Model-Based Systems Engineering

Transcription

1 Model-Based Systems Engineering Chris Paredis Systems ealization Laboratory Systems ealization Laboratory Product and Systems Lifecycle Management Center G.W. Woodruff School of Mechanical Engineering Georgia Institute of Technology

2 Presentation Overview Context: What is Model-Based Systems Engineering? Why MBSE? A simple model of knowledge and information in MBSE Which kinds of knowledge and information? Information schema Language mappings Abstractions and idealizations Analysis patterns Synthesis patterns Design Workflow Challenges and Summary 2 Systems ealization Laboratory

3 A Systems Approach for Product Development Technology Trends Miniaturization Embedded intelligence Networked connectivity Modularization One subsystem per function Standardized interfaces Systems Approach Increased number of functions New functions by integrating multiple subsystems New architectures (source: DaimlerChrysler) 3 Systems ealization Laboratory

4 Traditional Design Approaches Document-centric, systematic information transformations Generally, information is transferred manually between design steps and team members Output is large set of documents Planning & Task Clarification Conceptual Design Embodiment Design Inadequate for contemporary SE Detail Design 4 Systems ealization Laboratory

5 Challenges in Systems Engineering Multiple integrated functions Multiple engineering disciplines Multiple stakeholders Globally distributed, heterogeneous design teams Complex, emergent system behavior Large quantities of design knowledge and information Need Formal, Model-Based Approach 5 Systems ealization Laboratory

6 Presentation Overview What is Model-Based Systems Engineering? Why MBSE? A simple model of knowledge and information in MBSE Which kinds of knowledge and information? Information schema Language mappings Abstractions and idealizations Analysis patterns Synthesis patterns Design Workflow Challenges and Summary 6 Systems ealization Laboratory

7 Knowledge & Information in Systems Engineering PDM Coarse-grained Suppliers Not Computerinterpretable CAD1 CAD2 FEM Designers Not Interoperable Manufacturing Process Planning Implicit Analysts 7 Systems ealization Laboratory

8 Systems ealization Laboratory 8 A Workflow Perspective Product Perspective Workflow Perspective

9 Knowledge & Information in Systems Engineering Knowledge & Information epositories Designers Product Perspective CAD FEM Analysts Workflow Process Perspective Product Portfolio Planning Execution Perspective GID Knowledge equirements Definition Knowledge Information Information CAD PDM Process Planning Knowledge Information Maintenance & Support Suppliers Manufacturing Infrastructure: Security Notification Communication Visualization 9 Systems ealization Laboratory

10 Information Economics Information is only valuable to the extent that it leads to better decisions epresentation Modeling Computation Cost Benefit Better design decision higher utility Design esearch: Models, Methods, Tools for Shifting the Balance towards Higher Value 10 Systems ealization Laboratory

11 Presentation Overview What is Model-Based Systems Engineering? Why MBSE? A simple model of knowledge and information in MBSE Which kinds of knowledge and information? Information schema Language mappings Abstractions and idealizations Analysis patterns Synthesis patterns Design Workflow Challenges and Summary 11 Systems ealization Laboratory

12 Information Schema Traditionally: tool-specific proprietary schema CAD Finite element modeling Discrete event simulation Few formal representations in SE Standardization Islands of Knowledge & Information e.g.: STEP (ISO 10303) Standard for the Exchange of Product data Intent: educe the number of translators from N 2 to N Always lagging behind Overlap and inconsistencies Not all types of information have been modeled formally. The models are changing over time. 12 Systems ealization Laboratory

13 Example of Standardization: STEP 13 Systems ealization Laboratory

14 Federated Model (ussell Peak et al.) Abstraction Level Customer / Acquisitions Systems Engineering One single master model is impossible / impractical Development Process equirements Structures Electronics Human Interaction Legend Model interfaces: Associativities among domain-specific models & system-level models Fine-grained models: Information objects Parametric relations Domain Evolution Different levels of abstraction elationships must be finegrained Not file-based Models of varying abstractions and domains After Bajaj, Peak, & Waterbury Systems ealization Laboratory

15 SysML as a Federated Model for MBSE Visual information modeling for systems engineering Based on UML 2 Used to support system specification, analysis, design, verification and validation UML 2 SysML 15 Systems ealization Laboratory

16 GIT Testbed: Tools and Models Authoring in native tools Integration in SysML Different levels of abstraction Language Mapping Graph transformations (e.g., Modelica) Direct interface (e.g., Excel) 16 Systems ealization Laboratory

17 elated Work: Model Integration with SysML SysML is intended to unify the diverse modeling languages used by systems engineers. (SysML Specification, 2007) Constrained Objects (Peak et al., 2001, 2005) ModelicaML (Pop et al., 2007) UML H (Nytsch-Geusen, 2007) 17 Systems ealization Laboratory

18 elevant SysML Modeling Constructs Blocks Value types Part properties Value properties Stereotypes bdd Car Definition Car values mass: SI.Mass = 1500 suspension WheelSuspension «satisfy» values «moe» settlingtime: Time «requirement» eboundeq text = When disturbed by 0.1 m, the suspension shall settle to 5% of steady state in under 1 sec. «valuetype» SI.Mass unit = kg coil Coil Shock shock values springate: eal values dampingcoef: eal 18 Systems ealization Laboratory

19 The SysML Metamodel String sysmlimport generalization sysmlelement contain MS work by Tommy Johnson stereotype umlmultiplicity block sysmlpackage multi propassn type property constraintblock valuetype unitassn sysmlunit endassn sysmlconnector value part constraint sysmlparameter constraintproperty 19 Systems ealization Laboratory

20 The Modelica Metamodel modelicaimport composition extends class type modelicaconnector modelicatype modelicapackage componentef component unitassn modelicaunit equation eqnassn size modelicaparameter arraysize connectclause initialequation 20 Systems ealization Laboratory

21 The Correspondence Metamodel SysML Correspondence Modelica block block block2class class class block block2modelicaconnector class modelicaconnector valuetype block valuetype2modelicatype class modelicatype sysmlpackage sysmlpackage sysmlpackage2modelicapackage modelicapackage modelicapackage property property property2component component component property property2modelicaparameter component modelicaparameter constraint constraint constraint2equation equation equation constraint constraint2initialequation equation initialequation 21 sysmlconnector sysmlconnector sysmlconnector2connectclause connectclause connectclause Systems ealization Laboratory

22 Implementation in SD/Eclipse Created Java software that Queries Embedded Plus (E+) SysML CD model Transforms model in VIATA2 framework Generates Modelica code & simulates in Dymola 22 Systems ealization Laboratory

23 Implementation in SD/Eclipse Created Java software that Queries Embedded Plus (E+) SysML CD model Transforms model in VIATA framework Generates Modelica code & simulates in Dymola 23 Systems ealization Laboratory

24 Implementation in SD/Eclipse Created Java software that Queries Embedded Plus (E+) SysML CD model Transforms model in VIATA framework Generates Modelica code & simulates in Dymola 24 Systems ealization Laboratory

25 Implementation in SD/Eclipse Created Java software that Queries Embedded Plus (E+) SysML CD model Transforms model in VIATA framework Generates Modelica code & simulates in Dymola 25 Systems ealization Laboratory

26 Implementation in SD/Eclipse Created Java software that Queries Embedded Plus (E+) SysML CD model Transforms model in VIATA framework Generates Modelica code & simulates in Dymola 26 Systems ealization Laboratory

27 Implementation in SD/Eclipse Created Java software that Queries Embedded Plus (E+) SysML CD model Transforms model in VIATA framework Generates Modelica code & simulates in Dymola 27 Systems ealization Laboratory

28 Implementation in SD/Eclipse Created Java software that Queries Embedded Plus (E+) SysML CD model Transforms model in VIATA framework Generates Modelica code & simulates in Dymola 28 Systems ealization Laboratory

29 The Hydraulically Powered Excavator Example Demonstrates scalability of the integration approach epresents set of over 11,000 equations Depicts CD of earth-moving excavator Model includes hydraulic actuators & control circuitry (Fluid Power Library) rigid-body mechanics (User-defined external model) world reference model (Modelica Standard Library) operator command signals (User-defined external model) 29 Systems ealization Laboratory

30 The Hydraulically Powered Excavator Example Demonstrates scalability of the integration approach epresents set of over 11,000 equations Depicts CD of earth-moving excavator Model includes hydraulic actuators & control circuitry (Fluid Power Library) rigid-body mechanics (User-defined external model) world reference model (Modelica Standard Library) operator command signals (User-defined external model) 30 Systems ealization Laboratory

31 Presentation Overview What is Model-Based Systems Engineering? Why MBSE? A simple model of knowledge and information in MBSE Which kinds of knowledge and information? Information schema Language mappings Abstractions and idealizations Analysis patterns Synthesis patterns Managing Design Workflow Summary 31 Systems ealization Laboratory

32 Federated Model (ussell Peak et al.) Abstraction Level Customer / Acquisitions Systems Engineering Decomposability coupling between detailed models is limited Legend Development Process equirements Structures Electronics Human Interaction Model interfaces: Associativities among domain-specific models & system-level models Fine-grained models: Information objects Parametric relations Domain Abstraction hide the details Models of varying abstractions and domains 32 After Bajaj, Peak, & Waterbury Systems ealization Laboratory

