Sei sulla pagina 1di 57

Process Integration and Design

Optimization with modeFRONTIER


ESTECO North America

About ESTECO
Product Offerings

Established in 1999
Founded from
Academic + Industry
research funding
Global Presence

Over 300 companies


worldwide

modeFRONTIER

modeFRONTIER is a multi-objective optimization and design


environment, developed to allow easy coupling to almost
any computer aided engineering (CAE) tool, perform design
optimization and allow effective decision making.

modeFRONTIER
CAE Process
Automation

modeFRONTIER
CAE Process
Automation

Enables
Multi-Disciplinary Analysis

Process Integration
Process automation and
Standardization

DOE, Optimization
& RSM

Enables
Single/Multi-Objective
Optimization
Robustness/Reliability
Surrogate Models

Multi-Disciplinary Analysis and Optimization (MAO/MDO)

Design for Six Sigma (DFSS)

Data Plotting &


Analysis

Enables
Decision Making
Sensitivity Analysis

Interactive data analysis and


prediction

Process Integration

Process Integration and Optimization

Design Optimization
Problem Statement:

Challenges:

Design variables:

n dimensions

X = (x1, x2)

Multiple Objectives

Objective Function:

Constraints

Min F(X)

Black box analysis

Solution: Formal Optimization using modeFRONTIER

Parametric Optimization Approach

Design Variables:

Function Evaluation:

Output Variables

Objectives & Constraints

Defines problem size

(Black Box)

- Have as many

Defines problem size

Fewer the better

CFD, FEA, MBD etc

Fewer the better

Choose carefully

Excel, Matlab, ..

Discrete or
Continuous

CAE tools: ANSYS,


Adams, LSDyna etc

Influence on algorithm
choice

Influence on algorithm
choice

Influence on algorithm
choice

Design Variable consideration


C-Pillar Taper

Decklid Edge

Continuous variables:
Point coordinates
Process variables
Dimensions Shape variables

Discrete variables:
Components from a
catalogue
Material selection

Hood Radius

Body Taper

Body Taper Lower

Design Optimization

Design Optimization
F

Feasible Zone

Un-Constrained
Constrained
Optimum
Optimum
Unfeasible Zone

Max. F(x1 x2)


s.t. g(x1 x2) < ##.###

Design Optimization Derivative based


F

Max. F(x1 x2)


s.t. g(x1 x2) < ##.###

Design Optimization Derivative based


F
Unimodal, Convex problem.
Any starting point will lead to
real optimum.

Max. F(x1 x2)


s.t. g(x1 x2) < ##.###

Design Optimization

Multi-Modal problem
Global
Optima

Local
Optima

Recommendation:
Use Multiple Starting
points to find global
optima.

Design Optimization

BFGS - start

Design Optimization

BFGS 1st iter

Design Optimization

BFGS 2nd iter

Design Optimization

Total evaluations 16 (with central difference derivative calculation)

BFGS 3rd iter

Design Optimization

BFGS start (global)

Design Optimization

BFGS 1st iter (global)

Design Optimization

BFGS 2nd iter (global)

Design Optimization

BFGS 3rd iter (global)

Design Optimization

BFGS 4th iter (global)

Design Optimization

BFGS 5th iter (global)

Design Optimization

BFGS 6th iter (global)

Design Optimization

BFGS 7th iter (global)

Total evaluations 25 (with central difference derivative calculation)

Design Optimization

Simplex

Initial Triangle
Worst Point

Reflection

Reflection,
followed
by expansion

Reflection,
followed
by contraction

Design Optimization

Simplex

Simplex example

Design Optimization

Multi-Modal problem

Design Optimization

Genetic Algorithm

0-2

0-1

Survival of the fittest

0-3

0-4

0-5

1st Generation

Crossover

Mutation

1-1

1-2

1-3

1-4

1-5

New Generation

Design Optimization

Genetic Algorithm

Design Optimization

Genetic Algorithm

Design Optimization

NSGA -start

NSGA - test

NSGA 1st Gen

Design Optimization

NSGA - test

NSGA 2nd Gen

Design Optimization

NSGA - test

NSGA 3rd Gen

Design Optimization

NSGA - test

NSGA 4th Gen

Design Optimization

NSGA - test

NSGA 5th Gen

Design Optimization

NSGA - test

NSGA 6th Gen

Design Optimization

NSGA - test

NSGA 7th Gen

Design Optimization

NSGA - test

NSGA 8th Gen

Design Optimization

NSGA - test

NSGA 9th Gen

Design Optimization

NSGA - test

Design Optimization

After total of 100 evaluations ..

NSGA 10th Gen

Design Optimization

NSGA - After total of 200 evaluations .. (population 10, Generations - 20)

NSGA

Design Optimization

Multi Conflicting Objectives

Minimize F1

Minimize F2

Design Optimization

Multi-Objective

Pareto Optimum Conflicting objectives case

Example of Multi-objective Optimization

Goal: To design an internal combustion engine


with minimal NOX production and minimal fuel
consumption

Engine Optimization
4-stroke 4 cylinder 2.2 litre internal combustion
engine, simulated with GT-Power
4 variable parameters:
Variable

Minimum Value

Maximum Value

Step Size

Intake Cam Timing


(INTCAM)

214o

264o

5o

Exhaust Cam Timing


(EXHCAM)

101o

151o

5o

Intake Runner Length


(INTRUN)

100 mm

400 mm

10 mm

50% Burn Point


(TH50)

-10o

30o

2o

2 Objectives:
Minimize Fuel Consumption
Minimize NOX Production

Pareto Frontier
Dominated Point

Example showing
Trade-off Curve for
NOX and fuel
consumption (BSFC)

Non-Dominated Point

(can be improved
with respect to one
objective only at
the expense of the
other)

(can be improved
with respect to
both objectives)

Multi-Objective

NOX

Design Optimization

Pareto Frontier
Fuel Consumption

Pareto Frontier

Power

Example: Trade-off Curve for NOX (to minimize) and power


(to maximize). The curve shows the cost of doubling the
power.

Pareto Frontier
NOX

Design Optimization

Maximize: Vol. Eff.

Conflicting Multi Objective

Minimize: BSFC

Conflicting Multi Objective

BSFC

Design Optimization

Volumetric efficiency

modeFRONTIER - Interface

Graphics Settings
Workflow Details
Work-Flow Area
Overlook window

Properties

Logic Log

Summary

All Available
Nodes

Thank You !!

www.esteco.com
Contact email:
gokhale@esteco.com
na.support@esteco.com

Potrebbero piacerti anche