Fundamentals of Systems Engineering. Lecture 6 Design Definition and Multidisciplinary Design Optimization

Transcription

1 October 16, Fundamentals of Systems Engineering Lecture 6 Design Definition and Multidisciplinary Design Optimization Maj. Jeremy Agte

2 V-Model Oct 16, 2009 Stakeholder Analysis Requirements Definition System Architecture Concept Generation Systems Engineering Overview Cost and Schedule Management Verification and Validation Lifecycle Management Commissioning Operations Human Factors Tradespace Exploration Concept Selection System Integration Interface Management Design Definition Multidisciplinary Optimization System Safety

3 Outline NASA Design Definition Process Process Overview Slides from Lecture 5b on Stellar Multidisciplinary Design Optimization What it is and where it fits in Motivation Roots Method Limitations Future Trends

4 Design Solution Definition Process The Design Solution Definition Process is used to translate the outputs of the Logical Decomposition Process into a design solution definition Design Solution Definition: the specification of a rocket, for example Figure Example of a PBS

5 Design Solution Importance Define solution space What we wanted Develop design alternatives This image of a space shuttle has been removed due to copyright restrictions. Trade studies to analyze Alternate Design Cost, performance, schedule Select Design Solution Drive down to lower level Identify enabling products What we got This image of a space shuttle in a building complex has been removed due to copyright restrictions. Design with the end in mind!

6 Design Solution Definition Best Practice Process Flow Diagram Activities Output Input

7 Design Solution Definition Methods Analyzing each alternative to be able to select the preferred alternative Once an alternative is selected or baselined, the Design Solution Definition Process will be used for: Generating end products as a function of the hierarchy in the system structure The output end product Design Solution Definition will be used for conducting product verification

8 Design Solution Definition Important Design Considerations Capabilities Functions Priorities Reliability Maintainability Supportability System Performance System Availability Technical Effectiveness System Effectiveness Other Considerations Software System Safety Accessibility Information Assurance COTS Disposal Human Factors Environ. Constraints Producibility Operations Maintenance Logistics Process Efficiency Affordable Operational Effectiveness Life Cycle Cost/Total Ownership Cost

9 Design Solution Definition Summary The Design Solution Definition Process is used to translate the outputs of the Logical Decomposition Process into a design solution definition Alternative design solutions must be defined and analyzed to select the best alternative that satisfies technical requirements The form of the design solution definition depends on the product life cycle phase and its hierarchy in the system The Technical Data Package allows for the building, coding, reusing, or buying of products

10 Outline NASA Design Definition Process Process Overview Slides from Lecture 5b on Stellar Multidisciplinary Design Optimization What it is and where it fits in Roots Motivation Method Limitations Future Trends Some MDO Applications

11 Multidisclinary Design Optimization (MDO) What it is and where it fits in MDO defined as (AIAA MDO Tech Committee): an evolving methodology, i.e. a body of methods, techniques, algorithms, and related application practices, for design of engineering systems couple by physical phenomena and involving many interacting subsystems and parts. Emphasis: 1) still evolving, 2) involves many disciplines Conceptual Components of MDO (Sobieksi 97) Mathematical Modeling of a System Design Oriented Analysis Approximation Concepts System Sensitivity Analysis (less so w/ some methods) Classical Optimization Procedures Human Interface

12 Multidisclinary Design Optimization (MDO) What it is and where it fits in Conceptual Modeling & Optimization metric 2 metric 1 CDR 4/15 Optimal Design Decision Making Need Qualitative Quantitative SRR 2/12 Concept Synthesis Concept Screening Concept Selection SCR 3/19 Design Modeling & Optimization System Design metric 3 x 1 * variable x 1 Source: NASA Langley System Architecture System Design

13 MDO - Roots Topic MDO Early Years Schmit's 3 bar truss M Gen opt codes appear (Aesop, CONMIN) LaRC 1st MDO SST papers LaRC IPAD project LaRC AOO & MDOB & IRO Government-Sponsored MDO LaRC SST MDO project ARC ACSYNT & Applications EU MOB NATO AGARD, RTO M M M Theory, Methods and Frameworks, Tools and Companies Excel M Matlab M Mathematica M Integration VRD Integration Engineous Integration ALTAIR Genesis Integration Phoenix Concurrent Computing Linear decomp. M Opt Sensit M System Sensit M Approximations Approximation based decomp. Analytical Target Cascading (Michigan) Collaborative Optimization (Stanford) BLISS-LaRC CSSO-LaRC ND Visualization UofBuff Commercialization BLISS M Genetic Algorithms Optimality criteria (KKT) NASA Glenn NPSS Physical Programming (RPI) Isoperformance (MIT) MDO roots found in structural optimization Optimization algrthms in mainstream prgms More complex decomposition techniques appear Commercialization of multi-level algorithm

