design and optimization: status & needs

7
Design and Optimization: Status & Needs Dr. Wei Chen Associate Professor Integrated DEsign Automation Laboratory (IDEAL) Department of Mechanical Engineering Northwestern University [email protected] , (847)491-7019 Http://ideal.mech.northwestern.edu//

Upload: akeem-hardin

Post on 01-Jan-2016

26 views

Category:

Documents


2 download

DESCRIPTION

Design and Optimization: Status & Needs. Dr. Wei Chen Associate Professor I ntegrated DE sign A utomation L aboratory ( IDEAL ) Department of Mechanical Engineering Northwestern University [email protected] , (847)491-7019 Http://ideal.mech.northwestern.edu//. - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: Design and Optimization: Status & Needs

Design and Optimization: Status &

Needs

Dr. Wei ChenAssociate Professor

Integrated DEsign Automation Laboratory (IDEAL)Department of Mechanical Engineering

Northwestern [email protected], (847)491-7019

Http://ideal.mech.northwestern.edu//

Page 2: Design and Optimization: Status & Needs

Research Areas of IDEAL

Major Topics

•Robust Design & Optimization under Uncertainty (NSF)

•Metamodeling for Simulation-Based Design (Ford)

•Enterprise-Driven Multidisciplinary Decision Based Design (NSF)

•Model Validation (NSF)

Pratt & Whitney US Army

Ford

Motorola

GM

Industrial Applications

Alcoa, JD Power, etc.

Page 3: Design and Optimization: Status & Needs

State-of-the-Art: Efficient Probabilistic Optimization

MPP

f(u1, u2)

u1

u2

g=0

pdf

Opt 1Deter

RA Constr n

RA Constr 1

Opt 2Deter

RA Constr n

RA Constr 1

Cycle 1 Cycle 2MPP

Most Probable Point (MPP) Method for Efficient Reliability Assessment

Sequential Optimization and Reliability Assessment (SORA) Method

pdf of g

g0 gR

Red Area = Prob(ggR)=R

Inverse MPP Strategy

Page 4: Design and Optimization: Status & Needs

State-of-the-Art: Metamodeling Techniques for Simulation Based Design

CAE Model

Classification of Variables

Product/Process

ResponsesControlFactors

Noise Factors

A. Optimal Design of Experiments (DOE)

B. Sequential Metamodeling

0

0.5

1

0

0.5

10

200

400

C. Analytical Probabilistic Global Sensitivity

Analysis• Reduce the size of problem• Identify source for variance

reduction

E. Probabilistic Optimization

Confirmation & Metamodel Updating

D. Analytical Uncertainty Propagation

Page 5: Design and Optimization: Status & Needs

Demand

MARKETING TEAMCUSTOMER

PREFERENCES

Key Customer Attributes

1

GROUPS

EngineeringAttributes

2DesignOptions

ENGINEERING

State-of-the-Art: Multidisciplinary Decision-Based Design Framework

Total Lifecycle Cost

ACCOUNTING

3

CORPORATE

MANAGEMENT

Utility(Profit)

4

Optimized Design

and Price

5

Page 6: Design and Optimization: Status & Needs

Vision

Rapid concurrent design of material, product, and the associated manufacturing processes, optimizing quality, costs, and performance based on high fidelity modeling spanning the whole product realization and life cycle.

Product Design

Process Design

Material Design

Concurrent/Collaborative Optimal Material , Product, and

Process Design Decisions

Processing

Structure

Properties

PerformanceDesign Driven

Mapping Relation

Page 7: Design and Optimization: Status & Needs

Challenges/Research Thrusts

– Seamless communication with languages/representations across material scientists, product designers, and manufacturing process engineers.

– Problem decomposition/recomposition methods and modeling approach to reduce the interdependency (complexity) but to maintain the concurrency of subproblems.

– Rapid design optimization methods to employ high fidelity simulation programs that capture the life cycle requirements, with the consideration of uncertainty.

– Design synthesis methods to accumulate knowledge and experience that adapt to changes of design requirements.

– Adaptive design framework with decision support, knowledge accumulation, and support for incorporating business and costs modeling, for multi-level users with distributed, concurrent, and collaborative access.

– Model validation approach that requires the minimum amount of physical experiments and improves the confidence of using the result from optimization.