code generation at mechatronics · philips innovation services, sasg oct 2011 6 dynamics, actuators...

Oct, 2011

Visit us at www.innovationservices.philips.com

Code Generation at Mechatronics

Philips Innovation Services

Philips Innovation Services, SASG Oct 2011

Contents

• Introduction Mechatronics

• Mechatronic design challenge

• Control-design ‘old-style’

• Control-design with code generation

• Code details

• Questions

2


Philips Innovation Services Position in Philips

Non-Philips

Customers

Healthcare Consumer

Lifestyle

Lighting

Royal Philips Electronics

Corporate

Technologies

IP&S

Research

Incubators

Philips

Innovation

Services

3


Philips Innovation Services

6 open access labs

> 15,000 electronic instruments

> 19,000 m2 laboratories

Incl. 9,500 m2 clean rooms

High Tech Campus

Eindhoven - NL

Green house

Prototyping & manufacturing

4

Materials

Analysis

Measurement Lab

RF

Thin Film

Clean room

Reliability lab Electronic

Prototyping

Systems

Engineering

Life Science

facility

Device Processing

Facilities

Photonics

Facility

Measurement &

instrumentation

Experience

lab

E-Science

Support

Mechatronics

Competence center

EMC

Center

Consulting Competence

Center

650 technical specialists / 50 consultants

Philips Innovation Services, SASG Oct 2011 5

Mechatronics

• Provides its customers with drives,

control and mechanical construction for

smooth motion and high accuracy

positioning

• Is both active in equipment and

subsystem design, as well as in

supporting the development of consumer

and healthcare products.

• Staff: 150 Technical specialists,40 %

University degree / Ph.D.s, 50 %

Bachelors & Engineers, 10 % other

education

• Customers:


Dynamics, Actuators & Control

Advanced System Design

High precision motion stages

Project management & supply chain

Healthcare systems

Robotics

System architecture

Vibro-Acoustics & Low

Noise Design

Electrical

Systems Design

Metrology

TEMS

Vacuum Mechatronics

Mechanical

Design

Control

Equipment

Design

Product

Development

Tribology/Air

bearings

Electro

Mechanics

Dynamics

Production

Systems

Mechatronics Competences

Motion Control

Software

G

KFB,x

SPGx

Kfay,x

KFB,y

Kfax,x

Kfsy,x

Kfsx,x

snap

acceleration

MIMO FF


Contents





• Code details

• Questions

7


Reflective electron beam lithography

Challenge

• Rotary stage 1.2 m diameter

• High Accuracy positioning

Results

• MagLev rotary stage

• Performance well within specification

• Prototypes installed at customer

Customer Benefits

• Competitive platform for introduction direct

write technology

• Throughput wph/footprint

• Low energy dissipation (low forces)


Design Challenge: measure disc position

N

E

S

X

X

Z

Y

Reference and sensors

ZW

ZN/ZS

ZN

ZS

XE XW

YN

YS

XW XE


Design Challenge: Sensor transformations

N

E

S

X

X

Z

Y

Reference and sensors

ZW

ZN/ZS

ZN

ZS

XE XW

YN

YS

XW XE 4 x sensor Axial

4 x sensor Radial

Transformation

5DOF 8 5


Design Challenge: Sensor linearization required

4 x sensor Axial

4 x sensor Radial

Transformation

5DOF 8 5

Linearize Axi

Linearize Rad


Design Challenge: Ref surface is not flat

4 x sensor Axial

4 x sensor Radial

Transformation

5DOF 8 5

Linearize Axi

Linearize Rad

N

E

S

X

Y

ZN

ZS

XE XW

YN

YS

1e-006

2e-006

3e-006

4e-006

30

210

60

240

90

270

120

300

150

330

180 0


Design Challenge: Add calibration table

4 x sensor Axial

4 x sensor Radial

5DOF

Transformation

8 5

Linearize Axi

Linearize Rad

Axi Correction Table

Rad Correction Table

Angle Sensor


Design Challenge: Angle linearization required

4 x sensor Axial

4 x sensor Radial

5DOF

Transformation

8 5

Linearize Axi

Linearize Rad



Angle Sensor Linearize Angle


Design Challenge: Surfaces change with speed

6 x sensor Axial

6 x sensor Radial

5DOF

Transformation

8 5

Linearize Axi

Linearize Rad



Angle Sensor Linearize Angle

Periodic update Calc new

Corr tables


Implementation: measurement system

16

MAC_PosSensors

2

MAC_COG

1

cos0:0

sin0:0

SurfaceCorrection

DiskAngle SurfaceCorrection

MAC_Sensors

MAC_PosSensors

MAC_RadCor_Sensors

MAC_AxiCorSensors

Gain 1

gpMAC _M_SensorsToSRF * u

Gain

gpMAC _M_SRFtoCOG * u

zeros(3,1)

0:0

gpMAC _M_RotationPointShift0:0

gpMAC _SRF_Offsets0:0

Constant

gpMAC _M_UseSurfaceCorrections0:0

CaptureCorSignals

EQCP Tracer

DiskAngle

1

gsMAC_M_RadCorSensors

gsMAC_M_AxiCorSensors

gsMAC_M_PosSensors gsMAC_M_SensorsAfterSurfCor

gsMAC_M_SurfaceCor

gsMAC_M_Position


Implementation: sensor linearization

17

MAC_AxiCorSensors

3

MAC_RadCor _Sensors

2

MAC_PosSensors

1

Terminate ZN ZE

Terminate XE XW YN

I_Rad330

AnalogIn

I_Rad020

AnalogIn

I_PosZW

AnalogIn

I_PosZS

AnalogIn

I_PosZN

AnalogIn

I_PosZE

AnalogIn

I_PosYS

AnalogIn

I_PosYN

AnalogIn

I_PosXW

AnalogIn

I_PosXE

AnalogIn

I_Axi330

AnalogIn

I_Axi020

AnalogIn

double

double

double

gpMAC _M_DiffSensorOffsets0:0

Ax9

0:0Ax8

0:0Ax7

0:0Ax6

0:0

Ax5

0:0Ax4

0:0

Ax3

0:0Ax2

0:0

Ax11

0:0Ax10

0:0

Ax1

0:0Ax

0:0

gsMAC_M_CorrectedDiff


Implementation: Ref correction

18

SurfaceCorrection

1

mod

mod

Lookup Radial

0:0

Lookup Axial

0:0

double

2*pi

0:0

gpMAC _M_RadSensorPositions0:0

gpMAC _M_AxiSensorPositions0:0

Enable

DiskAngle

1


Contents





• Code details

• Questions

19


Motion Control Design flow ‘old-style’

Mech Designer Control Designer

User Spec

Concept design

and analysis Control design



Control model

Signal ▬ PID

Sensor Xform

Plant Filter




User Spec

Software Spec

Software Designer

Main()

{

doit();

}

Signal

▬

P

I

D

Sensor

Xfor

m

Plant

F

i

l

t

e

r




User Spec

Software Spec

Software Designer

Main()

{

doit();

}

Language

Barrier!

Signal

▬

P

I

D

Sensor

Xfor

m

Plant

F

i

l

t

e

r


Contents





• Code details

• Questions

24


Motion Control Design flow using code generation


User Spec

Concept design

and analysis Control design



Control model

Signal ▬ PID

Sensor Xform

Plant Filter




User Spec

Software Spec + mdl

Software Designer

Signal

▬

P

I

D

Sensor

Xfor

m

Plant

F

i

l

t

e

r

Signal ▬

PID

Sensor

Xform

Plant

Filter

Main()

{

doit();

}




User Spec

Software Spec + mdl

Software Designer

Signal ▬

PID

Sensor

Xform

Plant

Filter

Main()

{

doit();

}

Same

Language




User Spec

Software Spec + mdl

Software Designer

Main()

{

doit();

}

Same model

Allows fast

iterations Signal ▬

PID

Sensor

Xform

Plant

Filter


Contents





• Code details

• Questions

30


Code details

• Used in 10+ projects, various applications (RobotArm … REBL)

– Sample rates 500Hz … 20kHz

– # controlled axes 1 – 15

– No defects found yet!

• Control models only, no state diagrams

– General Matlab/Simulink blocks from library

– Hardware inputs and output blocks are custom designed

– Some custom blocks to interface with C++ code

31


Code details

– Efficiency of the generated code:

• Double precision math by default

• Simulink blocks that are not enabled are not evaluated

• 30k lines at 4kHz on 2.8GHz P4.

– RTW-generated code is not nice to read

• Each block in the model becomes a little code-block

• The program is an ‘endless’ list of code-blocks

• A signal with a name becomes a variable with same name

32


Code details: conclusion

• Pros

– No language barrier with other mechatronic disciplines

– Fast iterations in integration phase

– No bugs

• Cons

– Toolchain licensing

– Testing

– In Simulink variable scope is flat: use naming convention

• Note

• Some effort required for creating custom IO blocks

33


Questions

?

34

code generation at mechatronics · philips innovation services, sasg oct 2011 6 dynamics, actuators...

Documents