CYBER-INFRASTRUCTURE SUPPORT FOR ENGINEERING DESIGN

Transcription

1 CYBER-INFRASTRUCTURE SUPPORT FOR ENGINEERING DESIGN Perspectives from NSF ED2030 Workshop + + Jami J. Shah Mechanical & Aerospace Engineering, Arizona State University, Tempe 1

2 Industry representation Boeing Commercial, Boeing Defense, Intel, General Motors, John Deere, Ford, HP, Dassault, CPD Associates, IMTI Government agencies NSF, NASA, NIST Disciplines represented Mechanical, Aerospace, Civil, Industrial, EE, Computer Science, Chemical, Materials, Bio, Industrial Design Participants 2

3 Workshop Organization Pre-conference position papers posted on-line on-line chat for discussion and comments Plenary sessions understand the scope & objectives review current state of eng. design industry and technology forecasts Breakout sessions four focus groups, working in progressive stages: Ideation & design innovation Theoretical foundations of design Social aspects of design IT & computational tools for design 3

4 Industry forecasts Forecasts were delivered for the following sectors: - Aviation industry (Boeing) - Space industry (NASA JPL) - Automotive industry (GM) - Consumer electronics (HP) - CAD/CAM/PLM (CPD Associates) - Virtual prototyping (John Deere) Industry was reluctant to provide predictions beyond 10 years (except for space industry) Details can be found at: 4

5 Industry forecasts: common themes New developments in biology, bio-engineering, computers, etc. will directly affect what we design (products) and how we design (process, tools) Future products will be typified by complex structure and function but simpler usage (interface). Consumer products will be customized for each user customers will have direct involvement in product design Computer models will be embedded in delivered products for customization and soft reconfiguration Design teams, manufacturing resources and customers will be globally distributed Designs will be constrained by new and stricter regulations related to: emissions, noise, recycling, energy, global warming, terrorism Designs will be driven by public demand for: higher reliability, better safety Designs will be driven by the desire to seek products least vulnerable to international competition New markets in developing countries will offer new opportunities but require consideration of greater cultural and economic diversity Design services will also be outsourced overseas; US companies will do system of system integration 5

6 Vision of Computing Environments for ED 2030 In 2030 designers, distributed across disciplinary, geographical and temporal boundaries, will work synergistically within design environments that enable designers to focus on design issues without being distracted by the underlying computing infrastructure encompass all stages of the product life cycle operate at multiple levels of abstraction enhance the creativity of design teams leading towards innovation-guided design are capable of spawning tasks transparently; automatically refining designs or providing support, executing integrated multi-scale simulations/analyses facilitate communications with other human designers, web services, repositories, catalogs, and digital libraries as needed. enable the generation of solutions based on relationships among the design requirements and their physical actualization and context of use interact with designers in human terms and language; no distinction in humanhuman, human-computer, computer-computer communications. Future design tools and methods will not only support analysis and decision making from a technological point of view, but will also account for psychological, sociological, and anthropological factors based on fundamental understanding of these factors and their interaction. Boundaries between resources, people, and information/knowledge will disappear. 6

7 Major Recommendations at a glance Engineering innovation: Create a Division level initiative in developing the Science of engineering innovation directed at a fundamental understanding of the nature of engineering innovation to create new technologies, products, processes and services. Social-technical aspects: Create an agency wide, multidisciplinary program in the socio-technical aspects of engineering design; Basic knowledge regarding how humans and social dynamics influence product specifications, design process and decisions with multiple stakeholders. Design informatics: Focus the current Engineering Design Program on critical research areas listed above and theoretically founded design modeling, simulation and automation environments that represent functional intent, design choices, prediction of behavior. 7

8 Why invest in Cyber-infrastructure for Eng. Design? Engineering design is a discovery process that requires: exploration to develop a clear understanding of the problem, creativity and innovative skills to synthesize and predict the future outcome of a design. Experiential and analytical knowledge, methods, tools to predict and optimize the performance of a design past experience and access to past designs, catalogs to re-use past designs or to avoid past mistakes But The knowledge used during design (rationale) is currently not captured in any consistent form that it can be reused other then by the person who recorded it, if any records were kept at all Due to time and budgetary constraints exploration of design alternatives is often curtailed Poor integration of design & analysis tools and between disciplines, less time spent on optimization As products become more complex, traditional engineering methods, islands of automation and trial & error will not be sufficient Need computing infrastructure, computational tools and knowledge bases for Predictive Product Realization 8

9 Predictive Product Realization Progression of design and manufacturing towards predictive product realization needs to take place along two directions: enhancing our state of knowledge and increasing the level of automation. Better understanding of physical phenomenon will reduce trial and error in product realization. Increasing the level of automation will allow more alternatives to explored and more simulations to be carried out in a given amount of time. Will enable predicting the desired performance in future products at known cost. 9

