Strategic Engineering

Transcription

1 Strategic Engineering Designing Systems for an Uncertain Future Olivier L. de Weck Assistant Professor of Aeronautics & Astronautics and Engineering Systems Olivier L. de Weck, 2006 Page 1

2 Outline Customization of the F/A-18 Aircraft Introduction to Strategic Engineering Research Projects: Staged Deployment of Satellite Constellations Flexible Automotive Product Platforms Time Expanded Decision Networks (TDN) Engineering Education Olivier L. de Weck, 2006 Page 2

3 Customization of the F/A-18 Aircraft Olivier L. de Weck, 2006 Page 3

4 Mission (and Configuration) Change U.S. Navy Mission fighter and attack aircraft carrier based 3000 flight hours 90 min average sortie max 7.5g positive ~15 year useful life (1978) Swiss Mission interceptor land based 5000 flight hours 40 min average sortie max 9.0g positive ~30 year useful life (1993) Redesign (Switch) Standard U.S. Navy F/A-18 C/D Configuration Modified Swiss F/A-18 C/D Configuration Olivier L. de Weck, 2006 Page 4

5 F/A-18 Redesign Strategy 1. specify new Swiss mission usage spectrum 2. apply new spectrum to existing U.S. Navy Configuration 3. identify and prioritize hot spots that most need change 4. redesign and implement local changes at hot spots Olivier L. de Weck, 2006 Page 5

6 F/A-18 Wing Carry-Through Bulkheads Olivier L. de Weck, 2006 Page 6

7 F/A-18 Center Barrel Section Y488 Y470.5 Y453 Wing Attachment 74A Olivier L. de Weck, 2006 Page 7

8 F/A-18 Center Fuselage Buildup (1) Olivier L. de Weck, 2006 Page 8

9 Center Barrel Change Consequences Substitution from Aluminum to Titanium Intended Consequence: - Increased fatigue life of individual components from hours Unintended Consequences: achieved - Increased aircraft empty weight by ~O(100) lbs - Shifted C.G. of aircraft by ~ O(1) inch - Stiffened fuselage (1 st bending mode) ~(0.1) Hz - Rendered manufacturing processes obsolete not expected or wanted Olivier L. de Weck, 2006 Page 9

10 F/A-18 Complex System Change F/A-18 System Level Drawing Fuselage Stiffened Original Change Manufacturing Processes Changed Flight Control Software Changed Gross Takeoff Weight Increased Center of Gravity Shifted Olivier L. de Weck, 2006 Page 10

11 F/A-18 Lessons Learned Changes increased cost per aircraft by O(~$10M) Changing a system after its initial design is often required to accommodate new requirements expensive, and time-consuming if change was not anticipated in the original design Change propagation some changes are local and remain local other changes start local, but propagate through the system in complex, unanticipated ways switching costs include: engineering redesign cost, change in materials, manufacturing changes, change in operational costs Olivier L. de Weck, 2006 Page 11

12 Introduction to Strategic Engineering Olivier L. de Weck, 2006 Page 12

13 What about the Future? Typical Engineering Design Mindset: Give me a set of requirements today, a timeline and a budget and I will design and deliver the best possible product/system/project for you by tomorrow. 90% of thinking and design effort is spent on this But, in essence, we are always forecasting: what customers will require in 18 months what capacity our facility will need in 3 years what variants we will produce in 8 years how many missions we will fly in 12 years What if our forecast is wrong? (it usually is) Perhaps system will function technically. But system will not deliver optimal value, or architectural lock-in occurs, or it will fail financially if its configuration is not easily changed Olivier L. de Weck, 2006 Page 13

14 Example: Iridium Satellite System 'Motorola unveils new concept for global personal communications: base is constellation of low-orbit cellular satellites', Motorola Press Release on Iridium, London, 26 June Last week, Iridium LLC filed for bankruptcy-court protection. Lost investments are estimated at $5 billion. Wall Street Journal, New York, 18 August US (forecast) US (actual) Iridium Satellite Millions of subscribers Year Difficult to properly size capacity of large system Market assumptions can change when 7-8 years elapse between conceptual design and fielding ( ) Olivier L. de Weck, 2006 Page 14

