the
to
position
error
in known
as
sensitivity
matrix
For the
typical
sin(2 4 ) cos(+
+
coupler
3 ) 1, this sensitivity matrix
matrix (). For the typical four-bar
linkage
described
in
Figure
2
+
sin(
)

cos

31
4 ) , which describes how the
2
driving
stepper
motor.
The
matrix
term
(

four-bar linkage described

in Figure
1,
this
matrix is: (20)
sin(
=
)sensitivity
sin(+
sin(
4)
is:
2 34
3 )
2 4 ) cos(+
3 )in known as the sensitivity

sin2+ sin(
rotational error in the drive angle
propagates
to position
error
cos
sin(
sin(
2
3 )
4)

3
4
1, this sensitivity matrix
(20)
=

matrix
(). Forvariance
the typical
four-bar
linkage
in
Figure

sin(
The
statistical
of
position
error
can
be estimated
using
sin(+
described
sin(
44))cos(+
22
33)) the sensitivity matrix
cos
22

sin
+
is:

and
statistical
properties
of the rotational
sin(33error
44)) 2 as follows:
The the
statistical
variance
of
position
error can position
besin(
estimated
using the sensitivity matrix
=

(20)
sin(2 4 ) cos(+3 )
and the statistical properties ofthe
error

as
follows:
cosrotational
2 position
2
)
sin(+
sin(

)
2 344 ) 2using
3
The statistical variance
of position
error
be4 )sin(
estimated
sin(+

+can

2sin(
2
3)
2 the sensitivity matrix
2
= sin
2 sin
+rotational
(21)
sin(3 4)

and the statisticalproperties

position
as
sin(
follows:
(20)
= of
the
3)4 ) error22 2
sin(
4 2sin(+
2sin(
4 ) cos(+
)
3)
2
32
The statistical variance
of
position
error
can
be
estimated
using
the
sensitivity
matrix

2
(21)
= sincos
2 + 2 sin( )
3sin(
4 3 4 )2 2
Thethestatistical
variance
of
position
error
can
be
estimated
using
the
and
statistical properties
of
the
rotational
position
error

as
follows:
)
)
cos(+

sin(

)
sin(+

sin(

)
22
2 2 4 4
33

cos
22+

(22)

sin
(21)
22 =
22
sin(
sin(
sensitivity
S
and
the
statistical
the rotational
3)
4 )4 )properties
2
3
cos(+

sin(

)
The statistical matrix
variance
of
position
error
can
be
estimated
using
theofsensitivity
matrix

2
4
3
2
2
2

cos

(22)
sin(2 4 ) sin(+3 )
2
22
follows:
position
error2properties
2=as
sin(
4 )
3
sin

(21)
and the statistical
of the
rotational
position
error

as
follows:
2 sin( sin(
2
4 ) 3 ) 2 2
2 4 )3cos(+
2 = cos 2
22
(22)
sin( )
3

sin( ) sin(+ )2 2

2 )4cos(+ )3
2
4
3
=

sin 2+ sin(2
22 =

cos
222
sin(3 4 )
2

sin(
)

2 = cos 2

sin(2 4 ) cos(+3 ) 2
sin(3 4 )

22

(21)
(22)
(22)
2

where and are

of of
thethe
coupler
pointpoint
position
errors, errors,
and 2 is the
where
arethe
thevariances
variances
coupler
position
2
2
variance
thevariance
rotational
the drive angle.
stepper
motors,
22 is
nt
errors,
and
2 is of
the
of position
the coupler
point
position
errors,
and 2of
iserror
the
and
the
theinrotational
error For
in the
drive
angle.
Forreadily
2
2
estimated
from
the
known
properties
of
the
motor.
Stepper
motors
are
generally
rated as
stepper
motors,
readily estimated from the known properties
is readily
or
motors,For
2stepper
thestepper
drive angle.
motors, 2 isisreadily
being
accurate
to
a
given
percentage
of
the
step
size
(typically
3%
to
5%
in
a
common
of the
motor.
Stepper rated
motors
motors
are Stepper
generally
rated
as generally
sper
of the
motor.
motors
are
as are generally rated as being accurate to
stepper
motor), yielding bidirectional tolerance bounds on stepper motor position.
3% to
5%(typically
a common
eypically
of the step
size
3%
to 5% in aofcommon
a ingiven
percentage
the step size (typically 3% to 5% in a common
Conservatively,
the rotational position error can be considered to be uniformly
nds
stepperbounds
motor
position.
nal on
tolerance
on
stepper
motor
position.
stepper
motor),
yielding
bidirectional
tolerance
bounds
on stepper
distributed
over
the
error bounds,
and the variance
calculated
accordingly.
e considered
be considered
uniformly
sition
error canto motor
be
to
be
uniformly
position. Conservatively, the rotational position error can be
ulated
accordingly.
d the variance
calculated accordingly.
considered
to be uniformly
distributed
the error
bounds,
and
theUsing
In order to validate
the DLM model
for error over
propagation,
a case
study was
used.

geometry
as defined in Figure 1, the following dimensional values were used:
on,
casepropagation,
study wasthe
used.
Using
for aerror
a case
study was used. Using

=
40
mm,

= 120 mm, = 80 mm, = 100 mm, = 50 mm, and = 30o . The

valuesdimensional
were used: values
e dimensional
1, the following
were used:
driving
stepper
motor
was assumed to have a step size of 1.8/step, and an accuracy
o
ISSN:
5 of
= 50
==30
. The
mm,
=mm,
100and
mm,
50
mm,
and
2180-1053
= 30o . The Vol. 6 No. 1 January - June 2014
5%.
Under
the
conservative
assumption
of
a
uniformly
distributed
stepper
motor
error,
of have
1.8/step,
of and an accuracy of
to
a stepand
sizeanofaccuracy
1.8/step,
a variance of 2 = 8.127 107 rad2 was used.

where 2 and 2 are the variances of the coupler point position errors, and 22 is the
variance of the rotational error in the drive angle. For stepper motors, 22 is readily
estimated
knownof
properties
of the
motor.
Stepper
motors
where 2 and
2 arefrom
the the
variances
the coupler
point
position
errors,
andare
2generally
is the rated as
2
beingaccurate to a given percentage of the step size (typically 3%
to5%
in a common
2
variance
calculated
2 error 2in the drive angle.
variance ofstepper
the where
rotational
Forthe
stepper
motors,

is readily
2
2

accordingly.
and

are
the
variances
of
coupler
point
position
errors,
andpoint
22 ispositio
the
motor),
yielding
bidirectional
tolerance
bounds
on
stepper
motor
position.
2

where and are the variances of the coupler

estimated from
the
known
properties
of
the
motor.
Stepper
motors
are
generally
rated
as
2
Conservatively,
the
rotational
position
error
can
be
considered
to
be
uniformly
variance of the rotational
error inofthe
angle.error
For in
stepper
motors,
2 isForreadily
variance
thedrive
rotational
drive
angle.
stepper
being
accurate
to a given
percentage
of the
step
sizevariance
(typically
3% to 5%
athe
common
In
order
to
validate
the
DLM
model
for the
error
a in
case
study
distributed
over
the
error
and
calculated
accordingly.
estimated
from
the bounds,
known
properties
ofpropagation,
the
motor.
Stepper
motors
aremotor.
generally
ratedmotor
as
estimated
from
the
known
properties
of
the
Stepper
stepper
motor),
yielding bidirectional
tolerance
bounds
on stepper
motor
position.
was
used.
Using
geometry
as
defined
into
Figure
1,
the
following
beingthe
accurate
to a given
percentage
of
the
step
size
(typically
3%
tostep
5%size
in a (typically
common 3
being
accurate
a
given
percentage
of
the
Conservatively,
theto validate
rotationaltheposition
errorforcan
bepropagation,
considered ato
bestudy
uniformly
In order
DLM model
error
case
was used.
Using
dimensional
values
were
a=40
mm,
b=120
mm, bounds
c=80
mm,
d=100
stepper
motor),used:
yielding
bidirectional
tolerance
on stepper
motor
position.
stepper
motor),
yielding
bidirectional
tolerance
bounds
on st
distributedthe
overgeometry
the error bounds,
andin
theFigure
variance
calculated
accordingly.
as
defined
1,
the
following
dimensional
values
were
used:
Conservatively,
the
rotational
position
error
can
be
considered
to
be
uniformly
mm, e=50 =mm,and
=30.
The
driving
stepper
motor
was
assumed
Conservatively,
the
rotational
position
error
ocan be conside
40 mm, = over
120 mm,
= bounds,
80 mm, and
=the
100
mm, =
50 mm, accordingly.
and = 30 . The
the error
variance
calculated
distributed
over
the
error
bounds,
and
theUsing
variance
calculated
acco
to
havetoadriving
stepdistributed
size
ofmotor
1.8/step,
and
anhave
accuracy
ofstudy
5%.
theaccuracy
In order
validate
the DLM
model
error
propagation,
a case
wasUnder
used.
stepper
wasfor
assumed
to
a step
size
of
1.8/step,
and
an
of
the geometry
defined
Figure
the following
values
were
used:
conservative
assumption
of a1, uniformly
distributed
stepper
motor
5%.asUnder
thetoin
conservative
assumption
offor
adimensional
uniformly
distributed
stepper
error, Using
In order
validate the DLM
model
error propagation,
a case
studymotor
was used.
o error
to validate
DLM
for
2 = 80 mm, In
7
= 40amm,
= 120of
mm,
= order
1002 mm,
used.
= 50the
mm,
andmodel
= 30
. Thepropagation, a case
=
8.127

10
rad
was
used.
a variance
of

was
error,
variance

the geometry
as definedthe
in geometry
Figure 1, asthedefined
following Figure
dimensional
weredimensio
used:
1, the values
following
driving stepper motor was2assumed to have
a step size of
1.8/step, in
and an accuracy
of
o
.
= 40 mm, = 120 mm,

=
80
mm,

=
100
mm,

=
50
mm,
and

=
30
=
40 mm, distributed
= 120 mm,stepper
= 80motor
mm, error,
= 100 mm, = The
50 mm
5%. Under
the conservative
assumption
of
a to
uniformly
Equations
(21)
andwere
(22)
were
used
predict
position
error
variance
ashave
a function
theof 1.8/st
stepper
motor
assumed
to have
a step
size
ofvariance
1.8/step,
and
an of
accuracy
of
Equations
(21)
and
(22)
used
to
predict
position
error
as
2 driving
7
2 was
driving
stepper
motor
was
assumed
to
a step
size
a variance input
of 2angle
= 8.127

10
rad
was
used.
Under
a full
revolution
of the the
input
(in 1.8
increments).
Inuniformly
addition,
5%.input
the
conservative
assumption
of link
a uniformly
distributed
stepper
motordistribut
error,
2 over
5%.
Under
conservative
assumption
of
a
a function of the
angle

over
a
full
revolution
of
the
input
link
2
2 at
and2 the
position
MonteaCarlo
simulations
valueof
7
2 each
2 ,rad
variance
of 22 = using
8.127