14 MDO - Motivation Design systems are very complex and interconnected Structures Aerodynamics Propulsion Data Exchange Performance Control Everything affects everything else Source: NASA Langley

15 MDO - Motivation Simple example of interdependency Range (R) is the system objective Wing - structure P Loads Wing - aerodynamics P a = sweep angle Structure influences R: directly by weight indirectly by stiffness that affect displacements that affect drag a Displacements Loads & Displacements must be consistent R = (k/drag) LOG [( W o + W s + W f )/ (W o + W s )] Source: NASA Langley What to optimize the structure for? Lightness? Displacements = 1/Stiffness? An optimal mix of the two?

16 MDO - Motivation Today s design systems are subject to very stringent contraints Cost: the more complex the system is -> the more expensive Time: USAF (field as fast as possible), NASA (replace aging space systems) Environment: Aircraft noise restrictions, Auto exhaust, Improve fuel efficiency (minimum weight, max efficient engines) Further constrained in that these systems require large groups of people with a broad range of expertise Design teams divided into specialty groups These teams generally geographically seperated Teams prefer to optimize (design) in own expertise domain But their domains remain coupled by data exchange

17 MDO - Motivation Fusclage Group Electrical Group Equipment Group Controls Group Power Plant Group Aerodynamics Group Hydraulics Group Loft Group Production Engineering Group Stress Group MDO helps us get from this and still fully utilize all of today s modern computational tools. Image by MIT OpenCourseWare. to this

18 MDO - Method (1) Define overall system requirements (2) Define design vector x, objective J and constraints g, h (3) System decomposition into modules (4) Modeling of physics via governing equations at the module level - module execution in isolation (5) Model integration into an overall system simulation (6) Benchmarking of model with respect to a known system from past experience, if available (7) Design space exploration (DOE) to find sensitive and important design variables x i (8) Formal optimization to find min J(x) (9) Post-optimality analysis to explore sensitivity and tradeoffs: sensitivity analysis, approximation methods, isoperformance, include uncertainty

19 MDO Method: Bi-Level Integrated System Synthesis M, h t/c, h, M, AR W, Λ, S REF,S HT,AR M, h HT Propulsion X loc ={T} Y* D W BE Y^ Y^ h,m,,ar HT,S HT AR W,S REF,Λ,t/c ESF Y* Aerodynamics X sh - Variables X loc ={Λ HT,L W,L HT } Y* Y^ W T,Θ η H,C DMIN,M<1 Y^ t/c,s HT,AR W Λ,S REF,AR HT W E Structures X loc ={[t],[t s ], λ} W FO,W O N Z W FO, W O, N Z, W BE, C DMIN,M<1,η H Constants L Y* Y^ SFC L/D W T,W F Range R Y* Y^ T-throttle Λ HT - tail sweep L W -wing mom. arm L HT -tail mom. arm [t]-thickness array, size 1x9 [t S ]-thickness array, size 1x9 λ-taper ratio D-drag ESF-eng. scale fact. L-lift N Z -max. load fact. R-range SFC-spec. fuel cons. Θ-wing twist W E -engine weight W F -fuel weight W T -total weight AR W - wing aspect ratio AR HT - tail aspect ratio h-altitude M-Mach # S REF -wing surf. area S HT -tail surf. area t/c-thickness/chord Λ W -wing sweep } X LOC Y } X SH

20 MDO Method: Bi-Level Integrated System Synthesis Formulation of Design System: Supersonic Business Jet Example X sh -design variable shared by at least two subsystems X loc -design variable unique to a specific subsystem X loc -L HT,L W,Λ HT Aerodynamics X sh t/c, h, M, Λ, S, AR X loc -T Propulsion Y*-coupling variable input to particular subsystem Θ* L^ Y`s Y`s Y`s Y^-coupling variable output from a particular subsystem Θ^ L * Y`s Structures Range X loc -[t],[t s ],λ