15 Traditional (Systems) Engineering Customer Needs Requirements Definition Product System Y target Conceptual Design Subsystem Y target Preliminary Design Components Y target Detailed Design Marketing Systems Engineering Subsystem Development Component Design Component Testing Final Assembly Subsystem Integration Components Y actual System Functional Testing System Y actual Subsystem Y actual System Validation Fielding/ Launch System Operation Olivier L. de Weck, 2006 Page 15

16 Implicit Assumptions of TSE The customer knows what his/her needs are The requirements are known and time-invariant The system or product can be designed as one coherent whole and is built and deployed in one step There is only one system or product designed at once The system will operate in a stable environment as far as regulations, technologies, demographics and usage patterns are concerned Olivier L. de Weck, 2006 Page 16

17 But reality tells us that Customer knows some of his/her needs but not all The true requirements often change after the system is fielded and experience is gained Constraints on capital expenditures and operating budgets frequently only allow a piecemeal implementation Often multiple variants of a system must be designed and built, possibly based on some common standard Environment is not static, but dynamic macro economic/budgetary changes (e.g. prime interest rate) regulatory changes (e.g. new CAFÉ standards) new technologies emerge (e.g. hydrogen fuel cells for cars) demographic shifts (e.g. aging population in Western nations) changing customer preferences (e.g. weighting of fuel economy) disruptive events (natural, man-made) Olivier L. de Weck, 2006 Page 17

18 Strategic Engineering Strategic Engineering is the process of designing systems and products in a way that deliberately accounts for customization and future uncertainties such that their lifecycle value is maximized. Olivier L. de Weck, 2006 Page 18

19 Strategic Engineering Framework - CDI - Operate - RDI Operate - Time Baseline System Baseline System Gen 2 Baseline Gen 2 Baseline Variant B Variant B Variant B2 Variant B2 Variant C Variant C Variant C2 Variant C2 Space Development (Stage 1) Operations Development (Stage 2) Operations Olivier L. de Weck, 2006 Page 19

20 Alternatives 1. Ignore the future and design for `optimal immediate or short-term use (= TSE) 2. Come up with a `best guess of the most likely future scenario and design to it (= forecasting + TSE) 3. Develop a range of potential future outcomes and design such that the system will be optimal on `average across all future scenarios protected against the worst case scenario take advantage of the `best case scenario most flexible to adapt to any scenario flexible robust risk averse opportunistic Interested in how to do 3. Strategic Engineering Olivier L. de Weck, 2006 Page 20

21 Strategic Engineering Toolbox Traditional Systems Engineering Methods (QFD, DSM, ) Forecasting, Change Propagation Analysis System Architecting Principles, Illities Modularity, Flexibility, Scalability, Reconfigurability, Real Options in Projects Standardization Product/System Platforms Staged Development and Deployment Optimization: Dynamic Programming, Multiobjective, all these attempt to address part of the problem, when do these methods apply, is there a unifying framework? Olivier L. de Weck, 2006 Page 21

22 de Weck Research Approach theory Non-dimensional lifecycle analysis Time-expanded decision networks Meta-platforming Generic Lifecycle Cost Modeling Generic System Modeling (OPM) Comparative Analysis NASA: Launch Vehicle Selection & Evolution NASA: Interplanetary Supply Chain & Logistics DARPA/AFRL: Space Tug Mission Scenarios BP: Exploration & Production Standardization Iridium and Globalstar: Staged Deployment GM: Flexible Automotive Product Platforms ARM: Hydrogen Enhanced Combustion Engine BP: Commercial Office Building Staging application Olivier L. de Weck, 2006 Page 22

23 Staged Deployment of Satellite Constellations Funded by Alfred P. Sloan Foundation Reference de Weck, O.L., de Neufville R. and Chaize M., Staged Deployment of Communications Satellite Constellations in Low Earth Orbit, Journal of Aerospace Computing, Information, and Communication, 1, , March 2004 Olivier L. de Weck, 2006 Page 23