107trials
rad
used.
a10,000
variance
oferror
was
=
8.127
was
used. error
(in
1.8 increments).
In
addition,
Monte
Carlo
10,000
2values.
Equations
(21)
and
(22)
were
used
to
predict
position
variance
as10
ausing
function
ofthe
the
variance was computed from these simulatedsimulations
Figure
2 shows
path of the
a full
of the
input
(in link
1.8
increments).
In addition,
input angle
2 over
trials
at each
value
of revolution
(21)
, and
the
position
error
variance
coupler
for
one
full
revolution
of
thelink
crank
in
bothwas
the computed
open
and
2
Equations
and
(22) were
used
to
predict
position
error
variance
as
acrossed
function
of varian
the
Equations
andof(22)
werethe
used
to predict
position
error
Montethese
Carlo
simulations using
10,000
trials
atshows
each(21)
value
2 , and
position
error
configurations.
from
simulated
values.
Figure
2
the
path
of
the
coupler
for
revolution
of
the
input
link
(in
1.8
increments).
In
addition,
input angle 2 over a fullinput
over a 2full
revolution
of of
thethe
input link (in 1.8 in
angle 2 Figure
variance
computed of
from
these
simulated
shows
thecrossed
path
one
full was
revolution
thesimulations
crank
link
invalues.
both
the
and
Monte Carlo
using
10,000
trialsopen
at each
value
of 2 trials
, and the position error
Monte
Carlo
using
10,000
100
coupler for one full revolution
of the crank
link
in simulations
both the open
and
crossedat each value of 2 ,
variance was computed from
these
simulated
values.
Figuresimulated
2 shows values.
the pathFigure
of the 2
configurations.
variance
was
computed
from these
configurations.
80
coupler for one
full revolution
link in both
and incrossed
coupler of
for the
onecrank
full revolution
of the
the open
crank link
both the
configurations.
100
configurations.
60

Journal of Mechanical Engineering and Technology

80

60

40

20

-20

-40

-60

-80

40

20

-20

-40

-60

-80

-100
-50

100

100

80

80

60

60

40

40

20

20

-20

-20

-40
50

-40

100

150

Figure 2. Coupler curves for open and crossed configurations (units of mm)

-100
-50

-60

-60

-80

100
50
Figures 3 and 4 show 0the simulated
vs. predicted
values150of 2 and 2 using the open
Figure
2.
Coupler
curves
for
open
and
crossed
configurations
(units of100
-100
Figure
2. Coupler
curves
for
open
and
crossed
configurations
150 simulated vs.
50-100
0
-50
configuration
of the
four-bar
mechanism,
and Figures
5 andmm)
60 show the
100
50
-50
2 (units of
2 mm)
predicted values of

and

using
the
crossed
configuration
of
the
four-bar
Figure
and2.
configurations
(units
of
mm)
2. Coupler
curves for open
2crossed
2curves for open
Figure
Coupler
and
crossed
configurations
(u
Figures 3 and 4 show the simulated vs. predicted values of and using the open
mechanism, using the known kinematic solutions (Norton, 1999). In each case, the
2
configuration
of model
the four-bar
andposition
Figures
5predicted
andvariance.
6 show
theofsimulated
Figures
3 and
4 simulated
show
theFigures
simulated
vs.error
vs.
the open
accurately
predicts
the
Figures
3DLM
and
4 show
themechanism,
vs.
values
of 2vs.and
using
3predicted
and
4 show
thevalues
simulated
predicted
values
of 2 an
predicted values of 2 and 2 using the crossed configuration of the four-bar
configuration
of the four-bar
mechanism,
andfour-bar
Figures mechanism,
5and
and Figures
6 show
simulated
vs.6 s
using the open
configuration
of configuration
the four-bar
ofmechanism,
the
andtheFigures
5 and
mechanism, using
the known
kinematic
solutions
(Norton, 1999).
In each case, the of the four-bar
predicted
values
of predicted
2 predicted
and
2 values
using
valuesthe
of crossed
2 and 2configuration
using
the
crossed
configur
5DLM
andmodel
6 show
the
simulated
vs.
of
using
the

accurately predicts the position error variance.

mechanism, using
thefour-bar
known
kinematic
(Norton,
1999).solutions
In each (Norton,
case, the19
mechanism,
usingsolutions
the known
kinematic
crossed configuration
of the
mechanism,
using
the known
DLM model accurately predicts
the position
errorpredicts
variance.
DLM
model
accurately
the
position
error
variance.
kinematic solutions (Norton, 1999). In each case, the DLM model
-80

accurately predicts the position error variance.

ISSN: 2180-1053

Vol. 6

No. 1

January - June 2014

Optimal Design of A Stepper-Driven Planar Linkage Using Entropy Methods

-4

x 10 -4
x 10
12
12

simulated
simulated
predicted
predicted

10
10

2
2
)
(mm
2 P
2
x
)
P (mm

6
4

2
0

50
50

100
100

150
200
250
300
350
200
250
300
350
150
(degrees)
2
2 (degrees)
Figure 3. Variance of coupler x position (open configuration)
Figure 3. Figure
Variance
of coupler
x xposition
(open
configuration)
3. Variance
of coupler
position (open
configuration)
0

-4

x 10 -4
x 10

simulated
simulated
predicted
predicted

2
2
)
(mm
2 P
2
y
)
P (mm

5.5
5.5
5
5
4.5
4.5
4
4
3.5
3.5
3
3
2.5
2.5
2
2
1.5
1.5
1
1
0.5
0.5
0

150
200
250
300
200
250
300
150
(degrees)
2
2 (degrees)
Figure 4. Variance of coupler y position (open configuration)
Figure 4. Variance of coupler y position (open configuration)
50
50

100
100

350
350

Figure 4. Variance of coupler y position (open configuration)

ISSN: 2180-1053

Vol. 6

No. 1

January - June 2014

Journal of Mechanical Engineering and Technology

x 10

18

18

16

-4

simulated
simulated
predicted
predicted

14

12

12

10

2
2
P (mm )

x 10

16

14

2
2
P (mm )

-4

10
8
6

2
0

50

50

100

100

150

150

200

200

250

250

300

300

350

2 (degrees)
2 (degrees)
Figure
5.
Variance
of
coupler
x
position
(crossed
configuration)
Figure 5. Variance
coupler
x position
(crossed
configuration)
Figure 5.of
Variance
of coupler
x position
(crossed configuration)
x 10
9

x 10

-4

simulated
simulated
predicted
predicted

2
2
P (mm )

2
2
P (mm )

-4

4
3

1
0

50

50

100

100

150

150

200

200

250

250

300

300

350

350

2 (degrees)
2 (degrees)
Figure 6. Variance of coupler y position (crossed configuration)
Figure 6. Variance of coupler y position (crossed configuration)
Figure 6. Variance
of coupler y position (crossed configuration)

ISSN: 2180-1053

Vol. 6

No. 1

January - June 2014

350

Optimal Design of A Stepper-Driven Planar Linkage Using Entropy Methods

3.0

AN ENTROPY BASED APPROACH TO RELIABILITY

OPTIMIZATION

In
previous
section,
it was
demonstrated
that the DLM
method can
3.0theAN
ENTROPY
BASED
APPROACH
TO RELIABILITY
OPTIMIZATION
be used to establish an analytic expression for position error variance
In a
thestepper-driven
previous section, four-bar
it was demonstrated
that Consider
the DLM method
canwhere
be usedthe
to
in
mechanism.
the case
establish anmechanism
analytic expression
for position
varianceplace
in a stepper-driven
four-bar
is being
used toerror
precisely
the couplerfour-bar
point
mechanism. Consider the case where the four-bar mechanism is being used to precisely
at
a desired x-y location, with allowable position error described by
place the coupler point at a desired x-y location, with allowable position error described
bidirectional
tolerances
yTO
components
by bidirectional
tolerances
on on
theAPPROACH
xthe
andxyand
components
of position:of position:
3.0
AN ENTROPY
BASED
RELIABILITY
OPTIMIZATION
= , that
(23)
In the previous section, it
was
demonstrated
to
,
the DLM method can be used
establish
analytic expression
for position error
in a stepper-driven
four-bar
3.0
ANanENTROPY
BASED APPROACH
TO variance
RELIABILITY
OPTIMIZATION
mechanism. Consider the case
where=the
four-bar
,
,
mechanism is being used to precisely
(24)
place
couplersection,
point at ita desired
x-y location,that
withthe
allowable
positioncan
errorbedescribed
In
thethe
previous
was demonstrated
DLM method
used to
by
bidirectional
tolerances
on
thefor
x and
components
of position:
and
,
indicate
the yallowable
values
ofinposition,
, and
,
where
,
establish
an analytic
expression
position
error variance
a stepper-driven
four-bar
mechanism.
Consider
the
case
where
the
four-bar
mechanism
is
being
used
to
precisely
indicate
the
nominally
desired
position
for
the
coupler
point,
and

and

indicate
where Px,allow and Py,allow indicate the allowable values
of
position,

,onx-y
=the
location,
with
(23)
place
the
coupler
point
at a desired
allowable
position
error
described
,

bidirectional
tolerance
constraints
nominal
desired
position.
P
,
and
P
,
indicate
the
nominally
desired
position
for
the
coupler
des
y des
byx bidirectional
tolerances on the x and y components of position:
point,
and xtol and
y mechanism
indicate
bidirectional
constraints on(24)
the
,
tolerance
, =has
Since the position
of thetol
beenmodeled
probabilistically, the positioning
nominal
desired
position.
,
perspective; the optimal four-bar
(23)
problem can be interpreted
from=a,
reliability
andmaximize
, indicate
the allowable
of position,
,within
and ,
where
,
mechanism
will
the probability
that values
the coupler
point falls
the
tolerance
bound
placeofdesired
on
positioning
reliability.
This
=thereby
,
maximizing
coupler
(24)
indicate
the
nominally
position
for the
point,
and probabilistically,
indicate
Since
the
position
theposition,
mechanism
has
been modeled
,

and
problem
has tolerance
been problem
addressed
for
kinematic
chains
and roboticperspective;
manipulators
bidirectional
constraints
on
nominal
desired
position.
the
positioning
canopen-loop
bethe
interpreted
from
a reliability
by
employing
an
entropy
formulation
(Musto,
2002);
under
the
assumption
that
the
where
, and
,mechanism
indicate the allowable
values ofthe
position,
, and
,
the
optimal
four-bar
will maximize
probability
that
the
system
is
designed
to
nominally
hit
the
desired
x-y
position,
then
the
reliability
of
the
Since
the
position
of
the
mechanism
has
been
modeled
probabilistically,
the
positioning
indicate the
nominally
desired the
position
for the bound
coupler point,
position,
indicate
and thereby
coupler
point
falls
within
tolerance
placeand
onthe
positioning
system
can
be
expressed
as:
problem
can
be
interpreted
from
a
reliability
perspective;
optimal
four-bar
bidirectional tolerance constraints on the nominal desired position.
maximizing
problem
been
mechanism willpositioning
maximize thereliability.
probability This
that the
couplerhas
point
fallsaddressed
within the
2
2
2
2
for
open-loop
and
manipulators
by
employing
tolerance
bound kinematic
place
thereby
maximizing
reliability.
This
position,
=chains

robotic
modeled
positioning

(25)
Since
the position
of theon
mechanism
has
the
positioning
been
probabilistically,

problem
has
open-loop
kinematic
chains
robotic
manipulators
problem
canbeen
be addressed
interpretedfor
from
a reliability
perspective;
the
optimal
four-bar
an
entropy
formulation
(Musto,
2002);
under
the and
assumption
that
the
by
employing
an maximize
entropy
(Musto,
underx-y
the
assumption
thaterror
the
where
isisthe
reliability
offormulation
the
system.
In systems
where
the
mechanism
will
thepositioning
probability
that
the
coupler
point
fallsposition
within
the
system
designed
to
nominally
hit
the2002);
desired
position,
then
the
system
to
nominally
hit the
desired
x-y position,
then the
reliability
ofThis
the
varianceisisdesigned
known,
it has
been
shown
that
minimizing
an positioning
entropy
measure
associated
tolerance
bound
place
on
position,
thereby
maximizing
reliability.
reliability
of thecan
positioning
system
can be expressed as:
positioning
system
beisexpressed
as:to optimizing
with the has
position
equivalent
boundand
on robotic
system manipulators
reliability; a
problem
been error
addressed
for open-loop
kinematicachains
lower-bound
optimized
whenunder
the following
entropy-based
by
employingonan position
entropy reliability
formulationis (Musto,
2002);
the assumption
that the
2
2
2
2


the reliability of(25)
=
objective
system
is function
designedistominimized:
nominally
hit thedesired
x-y
position,
then
the

positioning system can be expressed as:
2
system.
2

where is the reliability of the positioning

In systems where the position error
= 2 +2 2 2
(26)
2
2
where
the reliability
of
thethat
positioning
systems
where
variance Ris is
known,
it has been
shown
minimizing
an
associated
=

entropy

Inmeasure
(25)

system.

withposition
the position
error
is equivalent
to optimizing
a bound
on system
reliability; a
the
error
variance
is known,
it has been
shown
that minimizing
This
approach
has
been
demonstrated
for
high
degree-of-freedom
systems,
whereerror
the
lower-bound
on
position
reliability
is
optimized
when
the
following
entropy-based
where

is
the
reliability
of
the
positioning
system.
In
systems
where
the
position
an entropy measure associated with the position error is equivalent
to
error
variance
must
be
determined
through
Monte
Carlo
simulation
(Musto,
2002).
objective
function
is
minimized:
variance is known,
it has on
beensystem
shown that
minimizing
an entropy measure
associated
optimizing
a
bound
reliability;
a
lower-bound
on
position
However,
when coupled
the DLM
method for astepper-driven
four-bar
linkages,a
with
the position
error iswith
equivalent
to optimizing
bound on system
reliability;
2following entropy-based objective
reliability
optimized
2the

this objectiveis
function
canreliability
be when
evaluated
analytically,
and used
as an objective
function
is
optimized

lower-bound
on
position
when
the
following
entropy-based
= 2 + 2
(26)
function
is minimized:
for designfunction
optimization.

objective
is minimized:
2 (21)degree-of-freedom
2 for high
Substituting
thehas
expressions
from Equations
and (22), and eliminating
constant
This
approach
been demonstrated
systems,thewhere
the

2
=
(26)
2 + for
2 Monte
,yields
the
following
cost
function
minimization:
term variance
error
must
be
determined
through
Carlo
simulation
(Musto,
2002).

However, when coupled with the DLM method for stepper-driven four-bar linkages,
2
2
can
sin(
) analytically,

sin(
cos(+
this objective
function
be
and
used
ansystems,
objective
function
2 evaluated
4 ) sin(+3for
2 4 )as
3)
This
approach
demonstrated
degree-of-freedom
where
the
has
sin been
high
cos
2

2+
sin(3 4 )
sin(3 4 )
This
approach
has
been
demonstrated
for
high
degree-of-freedom
for
design
optimization.
= must be determined
+
(27)
error variance
through
Monte
Carlo
simulation
(Musto,
2002).
2
2

systems,
where
the error
variance
must be
through
Monte
However, when
coupled
with the
DLM method
for determined
stepper-driven
four-bar linkages,
Substituting
expressions
Equations
(21) and and
(22),used
and as
eliminating
the function
constant
this
objectivethe
function
can befrom
evaluated
analytically,
an objective
termdesign
22 ,yields
the following cost function for minimization:
for
optimization.

ISSN: 2180-1053

Vol. 6

No. 1

January - June 2014

2
sin(2from
4 ) sin(+
sin(2 4 )eliminating
cos(+3 ) 2 the constant
Substituting
the sin
expressions
Equations
(21)
3)
2 +

cosand
2 (22), and

sin(3 4 )
sin(3 4 )
2
=
+
(27)

objective
function is aminimized:
with the position error is equivalent
to optimizing
bound on system reliability; a
lower-bound on position reliability is optimized when the following entropy-based
2
2

objective
function is Engineering
minimized: and Technology

Journal
of Mechanical
=

(26)

This approach
has when
been demonstrated
for the
highDLM
degree-of-freedom sys

Carlo simulation (Musto, 2002).

However,
coupled with
error
variance
must
be
determined
through
Monte
Carlo simulation
method
for stepper-driven
four-bar
linkages, this objective
function
This
approach
has been demonstrated
for high
where can
the
However,
whendegree-of-freedom
coupled with thesystems,
DLM method
for stepper-driven fo
be evaluated
analytically,
andobjective
used as
an objective
for design
error
variance must
be determined
through
Monte
Carlo
(Musto,
2002). and used as an ob
this
function
can simulation
be function
evaluated
analytically,
However,
when coupled with the
DLM
method
for
stepper-driven
four-bar
linkages,
optimization.
for design optimization.
2

this objective function can be evaluated analytically, and used as an objective function
for design optimization.

Substituting the expressions

fromtheEquations
(21) Equations
and (22),
Substituting
expressions from
(21)and
and (22), and elimina
2
term

,yields
the
following
cost
function
for
minimization:
,yields
the
following
cost
function
for
eliminating
the
constant
term

2
Substituting the expressions from Equations
(21) and (22), and eliminating the constant
2
minimization:
term 2 ,yields the following cost function for minimization:
=

sin 2 +

=
+

sin(2 4 ) sin(+3 ) 2

sin(3 4 )
2

sin(2 4 ) sin(+3 ) 2

sin(3 4 )
2
sin(2
2 4 ) cos(+3 )

cos 2

sin(3 4 )
2

sin 2 +

cos 2

(27)

This will be used in the formulation of the mechanism design problem

as a multivariable constrained optimization problem.
4.0

AN OPTIMAL DESIGN FORMULATION

In the previous sections, an analytic expression for the position

variance of a planar four-bar mechanism was determined, and a cost
function analogous to system reliability was introduced. In this section,
the design of step-driven four-bar mechanisms will be formulated as a
constrained optimization problem; the goal will be to design the fourbar mechanism to optimize the reliability of the positioning system
in the presence of uncertainty in the actuator (2); in essence, the goal
of the design will be to minimize the propagation of actuator error to
coupler point positioning error.
In this formulation, it is assumed that all of the system uncertainty is due
to the actuator error 2, and that the stepper resolution of the actuator is
known. Geometric errors in the link length, pivot point placement, etc.
are not considered; this is due to the fact that for a given mechanism,
these errors are fixed, and the effect of these errors can be removed
by calibration; only the actuator angle remains as a random parameter
once the system is built and calibrated. Design variables include the
length of the four links (a, b, c, d), the coupler length (e), the actuator
angle (2) and the coupler angle (). The angle of the ground link d
with respect to the x-axis could also be included as a design variable;
this would in essence have the effect of modifying the relative values of
xtol and ytol. In this development, the ground link will be assumed to be
coincident with the x-axis.
With these assumptions, the multivariable constrained optimization
problem can be formulated as:
10

ISSN: 2180-1053

Vol. 6

No. 1

January - June 2014

sin(2 4 ) cos(+3
sin(3 4 )
2

the coupler
ground link
d with
to theangle
x-axis(could
also be included as a design variable;
length
(e), respect
the actuator
2 ) and the coupler angle (). The angle of
this
would
in
essence
have
the
effect
of
modifying
the relative
values
andvariable;
. In
the ground link d with respect to the x-axis could also
be included
asof
a design
development,
the have
ground
will
assumed the
to
be
coincident
with
x-axis.
Optimal
Design
of A
Stepper-Driven
Planar
Linkage
Using
Entropy
Methods
and

this would
in essence
thelink
effect
ofbe
modifying
relative
values
of the

. In
this development, the ground link will be assumed to be coincident with the x-axis.
With these assumptions, the multivariable constrained optimization problem can be
formulated
as:
With
these
assumptions,
the multivariable constrained optimization problem can be
Find:
a,b,c,d,e,
,
2
formulated as:
Find: , , , , , 2 ,
To
minimize:
Find: , , , , , 2 ,
To minimize:
sin(2 4 ) sin(+3 ) 2
sin(2 4 ) cos(+3 ) 2
To minimize:
sin 2 +

cos 2

sin(3 4 )
sin(3 4 )
2
(, , , , , 2 , ) =
+
(28)
sin(
sin(22 4 ) cos(+3 ) 2
2 2 4 ) sin(+3 )

sin 2 +

cos 2

sin(3 4 )
sin(3 4 )
(, , , , , 2 , ) =
+
(28)
2
2

Subject to:
Subject
Subject to:to:
, cos 2 + cos( + 3 ) = 0
(29)
(29)
, cos 2 + cos( + 3 ) = 0
, sin 2 + sin( + 3 ) = 0
(30)
, sin 2 + sin( + 3 ) = 0
(30)
cos 2 + cos 3 cos 4 = 0
(31)
cos 2 + cos 3 cos 4 = 0
(31)
(32)
sin 2 + sin 3 sin 4 = 0
sin 2 + sin 3 sin 4 = 0
(32)
Equations (29) and (30) ensure that the mechanism will reach the nominal desired
position;
(32)that
enforce
the kinematic
equations.
It should
be
EquationsEquations
(29) and (31)
(30) and
ensure
the mechanism
willloop
reach
the nominal
desired
noted thatEquations
the(29)
link lengths
a, b,
c, and
d must
nonnegative
which
introduces
Equations
and
(30)
ensure
that
the
mechanism
will
position;
(31) and
(32)
enforce
the be
kinematic
loopvalues,
equations.
Itreach
shouldthe
be
additional
noted
that constraints.
the
link lengths
a, b, Equations
c, and d must(31)
be nonnegative
values, which
introduces
nominal
desired
position;
and (32) enforce
the kinematic
additional
constraints.
loop equations. It should be noted that the link lengths a, b, c, and d

must be nonnegative values, which introduces additional constraints.

In the practical implementation of this method, additional constraints

are likely. For example:

Nonzero lower bounds may be placed on link lengths, to
allow for practical geometric constraints such as bearing/pin
sizes, etc.
Upper bounds may be placed on link lengths, limiting the
overall size envelope for the mechanism
Constraints may be placed on the relative lengths of the
links in the mechanism, allowing for practical construction
and scale of the mechanism
A constraint may be added on the orientation a specific link,
if orientation is important to the positioning task (e.g. the
angle of the coupler link may be specified)
If desired, a constraint to enforce the Grashof condition and
allow for full rotation of the drive link can be formulated
(shortest link + longest link sum of remaining links)
Once formulated, a suitable computational algorithm can be used to
solve the multivariable constrained optimization problem, and yield
a mechanism design that is optimal from a reliability standpoint. The
addition of practical constraints will vary according to the usage of
ISSN: 2180-1053

Vol. 6

No. 1

January - June 2014

11

Journal of Mechanical Engineering and Technology

the linkage, but should be readily incorporated with an appropriate

computational algorithm. This will be demonstrated in the next section
in a mechanism design case study.
5.0

CASE STUDY

The DLM model and entropy-based optimization algorithm will be

demonstrated in a case study in this section. A four-bar linkage (as
shown in Figure 1), with a pivot point placed at the origin, will be the
assumed geometry for the case study. The mechanism will be designed
to achieve the following positioning task:
Px,des=120 mm, with an allowable tolerance of 1 mm
Py,des=180 mm, with an allowable tolerance of .1 mm
In addition to the constraints inherent to the optimization problem
described in the previous section, the following additional practical
constraints have been placed on the mechanism design:
The crank link length a will be at least 60% of the length of
the other links in the kinematic chain
The coupler link length e will be no longer than any other
link in the kinematic chain
The orientation of the coupler link (the quantity +3) will
be specified to be between 70 and 80
The relative lengths of the kinematic links a, b, c, and d will
be such that the mechanism meets Grashofs condition, with
a being the shortest link in the kinematic chain
With these constraints in place, a graphical solution which achieved
the these
nominal
desired
position
within
the
kinematic
constraints
was
With
constraints
in place,
a graphical
solution
which
achieved the
nominal desired
formulated.
The
graphical
solution,
which The
wasgraphical
used solution,
as an initial
position
within the
kinematic
constraints
was formulated.
which
was
used as an
initial
condition for is
optimization,
is summarized
condition
for
optimization,
summarized
in Tablein1.Table 1.
Table 1. Initial condition (graphical solution)

Table 1. Initial condition (graphical solution)

Parameter
a
b
c
d
e

2
3
4

Initial condition
91.0 mm
150.0 mm
147.7 mm
150 mm
150 mm
30
22
46
75

This graphical solution, which was obtained by traditional trial-and-error graphical

methods, was evaluated using the objective function in Equation (28); the value of the
12
ISSN:
Vol.was
6 found
No. 1 toJanuary
- June 52014
objective function
for 2180-1053
this initial design
be V=2.34x10
; additionally, use
of Equations (21) and (22) yielded standard deviations on positioning error to be

a
91.0 mm
b
150.0 mm
c
147.7 mm
Optimal Design of Ad Stepper-Driven
Planar Linkage Using Entropy Methods
150 mm
e
150 mm
30

22obtained by traditional trial-and2

This graphical solution, which
was
46
3 evaluated
error graphical methods, was
using the objective function in
75
4

Equation (28); the value of the objective function for this initial design
was found to be V=2.34x105; additionally, use of Equations (21) and (22)
This graphical solution, which was obtained by traditional trial-and-error graphical
yielded
deviations
on positioning
to be
=.102 mm
x value of the
methods, standard
was evaluated
using the objective
function inerror
Equation
(28);P
the
5
and
Py =.043
mm.
objective
function
for this initial design was found to be V=2.34x10 ; additionally, use

of Equations (21) and (22) yielded standard deviations on positioning error to be

.102 as
mman
and
=solution,
.043 mm. the optimization formulation offered in
= this
With
initial
the previous section, in addition to the additional practical constraints,
Withused
this astoanimprove
initial solution,
the optimization
formulation offered
in the previous
was
the design.
For the computational
algorithm,
the
section, in addition to the additional practical constraints, was used to improve the
Solver
tool
in
Microsoft
Excel
2010
was
used.
Specifically,
the
GRG
design. For the computational algorithm, the Solver tool in Microsoft Excel 2010 was
Nonlinear
optimization
was utilized.
After performing
the
used. Specifically,
the GRGalgorithm
Nonlinear optimization
algorithm
was utilized. After
optimization,
the mechanism
design detailed
in Table
2 resulted.
performing the optimization,
the mechanism
design detailed
in Table
2 resulted.For
For
comparison, thethe
initial
conditions
and values
of values
objectiveoffunction
and positioning
error
comparison,
initial
conditions
and
objective
function and
estimates are shown
well.
positioning
error as
estimates
are shown as well.
Table 2. Mechanism design

Table 2. Mechanism design

Parameter
a
b
c
d
e

2
3
4

Initial condition
91.0 mm
150.0 mm
147.7 mm
150 mm
150 mm
30
22
46
75
2.34x105
.102 mm
.043 mm

Final design
69.0 mm
115.0 mm
163.9 mm
163.9 mm
163.9 mm
-13
22
83
121
2.59x104
.109 mm
.010 mm

In order to validate the method, the design problem was reposed, with a change in the

In
order constraint
to validate
the
method, the design problem was reposed, with
tolerance
on
, for 1 mm in the previous case study to 0.01 mm in the
arevised
change
the tolerance
constraint
on Pconditions,
, for 1the
mm
in the previous
caseinstudy.
Starting from
the same initial
mechanism
detailed in
x des
Tablestudy
3 resulted.
case
to 0.01 mm in the revised case study. Starting from the same
initial conditions, the mechanism detailed in Table 3 resulted.

ISSN: 2180-1053

Vol. 6

No. 1

January - June 2014

13

Journal of Mechanical Engineering and Technology

Table 3. Mechanism design with

=1 mm,

=0.01 mm

,
Table 3. Mechanism design with,
Px,des=1 mm,
Py,des=0.01 mm

Parameter
Initial condition
Final design
a
91.0 mm
72.2 mm
b
150.0 mm
159.1 mm
c
147.7
mmwith , =1120.5
Table 3. Mechanism
design
mm, mm
, =0.01 mm
d
150 mm
120.6 mm
e
150
mmcondition
159.1
Parameter
Initial
Final mm
design
30
57
91.0 mm
72.2 mm
a
150.0 mm
159.1 mm
22
25
b2
147.7 mm
120.5 mm
46
13
c3
150 mm
120.6 mm
75
34
d4
5
150 mm7
159.1 mm
Ve
2.21x10
2.96x10
30
57

.102
mm
0.0002
mm

22 mm
25 mm
.043
.049
2
46
13
3
75
34
4
In each case, the optimization
algorithm
reduction
in the objective function;
5
V
2.21x107achieved a 2.96x10
each
case,
the optimization
algorithm
achieved
a reduction
in the
inIneach
case,
the mechanism
configuration
achieved
objective
function
minimization
by
.102 mm
0.0002 mm

reducing
error
in the direction
thatcase,
wasmm
held
the tighter
tolerance
constraint. achieved
objective
function;
in
thetomechanism
configuration
.043
.049 mm
each

objective function minimization by reducing error in the direction that

and
A
third
case
study
was
performed,
with the
tolerance
set to 0.1inmm
both ,
In
eachheld
case,to
thethe
optimization
algorithm
achieved
a reduction
the for
objective
function;
was
tighter
tolerance
constraint.

