Systems Engineering: Journey from Adolescence to Adulthood ( )

Transcription

1 Systems Engineering: Journey from Adolescence to Adulthood ( ) SWISSED16 The Swiss Society of Systems Engineering Day Kongresshaus, Zurich, 12 September 2016 Olivier L. de Weck, Ph.D. Professor of Aeronautics and Astronautics and Engineering Systems Editor-in-Chief of the journal Systems Engineering Adjunct Professor EPFL Space Center

2 What is special about 1991? Switzerland celebrated its 700th Birthday! I graduated from ETH Zurich Dept. IIIE Betriebs- und Produktionswissenschaften Switzerland planned to procure the F/A-18 Military Aircraft from the U.S. Navy / McAir The International Council for Systems Engineering (INCOSE) was founded V-Model of Systems Engineering Stakeholder Analysis Lifecycle Management Systems Engineering Overview Requirements Definition Commissioning Operations System Architecture Concept Generation Verification and Validation, Change Mgt Tradespace Exploration Concept Selection System Integration Interface Management The now famous V -Model was published Concurrent Engineering Multidisciplinary Optimization Is Systems Engineering any further along today? Olivier de Weck, September 2016 Page 2

3 Outline of this seminar qthe Swiss F/A-18 Story q Research in Change Propagation Analysis qmeta and Model-Based-Systems-Engineering (MBSE) q Utopia for Systems Engineering in 2041 Olivier de Weck, September 2016 Page 3

4 A Transatlantic Journey 1997 MIT Cambridge Boeing St. Louis NASA Goddard SFC JPL Pasadena NASA JSC Houston NASA KSC Florida EPFL Lausanne Fribourg 1987 ETH Zurich RUAG Aerospace Zuoz Size Difference is 200:1 Zermatt 4

5 F/A-18 Center Barrel Section Y453 Y470.5 Y488 Wing Attachment 74A Olivier de Weck, September 2016 Page 5

6 Swiss F/A-18 Experience Multidisciplinary systems like aircraft are very complex and highly coupled Typically optimized for mission performance Difficult to change design when requirements change during lifecycle Instigated changes Propagated changes Issue of Engineering Changes is critical: - some were anticipated (avionics, software) - others were not (structural airframe) - F/A-18 C/D system was redesigned for a mission that was not part of the original set of requirements for the U.S. Navy Page 6

7 F/A-18 Lessons Learned q Changes increased cost per aircraft by O(~$10M) Encountered some surprises along the way q Changing a system (or product) after its initial design is often required to accommodate new requirements expensive, and time-consuming if change was not anticipated in the original design à How flexible is the design? qchange propagation some changes are local and remain local other changes start local, but propagate through the system in complex, unanticipated ways q Can we predict why changes are initiated and how they propagate? Olivier de Weck, September 2016 Page 7

8 Outline of this seminar qthe Swiss F/A-18 Story q Research in Change Propagation Analysis qmeta and Model-Based-Systems-Engineering (MBSE) q Utopia for Systems Engineering in 2041 Olivier de Weck, September 2016 Page 8

9 Change Propagation Analysis in Complex Systems q Complex Sensor System q 46 Areas Long range sensor system, complex hardware, software, human operators Derivative of earlier generation 8 Year development program Hardware Software Program Documentation q System Map (graph) Interconnections between areas collaboration with Giffin M., de Weck O., Bounova G., Keller R., Eckert C., Clarkson P.J., Change Propagation Analysis in Complex Technical Systems, Journal of Mechanical Design, 131 (8), , August 2009 Olivier de Weck, September 2016 Page 9

10 Structure of the Data Set q Change Request Database technical, managerial, procedural track parent, child, siblings by areas with unique ID number chronologically numbered IDs Total of 41,500 change requests q Data Mining Procedure Export from DBMS to text file Written into MySQL database with Perl scripts Equivalent to a MS Word document with 120,000 pages Sorting, Filtering, Anonymizing Write simplified change request format (see right side) ID Number Date Created Date Last Updated Area Affected 19 Change Magnitude 3 Parent ID MAR-Y5 10-JAN-Y6 Children ID(s) 15678, Sibling ID(s) 9728 Submitter Assignees Associated Individuals Stage Originated, Defect Reason Severity Typical Change Request Completed? 1 eng231 eng008 eng231 eng018 Admin_001 Engineer_271 [blank], [blank] [blank] Olivier de Weck, September 2016 Page 10

11 Change Networks q Apply Graph Theory to extract networks of connected changes parent-child changes sibling changes q Most changes are only loosely connected 2-10 related changes q Some large networks emerged (rank) (connected changes) q Question: do these networks emerge from a single initial change? parent sibling Legend child sibling Change proposed Change rejected Change implemented No, Change Network Coalescence! Olivier de Weck, September 2016 Page 11

12 Change Propagation Network Network plot of largest change network in the dataset, with 2579 associated change requests. Created by Gergana Bounova Data from Monica Giffin (SDM) Olivier de Weck, September 2016 Page 12

13 Bi-Partite Graph Analysis (87-CR) Not all areas of the system are affected equally, there are hot spots. System Network Map Change Propagation Network Page 13

14 Change Propagation Index (CPI) q Classify each area Absorber, Carrier, Multiplier instigating area ΔDSM Change Propagation Frequency p ij change propagation probability c ( parent ) + c ( sibling) ij = ( j) C tot ij total completed changes in Area j Area receiving area N ( ) C () i = p C ( j) in ij tot j= 1 Also earlier work With Prof. Eun Suk Suh A change in Area 1 caused changes in Area 6 with a frequency of 4.17%. N ( ) C () i = p C () i out ji tot j= 1 CPI () i = C () i C () i out in C () i + C () i out in -1 <= CPI <= +1 Page 14

15 System Area Classification CPI Spectrum q Areas found to be strong multipliers 16: hardware performance evaluation 25: hardware functional evaluation 5: core data processing logic 32: system evaluation tools 19: common software services 3: graphical user interface (GUI) q Areas found to be perfect reflectors 27, 41: look like perfect absorbers but actually zero changes implemented despite numerous changes proposed = perfect reflectors Page 15

16 Change Request Generation Patterns Change Requests Written per Month system integration and test Discovered new change pattern: late ripple Number Written component design [Eckert, Clarkson 2004] subsystem design bug fixes major milestones or management changes Month Page 16

17 Outline of this seminar qthe Swiss F/A-18 Story q Research in Change Propagation Analysis qmeta and Model-Based-Systems-Engineering (MBSE) q Utopia for Systems Engineering in 2041 Page 17

18 Status quo approach for managing complexity SWaP used as a proxy metric for cost, and disincentivizes abstraction in design Cost Optimization System decomposed based on arbitrary cleavage lines... System Functional Specification MIL-STD-499A (1969) systems engineering process: as employed today System Layout Re-Design Conventional V&V techniques do not scale to highly complex or adaptable systems with large or infinite numbers of possible states/configurations Verification & Validation SWaP Optimization SWaP Optimization Power Data & Control Thermal Mgmt Subsystem Design Component Design Subsystem Testing Component Testing Resulting architectures are fragile point designs... and detailed design occurs within these functional stovepipes Unmodeled and undesired interactions lead to emergent behaviors during integration SWaP = Size, Weight, and Power V&V = Verification & Validation Desirable interactions (data, power, forces & torques) Undesirable interactions (thermal, vibrations, EMI)

19 Historical schedule trends with complexity

20 META Approach to 5x acceleration of SE Conventional Product-Development Design Flow e.g., MIL-STD-499 Requirements Definition Concept Design Preliminary Design Detailed Design System Integration T 3 META Product-Development Design Flow Design Time: T Design Time: 0.2T T 1 T 2 0.1T 3 Layer N Layer 2 Requirements Definition Design-Space Exploration Layer 1 T 1 10T 2 Layer N Layer 2 Abstraction-Layer Design C P (A P, Layer G P ) = C1 a ±C b ±! Composition C P (A P, G P ) Rules = C a ±C b ±! Composition C P (A P, G P ) = Rules C a ±C b ±! Composition Rules Complexity C 1 (A 1 ; G 1 ) C 3 (A 3 ; G 3 ) C 2 (A 2 ; G 2 ) C 1 (A 1 ; G 1 ) C 3 (A 3 ; G 3 ) C 2 (A 2 ; G 2 ) C 1 (AActor 1 ; G 1 ) C 3 (A 3 ; G 3 ) C 4 (A 4 ; G 4 ) Actor Library Actor Library C 4 (A 4 ; G 4 ) C 4 (A 4 ; G 4 ) Uncertainty Library - Structure, organization, dynamics n % n n 4 ( Adaptability C( n, A) = α i + ' β k a ijk * i=1 &' i=1 j=1 k=1 )* γe( A) - Switching cost for alternative architectures Performance 1 st Cost Meta-Language (Common Semantic Domain) Metrics Interactions System Validation Manufacturing Deployment T 4 System Validation - Confirmation 0.2T 4 Manufacturing Deployment 20

21 Vensim Model of META (5x) Process Levels of Abstraction META flag (on/off) Complexity Measure RDT&E (NRE) Cost Schedule Pressure Key META-related features shown in red Model Library Certificate of Completion de Weck O.L., Feasibility of a 5x Speedup in System Development due to META Design, Paper DETC , ASME 2012 International Design Engineering Technical Conferences (IDETC) and Computers and Information in Engineering Conference (CIE) Chicago, Illinois, August 12-15, 2012 Change Management 21

22 Benchmark Case Results (3,000 requirements ) Simulation Case Schedule to complete NRE $ to complete Idealistic Project months $27.9M Realistic Project w/changes 70 months $51.9M META-enabled project months $31.5M Spending Rate 6 M 4.5 M 3 M 1.5 M 0 META-enabled Idealistic Project (no changes) Time (Month) Spending Rate : META - enabled Spending Rate : Realistic (with changes) Spending Rate : Idealistic (no changes) Realistic project with changes Simulation Assumptions: All: Schedule Pressure = 1.5 META: 3-layers of abstraction (CB=9) META: C2M2L library coverage: 50% META: Novelty: 50% META: C2M2L library integrity: 80% Problems caught early: 70% Key Result: META speedup factor = 70/16=4.4 Confirmed that META speedup of 5x is possible but cost reduction is only 1.5 x! 22

23 META-Enablers Sensitivity Analysis is very revealing! [2 3 5] [ ] [ ] [ ] [ ] +,-./0'12'3405/,6718' 91:.;'+<4/,/-'E85.?/<5-' F6D.:C;.'A/.00C/.' Normalized Sensitivity Analysis: A 100% change in a META process parameter will cause a X % change on schedule and NRE cost GL.65'18'F6D.:C;.'M' GL.65'18'F6D.:C;.'!' GL.65'18'NC:?.5'M' GL.65'18'NC:?.5'!' [ ] [ ] 3/6D<5.65C/.'GHI;1/,718'J,5.' 9,H<BCB'3/6D<5.65C/.'KD/1C?DIC5' Baseline values In bold!"#$%$&'!"$$%$&'!($%$&'!)$%$&'!*$%$&'!#$%$&' $%$&' #$%$&' *$%$&' q Increasing Layers of Abstraction from 2 à 3 significantly improves schedule, there is much less benefit in going from 3 à 4 or from 3 à 5 q C2M2L Model Library Coverage (completeness) is key for both schedule and NRE q META ability to catch problems early has big budget impact q Schedule Pressure speeds up schedule also in META - but costs more 23

24 Validation: 777 Electric Power System (EPS) q Project Parameters from Hamilton Sundstrand: 5 Years Feb 1990 (project work authorized) Jan 1995 (final qualification test complete) Source control drawing was completed in 1993 This is the complete equipment spec. Number of customer requirements: ~1,500 Number of Change Request: ~300 Total number of major components: 33-2 Integrated Drive Generators (IDG); 1 auxiliary generator (APU driven); 3 Generator Control Units; 1 Bus Power Control Unit; 24 Current Transformers; 2 Quick attach/detach Units Ratio of systems people working CDR to people working Qualification test was 1:1.5 Approach: 1. Approximately simulate B777 EPS Program execution 2. Simulate META version of B777 EPS and see impact 24

25 Comparison of B777 EPS Program (actual vs. META) Comparsion B777 - Design and Integration, Validation and Completion 400 Specifications/Month 80 Specifications/Month 400 Requirements/Month 400 Requirements/Month 1 1 Completion Times: Actual: months META: 16 months (predicted) 200 Specifications/Month 40 Specifications/Month 200 Requirements/Month 200 Requirements/Month Specifications/Month 0 Specifications/Month 0 Requirements/Month 0 Requirements/Month 0 0 Design and Integration : B777-META Design and Integration : B777-actual Validation : B777-META Validation : B777-actual Certificate of Completion : B777-META Certificate of Completion : B777-actual B777 EPS- META B777 EPS - actual Rework mountain Requirements Defined Time (Month) Requirements 2,000 1,500 1, ~ 1,500 requirements Time (Month) Requirements Defined : B777-actual Requirements Defined : B777-META Cumulative Changes ~300 vs. 100 changes Specifications/Month 200 Specifications/Month Requirements/Month Requirements/Month Time (Month) Cumulative Changes : B777-META Cumulative Changes : B777-actual 25

26 A bitstream-programmable, foundry-style factory Paint & Finish Additive/Subtractive Manufacturing Welding Sheet Metal Fabrication Assumptions: 40k-60k ft2 total space for GCV-scale capability Need not be geographically co-located Custom components in-sourced to ifab network Unmodified COTS components out-sourced Drill & fill Wire bundles Robotics Fuels & Tribology Composites Autoclave Tape Laying Harness Buildup CNC Brake 3D Printer Paint Booth Automated Harness Loom Laser Cutter 6-Axis Robots Laser Sintering Welding Robots Fuel Cell Test Set Assembly Automated Storage and Retrieval Electronics Fabrication Swaging Press CNC CMM Anodizing Tank AGVs Articulating CMM CNC Tube Bender Machine Instructions (STEP-NC, OpenPDK) Dynamometer Logistics ifab Foundry Configuration Tube Bending Hydraulics & Pneumatics QA / QC Product MetaRepresentation 26

27 MBSE Case Study: REXIS q q q q One of six instruments on the OSIRIS-REX asteroid sample return Launch happened on Sept 8, 2016 Measures X-rays that are fluoresced from Bennu Fluorescent line energies depend on the atomic structure of the matter Provides a unique elemental signature Line strengths reflect element abundance Spectrometer SXM Credit: Mark Chodas 27

28 REXIS Design History Overview SysML Models created for SRR, SDR, and PDR 2014 SysML models created at SRR, SDR, and PDR From Fall 2011 through Spring 2012, REXIS team composed primarily of undergraduates With grad students and faculty mentors From Summer 2012 to present, REXIS team composed primarily of grad students With faculty mentors and undergraduate volunteers 28

29 REXIS Design History SRR - January 2012 PDR - January 2013 SDR - April 2012 CDR - February

30 REXIS Design History Statistics Parts per Assembly All assemblies experienced parts growth 30

31 REXIS Design History Statistics (cont.) Ports per Assembly All assemblies experienced interface growth 31

32 REXIS Design History Statistics (cont.) Ports Per Part in each Assembly Interfaces per part Average number of interfaces per part is invariant as design matures from SRR to CDR 32

33 Outline of this seminar qthe Swiss F/A-18 Story q Research in Change Propagation Analysis qmeta and Model-Based-Systems-Engineering (MBSE) q Utopia for Systems Engineering in 2041 Olivier de Weck, September 2016 Page 33

34 SE today and Utopia in 2041 q SE today Systems Engineering has transitioned to adulthood in 25 years SE has spread to many industries incl. medical devices etc. MBSE: Moving from documents to interactive models META: Design systems from libraries of components using compositional rules and guarantees of correctness. Speedup 4-5x. YES, we are much further along. q SE Utopia in 25 years We design elegant systems with only essential complexity Every physical system has a model-based digital twin There is a Nobel Prize awarded for Systems Engineering The 1 st, 2 nd,... law of systems science and engineering is well established and widely accepted (similar to thermodynamics) Olivier de Weck, September 2016 Page 34

35 1 st Law of SE: Conservation of Complexity f a k cr 6 P f r At a critical nodal degree <k> cr = 6 systems transition from a lower-complexity hierarchical to a higher-complexity distributed architecture. Essential complexity is conserved. Network resilience "&$% contour (f r vs. f a ) [Valente et al., 2004] Complexity+=+548+ [Whitney et al., 1999]

36 Systems Engineering: From Adolescence to Adulthood ( ) q Many thanks for your attention q Consider submitting a manuscript to the INCOSE Wiley Journal Systems Engineering Olivier de Weck, September 2016 Page 36