24 Design (Input) Vector X Astrodynamics Constellation Type: C Orbital Altitude: h Minimum Elevation Angle: ε min Design Space Polar, Walker 500,1000,1500,2000 [km] 2.5,7.5,12.5 [deg] Satellite Design Satellite Transmit Power: P t Antenna Size: D a Multiple Access Scheme MA: 200,400,800,1600, ,2.0,3.0 MF-TDMA, MF-CDMA [W] [m] [-] Network Network Architecture: ISL yes, no [-] X 1440 = C: 'walker' h: 2000 emin: Pt: 2400 DA: 3 MA: 'MFCD' ISL: 0 This results in a 1440 full factorial, combinatorial co design space Olivier L. de Weck, 2006 Page 24

25 Objective Vector (Output) J Performance (fixed) Data Rate per Channel: R=4.8 [kbps] Bit-Error Rate: p b =10-3 Link Fading Margin: 16 [db] Capacity C s : Number of simultaneous duplex channels Cost Lifecycle cost of the system (LCC [$]), includes: - Research, Development, Test and Evaluation (RDT&E) - Satellite Construction and Test - Launch and Orbital Insertion - Operations and Replenishment Olivier L. de Weck, 2006 Page 25

26 Multidisciplinary Simulator Structure Input Vector x Constants Vector p h, ε min m sat Constellation Spacecraft m sat Launch Module LV Cost T, p P, D, MA t m T p ISL sat a n spot Satellite Network Satellite Mass Number of Satellites Number of orbital planes ngw Link Budget Note: Only partial input-output relationships shown Rs Capacity nspot Number of spot beams ngw Number of gateways LV Launch vehicle selection Cs Output Vector LCC Olivier L. de Weck, 2006 Page 26 J

27 Governing Equations Satellite Simulator a) Physics-Based Models Energy per bit over noise ratio: (Link Budget) Eb PGrGt = N kl L T R 0 space add. sys. b) Empirical Models (Spacecraft) ( ) 0.51 m = P + m sat t prop Scaling models derived from FCC database Springmann P.N., and de Weck, O.L. A Parametric Scaling Model for Non-Geosynchronous Communications Satellites, Journal of Spacecraft and Rockets, May-June 2004 Olivier L. de Weck, 2006 Page 27

28 Lifecycle Cost [B$ FY 2002] waste Traditional Systems Engineering The traditional approach for designing a system considers configurations (architectures) to be fixed over time. Designers look for a Pareto Optimal solution in the Trade Space given a targeted capacity. If actual demand is below capacity, there is a waste Iridium actual Iridium simulated Globalstar actual Globalstar simulated under cap Pareto Front Global Capacity Cs [# of duplex channels] If demand is over the capacity, market opportunity may be missed Demand distribution Probability density function { } ( δ )d P a< D b = fd D 0 f ( D) for all D D f ( )dd 1 D δ = Olivier L. de Weck, 2006 Page 28 b a

29 Staged Deployment Adapt to uncertain demand with a staged deployment strategy: A smaller, more affordable system is initially built This system has the flexibility to increase its capacity if demand is sufficient and if the decision makers can afford additional capacity Economic Advantage Some capital investments are deferred to later The ability to reconfigure and deploy the next stage is a real option Olivier L. de Weck, 2006 Page 29

30 Step 1: Partition the Design Vector x flexible x base Astrodynamics Satellite Design Network Constellation Type: C Orbital Altitude: h Minimum Elevation Angle: ε min Satellite Transmit Power: P t Antenna Size: D a Multiple Access Scheme MA: Network Architecture: ISL Rationale: Keep satellites the same and change only arrangement in space Stage I Stage II C: 'walker' h: 2000 emin: Pt: 200 W DA: 1.5 m MA: 'MFCD' ISL: 1=yes x I base = x II base C: 'polar' h: 1000 emin: Pt: 200 W DA: 1.5 m MA: 'MFCD' ISL: 1=yes Olivier L. de Weck, 2006 Page 30