10 ENGINEERING DESIGN GENERIC PROCESS PROBLEM DEFINITION Design Objectives Tech specifications, QFD EMBODIMENT DESIGN Identification of key parameters Material selection Geometric & parametric design Engineering analyses Trade-off studies Economic analysis; Optimization PROTOTYPING & TESTING CONCEPTUAL DESIGN Data collection, research Functional analysis Concept generation Concept evaluation, selection DETAILED DESIGN Detailed layout CAD models, eng. drawings Tolerance analysis DfX Very little computational support for CAD tools for conceptual design Design history, rationale poorly archived Poor inter-operability between tools for embodiment and detailed design 10

11 The CAD Technology Gap HAVE WANT Geometry Physics Embodiment design Conceptual & functional Design Geometric model construction history support for capturing design intent, rationale, history Geometric constraint solvers Multi-modal solvers Disconnected use Shared use Generative design Generative + variant Std. part geometry & features catalogs Design Repositories User driven Knowledge driven Reactive, iterative Pro-active Limited customizability Customizable to domain, Company, designer CAD companies have very little R&D capacity CAD systems have always operated on Technology PUSH Most critical information is not archived,or not in digital format 11

12 The CAE/Simulation Technology Gap HAVE Long cycle times, labor intensive Requires specialists in the loop Analytically based CAE tools Hi-fidelity model Serial process Deterministic WANT Mesh-less analysis or intelligent tools to automate pre-process Usable by designers Analytical + experiential Multiple fidelity models Parallel process Stochastic Discipline specific analyses, conducted by specialists Each CAE tool requires different type of input Enormous amount of time spent in model preparation 12

13 The PLM/EIM Technology Gap HAVE PDM systems manage files Physical file exchange Not all CAD data can be exchanged today Limited inter-operability between CAE systems Coarse level of granularity; data exchange limited to low level data No constraints, relations between data in different applications WANT PDM systems must manage data inside the files Shared databases, knowledge bases Complete exchange of all product data Standards need to be extended to engineering analyses Multi level (adjustable) level of granularity Multi-domain constraint management, propagation mechanism Typically, companies have focused on managing Released designs Most design information is not even in digital format 13

14 Design Informatics: Research Engineering informatics: flexible, extensible ontologies for devices, functions, behavior, embodiment mechanisms to inter-relate ontologies information exchange standards information management at meaningful levels of granularity building blocks & standardized APIs to repositories and libraries Models to support Smart tools: knowledge capture/re-use, smart search engines Knowledge Representation for Design Synthesis fast approximate models to support preliminary design exploration tools for analysis with incomplete information/uncertainty Conjoint exploration of the Requirement Space and Solution Space Multi-scale models for shallow and-or deep granular searching Sophisticated knowledge representation structures to bridge CAD. CAE, AI 14

15 Design Informatics: Research - II Collaboration infrastructure multi-disciplinary synthesis and distributed collaboration organizational modeling; agents architecture intelligent constraint management, conflict resolution development and integration of multi-physics simulation tools with design synthesis tools for Virtual Prototyping Design Decision Support: Ability to make decisions at all phases of the design process: varying levels of abstraction, information completeness and uncertainty Methods appropriate for distributed, progressive, iterative processes with multiple decision makers and multiple disciplines 15

16 ED2030 relation to Atkins Report Atkins Report identified five types of services that are also relevant to ED: High performance computation services: Needed for multi-physics simulation, Data, info, knowledge management services: Design repositories with associated search engines for finding past similar designs Digital libraries for technical documents, books, design procedures, tutorials, algorithms Catalogs of standard parts, devices, materials for sale Design history, evolution archival Advanced PLM/EIM systems for Measurement, fabrication services: needed for product realization Interfaces, visualization services: Advanced design environments; immersive environements; exchange stds. Collaboration services Analysis/Consulting services for specialized engineering analyses Collaboration tools; infrastructure 16

17 Research Challenges How can we create tools (and representations) that: adaptively represent relevant physics at appropriate scale? integrate multi-domain knowledge in design libraries capable of supporting single and group users? help designers understand and manage design complexity? integrate the science and art of design have components that are interoperable and plug-able deliver the information at the right level of abstraction for all design activities? capture evolution and rationale of design? integrate analytical & experiential knowledge enable the access, reuse and adaptation of prior designs and processes? can represent a variety of types of requirements (e.g., qualitative and quantitative)? enable methods for discovering, representing, capturing knowledge? provide multiple views of information based on who is using it and when it is used? 17

18 Acknowledgements This report has freely used material from attendees of ED2030, particularly the Focus Group Leaders: Paul Wright, UC Berkeley Larry Gutkowski, HP Larry Leifer, Stanford Stephen Lu, USC David Trees, John Deere Carolina Cruz, Iowa State J. Vandenbrande, Boeing Co. Jon Cagan, CMU Joe Dondelinger, General Motors Susan Finger, Carnegie Mellon Supported by NSF DMII Website: