electrostatic actuated beam optimization case study · optimization problem definition what do we...

16
Coupling modeFRONTIER with ANSYS Electrostatic Actuated Beam Optimization Case Study www.ozeninc.com/Optimization [email protected]

Upload: hoangdat

Post on 12-Jul-2018

215 views

Category:

Documents


0 download

TRANSCRIPT

Coupling modeFRONTIER with ANSYS

Electrostatic Actuated Beam OptimizationCase Study

www.ozeninc.com/[email protected]

1

X

Y

Z

Electrost

atic clamp

ed beam an

alysis

SEP 17 2009

15:02:33

VOLUMES

TYPE NUM

2

Problem Description:

• An electrostatic structural analysis is performed to determine the deflection of a silicon beam for a MEMS switch.

• A clamped beam for an RF MEMS switch device is modeled to compute the center deflection for an applied voltage.

• Forces generated by the electrostatic field will bend the beam towards a ground plane.

Electrostatic Actuated Beam Model

3

FE Geometry & Boundary Definition

1

X

Y

Z

Electrostatic clamped beam analysis

SEP 17 2009

15:31:49

VOLUMES

TYPE NUM

U

1

X

Y

Z

Electrostatic clamped beam analysis

SEP 17 2009

15:31:49

VOLUMES

TYPE NUM

U

1

X

Y

Z

Electrostatic clamped beam analysis

SEP 17 2009

15:31:49

VOLUMES

TYPE NUM

U

Boundary Conditions

Surface Interface, Voltage and Ground Potential Locations

1

X

Y

Z

Electrostatic clamped beam analysis

SEP 17 2009

15:51:37

ELEMENTS

VOLT

FSIN

1

Optimization Problem Definition What do we want to improve?

Objectives

To determine the optimized geometric configuration of the Electrostatic Actuated Beam Model when subjected to a multi objective optimization:

Maximize Displacement, Minimize Electric Potential, Minimize Volume

Why to Maximize Displacement

• In order to maximize the performance of the system.

Why to Minimize Electric Potential• Higher operating voltages exponentially decrease the operating lifetime of the switch.

Why to Minimize Volume

• Minimize the Mass and Material Cost.

Create workflow in modeFRONTIER

• Define the Inputs and their Domains

• Set Ansys as an Application Node

• Set the Logic flow

• Set the Outputs

• Set the Objectives:

• Maximize Displacement

• Minimize Electric Potential

• Minimize Volume

Parameter Domain

Beam Length 135 - 160 mm

Beam Width 2.5 - 4 mm

Beam Height 1.5 – 2.5 mm

Voltage 80 – 140 V

Multi- Objective Electrostatic Actuated Beam Optimization Problem Definition In modeFRONTIERTM

1

Electrostatic clamped beam analysis

SEP 17 2009

16:00:56

ELEMENTS

BeamHeight

Multi- Objective Electrostatic Actuated Beam Optimization Problem Definition In modeFRONTIERTM

SOBOL as DOEMOGA-II as Scheduler

Multi Objective(functions to be maximized

or minimized)

Input variablesof the parametric

model

ANSYS

OutputVariables

Postprocessing – Bubble Plot

Maximize Displacement

Min

imiz

e E

lect

ric

Po

ten

tial

• The Bubble plot shown is a 3D view of the scatter.

• It compares the design points with respect to different factors (e.g. Displacement, Electric Potential, Volume).

• All the computed Design configurations are shown in the Bubble plot.

• Around 1000 configurations were computed

• ModeFRONTIER identifies designs that are Pareto (non-dominated)

• Total CPU time required for the optimization: circa 28 hours

ParetoFrontier

Postprocessing – Parallel Coordinate Chart

The Parallel Chart allows the viewing of all designs simultaneously, with one vertical axis for each variable or output

It is most useful for Filtering Designs, especially in cases with multiple, conflicting, objectives

Each Jagged Line across the Chart represents one Design Configuration

Moving the sliders up or down, hides all designs outside the range, allowing the selection of Designs of interest (RED LINE)

Optimum Design• The chart represents all the Pareto designs.

• By sliding the Volume Objective, the designs can be filtered.

Postprocessing – Parallel Coordinate Chart

• Sliding the other objectives will further filter designs which doesn’t satisfy the criteria.

Postprocessing – Parallel Coordinate Chart

• Final Design can be chosen from the parallel co-ordinate chart by further sliding the Max Displacement objective towards the higher end.

#371

Postprocessing – Parallel Coordinate Chart

ANSYS Postprocessing – Final Design

1

MN

MX X

Y

Z

Electrostatic clamped beam analysis

-.958366-.851881

-.745396-.638911

-.532425-.42594

-.319455-.21297

-.1064850

SEP 17 2009

15:05:03

NODAL SOLUTION

STEP=12

SUB =1

TIME=120

UY (AVG)

RSYS=0

DMX =.958366

SMN =-.958366

1

MN

MX X

Y

Z

Electrostatic clamped beam analysis

-.958366-.851881

-.745396-.638911

-.532425-.42594

-.319455-.21297

-.1064850

SEP 17 2009

15:05:03

NODAL SOLUTION

STEP=12

SUB =1

TIME=120

UY (AVG)

RSYS=0

DMX =.958366

SMN =-.958366

1

MN

MX X

Y

Z

Electrostatic clamped beam analysis

-.958366-.851881

-.745396-.638911

-.532425-.42594

-.319455-.21297

-.1064850

SEP 17 2009

15:05:03

NODAL SOLUTION

STEP=12

SUB =1

TIME=120

UY (AVG)

RSYS=0

DMX =.958366

SMN =-.958366

1

MNMX

X

Y

Z

Elect

rostat

ic cla

mped b

eam an

alysis

0 12

24 36

48 60

72 84

96 108

SEP 17

2009

15:10:

19

NODAL

SOLUTI

ON

STEP=1

2

SUB =1

TIME=1

20

VOLT

(AV

G)

RSYS=0

DMX =.

949166

SMX =1

08

1

MNMX

X

Y

Z

Electrostatic clamped beam analysis

012

2436

4860

7284

96108

SEP 17 2009

15:10:19

NODAL SOLUTION

STEP=12

SUB =1

TIME=120

VOLT (AVG)

RSYS=0

DMX =.949166

SMX =108

1

MNMX

X

Y

Z

Electrostatic clamped beam analysis

012

2436

4860

7284

96108

SEP 17 2009

15:10:19

NODAL SOLUTION

STEP=12

SUB =1

TIME=120

VOLT (AVG)

RSYS=0

DMX =.949166

SMX =108

Y-Displacement Electric Potential

Conclusions

• In few hours modeFRONTIER tested several configurations, the same task would have taken days for a single operator

• modeFRONTIER created an automatic procedure: once the parametric model is set, the optimizator will keep iterating it till it finds the best configurations

• modeFRONTIER finds the optimum solutions (pareto frontier), therefore the need of testing only the best configurations reducing the experimental phase and controlling the spending

Conclusions

• ModeFRONTIER found the optimum designachieving improvement for all the parameter specified except the Electric Potential Objective

• In this case the Displacement and Volume objectives are improved at the expense of Electric Potential

• Displacement increase of 14.2% from the initial design

• Electric potential increase of 1.84% from the initial design

• Volume reduction of 16.1% from the initial design

MCDM Multivariate AnalysisStatistical Analysis

Process Integration

Response Surface Tool

Design of Experiments Optimization Algorithms Robust Design

modeFRONTIER Capabilities

Stay Ahead During Challenging Times

• To learn more about how Ozen Engineering can help you incorporate simulation into your design and testing processes, please visit us at www.ozeninc.com

• For more Design Optimization case studies visit: http://ozeninc.com/OptimizationCaseStudies

• If you would like us to create a demo for your specific case or for any other question, please contact us at: [email protected]