31 Step 2: Search Paths in the Trade Space h= 400 km ε= 35 deg N sats =1215 Lifecycle cost [B$] h= 2000 km ε= 5 deg N sats =24 family h= 800 km ε= 5 deg N sats =54 h= 400 km ε= 20 deg N sats =416 h= 400 km Constant: Pt=200 W DA=1.5 m ISL= Yes ε= 5 deg N sats =112 Total: 40 Paths System capacity Olivier L. de Weck, 2006 Page 31

32 Step 3a: Model Uncertainty [GBM] Δ D = t t D μ Δ + σε Δ ΔD E = μ Δ t D ΔD var D = 2 Δ σ t Demand [N users ] x 10 5 Geometric Brownian Motion Model GBM model, Δt = 1 month, D o = 50,000, μ = 8% p.a., σ = 40% p.a. 3 scenarios are shown D - demand Δt time period ε- SND random variable μ, σ - constants Time [years] Demand can go up or down between two decision points Infinitely many scenarios can be generated based on this model Olivier L. de Weck, 2006 Page 32

33 Step 3b: Binomial Lattice Model u = e σ Δt d p = 1 = u e u uδt d d p p p Total 2 5 =32 scenarios Discretized Random Walk 1-p (1-p) Sample scenario ( ) n k k Pi () = p 1 p Olivier L. de Weck, 2006 Page 33

34 Step 4: Calculate cost of paths We compute the costs of a path with respect to each demand scenario We then look at the weighted average of every allowable path for cost over all scenarios Decision rule: We always adapt to demand when demand exceeds capacity The costs are discounted: the present value of LCC is considered Cap 2 Cap 1 Costs Initial wait Deploy wait deployment 2 nd stage wait Olivier L. de Weck, 2006 Page 34

35 Step 5: Identify optimal path System Lifecycle Cost [B$] A2 A3 A4 Best Path LCC of rigid design Traditional design For a given targeted capacity, we compare our solution to the traditional approach Our approach allows large savings (30% on average) 10 0 A Capacity [thousands of users] E [LCC(path j )*]= Best Deployment Strategy ( i ) E LCC( path ) = p LCC scenario n j i path i= 1 E[ΔLCC]=$650 million value of real option j Olivier L. de Weck, 2006 Page 35

36 Takeaway from Satellite Project Identified best initial configuration, as well as potential growth stages Stage A1 21 satellites 3 planes h=2000 km Stage A2 50 satellites 5 planes h=800 km Stage A3 Stage A3 112 satellites 8 planes h=400 km Previous work focused on optimal coverage for static requirements only, arrive at very different solution Requires extra upfront investment (e.g. extra fuel, tunable antenna patterns), technical details remain Olivier L. de Weck, 2006 Page 36

37 Flexible Automotive Product Platforms sponsored by General Motors Suh E.S., de Weck O.L., Chang D., Flexible Product Platforms: Framework and Case Study, Research in Engineering Design, submitted Nov.2, 2005 Olivier L. de Weck, 2006 Page 37

38 Research Context & Questions Sharp increase in number of models (variants) offered in the U.S. automotive market [Detroit News, Jan 2005]: 1947: : : 277 (estimate) Sales volumes per variant drop on average Market fragmentation Platform strategy adopted by most manufacturers Many uncertainties: - Styling & performance preferences shifting, regulations, new technologies future sales volumes are uncertain - How to design platforms to be flexible to respond to future developments? Model 3-4 years Model 3-4 years Model 3-4 years Platform ~ year life Olivier L. de Weck, 2006 Page 38

39 Typical Vehicle Architecture (Platform) General Motors Unique Carryover Modified Common Platform Traditional product platform concept: Unique Elements: Variant-specific customized elements Common Elements: Commonly shared elements among product family Rise of new elements class Flexible ( Cousin ) Elements: Elements used (with modification) in more than one variant to satisfy variant-specific requirement Olivier L. de Weck, 2006 Page 39

40 Change Propagation Analysis Design Automotive Platforms to accommodate future changes in styling and demand of individual variants Identify flexible elements Developed 7-step process BIW Change Propagation Network Key Design Variables Body-in-White Platform Olivier L. de Weck, 2006 Page 40