33 Abstraction: Integrating Black Box Models epresents a pre-existing model using SysML constructs Hides details of the model & only exposes important aspects of model The «external» stereotype Black Box Model Modelica-specific system nodes & the «connectclause» stereotype Metadata Only important properties Black Box interface 33 Systems ealization Laboratory

34 MIA Integrating Black Box Models bdd External MSD Definition epresents a pre-existing model using SysML constructs «external» ExternalMSD MSLMechanicalFlange Hides details of the model & only exposes important aspects of model mass «external» spring «external» The «external» MSLSlidingMass stereotype MSLSpring values Modelica-specific m: MSLSIMass system nodes & the url: String = /Modelica 2.2.1/ damper «external» «connectclause» fqn: String = Modelica.Mechanics. stereotype MSLDamper Translational.SlidingMass Black Box Model mime: String = model/modelica parts flange_a: MSLMechanicalFlange flange_b: MSLMechanicalFlange Metadata ground Only important properties «external» MSLFixed 34 Black Box interface Systems ealization Laboratory

35 MIA Integrating Black Box Models par ExternalMSD epresents a pre-existing model using SysML flange_a: MSLMechanicalFlange constructs Hides details of the model & only exposes important «connectclause» «connectclause» aspects of model The «external» stereotype spring: MSLSpring mass: MSLSlidingMass flange_b: MSLMechanicalFlange damper: MSLDamper Modelica-specific flange_a: MSLMechanicalFlange system nodes & the «connectclause» stereotype flange_b: MSLMechanicalFlange Black Box Model «connectclause» Metadata Only important properties ground: MSLFixed flange_b: MSLMechanicalFlange flange_a: MSLMechanicalFlange flange_b: MSLMechanicalFlange «connectclause» 35 Black Box interface Systems ealization Laboratory

36 Presentation Overview What is Model-Based Systems Engineering? Why MBSE? A simple model of knowledge and information in MBSE Which kinds of knowledge and information? Information schema Language mappings Abstractions and idealizations Analysis patterns Synthesis patterns Managing Design Workflow Summary 36 Systems ealization Laboratory

37 Idealizations Work by Fenves, Choi, et al. 37 Systems ealization Laboratory

38 Idealizations Different from just abstraction: Additional knowledge is necessary Appropriate idealization is context dependent Most common for finite element modeling: Transform design model into corresponding analysis model Modify geometry Eliminate details Impose symmetry Determine dimensionality Define materials model Model loading conditions Generate mesh 38 Systems ealization Laboratory

39 Methods for Generating Finite Element Models Direct FEA model creation Import design from CAD tool Modify geometry (as per idealizations) Select elements to mesh (as per idealizations) Mesh Script-based FEA model structure creation Author a solver-specific script Idealized geometry, material properties, load, boundary (and/or initial conditions) Element, mesh size, un script for different instances 39 Systems ealization Laboratory

40 Methods for Generating Finite Element Models Multi-representation architecture-based method 4 stepping-stone model structures Models related by parametric constraints (per idealization) static Analysis model = assembly of reusable building blocks Solution-approach and method specific analysis models (Peak, Fulton 1998) 40 Systems ealization Laboratory

41 Transform Design Model into Analysis Model 41 Systems ealization Laboratory

42 Presentation Overview What is Model-Based Systems Engineering? Why MBSE? A simple model of knowledge and information in MBSE Which kinds of knowledge and information? Information schema Language mappings Abstractions and idealizations Analysis patterns Synthesis patterns Design Workflow Challenges and Summary 42 Systems ealization Laboratory

43 Scope of the Problem Behavior Aspect Modeling Simulation Cost Aspect System Architectures Modeling Simulation Evaluation of Preferences Various Topologies eliability Aspect Multi-Attribute Utility Theory Modeling Simulation Can we reduce the cost of modeling through model composition? 43 Systems ealization Laboratory

44 Analysis Patterns System is composition/configuration of subsystems or components The subsystems and components are reusable building blocks We need to predict the behavior of every configuration considered in design System Configuration Component Component Model Component Model Model Composition System Model 44 Systems ealization Laboratory

45 SysML Internal Block Diagram Composition of Behavior Model Corresponding System-level Dynamic Model System Architecture 45 Corresponding Algebraic Model slidingmass.vel = vmax; slidingmass.vel = cylinder.piston.vel; cylinder.porta.q = cylinder.piston.vel*cylinder.area; cylinder.porta.q = pump.portp.q; pump.shaft.omega*pump.displ / (2*pi) = pump.portp.q; motor.shaft.omega = pump.shaft.omega; Systems ealization Laboratory

