Adaptive Make Radical Advances in System Design & Manufacturing

1. Adaptive Make Radical Advances in System Design & Manufacturing Paul Eremenko eremenko@alum.mit.edu fmr. Deputy Director/Acting Director Tactical Technology Office • Briefing prepared for MIT Club of Washington, DC • 31st 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.

2. The six frigates (1794) “… 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

3. USS Philadelphia, Tripoli Harbor, February 16, 1804

4. 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

5. F-111 Aardvark (1961)

6. F-35 Joint Strike Fighter (2001)

7. Technical spec length over time Year of Entry into Service

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. Exponential trends in component technologies Embedded Software Size

13. Complexity—a driver of cost & schedule growth?

14. Evidence for a causal relationship

15. The problem is in the “seams” Between stages of production → ← Between system components Source: MIT ESD (deWeck et al., 2008) Between people & organizations → Source: DDR&E/SE (Flowe et al., 2009)

16. Systems engineering—a method for managing complexity

17. Systems engineering process unchanged for 50 years Engineering Change Requests (ECRs) per Month of Program Life Mariner Spacecraft (1960s) Modern Cyber-Electromechanical System (2000s) 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. Giffin M., de Weck O., et al., Change Propagation Analysis in Complex Technical Systems, J. Mech. Design, 131 (8), Aug. 2009.

18. The Consequences

19. Cost growth $1 quintillion $1 quadrillion $1 trillion $1 billion $1 million $1 thousand Entire GNP to buy one airplane. gross national product defense budget Entire Defense budget to buy one airplane. SPAD B-52 F-14 P-61 P-51 P-39 Morse JN-4A F-18 F-35 F-15 Standard E-1 F-16 F-18 A-10 Wright Model A DH-4 1900 1950 2000 2050 2100 2150 Year of Entry into Service Source: Norm Augustine, Augustine’s Laws, 6th Edition, AIAA Press, 1997.

20. Schedule growth Sources: Air Force Museum, Air Force Magazine, GlobalSecurity, DAMIR database, and other sources

21. Decrease in aircraft program new starts Source: Anton, P., Wind Tunnel and Propulsion Test Facilities, RAND TR-134, 2004

22. Decrease in industrial output Source: Booz & Company analysis

23. Production facility closures Source: Booz & Company analysis

24. Failure to replenish the talent pool Ideal Employer Ranking and Source: Booz & Company analysis of data from AIAA and Universum Ideal Employer Rankings Survey

25. Enter DARPA

26. What is DARPA? 1957 1958 Sputnik Dwight D. Eisenhower 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.

27. What makes DARPA unique? ► 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

28. TTO’s history of technology demonstrators Space Systems 1997 1997 2003 2007 2006 1985 1995 1990 Pegasus Taurus MiTEX GLOMR DARPASAT Falcon SLV Orbital Express Aero-/Hydro-Dynamic Systems 2011 1982 1984 1998 2002 2005 1977 1990 Have Blue Tacit Blue Sea Shadow X-31 Global Hawk X-45/46/47 A-160 HTV-2 Ground/Soldier Systems 2008 1973 1987 2002 2003 2003 1978 M16 Tank Breaker SSN Talon Boomerang Netfires Big Dog Tactical Technologies 1984 1990 1995 2004 2011 NAVLAB ALG AM3 UGCV DTC

29. Approaches for tackling complexity Simplify Modularize ??? Disaggregate

30. Raising the level of abstraction—VLSI design Transistor model Capacity load Gate level model Capacity load System-on-chip Design Framework Wire load IP block performance Inter IP communication performance models Abstract IP blocks RTL Abstract Cluster RTL clusters increasing abstraction Abstract Cluster Cluster SW models Transistors per chip Speed (Hz) Feature Size (µm) Daily engineer output (Trans/day) Develop- ment time (mo) 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

31. Foundry-style manufacturing—integrated circuits • 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. • 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. • 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. Continues to enable, cost-effective custom VLSI products: Generating new markets & new companies including Apple, Silicon Graphics, Cadence, Jazz, TSMC, Broadcom, Nvidia and Qualcomm.

32. Raising the level of abstraction—synthetic biology minimal bacterium yeast 1011 DARPA annual budget 1010 109 108 Effort (total $ * yrs to develop) [$*yr] 107 LF: after 6 mos 106 Living Foundries 105 metabolic engineering 104 genome rewrite complex genetic circuits 103 1 10 100 1,000 10,000 100,000 Complexity (# genes inserted/modified)

33. Foundry-style manufacturing—proteins • 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. • Biology provides the design rules and models • Vaccine implementation: Only the relevant genetic sequence of bug required, not entire virus. • The tobacco plant is the ‘protein foundry.’ • Vaccine implementation: Redirection of tobacco plant protein production results in candidate protein synthesis. 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 • 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