.
In
this
case,
with
the
error
in
x
and
y
directions
weighted
equally
in
the
objective
in,
each case, the mechanism configuration achieved objective function minimization by
function,
the results
in Table
3 were
reducing
theshown
direction
that was
heldobtained.
to
the tighter
tolerance constraint.
A thirderror
caseinstudy
was performed,
with
the tolerance
set to 0.1 mm for

both
, and was
Py,des
In this with
case,
with
the
error
inmm
x and
y directions
4..Mechanism
Design
,
=
0.1
mm
,
x des study Table
A
thirdPcase
performed,
thewith
tolerance
set
to=0.1
for both
, and
equally
in
the
objective
function,
the
results
shown
in
Table
weighted
.
In
this
case,
with
the
error
in
x
and
y
directions
weighted
equally
in
the
objective
,
Parameter
Initial condition
Final design
3 were obtained.
function,
the results shown
in
Table
3
were
obtained.
A
91.0 mm
84.6 mm

In

B
150.0 mm
141.0mm
C
147.7
mm
143.9
mm
Table
4.
Mechanism
Design
with

0.1=0.1
mm mm
,
,
Table 4. Mechanism Design with Px,des
=P=
,des
D
150 mm
144.0 ymm
E
150condition
mm
144.0design
mm
Parameter
Initial
Final
30mm
27mm
91.0
84.6
A
150.0
141.0mm
22mm
31
B2
147.7
143.9
46mm
43mm
C3
150
144.0
75mm
78mm
D4
5
150 mm6
144.0x10
mm
VE
1.49x10
4.86
30
27

.102
mm
.054
mm

22
31
.043
mm
.032
mm
2
46
43
3
75
78
4
each of the three case studies,
the following
V
1.49x106 can be shown:
4.86 x105
.102 mm
.054 mm

The optimized results,

when substituted
into Equations
.043 mm
.032 mm (4) and (5),

yield the
nominal desired position of , =120 mm, , =180 mm. This is the
of the
kinematic
constraints
in be
theshown:
optimization algorithm.
In eachconsequence
of the three case
studies,
the following
can
In of
each
thecase
optimized
mechanism
results in
In each
thecase,
three
studies,
the following
cananbeimproved
shown:design (as
by theresults,
improvement
in the objective
function)
the yield
nominal
indicated
The optimized
when substituted
into Equations
(4) over
and (5),
the
graphical
solution.
This indicates
more
accurate
positioning
of mm.
the mechanism
in
nominal
desired
position
of

=120
mm,

=180
This
is
the
,
,
presence
The optimized
results, when substituted into Equations
the
of
drive
angle
error.
consequence
of the
constraints
in the
optimization
algorithm.
(4) and
(5),kinematic
yield the
nominal
desired
position
of Px,des=120
In each case, the optimized mechanism results in an improved design
(as
mm,
P
,
=180
mm.
This
is
the
consequence
of
the
kinematic
des improvement in the objective function) over the nominal
indicated by ythe
constraints
in the
optimization
algorithm.
graphical
solution. This
indicates
more accurate
positioning of the mechanism in
presence
In each
case,
the
of drive
anglethe
error.optimized mechanism results in an

improved design (as indicated by the improvement in the

14

ISSN: 2180-1053

Vol. 6

No. 1

January - June 2014

Optimal Design of A Stepper-Driven Planar Linkage Using Entropy Methods

objective function) over the nominal graphical solution. This

indicates more accurate positioning of the mechanism in the
presence of drive angle error.
These case studies indicate that more the DLM method can be used to
optimize the positioning reliability of stepper-driven mechanisms.
6.0 CONCLUSIONS
In this paper, the DLM method has been extended to use in the
optimization of stepper-driven mechanisms. The method was shown
to accurately predict drive error propagation, through comparison to
Monte Carlo simulation. When coupled with entropy-based system
reliability optimization methods, it was shown that optimal mechansims
could be synthesized in response to various positioning constraints.
In this new method, only drive angle error was considered. This
method could be augmented to include additional uncertainties in
design, such as tolerances on link lengths, clearance in bearings, and
planar misalignments. Work is underway in this area.
REFERENCES
Avils, R., Vallejo, J., de Bustos, I. F., Aguirrebeitia, J., & Ajuria, G. (2010).
Optimum synthesis of planar linkages using a strainenergy error
function under geometric constraints. Mechanism and Machine Theory,
45(1), 6579.
Huang , X., & Zhang, Y. (2012). System reliability analysis for planar
mechanisms using matrix method. Advanced Materials Research, 452453, 11901194.
Huang, X., & Zhang, Y. (2010). Reliability sensitivity analysis for rack-andpinion steering linkages. Journal of Mechanical Design, 132(7), 071012.
Leishman, R. C., & Chase, K. W. (2010). Direct linearization method kinematic
variation analysis. Journal of Mechanical Design, 132(7), 071003.
Musto, J. C. (2002). A two-level strategy for optimizing the reliability of
redundant inverse kinematic solutions. Journal of Intelligent and Robotic
Systems, 33(1), 7384.
Norton, R. L. (1999). Design of Machinery (2nd ed.). New York, USA: McGrawHill.

ISSN: 2180-1053

Vol. 6

No. 1

January - June 2014

15

Journal of Mechanical Engineering and Technology

Pandey, M. D. & Zhang, X. (2012). System reliability analysis of the robotic

manipulator with random joint clearances. Mechanism and Machine
Theory, 58, 137152.
Wittwer, J. W., Chase, K. W., & Howell, L. L. (2014). The direct linearization
method applied to position err in kinematic linkages. Mechanism and
Machine Thoery, 39, 681693.
Zhang, X. & Pandey, M. D. (2013). An efficient method for system reliability
analysis of planar mechanisms. Proceedings of the Institution of
Mechanical Engineers, Part C: Journal of Mechanical Engineering Science,
227(2), 373386.

16

ISSN: 2180-1053

Vol. 6

No. 1

January - June 2014