Post on 24-Apr-2018
transcript
Resistance vs. strain
5µm
Silver in elastic matrixHigh ∆R/RRepeatable, no drift or offset
STRAIN SENSORS
Pixel color - tuning applied control voltage
1.5V operationModulate transparencyContrast – oxidized/reduced state of PEDOT
PIXELS
AccessibilityFreedom of Form
IntegrationVTT, Finland
CSAIL,MIT
Rethink Robotics
ADVANCED MANUFACTURING
A unit microstructure cell
Base materials
Heterogeneous material
1.E-04
1.E-03
1.E-02
1.E-01
1.E+00
10 100 1000
Density (kg/m3)
Youn
g’s
Mod
ulus
(G
Pa) ?
What is the range of achievable physical properties (gamut)?
MAPPING MICROSTRUCTURES TO MATERIAL PROPERTIES
TRADITIONAL MATERIALS
Range of mechanical material properties for traditional homogeneous materials such as foams, metals and polymers.
Finite element solvers with multi-grid preconditioners to simulate multi-physics problems
PHYSICS SIMULATION
Stretch
Shear
Young’s Modulus
Pois
son’
s ra
tio
A microstructure
MAPPING MICROSTRUCTURES TO MATERIAL PROPERTIES
• Microstructure samples
• Compute level set
• Find random seeds near the level set boundary
COMPUTING MICROSTRUCTURE GAMUT
• Microstructure samples
• Compute level set
• Find random seeds near the level set boundary
• Find gradient towards outside of gamut
COMPUTING MICROSTRUCTURE GAMUT
• Microstructure samples
• Compute level set
• Find random seeds near the level set boundary
• Find gradient towards outside of gamut
• Discrete and continuous sampling
COMPUTING MICROSTRUCTURE GAMUT
• Microstructure samples
• Compute level set
• Find random seeds near the level set boundary
• Find gradient towards outside of gamut
• Discrete and continuous sampling
• Update level set
COMPUTING MICROSTRUCTURE GAMUT
GAMUT FOR MICROSTRUCTURES WITH CUBIC SYMMETRY
Relative Young’s Modulus (log scale)
Pois
son’
s ra
tio
DISCOVERY OF MICROSTRUCTURES
[Bickel et al. 2010] [Shim et al. 2013] [Babaee et al. 2013][Kadic et al. 2012]
[Meza et al. 2014] [Kadic et al. 2016][Clausen et al. 2015] [Volgiatzis and Chen 2016]
COMPUTATIONAL FRAMEWORK
Nonlinear dimensionalityreduction
Skeleton
Distribution of a family
Family representative
Fitted structure
Input: Estimate Gamut
Identify Families Fit Templates Reduce Parameters
Automated discovery of new microstructural materials
STEP 2: FIT MICROSTRUCTURE SKELETONS
θl1l
2
Extracting a skeleton
Template definition given a skeleton
Representative structure
Skeleton Graph
Beam on graph edges
STEP 3: REDUCE TEMPLATE PARAMETERS
Pois
son’
s ra
tio
Range of material properties for Family 4 Reduced parameter directions
Log G
Poiss
on’s
ratio
Log E
Reducing joint beam thickness decreases Young’s modulus E and Poisson’s ratio.
EXAMPLE: PARAMETER REDUCTION
Log G
Poiss
on’s
ratio
Log E
Shifting joint location outwards increases shear modulus G and Poisson’s ratio.
EXAMPLE: PARAMETER REDUCTION
RESULT: DISCOVERY OF NEW AUXETIC MATERIALS
2 beams 3 beams 4 beams 7 beams6 beams
21
4
5Log G
Poi
sson
’s
rati
o
3
Five families of new microstructural materials with extremal auxeticproperties
DISCOVERED AUXETIC MECHANISMS
Slanted column Rotating triangle
CompressionCompression
Beam deformation
Joint rotation
CHALLENGES
Software: SIMP Topology OptimizationUp to millions of elementsDifficult to handle multiple materials
Hardware: Object-1000 Plus• Up to 39.3 x 31.4 x 19.6 in. • 600dpi (~40 microns) • 5 trillion voxels
𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑛𝑛′𝑃𝑃 𝑅𝑅𝑅𝑅𝑅𝑅𝑃𝑃𝑃𝑃
Young’s Modulus𝑆𝑆𝑆𝑆𝑆𝑅𝑅𝑆𝑆 𝑀𝑀𝑃𝑃𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑃𝑃
Base materials
Force
GripDesign Goal Continuous Optimization
Material Property Space
Continuous Representation
Fabrication
TWO-SCALE TOPOLOGY OPTIMIZATION
TWO-SCALE TOPOLOGY OPTIMIZATION
𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑛𝑛′𝑃𝑃 𝑅𝑅𝑅𝑅𝑅𝑅𝑃𝑃𝑃𝑃
Young’s Modulus𝑆𝑆𝑆𝑆𝑆𝑅𝑅𝑆𝑆 𝑀𝑀𝑃𝑃𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑃𝑃
Base materials
Force
GripDesign Goal Continuous Optimization
Material Property Space
Continuous Representation
Fabrication
SUMMARY: THREE FUNDAMENTAL QUESTIONS
1. How to compute the space of possible material properties?2. How to discover material families with extreme properties?3. How to synthesize designs with desired function?
OUTLOOKMulti-material 3D printing will become an important manufacturing method
Prototyping, Presentation,
Education 44%
Manufacturing 27%
Functional Parts 28%
Other 1%
Source: Wohlers Report
OUTLOOKComputational methods will redefine the design process
• Engineers will specify material properties and not materials• Engineers will specify product function and not individual elements
Coef
ficie
nt of
Exp
ansio
n
Relative Stiffness
OUTLOOKComputational methods will redefine the process of scientific discovery
• Material discovery will become more automated, relying on physical simulation, data generation, and machine learning
Identify familiesEstimate gamut Fit Templates Find mechanismNonlinear dimensionalityreduction
SkeletonFamily representative
Fitted structure
OUTLOOKComputational methods will redefine cyber-physical system design and mechanism understanding
• Agile drones, fast walkers, efficient swimmers• Discover the optimal underlying mechanisms
? ? ?
??
?
?
?
OUTLOOKComputational methods will guide the development of new manufacturing processes
Material Property GamutBase Materials Manufacturing