41 Embed Flexibility *Assume it meets quality, manufacturing, and safety requirements H122 W27 Flexible/Unique Upper Passenger Compartment H50 L48 Common Lower Front Passenger Compartment Inflexible BIW Design Unique Critical Components (Example) Body Outer Panel Flexible Lower Rear Passenger Compartment Flexible BIW Design Unique Unique Common Unique Body Inner Panel Flexible Common (Blanking) Olivier L. de Weck, 2006 Page 41

42 Cost of Design Alternatives Design Inflexible BIW Flexible BIW Component Fabrication Inflexible Flexible BIW Assembly Line Inflexible Flexible Design Initial Investment (Line + Tooling) Refurbish Cost (Every 5 Years) Switch Cost (Styling Only) Switch Cost (Styling + Length) Inflexible BIW Flexible BIW Above Belt Line H122 Length Change Forecast: Profit Differenct (Inflexible - Flexible 25,000 Trials Frequency Chart 24,973 Displayed H W L Normalized Profit Olivier L. de Weck, 2006 Page 42

43 Takeaway Automotive Platforms Product Platforms. Bandwidth can be increased by carefully embedding flexibility in the design Key is to propagate exogenous, functional uncertainties into design variables and find critical physical components Critical components are those that are change multipliers, or whose change would cause large switching costs Design for flexibility might cause larger upfront investment and larger variable costs Crossover between rigid and flexible design as a f(uncertainty) typically occurs Olivier L. de Weck, 2006 Page 43

44 Wrap-Up Olivier L. de Weck, 2006 Page 44

45 Time-expanded Decision Networks wait switch state node chance node decision node wait switch start end Period 1 Period 2 Period N Olivier L. de Weck, 2006 Page 45

46 Path Optimization in TDN For each uncertain scenario, find the optimal path through the TDN example max NPV, min LCC, start end Period 1 Period 2 Period N Olivier L. de Weck, 2006 Page 46

47 Principles of Strategic Engineering A rigid design will be optimal (max NPV) if future events unfold exactly as forecasted A robust design can minimize the standard deviation of outcomes (reduce risk), but will usually also lower the expected NPV and max achievable NPV The larger the degree of uncertainty, the more valuable flexibility will be. Flexible designs can increase the E[NPV], while limiting downside and maximizing upside The larger the switching costs from one configuration to another the more likely that the current system will be continued due to architectural lock-in, despite operational sub-optimality Olivier L. de Weck, 2006 Page 47

48 Strategic Engineering Map Degree of NPV Uncertainty σ E[NPV] Flexible Design we can adapt we are betting the farm Strategically Redesign Optimize for Expected Requirement we know what s coming Robust Design we will be ok no matter what Relative Switching Costs ΔC/LCC r Olivier L. de Weck, 2006 Page 48

49 Future Work: Where do various systems fall? Degree of NPV Uncertainty σ E[NPV] wireless sensor networks communication satellites automotive platforms? commercial aircraft consumer products water supply system highway infrastructure Relative Switching Costs ΔC/LCC Olivier L. de Weck, 2006 Page 49

50 The migration of strategic thinking Warfare ~500 A.D. Sun Tzu The Art of War Carl von Clausewitz ( ) Management since ~1960s Michael E. Porter Competitive Strategy: Techniques for Analyzing Industries and Competitors Engineering since 2000? target domain: Army Firm System/ Product Olivier L. de Weck, 2006 Page 50

51 Last Slide Engineering Education Teaching Pedagogy Courses Age Dissemination Outlets Eng. Systems Studies Satellite Constellations (exits) Automotive Platforms (new) Oil & Gas Exploration (new) Active Learning City Planning Game (exists) Auto Market Simulator (new) Others (TBD) Future Work Strategic Engineering in additional Industries Comparative and Non-dimensional Analysis - Focus on TDN (U) Eng. Design & Rapid Proto. Graduate Courses Multi Sys Des Opt ESD.71 Eng Sys Analysis SDM Program ESD.34 Sys Architecture ESD.36 Sys Project Mgt OpenCourseWare Engineering Systems Learning Center Others MIT Professional Institute Seminars, Workshops, Olivier L. de Weck, 2006 Page 51

