An Overview of the Robust Design Computational System (RDCS) A Collection of Tools To Enable Low Risk Designs Dr. Raj Rajagopal Technical Fellow Boeing - Rocketdyne Propulsion and Power - PowerPoint PPT Presentation

Loading...

PPT – An Overview of the Robust Design Computational System (RDCS) A Collection of Tools To Enable Low Risk Designs Dr. Raj Rajagopal Technical Fellow Boeing - Rocketdyne Propulsion and Power PowerPoint presentation | free to download - id: 41985e-NGJiO



Loading


The Adobe Flash plugin is needed to view this content

Get the plugin now

View by Category
About This Presentation
Title:

An Overview of the Robust Design Computational System (RDCS) A Collection of Tools To Enable Low Risk Designs Dr. Raj Rajagopal Technical Fellow Boeing - Rocketdyne Propulsion and Power

Description:

... Taguchi analysis (orthogonal array) Probabilistic analysis (Monte Carlo ... The mathematical optimization design process is one ... cost and schedule constraints ... – PowerPoint PPT presentation

Number of Views:208
Avg rating:3.0/5.0
Slides: 25
Provided by: DavidBC3
Learn more at: http://www.rocketdynetech.com
Category:

less

Write a Comment
User Comments (0)
Transcript and Presenter's Notes

Title: An Overview of the Robust Design Computational System (RDCS) A Collection of Tools To Enable Low Risk Designs Dr. Raj Rajagopal Technical Fellow Boeing - Rocketdyne Propulsion and Power


1
An Overview of the Robust Design Computational
System (RDCS) A Collection of Tools To Enable
Low Risk DesignsDr. Raj RajagopalTechnical
FellowBoeing - Rocketdyne Propulsion and Power
  • Training Workshop on Nondeterministic Approaches
    and Their Potential for Aerospace Systems
  • May 30 - 31, 2001
  • NASA Langley Research Center
  • Hampton, Virginia

2
Fundamentals that Drove the RDCS DevelopmentThe
Launch Market Business Case
  • Cost is an independent variable
  • Increased emphasis on performance at an
    affordable cost
  • The competition is global
  • Customer satisfaction demands robust products
    with high quality and reliability
  • Time to market is critical

3
Low-Cost Development Capability Needed
HISTORY
THE NEEDED FUTURE
Dominated By Cost of Corrective Actions
COST
COST


Certified Product
Certified Product
TIME
TIME

ADVANCED PROPULSION SSME
2.5B F-1
2.4B J-2
1.7B F-100
2.0B AUTOMOBILES 96 FORD TAURUS 2.8B
Technology for Low-Cost Development
LESS THAN 500M
4
Fundamentals that Drove the RDCS Development
Aerospace Design Shop
  • Consequence of failure high
  • risk f(probability of failure, cost of failure)
  • Cutting edge technologies
  • Compute intensive analysis
  • Limited quantity or one of a kind hardware
  • Weight critical
  • Geographically dispersed design teams

5
Fundamentals that Drove the RDCS Development
Automation Vs Process Improvement
  • Automation is normally speeding up the design
    process
  • Traditional CAE
  • Partial but not the full answer
  • Improvement in design quality requires a more
    fundamental change
  • Quality is meeting contractual/customer
    expectations of the product performance
    consistently every time
  • It is impossible to think quality improvement
    without considering variability
  • Reliability is a significant attribute
    contributing to the quality

6
Fundamentals that Drove the RDCS Development A
Quality Scenario
Higher performance operating point but sensitive
to variations
Small Variation
Large Variation
Performance
Lower performance operating point but relatively
insensitive to variations
Input Variable
7
Fundamentals that Drove the RDCS Development
Multi-Disciplinary Approach a Must
INTEGRATED CALCULATION OF
Capability
RISK RELIABILITY
ROBUSTNESS LIFE
Frequency
Performance Parameter
8
Reliable Product Designs Can be Achieved Using a
Variety of Approaches
  • A robust design is one wherein the operating
    point for the controllable design variables are
    optimized such that design performance measures
    are less sensitive to the random factors that
    affect the performance.
  • A robust design can be achieved
  • By appropriately choosing the nominal design
    point that yields desired insensitivity to the
    random variables or
  • Controlling the variations in random variables by
    a tighter tolerance at an additional cost or
  • By a combination of both approaches
  • RDCS helps achieve robust design
  • Provides tools that facilitates understanding of
    the design space in a systematic manner. ( E.G
    sensitivity analysis, design scan, response
    surface )
  • Deterministic and probabilistic