21 MDO Method: Bi-Level Integrated System Synthesis Subsystem Optimization (SSOPT) AR θ w X loc L W,L HT,Λ HT Aerodynamics L D L/D SSOPT Formulation Given: Q = {[X sh ],[Y*],[w]}, minimize: f ( w, Y^(X loc, X sh, Y*)) by varying: [X loc ]. Satisfy: g(x loc ) 0 h(x loc ) = 0 and [X loc,lb ] [X loc ] [X loc,ub ], and retrieve: [X loc ] and [Y^] at optimum f = w 1 Y^1 + w 2 Y^2 + w 3 Y^3 = n i= 1 w Y ^ i i where n = # of Y^ outputs

22 MDO Method: Bi-Level Integrated System Synthesis Subsystem Optimization (SSOPT) X SH Y * w f = w 1 Y^1 + w 2 Y^2 + w 3 Y^3 = n i= 1 w Y ^ i i where n = # of Y^ outputs Have series of approximation models, one for each Y^ output

23 MDO Method: Bi-Level Integrated System Synthesis Subsystem Optimization (SSOPT) SubSys 1 System-Level Optimization These make up an approximated subsystem which is then sent to the system-level optimization

24 MDO Method: Bi-Level Integrated System Synthesis System Optimization (SOPT) Y * Y^ Y^ Y * X sh, Y*, w SubSys 1 SubSys 2 Y`s Y`s Y`s Y`s SubSys 3 SubSys 4 SOPT Formulation Given: approximation models for optimized subsystem outputs, minimize: F (X sh, Y*, w), by varying: Q = {[X sh ],[Y*],[w]}. Satisfy: c = [Y*]-[Y^] = 0, [X sh,lb ] [X sh ] [X sh,ub ], [Y* LB ] [Y*] [Y* UB ], and [w LB ] [w] [w UB ], and retrieve: [X sh ],[Y*],[w], and F at optimum SOPT Objective Function ^ F = Y o Y^o

25 BLISS Cycle #

26 BLISS Cycle #

27 MDO - Challenges Fidelity/expense of disciplinary models Fidelity is often sacrificed to obtain models with short computation times. Complexity Design variables, constraints and model interfaces must be managed carefully. Communication The user interface is often very unfriendly and it can be difficult to change problem parameters. Flexibility It is easy for an MDO tool to become very specialized and only valid for one particular problem. How do we prevent MDO codes from becoming complex, highly specialized tools which are used by a single person (often the developer!) for a single problem?

28 MDO - Challenges Fidelity vs. Expense High fidelity (e.g. CFD,FEM) can we do better? how to implement? Intermediate fidelity (e.g. vortex lattice, beam theory) Empirical models Fidelity Level Level of MDO Increasing difficulty can the results be believed? Trade studies Limited optimization/iteration Full MDO Image by MIT OpenCourseWare. from Giesing,

29 MDO - Challenges Breadth vs. Depth High fidelity (e.g. CFD,FEM) is design practical? how to implement? Intermediate fidelity (e.g. vortex lattice, beam theory) Empirical relations Disciplinary Depth Increasing difficulty System Breadth can the results be believed? Focus on a subsystem All critical constraints Complete system Image by MIT OpenCourseWare

30 MDO Future Trends Vertical Growth: Tread New Grounds Open up new areas e.g. Product family optimization Adaptation to flexible requirements Dream Growth: Qualitatively New and Powerful Capabilities Horizontal growth: More of the Same Dream Growth Utopian Long Term Deepen current areas Increase dimensionality Extend applications

31 MDO - Summary Multi-Disciplinary Optimization is a rapidly developing engineering discipline Still evolving Requires a paradigm shift from traditional design practices Several methods are available for use in MDO Choice of method depends on design problem Nearly all can still be further improved Newfound applications for MDO are multiplying quickly Modeling the design system often the significant challenge Rapid increases in computing power greatly enhance MDO capability Future developments are directed in two orthogonal, but mutually reinforcing directions Horizontal for more capability Vertical for new and innovative solutions

32 MIT OpenCourseWare Fundamentals of Systems Engineering Fall 2009 For information about citing these materials or our Terms of Use, visit: