Computational and C D R Theoretical Problems in Modern Rapid - - PowerPoint PPT Presentation

C D R

Computational and Theoretical Problems in Modern Rapid Prototyping

Mark R. Cutkosky Stanford Center for Design Research

http://cdr.stanford.edu/interface

C D R

NAS Math. Modeling Forum 5/10/99 -mrc

2

Outline

• Introduction to Layered Manufacturing

– Commercial and research processes – Enabling factors (why now)

• Capabilities and opportunities

– (Almost) arbitrary geometry – Functionally graded materials – Integrated assemblies, “smart parts”

• Computational challenges

– Huge design space – Analysis – Process planning and control

• Summary

C D R

NAS Math. Modeling Forum 5/10/99 -mrc

3

Traditional manufacturing: a sequential process of shaping and assembly

C D R

NAS Math. Modeling Forum 5/10/99 -mrc

4

Layered Manufacturing: commercial example

Laser UV curable liquid elevator Formed

• bject

Photolithography process schematic Sample prototype (ME210

power mirror for UT Auto) http://me210.stanford.edu

C D R

NAS Math. Modeling Forum 5/10/99 -mrc

5

Almost arbitrary 3D Geometries

Loop Tile -- dense tiling of 3D space. (Carlo Sequin, U.C.B.) Minimum toroidal saddle surface (C. Sequin) Tilted frames (RPL)

C D R

NAS Math. Modeling Forum 5/10/99 -mrc

6

From RP and CNC to . . .

2000 1970 1990 Shape Deposition Manufacturing ( SDM) RP CNC

C D R

NAS Math. Modeling Forum 5/10/99 -mrc

7

Deposit (part) Shape Embed Deposit (support) Shape Part

• Support

Shape Deposition Manufacturing

(CMU/SU)

C D R

NAS Math. Modeling Forum 5/10/99 -mrc

8

SDM#1: Injection mold tooling

(SU RPL)

C D R

NAS Math. Modeling Forum 5/10/99 -mrc

9

SDM #2: Frogman (CMU)

• Example of polymer component with

embedded electronics

C D R

NAS Math. Modeling Forum 5/10/99 -mrc

10

SDM #3: Ceramic parts

(RPL)

Alumina vane Silicon nitride pitch shaft

Alumina turbine wheels

C D R

NAS Math. Modeling Forum 5/10/99 -mrc

11

SDM for integrated assemblies

Motor Leg links Shaft Shaft coupling

Body frame Lift pot Knee pot Hip pot Abduct pressure sensor Lift pressure sensor Extend pressure sensor Gears Actuators

Motivation: Building small robots with prefabricated components is difficult... and results are not robust.

C D R

NAS Math. Modeling Forum 5/10/99 -mrc

12

Designer composes the design from library

• f primitives,

including embedded components

Steel leaf spring Piston Outlet for valve Valve Primitive Circuit Primitive Inlet port primitive Part Primitive

SDM #4: Robot leg with embedded components

(http:cdr.stanford.edu/biomimetics)

C D R

NAS Math. Modeling Forum 5/10/99 -mrc

13

Internal components are modeled in the 3D CAD environment.

Steel leaf-spring Piston Sensor and circuit Spacer Valves

Components are prepared with spacers, etc. to assure accurate placement.

Robot Leg design (cont’d.)

C D R

NAS Math. Modeling Forum 5/10/99 -mrc

14

The output of the software is a sequence of 3D shapes and toolpaths.

Robot Leg: compacts

Support Part Embedded components

C D R

NAS Math. Modeling Forum 5/10/99 -mrc

15

A snapshot just after valves and pistons were inserted.

Steel leaf-spring Piston Sensor and circuit Valves

Robot Leg: embedded parts

C D R

NAS Math. Modeling Forum 5/10/99 -mrc

16

Finished parts ready for testing

Robot Leg: completed

C D R

NAS Math. Modeling Forum 5/10/99 -mrc

17

Layered Manufacturing: is it a new manufacturing paradigm?

Photo-sculpture studio (1860) Laminated manufacturing (1892-1940s) Laser-based photolithography (1977)

[Source: Beaman 1997]

C D R

NAS Math. Modeling Forum 5/10/99 -mrc

18

A process enabled by computing...

3D solid model

CAD

slicing trajectory planning material addition process process planner fabrication machine

C D R

NAS Math. Modeling Forum 5/10/99 -mrc

19

Summary of layered manufacturing processes

Commercial

• Photolithography
• Fused deposition
• Laser sintering
• Laminated paper

Research

• Selective laser sintering

(UT Austin)

• 3D printing (MIT)
• Shape deposition

manufacturing

(CMU/Stanford) “Look and feel” prototype Complex 3D shapes direct from CAD model Engineering materials (metals, ceramics, strong polymers) Graded materials Embedded components Not quite direct from CAD model...

C D R

NAS Math. Modeling Forum 5/10/99 -mrc

20

Layered manufacturing results in a huge space of possible designs:

• Ability to create arbitrary 3D structures with

internal voids

• Ability to vary material composition throughout

the structure

• Ability to embed components such as sensors,

microprocessors, structural elements.

What kind of design environment will help designers to understand and exploit the potential of layered manufacturing?

C D R

NAS Math. Modeling Forum 5/10/99 -mrc

21

Ability to create arbitrary 3D structures with internal voids (homogeneous materials)

Shape optimization example:

Find the minimum-weight shelf structure, bounded by box B, that supports load W without failing. B W

Space within B is divided into N cells, each of which can be filled or empty. Number of unique designs ≈ 2N

Rapid Prototyping Workshop 5/99 -mrc

C D R

NAS Math. Modeling Forum 5/10/99 -mrc

22

Ability to vary material composition

Support structure deposition heads

Deposition heads can be controlled to deposit varying amounts of each material* as the part is

• built. Total material

composition varies throughout the part.

Volume fractions always add to unity* *void, or empty space, is treated as a special case of material

C D R

NAS Math. Modeling Forum 5/10/99 -mrc

23

Material composition: product space

Product Space: )! 1 ( ! )! 1 ( 1 1 − − + =         − − + m r m r m m r C

urethane glass void teflon

m = number of materials (including void) vi = volume fraction of each material

r = deposition mixture resolution

1

1

=

∑

= m i i

v

Example: urethane, glass fibers, teflon, and void, controlled to a resolution of 10% volume fraction ⇒ 286 unique mixtures possible.

C D R

NAS Math. Modeling Forum 5/10/99 -mrc

24

Design space with arbitrary geometry and heterogeneous materials (E3 × Tm)

Shape + material optimization: Assume m possible materials, (including void) with a mixture resolution of r. B W Space within B is discretized into N cells, each of which can be filled with a unique mixture of materials. Number of unique designs ≈

        − − + 1 1 m m r C

N Example: 10×10×10 cells, 4 materials, 10% mixture resolution ⇒ 2861000 designs!

Rapid Prototyping Workshop 5/99 -mrc

C D R

NAS Math. Modeling Forum 5/10/99 -mrc

25

Toward a design environment for layered manufacturing

• The design space is huge.
• But there are significant constraints

associated with the manufacturing processes.

• Therefore, provide an environment that

combines manufacturing analysis, design rules, and design libraries to help designers explore the full potential of layered manufacturing.

C D R

NAS Math. Modeling Forum 5/10/99 -mrc

26

Computational issues #1: Process Planning

• Process constraints
• Manufacturability
• Support structures
• Deposition method
• Deposition parameters
• Path planning
• Machining method
• Tool selection
• Machining parameters
• Path planning

Decompose Deposit Machine Decompose Deposit Machine Input (source: J.S. Kao SU RPL)

C D R

NAS Math. Modeling Forum 5/10/99 -mrc

27

Decomposition into ‘compacts” and layers

C D R

NAS Math. Modeling Forum 5/10/99 -mrc

28 (1) (2) (3) (4)

primary material support material

Decomposition based on process sequence

(5) (6) (7) (8)

C D R

NAS Math. Modeling Forum 5/10/99 -mrc

29

Definitions: Compact [Merz et al 94]

• 3-D volume with no overhanging features
• Rays in growth direction enter only once
• Compacts correspond to SDM cycles
• (c) OK
• (b) OK

( )

[ ]

∀ ∀ ∀ ∃ ∃ ∈ ↔ ≤ ≤ x y z z z x y z a z z z

1 2 1 2

, ,

z1 z2

C D R

NAS Math. Modeling Forum 5/10/99 -mrc

30

Decomposition algorithms

Locate silhouette edges, split surfaces

D E F

Extrude concave loops Merge compacts

(source: J.S. Kao SU RPL)

C D R

NAS Math. Modeling Forum 5/10/99 -mrc

31

Deposition Process Planning (RPL)

6” 3/8” laser beam metal powder

• Thermal Stresses Develop due to:
• Temperature gradients
• Differences in expansion coefficient
• Thermal Stresses Cause:
• Part inaccuracy
• Delamination

cooling deflection

• Solutions
• Develop optimal deposition path

and process parameters to minimize thermal stresses

• Tailor alloy to maintain desirable

properties while minimize thermal expansion coefficient

C D R

NAS Math. Modeling Forum 5/10/99 -mrc

32

Problems with automated process planning

• finite thickness of support material
• finish on unmachined surfaces
• warping and internal stresses
• decomposition depends on

geometry, not on intended function

C D R

NAS Math. Modeling Forum 5/10/99 -mrc

33

Design by Composition (M. Binnard)

Users build designs by combining primitives with Boolean operations

– Primitives have high-level manufacturing plans – Embed components and shapes as needed Primitives merged by designer Manufacturing plans merged by algorithm

C D R

NAS Math. Modeling Forum 5/10/99 -mrc

34

Decomposed Features

SFF/SDM VLSI

Boxes, Circles, Polygons and Wires SFF/SDM Design Rules Mead-Conway Design Rules

λ 2λ 2λ λ

Wc/λ >= 2

Minimum gap/rib thickness d d ∆d (top view) a) Generalized 3D gap/rib d(α1,α2) (side view) b) d(α1,α2) Minimum feature thickness d(m1,m2,m3) (side view) e) m1 m2 m3 d(m1,m2,m3,α1,α2) m1 m2 m3

Toward a mechanical MOSIS?

Primitive = Compact Set + Precedence Graph

• Set of valid compacts
• No intersections
• Fills the primitive’s projected

volume

Primitive Compact set Compact precedence graph

• Acyclic directed graph
• Link for every non-

vertical surface

C D R

NAS Math. Modeling Forum 5/10/99 -mrc

36

Merging Algorithm Example

intersection compacts non-intersecting compacts

A B

• A

B C=A ∪ B

C D R

NAS Math. Modeling Forum 5/10/99 -mrc

37

CAD MODEL CAD MODEL DESIGN DECOMPOSITION DESIGN DECOMPOSITION DESIGN BY COMPOSITION DESIGN BY COMPOSITION

LIBRARY:

Decomposed Designs & primitives

COMPACT SET CPG

SEQUENCE & TOOL PATH PLANNING SEQUENCE & TOOL PATH PLANNING

re-analysis (if needed)

Combining composition and decomposition

C D R

NAS Math. Modeling Forum 5/10/99 -mrc

38

A need for integrated mechanical, thermal and electrical analysis

VuMan (CMU) mechanical, thermal analysis

C D R

NAS Math. Modeling Forum 5/10/99 -mrc

39

Summary

Emerging layered manufacturing processes such as SDM:

– are made feasible by recent advances in desktop computing and solids modeling – afford a huge design space (E3 × Tm) – provide a rich area for geometric reasoning and process planning – present formidable challenges in analysis, process planning and control to achieve consistent, high-quality parts

C D R

NAS Math. Modeling Forum 5/10/99 -mrc

40

Acknowledgements

Thanks to the members of the Center for Design Research and the Stanford Rapid Prototyping Lab for their work in generating the results and ideas described in this presentation. This work has been supported by the National Science Foundation (MIP-9617994) and by the Office of Naval Research (N00014-98-1-0669)