9
Requirements for a Modern Design Framework
  • It must have an architecture that is an enabler
    of
  • Exploring the design space
  • Automated design explorations
  • Multidisciplinary system models
  • Parametric concepts
  • Interfaces with COTS (commercial Off The shelf)
    and custom codes
  • Computational efficiency
  • Distributed collaborative computing/engineering
  • Suite of design procedures
  • Wide range of design tools from traditional past
    design practices to more modern design procedures

10
RDCS ToolAn Instance of Modern Design Framework
Ver. 1 Release
Beta Release 1999
Min Cost, Weight Max Reliability
Alpha Release 1997
Reliability Based Ranking
Risk Reliability
Min cost, Weight Max Performance
Robustness Nominal Design Point
Design Space Exploration Response Surface
Probabilistic Sensitivities Scans
Probabilistic Optimization
Sensitivity Variable Ranking
Probabilistic Analysis
Deterministic Optimization
Typical Case Worst Case
Design Scans
Taguchi
Sensitivity Analysis
Deterministic Design
Flexibility To Approach Product Design Many Ways
RDCS System Director
PARAMETRIC MATH MODEL
Dynamic Analysis
Logistics Field Support
Cost Analysis
Mechanical Design
Risk/Life Management
Stress Analysis
Aerodynamics
Manufacturing
Capture analysis design process
Rapid parallel computing
1
11
Design Space Is Same For All The Design
Processes
Design Space Scan
Deterministic Optimization
Deterministic Designs
All Design Processes Operate on the Same Design
Space
Nominal High Low Estimated Worst Case
Optimization path o-o- Constraint Surface - - - -
Partial Scan Full Scan
Approximate Probabilistic Methods
Monte Carlo Simulation
Robust Design
(A Priori Unknown)
(A Priori Unknown)
Variables x1, x2
Response
response distribution at C2
........
Variable Bounds (Design Window )
Noise Distribution
C1
C2
Control Variable
Noise Variable
Most Probable Points
12
Available Options and Tools Within RDCS
13
Why Do We Need Distributed Computing?
  • Understanding of the design space comes at
    increased computational cost
  • Sensitivity analysis - finite difference(2n)
  • Design scan - (partial, factorial, DOE) ( nm)
  • Response surface ( n sampling points)
  • Optimization (iterative)
  • Taguchi analysis (orthogonal array)
  • Probabilistic analysis (Monte Carlo simulation or
    other)
  • Probabilistic sensitivity analysis (variability
    sensitivity)
  • Probabilistic optimization
  • All involve one to three orders of magnitude more
    function evaluations than the traditional
    approach
  • A design framework with out a robust distributed
    computing element is not a scaleable practical
    tool for design shops

14
Distributed Computing And Internet Technologies
  • Distributed Computing is the basic element of
    internet service model
  • Collaborative computing
  • virtual design teams
  • Geographically separated
  • Web enabled design environment
  • Modern tools makes it easier to implement a
    distributed computing model
  • CORBA, JAVA/RMI/IIOP
  • Web Browser, Applets, Servlets
  • Internet technologies provide the glue to
    integrate
  • Heterogeneous computing platforms
  • UNIX, PC/NT, LINUX

Client 1
Service or Server
Client 2
Internet Network Connection
15
Graphical Math Model Construction and Display
  • Bottoms Up or Top Down Construction

16
Deterministic Design Concepts
  • Deterministic typical worst case design provides
    information about margins but frequently more
    information is needed

17
Understanding the Design SpaceDesign Sensitivity
Provides Variables Ranking
18
Understanding the Design SpaceDesign Scan View
of the Design Space
  • Several Sampling Schemes Exist - Partial, Full
    factorial, Design of Experiments

19
Understanding the Design Space Response Surface
  • Two Way Design variable Interaction

20
Understanding the Design Space Deterministic
Optimization
  • Minimizing an objective function subject to
    constraints

21
Understanding the Design Space Taguchi View of
Design
  • A nominal operating design that is insensitive
    to noise - Signal to noise ratios

22
Understanding the Design Space Probabilistic
Analysis
  • Provides the effect of variability and/or
    uncertainties on design performance

23
RDCS Applications
Space
Automotive
Air Frame
Rocket Propulsion
24
Concluding Remarks
  • The concept of design framework is very
    successful
  • Dramatic improvements in multi-disciplinary
    analysis / design cycle time
  • A road map for achieving robust designs
  • Cost avoidance because of design space
    exploration
  • Demonstrated the value of introducing the NDA
    approaches progressively.
  • NDA is a critical technology but it is one of the
    many other technologies to achieve reliable
    designs
  • Use of the tool popular with advanced design
    groups, but, making significant in-roads in to
    product teams
About PowerShow.com