Sie sind auf Seite 1von 16

8/25/2011

ME392C (19060) Design Optimization and Automation

Matt Campbell 222-9122 mc1@mail...

Day Wed., Aug. 24 Fri., Aug. 26 Mon., Aug. 29 Wed., Aug. 31 Fri., Sept. 02 Mon., Sept. 05 Wed., Sept. 07 Fri., Sept. 09 Mon., Sept. 12 Wed., Sept. 14 Fri., Sept. 16 Mon., Sept. 19 Wed., Sept. 21 Fri., Sept. 23 Mon., Sept. 26 Wed., Sept. 28 Fri., Sept. 30 Mon., Oct. 03 Wed., Oct. 05 Fri., Oct. 07 Mon., Oct. 10 Wed., Oct. 12

Topic Introduction Constraints: Well-constrained no class Constraints: Well-constrained Monotonicity Analysis no class Monotonicity Analysis Monotonicity Analysis Karush-Kuhn-Tucker Conditions no class Karush-Kuhn-Tucker Conditions Karush-Kuhn-Tucker Conditions Penalty Functions 1-D search methods 1-D search methods

Reading Assignment Section 1 -2.4 (pp. 1-25)

Assignment HW #1 due 8/31

Papalambros, Panos Y. Principles of Optimal Design: Modeling and Computation pp. 99-122; 232-240 HW #2 due 9/16

Arora, Jasbir, Introduction to Optimum Design, Sections 4.3-4.4 p.103-143

Vanderplaats, GN Numerical Optimization Onwubiko, Chinyere, Introduction to Engineering Design Optimization Sections 3.3; pp. 80-83; 88104 Exam #1

HW #3 due 9/26

Gradient Based Search Methods (Quasi-)Newton Methods (Quasi-)Newton Methods GRG Method GRG Method Active Set Strategy Sequential Quadratic

Belegundu, Ashok D., Optimization Concepts And Applications In Engineering, pp.63-81 Sections 3.4-3.8

HW #4 due 10/10

Belegundu, Ashok D., Optimization Concepts And Applications In Engineering, pp. 176-188 Sections

8/25/2011

28 Fri., Sept. 30 Mon., Oct. 03 Wed., Oct. 05 Fri., Oct. 07 Mon., Oct. 10 Wed., Oct. 12 Fri., Oct. 14 Mon., Oct. 17 Wed., Oct. 19 Fri., Oct. 21 Mon., Oct. 24 Wed., Oct. 26 Fri., Oct. 28 Mon., Oct. 31 Wed., Nov. 02 Fri., Nov. 04 Mon., Nov. 07 Wed., Nov. 09 Fri., Nov. 11 Mon., Nov. 14 Wed., Nov. 16 Fri., Nov. 18 Mon., Nov. 21 Wed., Nov. 23 Fri., Nov. 25 Mon., Nov. 28 Wed., Nov. 30 Fri., Dec. 02

Gradient Based Search Methods (Quasi-)Newton Methods (Quasi-)Newton Methods GRG Method GRG Method Active Set Strategy Sequential Quadratic Programming Sequential Quadratic Programming Direct Search Methods Direct Search Methods Uninformed Search Uninformed Search Informed Search Informed Search Representing the Tree Graph Grammars Branch and Bound Random Hill Climbing Simulated Annealing Simulated Annealing Genetic Algorithms No class No class

Belegundu, Ashok D., Optimization Concepts And Applications In Engineering, pp.63-81 Sections 3.4-3.8

HW #4 due 10/10

Belegundu, Ashok D., Optimization Concepts And Applications In Engineering, pp. 176-188 Sections 5.9-5.10 HW #5 due 10/21 Bertsekas, Dimitri P., Nonlinear Programming, pp. 159-166, Section 1.8 Exam #2 Russell, Stuart J., Artificial Intelligence: Modern Approach, pp. 73-81; 92-110 Section 3.5, 4.1-4.3 Winston, Patrick Henry, Artificial Intelligence Section 5.5 Arora, Jasbir, Introduction to Optimum Design, pp. 513-525 Section 15.1-15.7 Mitchell, Melanie, An Introduction to Genetic Algorithms Exam #3

HW #6 due 11/11

Literature Review Project due 11/28

Project Presentation

8/25/2011

DAC-1 Design Optimization Algorithms DAC-2 Application-Tailored Optimization Methods


DAC-3 Keynote DAC-4 Celebration of K.K. Choi's 65th Birthday DAC-5 Formulation of Mass Customization Problems DAC-6 Q&A with Design Engineering Division's Journal Editors

DAC-13 Panel: Environmental Policy in Vehicle Design

DAC-14 Multidisciplinary Design Optimization (MDO) DAC-15 Metamodel-Based Design Optimization (MBDO)
DAC-16 Simulation-Based Design Under Uncertainty DAC-17 Geometric Modeling and Algorithms for Design and Manufacturing DAC-18 Design and Manufacturing of Systems

DAC-7 Multiscale Mechanics and Design Optimization of Cellular Materials DAC-8 Decision Making in Engineering Design
DAC-9 Design for Market Systems

DAC-19 Product Design and Exploration


DAC-20 Design of Multiscale Engineering Systems

DAC-10 Applying Large Scale Demand Models in Design Optimization


DAC-11 Design for the Developing World

DAC-21 Multi-Objective Optimization and Sensitivity Analysis


DAC-22 Product Family and Product Platform Design DAC-23 Panel: Design Frontiers

DAC-12 Design and Optimization of Sustainable Energy Systems

8/25/2011

E.D.A.

Bell Labs, ca. 1980

Source: Wikipedia http://en.wikipedia.org/wiki/Moore%27s_law

Automate the design of 1000s of transistors Algorithmic description


(e.g. C++)

Using templates and rules


Algorithms changed to machine code Machine code changed to logic Logic changed to circuit Circuit changed to masks

High level synthesis Register Transfer Model (HDL) Logic synthesis Gate netlist Layout synthesis Layout / mask data

8/25/2011

Circuit Level
Behavior A LOW LOW HIGH HIGH B LOW HIGH LOW HIGH C HIGH HIGH HIGH LOW
A C

Structure
vdd!

Geometry

HIGH

LOW
B

HIGH
gnd!

HIGH: True, high voltage near to VDD LOW: False, low voltage near VSS (gnd)

Example DE-Solver
Solver for a differential equation of the form

with explicit Euler method. Auxiliary variables:

Iterative computation: x1 = x+dx u1 = u+du = u+u*dx = u3*x*u*dx3*y*dx y1 = y+dy = y+u*dx until convergenge condition x1 < xe is met.

8/25/2011

Example DE-Solver
C-Code:

float diffeq (float x, float y, float u, float dx, float a) { float x1,u1,y1; repeat { x1 = x+dx; u1 = u3*x*u*dx3*y*dx; y1 = y+u*dx; x=x1;u=u1;y=y1; } until (x1 < a); return y; }

Intermediate Code Generation


Compiler generated temporary variables: t1
t7
Three address code C-Code section repeat { x1 = x+dx; u1 = u3*x*u*dx3*y*dx; y1 = y+u*dx; x=x1;u=u1;y=y1; } until (x1 < a); B1: x1 := x+dx; t1 := y*dx; t2 := 3*t1; t3 := u*dx; t4 := x*t3; t5 := 3*t4; t6 := u-t5; u1 := t6-t2; t7 := u*dx; y1 := y+t7; x:=x1; u:=u1; y:=y1; if x1 >= a goto B1;

8/25/2011

Data flow analysis


Three address code t1 t2 t3 t4 t5 t6 u1 = = = = = = = 3*y; dx*t1; u*dx; 3*x; t3*t4; u-t5; t6-t2; Data flow graph

x u

dx 3

t4


t5 t3 u t6 u1

t1

t2

3 x1 Implementation

dx

Data flow graph

*
dx 3 y Multiplier1

*
Multiplier2

*
Multiplier3 dx

x u

t4


t5 t3 u t6 u1

t1

*
Multiplier4 u

*
Multiplier5

t2

Subtractor1

Subtractor2 u1

8/25/2011

Circuit Level
Behavior A LOW LOW HIGH HIGH B LOW HIGH LOW HIGH C HIGH HIGH HIGH LOW
A C

Structure
vdd!

Geometry

HIGH

LOW
B

HIGH
gnd!

HIGH: True, high voltage near to VDD LOW: False, low voltage near VSS (gnd)

Section 1.
1. Design as Search 2. Representing Candidates by a Fixed Set of Design Variables 3. Representing Candidates by a System of Production Rules 4. Challenges
1. 2. 3. 4. Generation Guidance Evaluation Representation

8/25/2011

Design as Search
four main challenges
how to represent the set of all possible designs? how to generate candidates based on that representation? how to evaluate the quality of each candidate? how to guide the search to better solutions?
User Defined: Problems Description constraints & objectives

Representation

Generate

Search Process
Guide

Evaluate

Final Design

ASME Design Automation Committee


promote research and the dissemination of knowledge in the areas of: Design Representation
CAD, virtual and physical prototyping, knowledge based systems, product data management

Design Optimization (Generation and Guidance)


Structural, topological, multi-disciplinary, heuristic and deterministic, under uncertainty

Design Evaluation
Modeling, simulation, approximation

Design Integration
Integration, decomposition, collaboration

8/25/2011

How used?
User Defined: Problems Description constraints & objectives

Representation

Generate

Search Process
Guide

Evaluate

Final Design

Representation
Is it a fixed set of variables? Is it possible to capture different technological solutions to the same sub-problem? What are the basic building blocks in the design problem? How are interrelations between basic building blocks going to be handled? Can you live with a representation that leaves out a large set of feasible solutions? (or Can you live with a representation that includes numerous infeasible solutions?)

10

8/25/2011

Section 1.
1. Design as Search 2. Representing Candidates by a Fixed Set of Design Variables 3. Representing Candidates by a System of Production Rules 4. Challenges
1. 2. 3. 4. Generation Guidance Evaluation Representation

physical world

Analysis
answer

need

Design
candidate designs

11

8/25/2011

Two Representations
need

1,1 1,2 2,1 2,2 ,1 ,2

1, 2, ,

The space is an n-tuple of design variables a vector of fixed length

The space is a system of options that creates a large decision tree.

: 1 = 0 hp x = 0 : 1, < 1 < 1, , < < ,

12

8/25/2011

knowns, givens, constants F max & Fmin (the predicted maximum and minimum force the spring should provide) xmax & xmin (the predicted maximum and minimum defections) L a,max (the maximum allowable assembled length) Ds (the diameter of the umbrella shaft) Dmax (the max allow able diameter)

design variable s D (mean coil diameter) d (wire diameter) N (number of active coils) L f (free length) end type (ground? squared?) material (likely steel; (density), E (modulus), S y (yield strength) , S UT (ult imate strength) are the relevant material properties needed)

dependent parameters D i (inner diameter; = D d) D o (outer diameter; = D + d) C (spring index; = D/d) k (spring constant; 4 3 = d G/8ND (shear stress; 2 = (1+d/2D)(8FD/ d ) Nt (total # of coils; = N + <end type>) Ls (shut length) La (assembled length)

performance parameters / issues $ (cost; lower is better) m (mass; lower is better) NFY (safety factor for yielding failure; should be above 2) NFE (safety factor for fatigue; should be above 2) NFB (safety factor for buckling; should b e above 2) corrosion resistant?

Example Design Problem

Design a flat belt belt drive for an input of 3,000 rpm and an output 750 rpm and a total power of 50 hp.

13

8/25/2011

Two Representations
need

1,1 1,2 2,1 2,2 ,1 ,2

1, 2, ,

The space is an n-tuple of design variables a vector of fixed length

The space is a system of options that creates a large decision tree.

rule #1 rule #2

rule #3

rule #1 rule #3 rule #1 rule #2 rule #2

rule #1

rule #3

14

8/25/2011

Component Selection
Component selection in engineering design is a process in which an assembly of pre-defined component types is given and a choice of specific components is desired that satisfies a set of design requirements and constraints.

Search Process
CFG Battery 1 Motor 1 Motor 2 . Motor N Motor 1 Shaft 1 Shaft 2 Motor 2 . Motor N Battery 2

Gear 1. Gear N

Root of the tree is a complete CFG of component types Iterative process of replacing each component type in the CFG with an artifact from the database

15

8/25/2011

Challenges in Using Computational Search for Design


Generation and Guidance Evaluation Representation
User Defined: Problems Description constraints & objectives

Representation

Generate

Search Process
Guide

Evaluate

Final Design

16

Das könnte Ihnen auch gefallen