46 Multi-Aspect Component Models MAsCoMs Define relationships between all the models related to a particular component or subsystem Model context Aspects Discipline Lifecycle phase Discretization Time Space Math Formalism epresentation Syntax elief Model Valve FP component Check Model Const Displ. Model Taxonomy of Components Pump Var Displ. Model Model Model Model Hydraulic Thermal Algebraic DAE Multi-Aspect Component Model Taxonomy of Aspects Discipline Formalism Behavior 46 Systems ealization Laboratory

47 Composition Process MAsCoM epository Automated Composition System Models Aspect C Aspect B Aspect A Search & Optimization 47 Systems ealization Laboratory

48 Multi-Aspect Component Models in SysML Composition of reliability models Characterization of analysis model of a component 48 Systems ealization Laboratory

49 Presentation Overview What is Model-Based Systems Engineering? Why MBSE? A simple model of knowledge and information in MBSE Which kinds of knowledge and information? Information schema Language mappings Abstractions and idealizations Analysis patterns Synthesis patterns Design Workflow Challenges and Summary 49 Systems ealization Laboratory

50 System Synthesis Patterns Electric or IC prime mover Voltage- Current PM Torque 50 Pump Flow Press. Transmission line & valve Flow Press. The system consists of a series of transformation of power variables Power is either converted to another useful form or waste heat Impedance is modified (unit force/unit flow) Power is controlled Function is achieved Motor or cylinder PM-Torque or Velocity- Force Coupling mechanism PM- Torque Load Systems ealization Laboratory or Velocity- Force

51 Different Architectures Key issues: Hardware + Controllers Compatibility constraints Hidden dependencies Best solution depends on the preferences of the designer Each concept is a large set of design alternatives 51 Systems ealization Laboratory

52 Graph Grammars Capture application domain knowledge in a formal grammar Grammar describes plausible configurations Ideally optimal configurations Even feasible configurations may too complex to define in grammar Explore design space by applying transformation rules in some randomized fashion Which knowledge to model in terms of transformations? Which knowledge to model in terms of analyses? 52 Systems ealization Laboratory

53 Capturing Fluid Power Knowledge in Graph Transformations equirements & Objectives SysML system alternative ibd [Block] Hydraulic_Subsystem[ Schematic ] pump : FDpump discharge : FlowPort inputshaft : FlowPort housing : FlowPort suction : FlowPort tank-to-pump : Line b : FlowPort a : FlowPort pump-to-valve : Line a : FlowPort b : FlowPort valve : 4port3wayServoValve portp : FlowPort portt : FlowPort cylb : FlowPort cyla : FlowPort Topology Generation using Graph Transf System Alternatives MAsCoMs Model Composition using Graph Transf SysML tank : Tank sump : FlowPort return : FlowPort filter-to-tank : Line b : FlowPort a : FlowPort filter : Filter in : FlowPort out : FlowPort valve-to-filter : Line a : FlowPort b : FlowPort actuator : Double-ActingCylinder housing : FlowPort rod : FlowPort b : FlowPort a : FlowPort valve-to-cylp1 : Line a : FlowPort b : FlowPort valve-to-cylp2 : Line a : FlowPort b : FlowPort System Behavior Models Executable Simulations SysML Model Translation using Graph Transf Dymola Dig Cycle Traj Sw ing Boom Arm Bucket hydraulics behavior model y world x simulation configuration Simulation Configuration using Graph Transf Design Optimization 53 ModelCenter Systems ealization Laboratory

54 Design Workflow Ideally all models would always be consistent In practice, changes need to be propagated carefully e.g., large parametric assembly model, ECAD-MCAD integration Traceability is crucial, but automated propagation is not desirable Propagation is controlled at decision points Human decision-maker needs to remain in the loop But in the end, all models should be consistent How does one model the application of transformations? 54 Systems ealization Laboratory

55 Challenges and Summary Evolving representations maintenance of transformations, schema, instances Maintaining multiple views bi-directional Modeling idealizations as graph transformations Capturing knowledge context of synthesis rules Hidden dependencies system-level interfaces are only models Workflow and consistency management 55 Systems ealization Laboratory

56 Acknowledgments Graduate Students Manas Bajaj Jonathan Jobe Tommy Johnson Aleks Kerzhner ichard Malak oxanne Moore Stephanie Thompson Lina Tucker Collaborators oger Burkhart (Deere) Sandy Friedenthal (LMCO) Leon McGinnis (GT) ussell Peak (GT) Sponsors: National Science Foundation Grant DMI National Science Foundation, Center for Complex and Efficient Fluid Power Deere & Co. Lockheed Martin Questions? 56 Systems ealization Laboratory