Adaptive Make Radical Advances in System Design & Manufacturing
description
Transcript of Adaptive Make Radical Advances in System Design & Manufacturing
Adaptive Make Radical Advances in System Design & Manufacturing
Paul [email protected]
fmr. Deputy Director/Acting DirectorTactical Technology Office
Briefing prepared for MIT Club of Washington, DC31st Annual Seminar Series—Rebuilding U.S. Manufacturing
March 12, 2013
The views expressed are those of the author and do not reflect the official policy or position of the Department of Defense or the U.S. Government.
The six frigates (1794)
2
“… the sum of $688,888… to provide, equip and employ, four ships to carry forty guns each, and two ships to carry thirty-six guns each….”
--An Act to Provide a Naval Armament, March 27, 1794
USS Philadelphia, Tripoli Harbor, February 16, 1804
3
B-52 Stratofortress (1946)
"It is desired that the requirements set forth be considered as a goal and that the proposal be for an interim airplane to approximate all requirements, except that emphasis must be placed on meeting the high speed requirement... It is the intent that design proposals should present the best possible over-all airplane..." --Directive letter inviting design proposals for the B-52 bomber, February 13, 1946
F-111 Aardvark (1961)
5
F-35 Joint Strike Fighter (2001)
6
Technical spec length over time
Year of Entry into Service
7
Exponential trends in component technologies
8
Exponential trends in component technologies
9
Exponential trends in component technologies
10
Exponential trends in component technologies
11
Exponential trends in component technologies
12
Embedded Software Size
Complexity—a driver of cost & schedule growth?
13
Evidence for a causal relationship
14
The problem is in the “seams”
Between stages of production →
← Between system components
Between people & organizations →
Source: MIT ESD (deWeck et al., 2008)
Source: DDR&E/SE (Flowe et al., 2009)
15
Systems engineering—a method for managing complexity
CostOptimization
Power Data & Control Thermal MgmtSWaP
Optimization
SWaPOptimization
System FunctionalSpecification
. . .
. . .
SubsystemDesign
ComponentDesign
SystemLayout
Verification & Validation
ComponentTesting
SubsystemTesting
SWaP = Size, Weight, and PowerUndesirable interactions (thermal, vibrations, EMI)Desirable interactions (data, power, forces & torques)
V&V = Verification & Validation
System decomposed based on arbitrary cleavage lines . . .
Conventional V&V techniques do not scale to highly complex or adaptable systems–with large or infinite numbers of possible states/configurations
SWaP used as a proxy metric for cost, and dis-incentivizes abstraction in design
Unmodeled and undesired interactions lead to emergent behaviors during integration
. . . and detailed design occurs within these functional stovepipes
MIL-STD-499A (1969) systems engineering process: as employed today
Re-Design
Resulting architecturesare fragile point designs
16
Systems engineering process unchanged for 50 years
Giffin M., de Weck O., et al., Change Propagation Analysis in Complex Technical Systems, J. Mech. Design, 131 (8), Aug. 2009.
Engineering Change Requests (ECRs) per Month of Program Life
From Project Inception through Midcourse Maneuver, vol. 1 of Mariner Mars 1964 Project Report: Mission and Spacecraft Development, Technical Report No. 32-740, 1 March 1965, JPLA 8-28, p. 32, fig. 20.
Mariner Spacecraft (1960s) Modern Cyber-Electromechanical System (2000s)
17
The Consequences
18
Cost growth
19
$1 quintillion
$1 quadrillion
$1 trillion
$1 billion
$1 million
$1 thousand
1900 1950 2000 2050 2100 2150Year of Entry into Service
Wright Model A
Morse JN-4A Standard E-1
DH-4
SPAD P-39P-51
P-61
F-15
F-14
F-18F-16
A-10
F-35F-18
B-52
GROSS NATIONAL PRODUCT
DEFENSE BUDGETEntire Defense budget to buy one airplane.
Entire GNP to buy one airplane.
Source: Norm Augustine, Augustine’s Laws, 6th Edition, AIAA Press, 1997.
Sources: Air Force Museum, Air Force Magazine, GlobalSecurity, DAMIR database, and other sources
Schedule growth
20
Decrease in aircraft program new starts
Source: Anton, P., Wind Tunnel and Propulsion Test Facilities, RAND TR-134, 2004
21
22
Decrease in industrial output
Source: Booz & Company analysis
Source: Booz & Company analysis
Production facility closures
23
and
Ideal Employer Ranking
Failure to replenish the talent pool
Source: Booz & Company analysis of data from AIAA and Universum Ideal Employer Rankings Survey
24
Enter DARPA
25
1957
Sputnik Dwight D. EisenhowerDwight D. Eisenhower
1958
What is DARPA?
The Defense Advanced Research Projects Agency (DARPA) was established in 1958 to prevent strategic surprise from negatively affecting U.S. national security and create strategic surprise for U.S. adversaries by maintaining the technological superiority of the U.S. military.
26
► Singular mission to create & prevent strategic surprise
► Not a requirements-driven culture
► Acceptance of risk (technical & process) and failure
► Projects of finite duration culminating in demonstration
► Tenures of finite duration; broad aperture for recruitment
► Agile organization; no labs; no facilities; flexible
contracting
► Varied transition paths—sometimes very indirect
What makes DARPA unique?
27
Orbital ExpressMiTEXPegasus Taurus
1997
DARPASAT Falcon SLVGLOMR
1998
Tacit Blue X-45/46/47Global HawkX-31
199019821977 2002
Have Blue A-160 HTV-2
2011
Sea Shadow
1984 2005
2002
Boomerang
2003
Tank BreakerM16 SSN
1987
NetfiresTalon
20031973 1978
Big Dog
2008
200720061990 1995 20031985
1997
AM3ALG DTCUGCVNAVLAB
1984 2011200419951990
Aero-/Hydro-Dynamic Systems
Space Systems
Ground/Soldier Systems
Tactical Technologies
TTO’s history of technology demonstrators
28
Approaches for tackling complexity
Simplify
Disaggregate
Modularize
??? 29
Raising the level of abstraction—VLSI design
Daily engineer output(Trans/day)
Develop-ment time (mo)
IP block performance Inter IP communication performance models
incr
easi
ng a
bstr
actio
n
Cluster
AbstractCluster
AbstractRTL RTL
clusters
Abstract
Cluster SWmodels
IP blocks
Transistor model Capacity load
Gate level model Capacity load
System-on-chip Design Framework Wire load
30
Transistors per chip
Speed (Hz)
Feature Size (µm)
Sources: Singh R., Trends in VLSI Design: Methodologies and CAD Tools, CEERI, Intel, The Evolution of a Revolution, and Sangiovanni-Vinventelli, A., Managing Complexity in IC Design, 2009
Foundry-style manufacturing—integrated circuits
An approach to VLSI chip design that separates design from manufacturing (Mead & Conway, 1979).
Design implementation:
Use of simplified device & component models that trade some performance for automation of design.
Design rules that are independent of and scalable with process technologies.
Semiconductor manufacturing facility becomes the semiconductor foundry.
Semiconductor product implementation: Chip prototypes are manufactured in silicon foundries using the same tools, fabrication processes and materials used for high-volume chip manufacturing… no seams.
The result:Moved from hundreds of chip designers using vertically-integrated, captive semiconductor facilities to tens of thousands of designers using pure-play semiconductor foundries to create thousands of products.
Continues to enable, cost-effective custom VLSI products: Generating new markets & new companies including Apple, Silicon Graphics, Cadence, Jazz, TSMC, Broadcom, Nvidia and Qualcomm.
31
1 10 100 1,000 10,000 100,000 Complexity (# genes inserted/modified)
1010
1011
109
108
107
106
105
104
103
Eff
ort
(to
tal $ *
yrs
to d
evelo
p)
[$*y
r] yeastminimal bacterium
DARPA annual budget
Living Foundries
genome rewritecomplex genetic circuits
metabolic engineering
LF: after 6 mos
Raising the level of abstraction—synthetic biology
32
Foundry-style manufacturing—proteins
Biology provides the design rules and models
Vaccine implementation: Only the relevant genetic sequence of bug required, not entire virus.
DARPA Blue Angel program enabled… •A 4 site manufacturing platform in the USA capable of meeting phase 1 appropriate FDA requirements for vaccine production.•3 Investigational New Drug Applications with the FDA•3 Phase 1 clinical trials
Texas A&M University (TAMU)-Caliber example:Growth room is approximately the size of half a football field at four stories tall (150 feet x 100 feet x 50 feet high) Total number of plants: 2.2 million
The result today…Rapid, adaptive platform. Tobacco plant production may result in more rapid production cycles (< 30 days) and less facility expenditures to increase capacity once an FDA approved product is available.
33
Aerospace state-of-the-art
34
Image courtesy of Dassault Systemes
Dassault Falcon 7XTwo-fold schedule compression for new business jets through faithful application of a digital master model with QA/QC feedback by tail number
Lockheed Martin F-35Shimming and ‘drill and fill’
approach significantly worsens production learning effects,
leading to delays and cost growth*
Image courtesy of Lockheed Martin
* GAO-10-382: Joint Strike Fighter – Additional Costs and Delays Risk Not Meeting Warfighter Requirements on Time, Mar 2010
34
Adaptive Vehicle Make
35
Improving designer productivity by abstractionH
iera
rcha
l Abs
trac
tion
System
Subsystems
Assemblies
Components
Static Models
Qualitative Models
Linear / ODE
Nonlinear / PDE
Relational Models
~1010
~105
~102
~10
Model Abstraction
# of
Des
ign
Alte
rnati
ves
36
Integration of formal semantics across domains
META Semantic Integration
Formal Verification
•Qualitative reasoning•Relational abstraction•Model checking•Bounded model checking
Distributed Simulation
•NS3•OMNET•Delta-3D•CPN
EquationsModelica-XML
FMU-MES-functionFMU-CS
High LevelArchitectureInterface (HLA)
Composition•Continuous Time•Discrete Time•Discrete Event
•Energy flows•Signal flows•Geometric
Hybrid Bond Graph
ModelicaFunctional Mock-up
Unit
Embedded Software Modeling
TrueTimeSimulink/Stateflow
Stochastic Co-Simulation
•Open Modelica•Delta Theta•Dymola
37
Verification of design correctness
• Models are fully composable• Simulation trace sampling to verify
correctness probability• Application of probabilistic model
checking under investigation• 10^2 10 designs
Component Models•Modelica•State Flow•Bond Graphs•XML•Geometry
Sem
an
ticIn
teg
ratio
n
• Static constraint application• Manufacturability constraints• Structural complexity metrics• Info entropy complexity metrics• Identify Pareto-dominant designs• 10^10 10^4 designs
Static Trade Space Exploration Qualitative Reasoning
• Qualitative abstraction of dynamics• Computationally inexpensive• Quickly eliminate undesirable designs• State space reachability analysis• 10^4 10^3 designs
Relational AbstractionLinear Differential Equation Models
• Relational abstraction of dynamics• Discretization of continuous state space• Enables formal model checking• State-space reachability analysis• 10^3 10^2 designs
• Generate composed CAD geometry for iFAB
• Generate structured &unstructured grids
• Provide constraints and input data to PDE solvers
• Couple to existing FEA, CFD,EMI, & blast codes
• 10 1 design
CAD & Partial Differential Equation Models
Embedded Software Synthesis
• Auto code generation• Generation of hardware-
specific timing models• Monte Carlo simulation
sampling to co-verify• Hybrid model checking
under investigation
Phys
ical
Softw
are
Compu
ting
A
B
38
Modeling shows promise for 5X time compression
400 Req/M 40010 Arch/M 80
400 Spec/M 6002000 Tests/M 600 Req*/M 800
200 Req/M 2005 Arch/M 40
200 Spec/M 3001000 Tests/M 300 Req*/M 400
00 12 24 36 48 60 72 84 96 0 12 24 36 48 60 72 84 96
Time (Month) Time (Month)Requirements Elicitation : METAm-on/off-with-changeConcept Exploration : METAm-on/off-with-changeDesign and Integration : Metam-on/off-with-changeVerification : METAm-on/off-with-changeValidation : METAm-on/off-with-changeCertificate of Completion : METAm-on-with -change
Requirements/MonthArchitectures/MonthSpecifications/Month
Tests/MonthRequirements/Month
META Design FlowTraditional Design Flow
Source: Olivier de Weck, MIT39
Digitally-programmable manufacturing network
Manufacturing Process Model LibraryManufacturing Process Model Library
Constraintsfrom Selected Configuration
META Design
Static Process Mapping SequencingFoundry Trade Space
Exploration
Kinematic Machine Mapping
Topological Decomposition
Kinematic Assembly Mapping
SchedulingCNC Instructions
Human Instructions
*Manufacturing ConstraintFeedback to META Design Rock Island Arsenal Bldg
299 Final Assembly
*
*
40
Unfolded (unstable)
Folded (stable)
41
Expanding the talent pool: FoldIt
Sources: Fold it, Katib et al, Crystal structure of a monomeric retroviral protease solved by protein folding game players., Nature Structural and Molecular Biology 18, 1175–1177, 2011
Purpose: Research mobilization, self-organization potential of social networks & crowd sourcing
Challenge: Locate 10 large, red weather balloons at undisclosed locations across the United States on Saturday, December 5, 2009
Result: All 10 balloons located in 8 hours 52 minutes•Winner = MIT Red Balloon Challenge Team
• recursive incentive scheme with human data validation•4,368 total registrants
12:56
10:08
11:5411:27
14:20
11:1111:32
15:57
13:01
12:14
Balloon Locations/Time First Submitted
*Balloons Went Up at 10:00 EST*
Union Square, San Francisco
ALICE WINS $750BOB WINS $500CAROL WINS $1,000DAVE WINS $2,000
42
Expanding the talent pool: Red Balloon Challenge
Source: Complexity and Social Networks Blog
Expanding the talent pool: VehicleFORGE
43
FANG 1: Mobility/Drivetrain• Vehicle drivetrain to meet IFV speed, efficiency,
terrain, reliability objective• Model library spanning multiple instances of all
components necessary to build system, along with curation capability to accept new models
Fast Adaptable Next-Generation Ground Vehicle (FANG)
FANG 3: Full Amphibious IFV
• Complete IFV based on core USMC ACV objectives and distilled requirements
FANG 2: Chassis/Structure
• Chassis and armor design to meet principal IFV-like survivability objectives
• Model library to include: • Materials and joining mechanisms for hull• Novel configs (monocoque, v-hulls)
Purp
ose
Tim
ing
Priz
eSc
ope
• Demo META design flow functionality• Establish collaborative design environment• Demo foundry configuration functionality• Test breath/depth of component model library
• Design: Jan-Apr 2012• Build: May-Jul 2012
• Demo META advanced design analysis tools• Enable vendor-driven component models• Expand AVM collaborative design community• Resolve handling of sensitive design features
• Design: Oct 2012 – Jan 2013• Build: Feb-Apr 2013
• Demo 5-fold reduction in development time• Produce amphibious IFV to USMC req’ts• Bolster industry adoption of META approach • Establish self-sustaining model libraries
• Design: Jun 2013 – Jan 2014• Build: Feb-Jul 2014
• $1M for winning design • $1M for winning design • $2M for winning design
• Complete basic functional development of META tools
• Establish FANG Challenge administration/rules/mechanisms
• Generation of timely TDP for iFAB Foundry build
• Executing entire make process within 3 years (1/5th of SoA timescale)
• Developing/exiting a design community large enough to enable 25+ competitive teams
• Integrating critical GOTS components (i.e. weapons systems, turret, C4ISR, etc.)
• Handling and incorporation of security sensitive data and analysis into challenge execution
• External vendor component model contributions• Standing with PM-AAA ACV Hull Survivability
Demonstrators (HSDs)
Prog
ram
Fo
cus
44
www.darpa.mil
45