52 Backup Charts Olivier L. de Weck, 2006 Page 52

53 F/A-18 Change Propagation Network Object Process Diagram (OPD) Olivier L. de Weck, 2006 Page 53

54 Local Change to affect crack growth Metal Fatigue Paris Law: da m C K dn = Δ π a Δ K =Δσ πa sec w stress Δσ Center Cracked Panel 2a w=6 Crack length a [inch] σ max Δσ σ min R=0 Critical Load Number of Cycles N Load cycles N [-] x 10 4 Initial Crack Length a o [inch] Isoperformance Curve: Requirement=CGL=Nc: Parameter Bounding Box Performance Jz = Nc =25000 cycles to failure C=4e-9 m= Stress Amplitude Δσ [ksi] Olivier L. de Weck, 2006 Page 54

55 F/A-18 Avionics Suite Olivier L. de Weck, 2006 Page 55

56 Existing Big LEO Systems Iridium Globalstar Time of Launch Number of Sats Constellation Formation polar Walker Altitude (km) Sat. Mass (kg) Transmitter Power (W) Multiple Access Scheme Multi-frequency Time Division Multiple Access Multi-frequency Code Division Multiple Access Individual Iridium Satellite Single Satellite Capacity Global Capacity Cs 1,100 duplex channels 72,600 channels 2,500 duplex channels 120,000 channels Type of Service voice and data voice and data Average Data Rate per Channel 4.8 kbps 2.4/4.8/9.6 kbps Total System Cost $ 5.7 billion $ 3.3 billion Current Status Bankrupt but in operation Bankrupt but in operation Individual Globalstar Satellite Olivier L. de Weck, 2006 Page 56

57 CPF Satellite System Economics = k I T i= 1 Lifecycle cost C s T T i= 1 C ops, i L f, i Number of billable minutes C s = 100,000 [#ch] 6 N u = 310 L f = A = 1, 200 [min/y] u Numerical Example: I = 3 [B$] k = 5 [%] C ops = 300 [M$/y] T =15 [y] CPF I k C ops C s L f N u A u T CPF = 0.20 [$/min] L Cost per function [$/min] Initial investment cost [$] Yearly interest rate [%] Yearly operations cost [$/y] Global instant capacity [#ch] Average load factor [0 1] Number of subscribers Average user activity [min/y] Operational system life [y] f Nu Au = min C 1.0 But with N u = 50,000 CPF = [$/min] Non-competitive Olivier L. de Weck, 2006 Page 57 s

58 Strategic Building Architecture BP Exploration Headquarters, Aberdeen, Scotland Source: J. Fernandez, MIT Olivier L. de Weck, 2006 Page 58

59 Benchmarking Benchmarking is the process of validating a simulation by comparing the predicted response against reality. Number of simultaneous channels of the constellation Benchmarking Result 1: Simultaneous channels of the constellation 140, , ,000 80,000 60,000 40,000 20, Iridium 2 Globalstar Iridium and Globalstar actual or planned simulated Lifecycle cost (billion $) Benchmarking Result 2: Lifecycle cost actual or planned simulated 1 Iridium 2 Globalstar Iridium and Globalstar Benchmarking Result 3: Satellite mass Benchmarking Result 4: Number of satellites in the constellation Satellite mass (kg) 1, , , Iridium Globalstar Orbcomm SkyBridge Iridium, Globalstar, Orbcomm, and SkyBridge actual or planned simulated Number of satellites in the constellation Iridium Globalstar Orbcomm SkyBridge Iridium, Globalstars, Orbcomm, and SkyBridge actual or planned simulated Olivier L. de Weck, 2006 Page 59

60 Platform Leverage Increases Average Vehicle Models per Platform 6 Models/Platform DCX Ford Honda Toyota VW Year Source: Price Waterhouse Coopers, 2003 Olivier L. de Weck, 2006 Page 60