Section 8.|
Step-by-Step: Flexible Gripper
Section 8.1 Step-by-Step: Flexible Gripper 283,
h,
\
8.1-1 About the Flexible Gripper
In Section 7.1, we introduced a flexible gripper [1], created
a line model for it, and performed a static structural
simulation to assess the GA (geometric advantage, defined
as the ratio of horizontal output displacement to input
displacement) of the design. For that particular design, the
GA value is 1.04 (7.1-16). In this section, we want to
improve the GA value by adjusting the shape of the flexible
gripper.
The shape of the gripper consists of ewo spline
curves defined by 7 key points [2]. Positions of the PI. P4,
P5,and P7 are fixed (cannot be changed) due to constraints
imposed by some functional requirements. The positions
of the P2, P3,and P6 are free to be changed.
‘The idea is to fix the X-coordinates of these three
points and adjusting their Y-coordinates to achieve a better
GA value, The allowable adjustment range on both sides of
the Y-coordinates is 10 mm for P2.20 mm for P3,and 5 mm
for P6, Besides the GA value. we also have to watch out
the maximum stress, which should not exceed 15 MPa for
a reliability consideration.
‘We will use , a native application of
the Workbench, to complete the task.
The unit system used in this exercise is mm-s-N (to
set up the unit system, see Section 3.3-5).
Keep the Number of Design Variables Minimum
(\ ,
[1] The %
gripper.
[2 The shape is P4(~35.160)
defined by 7 key
points >
____/
P3(-69,120)
P5(-~-34,100)
or)
2(-90,40) ‘
a.
+
“The design parameters that can be changed to improve a design are called design variables (in this case, X-coordinates
of P2, P3, and P6). The overall computing time of an optimization project will increase dramatically az the number of
design variables increases. Therefore it is important co consider
rniaimum number of design variables
In this case, we could have chosen both X-caordinates and Y-coordinates of P2, P3,P6 as design variables, That
would increase the number of design variables from 3 to 6, without earning any benefits at all, since the X-coordinate
and the ¥-coordinate of a point are not independent co each other,284 Chapter 8 Optimization
8.1-2 Resume the Project "Gripper"
Rene a ahem mont 7 Unda ome lis
etter ees EE
or
8 Fateh ane nataore
[1] Launch Bae
sai Gains
, 1 trowenaten
open the eo , ere tensinnantsy |
"Gripper." [| enw i
Fes rosso || :8= a mimeat come
Section 7.1, ftemmvewenters |] 6 care “4 coeeees
Ja rn e fl eee : + copes ora es
|@scosmef [3] Pulldown
[Greene] select . %
| Scream eam } ( 12] Select either
< < or
tonne, mm, 5, N> unit
system
Hoe Manage
Sopeeace
Regensl snd engine ope:
Grophas reaton
[8] Turning on this
[aoc | has the sme effect
lemmas as che property cet |
ewan | pw TE |
J exceptivisa |
ermanent sein
pe B)
f [A] Select
|
You can define as mary parameters as you want in a CAD system, such as DesignModeler, or Pro/Engineer. However,
only those parameters with names of prefix or suffix specified in the are exposed
(sible) co applications of the Workbench. The default prefix or suffix is "DS" (case insensitive, DS stands for Design
Space, an obsolete name for the Workbench). For example, by default. following parameter names are exposed to the
Workbench: DS_radius. ds_radius, Radius_DS, Radius_ds. where the underscore can be dropped
{In our case, we leave a blank for the , That means any parameters defined in a
CAD system are exposed to applications of the Workbench. Note that the changes of settings in the are
ent: they are effective even for other project. So, remember to change back if needed.
permaSection 8.1 Step-by-Step: Flexible Gripper 285,
8.1-3 Define Parameters in DesignModeler
Start up by double-clicking in the system. Highlight in
the tree outline and switch to mode.
[5] Type “Y2"
ce as the
oMae deg aewe seis See eRm@al paramerer
ic teas ine Gaete BD aad + oe
Some sa *
2
vertical dimension of P2 (40
mm). Note that your
| dimension name may not be
the same here (here,V7).
(
| [A] Click the box before the
i [6] Click
—= Le
| (71A"D" ismarked on | ‘\\
| the box to indicate that
itisa"driving | \
parameter" ks valve |
controled bythe
program now and
cannot be changed by [
you unless you remove
” by clicking it
agin
Repeat steps [4-6] for the vertical dimensions of P3 and Pé [8,9]
Remember that your dimension names may not be the same as they CES
Details of Stet
appear here (here, they areVB and V9 respectively)
Shah Sethi
‘Shati Vistitty ‘Show Snteh
Shove Contsinn’ io
7 | Dimensions: 12
3 e {11] Double. | u Oven
i click to open @ rm
2 Engneering Dats 4, J . | iB rn
ean Ae) | a Bs
a 35 wn
4G Hour _ |
tae (a) Chick the ee pe
sw, , ( ick the) | oa a
e Pro, ~, | box before the | ve ae
Sf [10] A bus is} | vertical dimension | van 10 own
7° added to the project schematic. | | "Srp3 (120mm). | vi Svan
| |
2 i] Fxposed parameters can be
accessed from this "bus.” We
| will explore ic later after we
define more parameters in|
\
‘Onm
1m
em
[9] Click the box before the vertical
dimension of P6 (60 mm) and give the
name "Y6." Close .286 Chapter 8 Optimization
8.1-4 Define Parameters in
[7] Hightignt , and in the Details
Fume | (OMe || ONE ae hogan
eae ete Controlled> for . This will save some ‘. |
i cise | | "
[i serene
Tessin
1 2m (KW
Yemen [] Highlight
“Ssetcsemeaal.
& fsnesan
te |
it An
1 seen wma
oe
She Sea | Ree | ee
{ oars ' sia
: | Ge ee
i | sSse i
aoe
| [2] Click the box before [6] Click the
|
!
the , A "P” r bax before the
| is marked on the box to { [4] Click the box } tamu
| indicate that ieis a | [before the |
Cr {stinrume.
8.1-5 Explore the Parameter Sets
r
| to |
| reset a “standard”
workspace.
— ( pytieae
|: several window
[1] Doubleick |} panes nthe
| SParameter Sec>, | | workspace. Each
| can be turned on!
a off from this menu.
, °Section 8.1, Step-by-Step: Flexible Gripper 287
[4] The lises
all parameters and
charts (if any).
( (ry The
lists
series of “design poines.”
The only design pore now |
ie che currant design
CURRY
[oe son Te teas te
me oe. sare sae
ifr iTt
freemen catenn?
Bl ownnecrs?
[9] Click when finish
the exploration.
[5] The | [6] The
°
@Return to Project }
[8] The lists the
tools (for creating charts)
available for the parameter
highlighted in the
\ .
displays the chart
| displays properties of
ihe parameter highlighted in the
highlighted in the
J { .
. There is
no charts yet in the
now.
Input Parameters and Output Parameters
In optimization jargon, input parameters are also called design variables, and output parameters are also called state
variables. Initial values of input parameters are set up by the user, and the Workbench subsequently change these
values to improve the design. Values of output parameters are calculated from the simulation applications (eg,
). In our case the input parameters are Y2,Y3, and Y6, and the output parameters are the maximum
horizontal displacement, the minimum combined stress, and the maximum combined stress. Note chat the input
parameters are not necessarily defined in . For example, the loads defined in could be
defined as input parameters.
Design Space
The space spanned by the input parameters is called the design space, In our case, the space spanned by parameters
Y2.Y3, and Y6 is the design space, which is a three-dimensional space. The optimization process can be thought of a
process of searching a best point in the design space.
Design Point
Any point in the design space can be a design point. The
[5] lists a series of design points with
Which the simulations have been carried out. In our case, we have only one design point:Y2 = 40,3 = 120,Y6 = 60.
This design point is now called current design.
Remark
The Workbench refers the parameters as PI, P2,P3 (norvitalic), etc. [4]. In our ease of flexible gripper. we use P2,P3.
and Pé (italic) for three control points of the shape. Be careful not to mix up these two sets of notations288 Chapter 8 Optimization
8.1-6 Create a System
[Sent nero
| Gives
acres
Bi merestenncr
(Gece Seomeeinte
Bisa
{Gemma
oN RE, ones
aE tretewress
1a ae
smn
295 roma
[2] This link indicates |
that the parameter set |
is accessible.
[1] Expand and
double-click . re =
— [3] Double-click
— .
There are three cells, or steps,in a system.
Design of Experiments
In this step, simulations with several design points are performed. The design points are chosen according to a method
of Design of Experiments (DOE). in our case, total of 15 design points (including current design) will be chosen
(8.1-7[II]). That implies a total of 15 times of simulation similar to Section 7.1 will be performed, The output
parameters are recorded. The values of output parameters are also called the responses.
Response Surface
In this step, the design points and their response values are used to best-fit mathematical functions (typically second-
order polynomials). The functions are called response surfaces. Each response value (output parameter) has
response surface, so three response surfaces are constructed in our case. The response surfaces are used to predict
the output values for any point in the design space. Any point in a response surface is called a response point.
Optimization
In this step, the goal or constraint of each output parameter is specified. In our case, the goal is to maximize the
horizontal displacement, subject to a constraint on the combined stress that its absolute value does not exceed 15
MPa. Knowing the goals and constraints of the output parameters, searches through the response
surfaces and suggests candidate designs for you, The provides three algorithms for searching the
response surfaces. The MOGA (multi-objective genetic algorithm) deals with both single- and multi-objective
problems, The NLPOL (nonlinear programming by quadratic Lagrangian) deals with single-objective problems.
Screening method divides the design space into grids and exhaustively searches the design space; the method can deal
with problems involving discrete design variables.Section 8.1 Step-by-Step: Flexible Gripper 289
8.1-7 Generate DOE Table and Results
[1] In the
, highlight
[5] In the > ae
ee) , highlight
: highlight | +
IY. _ epee
\/ i
DB iw bec [6] Type lower
8 enn and bound and upper
ie: toran nine ret bound in
,
[4] Type tower
bound and upper
bound in
.
[10] In the toolbar, click
,
[7) In the , 99 Preview Design of Experitents
light
fo ennnid [8] Turn on |
. a290 Chapter 8 Optimization
The DOE Table Pay)
With highlighted [7]. the
shows details of how the experimental array 10 1
is generated. The default fe E
. We change the to [9], just because it is easier to
understand for a newcomer of Design of Experiments
(DOE). dt 4
The experimental array [1 1] is a typical Face-
Centered Central Composite Design. The 5 design points 8 PuY2)
in the design space are graphically shown on the right [12].
‘With these design points, it is possible to construct a
response surfaces of full second order polynomials. 2 3
The effect of turning on [8] is to save the results of the simulation _P3('6) sage oe
of each design point. Otherwise, would the design
preserve only che values of ouput parameters. It may take space.
imuch disk space, but, in case we need these data, we don't
need to rerun the project.
# Update Design of Experiments(
[14] le takes a while to
finish the simulations of 15
design points. You can click
to watch
the progress report.
[13] Save project. In
the toolbar click
[15] The results of 15 design
points show the horizontal
displacement can be as high as
58 mm among 15 design points.
The next step is to build
response surfaces using these
design points and search the
response surface to find an
even higher value.
TE
1
=} [16] Click .
> Return to Project
SRT SRA RECT Co iSection 8.1 Step-by-Step: Flexible Gripper 291
8.1-8 Build Response Surface
eae
Capikaansias 7
1 FA hor Orta nanan Mnban
[1] Double-click
‘.
[2] Turn on these
‘options to preserve
[3] Click . It |:
takes just a few
seconds to build the
response surfaces.
"Baoan lm ane
Thema nine Ss ora
[6] Leave them as
default for now. You
can try other
configuration later
ae [4] To visualize the B
response surface,
highlight .
[2] Set up objective and
constraints, Note that the
Unit of stress is MPa.
(uywieh
in the
highlighted,
select for
in the
.
DTT peice ote =
ef earth} neo +
ry ‘Soe of Goneratod Senne Set 8 SR eS
rake 0
. coontne 8 3707 mon
7 Update Optimization
{2] Click .
SERENE 3) The best ic finds is 55.83! mm.
| Note that sears (*) are used to give
suggestion: the more stars, the better;Section 8.1 Step-by-Step: Flexible Gripper 293,
8.1-11 Perform Optimization Using MOGA Method
[1] Select
for , ‘Optimization>.
ete fad Sos
Hantro SngePer Raion
‘am Ao PotoFeeiage 70
[iets
SEES
aaron
[3] The best it finds
is 56.828 mm.
Yaies
Method»
—
{1 Select == |
tune’ tt Saas
nea (sto) |
Se of Gerad ance Set 0
1 it oe
ead Jee cased Lime ten | Codra tne | Cand apts
ay Monreal eum or
pe [3] The best it finds e
- Neoen| is 58.144 mm,
AY RRR294 Chapter 8 Optimization
8.1-13 Update Current Design
v
So far, the best GA we can find is 1.16 (58.144/50) (8.1-12[3]), which is z
close to one of the design point in the DOE table (8.1-7[15]). We want v
to update the current design with this design point, so that we can start va
from this point for another iteration to improve the design. 6 Soktion va
7D ens es
2896) Paeneters ay
Skate Srna N55) i
r
[1] Click is
replaced by .
[3] Right-click
f and select ,
oe
[6] Click .
return to Project
{5] Right-click
and select
i] .Section 8.1 Step-by-Step: Flexible Gripper 295
8.1-14 Check the Current Design
= Details of Sketehl
‘Slawh [Sine
Slotch Vistiity Show Shatch
Show Conctiaints? |No
|=: Dimensions: 12
[1] Double-click to
open
DesignModeler.
5 @ sep ae
6 |G soon Ya
7 @ Rents Ye
‘Stati Structural (ANSYS)
[3] Close
| DesignModeler and
open
.
geometry is
updated
| A: Statie Structural (ANSYS)
X foss- Dietonal Defoonation
|. Tope: Digestion Defomaton (1 Axis)
nit
‘Global Coonan System
[4] The results is
updated. aaa
0.14871 Min
Wrap Up
Close . Save the projec and exit . As an exercise (83-2), you may perform another
iteration of optimization using the updated design as “central design.” Choose appropriate ranges of adjustment for
the design variables,