34. Aerospace state-of-the-art Dassault Falcon 7X Two-fold schedule compression for new business jets through faithful application of a digital master model with QA/QC feedback by tail number Image courtesy of Dassault Systemes Lockheed Martin F-35Shimming and ‘drill and fill’ approach significantly worsens production learning effects, leading to delays and cost growth* * GAO-10-382: Joint Strike Fighter – Additional Costs and Delays Risk Not Meeting Warfighter Requirements on Time, Mar 2010 Image courtesy of Lockheed Martin 34

35. Adaptive Vehicle Make

36. Improving designer productivity by abstraction Components ~10 Assemblies ~102 Hierarchal Abstraction # of Design Alternatives Subsystems ~105 System ~1010 Qualitative Models Nonlinear / PDE Relational Models Static Models Linear / ODE Model Abstraction

37. Integration of formal semantics across domains • Composition • Continuous Time • Discrete Time • Discrete Event • Energy flows • Signal flows • Geometric META Semantic Integration Simulink/ Stateflow Embedded Software Modeling Hybrid Bond Graph Modelica TrueTime Functional Mock-up Unit Equations Modelica-XML FMU-ME S-function FMU-CS High Level Architecture Interface (HLA) • Distributed Simulation • NS3 • OMNET • Delta-3D • CPN • Formal Verification • Qualitative reasoning • Relational abstraction • Model checking • Bounded model checking • Stochastic Co-Simulation • Open Modelica • Delta Theta • Dymola

38. Verification of design correctness Qualitative Reasoning Static Trade Space Exploration • Component Models • Modelica • State Flow • Bond Graphs • XML • Geometry Semantic Integration • Qualitative abstraction of dynamics • Computationally inexpensive • Quickly eliminate undesirable designs • State space reachability analysis • 10^4  10^3 designs • Static constraint application • Manufacturability constraints • Structural complexity metrics • Info entropy complexity metrics • Identify Pareto-dominant designs • 10^10  10^4 designs Embedded Software Synthesis • Auto code generation • Generation of hardware-specific timing models • Monte Carlo simulationsampling to co-verify • Hybrid model checkingunder investigation Linear Differential Equation Models Relational Abstraction Physical A Software Computing CAD & Partial Differential Equation Models B • 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 • Relational abstraction of dynamics • Discretization of continuous state space • Enables formal model checking • State-space reachability analysis • 10^3  10^2 designs • Models are fully composable • Simulation trace sampling to verifycorrectness probability • Application of probabilistic modelchecking under investigation • 10^2  10 designs

39. Modeling shows promise for 5X time compression Traditional Design Flow META Design Flow 0 Time (Month) Time (Month) Requirements Elicitation : METAm-on/off-with-change Requirements/Month Concept Exploration : METAm-on/off-with-change Architectures/Month Design and Integration : Metam-on/off-with-change Specifications/Month Verification : METAm-on/off-with-change Tests/Month Validation : METAm-on/off-with-change Requirements/Month Certificate of Completion : METAm-on-with -change Source: Olivier de Weck, MIT 0 0 12 12 24 24 36 36 48 48 60 60 72 72 84 84 96 96

40. Digitally-programmable manufacturing network * Foundry Trade Space Exploration Static Process Mapping Sequencing META Design Constraintsfrom Selected Configuration Manufacturing Process Model Library CNC Instructions Kinematic Machine Mapping Scheduling Topological Decomposition Human Instructions Kinematic Assembly Mapping * *Manufacturing Constraint Feedback to META Design Rock Island Arsenal Bldg 299 Final Assembly

41. Expanding the talent pool: FoldIt Unfolded (unstable) Folded (stable) 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

42. Expanding the talent pool: Red Balloon Challenge • 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 incentivescheme with human data validation • 4,368 total registrants Union Square, San Francisco 12:56 10:08 12:14 13:01 11:54 11:32 11:11 11:27 Balloon Locations/ Time First Submitted ALICE WINS $750 BOB WINS $500 CAROL WINS $1,000 DAVE WINS $2,000 15:57 14:20 *Balloons Went Up at 10:00 EST* Source: Complexity and Social Networks Blog

43. Expanding the talent pool: VehicleFORGE

44. Fast Adaptable Next-Generation Ground Vehicle (FANG) 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 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) FANG 3: Full Amphibious IFV • Complete IFV based on core USMC ACV objectives and distilled requirements Scope • Demo META design flow functionality • Establish collaborative design environment • Demo foundry configuration functionality • Test breath/depth of component model library • Demo META advanced design analysis tools • Enable vendor-driven component models • Expand AVM collaborative design community • Resolve handling of sensitive design features • 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 Purpose • Design: Jan-Apr 2012 • Build: May-Jul 2012 • Design: Jun 2013 – Jan 2014 • Build: Feb-Jul 2014 • Design: Oct 2012 – Jan 2013 • Build: Feb-Apr 2013 Timing Prize • $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 • 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) • 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.) Program Focus