Java程序辅导

Knowledge Management in 
Simulation-based Design
Manas Bajaj
(gtg628c@prism.gatech.edu)
Georgia Tech
Copyright © 1993-2005 by Georgia Tech Research Corporation, Atlanta, Georgia 30332-0415 USA.  All Rights Reserved.
Developed by eislab.gatech.edu.  Permission to use for non-commercial purposes is hereby granted provided this notice is included.
SRL Weekly Meeting
MaRC 201
2nd Aug, 2005
Systems Realization Laboratory
2Georgia Tech  ♦ Engineering Information Systems Lab ♦ eislab.gatech.edu© 1993-2001 GTRC
Simulation-based Design
‹ Simulation-based Design is a process in which 
simulation is the primary means of supporting 
design synthesis and evaluation
‹ Ryder’s 6 steps to analysis problem solving
‹ Classes of Analysis Problems: 
Original, Adaptive, Routinized
Formulation Evaluation
Analysis
Model Î Surrogate of a real object
3Georgia Tech  ♦ Engineering Information Systems Lab ♦ eislab.gatech.edu© 1993-2001 GTRC
Classes of Analysis Problems [Peak et al.]
Analysis Class
Aspect Example Original Adpative Routinized
Inputs
Design Families* Multi-layer PWBs 9 9
Design Instances PWB #95145 Several Several One - Many
Design Variations Re-order layer stackup Several - Many Several - Many Several
Solution Method
Develop new method 9
Use established method 9 9
Analysis Procedure
Develop procedure PWB warpage analysis 9 9
Define analysis criteria Tmax = avg T of chip 9 9
Define idealizations, Γ: 9 9
Boundary conditions Uniform temperature
Analytical body types Plane strain body
Geometric simplifications Total thickness
Material models Linear elastic
Validate procedure 9 9
Measure samples Shadow moire'
Correlate with measurements
Use established procedure IPC-D-279 PTH fatigue 9
Outputs
Validated solution method 9
Validated analysis procedure 9 9
Sensitivity studies 9 9
Example datasets 9 9
Analysis results & design impact 9 9 9
Who Senior Analyst Analyst Designer
Focus R&D R&D Regular Usage
* Design = product or process
FEM, FDM, …
4Georgia Tech  ♦ Engineering Information Systems Lab ♦ eislab.gatech.edu© 1993-2001 GTRC
Research Issues
‹ Current State-of-the-Art in SBD
– Establishing relationships between design and 
analysis-related models
» applicable only to routinized analysis
– Lack of a formal methodology to use domain knowledge 
to DESIGN analysis-related models
» Ad-Hoc at the detailed level
» Formalized at a very abstract level
– Approaches:
» Composable Simulation, Multi-Representation 
Architecture, Multi-Aspect Information Structure, STEP 
Information Exchange, Master Product Models,…
5Georgia Tech  ♦ Engineering Information Systems Lab ♦ eislab.gatech.edu© 1993-2001 GTRC
My Research Contribution
Adaptive and Original Analyses (focus); Routinized Analyses (supported)
‹ Context
Managing product related knowledge during SBD
‹ Research Gap
- Configuring and connecting product design and analysis-related models is  primarily 
suitable for routinized analysis
- Analysis-model design is formalized only at a very abstract level and ad-hoc at the 
detailed level
‹ Primary Research Question
How can we design / compose analysis-related models from a given product design 
model for original / variant analysis?
‹ Primary Hypothesis
By formalizing the model states and transformations in SBD using the paradigm of 
graph-based meta-representation and graph grammars, we can design 
analysis-related models for supporting original and variant analysis
Analysis Problem 
Statement Solution Approach
Analytical Product 
Representation and 
Transformations
6Georgia Tech  ♦ Engineering Information Systems Lab ♦ eislab.gatech.edu© 1993-2001 GTRC
My Vision
Fancy….but exciting !
Product Design
Process Design
Knowledge 
Representation Design
Focus on Design States
Focus on designing the path to achieve design states
Focus on designing the knowledge models to represent 
and evaluate products and process alternatives
+
Models representing all states and 
abstractions of a product
Transformations between model 
states (process-centric)
7Georgia Tech  ♦ Engineering Information Systems Lab ♦ eislab.gatech.edu© 1993-2001 GTRC
R
R
R
Information-Driven Product Development
R
R
R
R
R
DesignersSuppliers
R
R
R
R
R
R
R
R
R
R
R
R
R R
R
R
R
R
Manufacturing
AnalystsImplicit
Not Computer-
interpretable
Not 
Interoperable
Coarse-grainedPDM
CAD1
CAD2
FEM
Process
Planning
8Georgia Tech  ♦ Engineering Information Systems Lab ♦ eislab.gatech.edu© 1993-2001 GTRC
My Research Contribution (cont.)
Direct Impact of 
my Research
9Georgia Tech  ♦ Engineering Information Systems Lab ♦ eislab.gatech.edu© 1993-2001 GTRC
Items of Acquaintance
‹ Domains
– SBD
– Knowledge Representation
– Open Standards (ISO STEP, OMG SysML, OMG UML,…)
‹ Tools
– Open-Standard Tools
» STEP-Book, IDA-STEP, JSDAI, Artisan (UML / SysML), …
– Knowledge Modeling Tools
» Protégé, CLIPS, JESS
– Programming
» ECLIPSE (IDE for Java, C++, XML, UML, SysML,…); 912 plugins so far
– Notes / e-binder
» MS OneNote
– Versioning
» CVS
– Issue Resolution
» Bugzilla
Nsikan’s Research
Overview
SRL Weekly Meeting
MaRC 201
2nd Aug, 2005
11Georgia Tech  ♦ Engineering Information Systems Lab ♦ eislab.gatech.edu© 1993-2001 GTRC
Overview
‹ Problem
– Semantic retrieval methods are needed to support retrieval of 
archived parametric FEA models and supporting information from 
large online repositories
‹ Research Issues 
– Representation
» Heterogeneity
» Expressivity
– Retrieval algorithms
» Precision
» Computational efficiency
» Formalizing relevance determination
12Georgia Tech  ♦ Engineering Information Systems Lab ♦ eislab.gatech.edu© 1993-2001 GTRC
Common Scenario
‹ Reuse of FEA model as reference model for new 
variation of an existing case due to
– Need for data about new components
– Need for additional details
- Search in which no exact match 
to the physical problem faced 
exists in archive
- Existing FEA model is iteratively
adapted
A
B
13Georgia Tech  ♦ Engineering Information Systems Lab ♦ eislab.gatech.edu© 1993-2001 GTRC
Concerns
‹ Impact of addition of new information on modeling strategy
– New implementation tasks generated 
‹ Amount of reusable data
– Additional searches required for missing data
‹ Relevance of model’s assumptions to new context
‹ Calls for development of semantic retrieval method, but
– What needs to be captured? How?
– How should these representations be manipulated?
14Georgia Tech  ♦ Engineering Information Systems Lab ♦ eislab.gatech.edu© 1993-2001 GTRC
In Conclusion
‹ Contribution 
– Semantic retrieval methodology
– Assess usefulness of description logic inference for semantic retrieval of FEA 
models
‹ Skills 
– Software
» Matlab, CATIA
– Programming
» C, Java, SVG, XML, Knowledge-based 
– Other
» Description Logic
‹ Similar folks
– Sungshik, Greg, Manas
Sungshik’s Research
Overview
SRL Weekly Meeting
MaRC 201
2nd Aug, 2005
16Georgia Tech  ♦ Engineering Information Systems Lab ♦ eislab.gatech.edu© 1993-2001 GTRC
Design for manufacture (DFM)
‹ Designing part having manufacturability in mind
‹ DFM that requires geometric tailoring
– Produces an existing part using different technology and material 
– Retains important mechanical characteristics by tailoring the 
geometry
– Procedure
» Approximate geometry and design requirements are provided 
by the designer
» Manufacturing process planning engineer determine process 
parameters and solid geometry that satisfy design 
requirements
17Georgia Tech  ♦ Engineering Information Systems Lab ♦ eislab.gatech.edu© 1993-2001 GTRC
Challenge
‹ Competing nature of design requirements
‹ Coupling conditions in process parameters
‹ Modeling the design requirements in terms of process 
parameters
‹ Formulating and developing solution strategies for DFM 
problem 
‹ DFM problem:
– A formulated problem that can be solved to provide solid geometry 
and process parameters that satisfy design requirements
18Georgia Tech  ♦ Engineering Information Systems Lab ♦ eislab.gatech.edu© 1993-2001 GTRC
Goal
‹ To support formulating and developing solution 
strategies for DFM problem by allowing distributed 
engineers to share the previously formulated DFM 
problem
19Georgia Tech  ♦ Engineering Information Systems Lab ♦ eislab.gatech.edu© 1993-2001 GTRC
Approach
‹ Complete automation of reasoning to determine process 
parameters and geometric variable is not feasible
‹ Reason to the point where we can determine ranges for 
process parameters and retrieve DFM problems that 
posses same reasoning and ranges for process 
parameters
Open repository 
for DFM 
problem
Distributed 
engineers Storing and 
retrieving DFM 
problems
Description logics will be 
used to enhance the 
reasoning capability for 
determining relevant DFM 
problem