Vous êtes sur la page 1sur 46

Multi-Level Design Process For 3-D Preform Shape Optimization In Metal Forming Using The Reduced Basis Technique

Department of Mechanical and Materials Engineering Wright State University Dayton, OH 45435

CDOC

Presentation Outline
Research objectives Overview of forging process Need for preform shape optimization Research challenges Reduced basis design approach Case studies Summary

CDOC

Research Objectives

CDOC

Develop a methodology for 3-D preform shape optimization in forging

Identify 3-D preform shape parameters Define optimization design parameters Define finite element-based objectives and constraints Establish explicit relationship between design parameters and objectives and constraints

Enable preform design for complex 3-D forging components Develop a computationally feasible technique Improve product quality

Introduction to Metal Forming Process


Initial billet Preform Blocker Finisher

CDOC

Trimming

Intermediate shapes
Tradeoff designs

Forming process
Quality check

Robust design

0 .0 6 0 .0 5 0 .0 4
Frequency

Heat treatment process

0 .0 3 0 .0 2 0 .0 1 0 -1.79 -1.53 -1.27 -1.01 -0.75 -0.49 -0.23 0.03 0.29 0.55 0.81 1.07 1.33 1.59
P e rce n tag e c h a n g e in d im e n sio n

Machinable

Rejected

Temperature distribution

Need for Preform Shape Design

CDOC

Flash

Underfill

Reduced underfill

Initial shape

Blocker shape

Preform shape

Blocker shape

Minimize load Minimize material waste More uniform Material flow

Minimize geometric variation Decrease production cost Increase die life

Conventional Preform Design Techniques

CDOC

Preforms for steel finished forgings

Design guidelines Empirical relations Computer-aided design

No preform

h=b

Knowledge-based approach
Finite element approach
h = 2b

Forward simulations Backward simulations


Upset stock Preforms Finished forgings

h = 3b

Design Issues
2D Assumptions for 3D parts
Plane-strain - No deformation in out-of-plane direction Axisymmetric - Material flow is radial

CDOC

Practical forgings are neither axisymmetric nor plane-strain Require large number of parameters Require large number of simulations Long computational times

Shape parameters

Preform shapes

Shape Optimization Methodology


Identify initial preform shape Parameterize preform shape Determine critical optimization parameters Construct surrogate models

CDOC

Reduced basis technique


Obtain basis shapes Employ design variable linking

Generate DOE points using Latin

Hypercube sampling techniques Conduct forging simulations of the DOE billets Extract FEM output data

Optimize

Reduced Basis Technique


Innovative technique for defining preform shapes

CDOC

Axisymmetric 3-D metal hub

Finite element model of 2-D section

Reduced Basis Technique


Innovative technique for defining preform shapes Generate initial guess preform basis shapes
Basis 1 [Y1] Basis 2 [Y2] Metal hub Basis 3 [Y3]

CDOC

X and Y co-ordinates of boundary points define basis vectors


(xi, yi)
x1 y 1 Y1 = x n yn

Basis shape boundary points

Basis Vectors
Need large number of boundary points
Increases length of the basis vector Require no extra computational cost
Basis 1 [Y1]
Desired boundary

CDOC

Basis 2 [Y2]

Basis 3 [Y3]

False boundary

Define all of the basis shapes similarly

Basis Vectors
Need large number of boundary points
Increases length of the basis vector Require no extra computational cost
Basis 1 [Y1] Basis 2 [Y2] Basis 3 [Y3]

CDOC

Perform Gram-Schmidt orthogonalization


Produces independent basis shapes

Design Parameters Definition


Weighted combination of orthogonal basis shapes
Y = Y + a i Yi
c i=1 n

Basis shapes

Forged parts

Design Parameters Definition


Weighted combination of orthogonal basis shapes
Y1

Y = Y + a i Yi
c i=1

a1
Y2
a1 0.2 a2 0.5 a3 0.6

a2

Y3

a1

a2

a3

a3
Forged part 0 ai 1 Basis shapes

Design Parameters Definition


Weighted combination of orthogonal basis shapes
Y1

Y = Y + a i Yi
c i=1

a1
Y2
a1 0.6 a2 0.2 a3 0.3

a2

Y3

a1

a2

a3

a3
0 ai 1 Forged part Basis shapes

Design Parameters Definition


Weighted combination of orthogonal basis shapes
Y1

Y = Y + a i Yi
c i=1

a1
Y2
a1 0.6 a2 0.1 a3 0.8

a2

Y3

a1

a2

a3

a3
0 ai 1 Basis shapes Forged part

Design Parameters Definition


Weighted combination of orthogonal basis shapes
Y1

Y = Y + a i Yi
c i=1

a1
Y2

a2

DOE points
Resultant preform shapes Y3

a3
0 ai 1 Basis shapes

Scaling maintains constant volume for resultant preforms Weights (ai) are the design parameters

Construction of Surrogate Model


Variable 2

CDOC

Generate Latin Hypercube sampling points Perform finite element simulations Obtain objectives and constraints

Variable 1 Max Underfill

Med

Strain variance

Min

Underfill

Construct response surface model

Optimization Statement
Design variables

CDOC

Weighting factors of reduced basis technique (ai)

Cost function

Minimize strain variance f(ai)

Subject to

Underfill g(ai) 0

Side bounds on weights

0 ai 1

Design Optimization
b1 b2 h1 h1 = 1.25 x b1 h2 = 1.50 x b2 h2

CDOC

Rail section

Symmetry axis Cross-sectional

view

Four basis shapes generated


Basis 1 [Y1] Basis 2 [Y2] Basis 3 [Y3] Basis 4 [Y4]

Basis shapes Finite element simulations Final forged parts

Results
Parameter a1 a2 a3 a4 Initial value 0.2 0.6 0.4 0.3 Optimum values 0 1 0.758 0

CDOC

Optimized billet

Iteration history of objective and constraint functions


0.4 0.35 0.3

Strain variance Underfill

Final forged part

Response

0.25 0.2 0.15 0.1 0.05 0 -0.05 0 2 4

Strain variance : 0.0647


6 8

Flash : 3 %

Iteration number

Need For Multi-Level Design Process

CDOC

Single level design requires appropriate starting basis shapes


Basis 1 [Y1] Basis 2 [Y2] Basis 3 [Y3] Basis 4 [Y4]

Simple basis shape

Viable basis shapes

Inappropriate basis shape

No information available for a new product Enable design with geometrically simple basis shapes

Multi-Level Optimization Routine


Generate starting guess shapes Define design parameters (Level L) Obtain design points Build new basis shapes (L=L+1)

CDOC

Reduced basis method DOE techniques

Conduct FEM analysis using DEFORM-3D Obtain objective and constraints Generate surrogate model using response surface method Redesign using optimization algorithm No Constraint satisfied? Yes Optimum preform

Case Study 1 (Level 2) 1)


Multi-Level design optimization
Basis 1 [Y1] Basis shapes Basis 2 [Y2] Basis 3 [Y3]

CDOC

Rail section

Final forged part Best shape Best weights


a1 0.769 a2 0 a3 0.449

Final forged part

Case Study 1 (Level 3) 2)


Basis 1 [Y1] Basis 2 [Y2] Basis 3 [Y3]

CDOC

Basis shapes

Rail section

Final forged part

Best shape

Best weights
a1 0 a2 1 a3 1

Final forged part

Case Study 1(Level 3)


Basis 1 [Y1] Basis 2 [Y2] Basis 3 [Y3]

CDOC

Rail section

Basis shapes

Final forged part Optimum shape

Optimum weights
a1 0 a2 1 a3 0.7246

Final forged part

Complete die fill Flash: 3 %

Result Comparison
Single-level
Basis 1 [Y1]

CDOC

Multi-level design scheme Single-level design scheme


Basis 21 [Y1] Basis [Y2] Basis 3 [Y3] Basis 2 [Y2] Basis 4 [Y4] Basis 3 [Y3]

Strain variance Single-level Multi-level 0.0648 0.0710

Multi-level

Multi-level design scheme leads to optimum billet Computational time increases Expert knowledge can be used for single-level design scheme
Optimum billet Optimum billet

Case Study - 2
h/b ratio is one Three simple billet shapes as basis shapes All basis shapes give underfill Flash: 3 % Quarter model for forging simulation
Zone A
3xh 2.2xh 1.5xh

CDOC

3-D Metal hub

Zone B h b h=b

Basis 1 [Y1]

Basis 2 [Y2]

Basis 3 [Y3]

Forged part

[Y1] and [Y2] give underfill at Zone A and [Y3] at Zone B

Results
Three design variables 15 DOE points generated None give complete die fill
Optimum weights
a1 a2 0.1036 0.7165 a3

CDOC

0.6158

Preform shape

Forged part with complete die fill

Achieved optimum shape in single level Requires multi-level design process for higher h/b ratios

Case Study - 3
3-D Metal hub with h/b = 2 Allowable flash percentage: 2% Quarter model assumed for forging simulations

CDOC

3-D Metal hub

Quarter models with section view (h/b = 2)

Level 1
Three basis shapes selected in Level 1
Basis 1 [Y1] Basis 2 [Y2] Basis 3 [Y3]

CDOC

Basis shapes (quarter models)

Fifteen DOE points for building the RSM

Level 1 Results
Level 1 best shape : Basis 3
a1 0 a2 0 a3 1

CDOC

Top die profile Underfill

Forged part

No other basis shape combinations give less underfill Need to satisfy the underfill constraint Tapering profile of the Basis 3 is crucial

Level 2
Basis 1 is the Level 1 best shape Variation of Basis 1 form Basis 2 and Basis 3
Basis 1 [Y1] Basis 2 [Y2] Basis 3 [Y3]

CDOC

Basis shapes (quarter models)

Basis 2 and Basis 3 have opposing profiles

Level 2 Results
No underfill Flash volume: 2% Basis 2 has the maximum contribution
Optimum weights
a1 0.6 a2 1.0 a3 0.3

CDOC

Performance characteristics
Level/Response Level 1 Level 2 Underfill(cm3) 139.60 0

Preform shape Strain Variance 0.0328 0.0404

Forged part with complete die fill

Case Study - 4 (Steering Link)


High volume forged component Huge material waste (30%) occurs
Side view Front end

CDOC

Rear end

Top view Isometric view

Steering link

Cross-sections vary along all three axes

Level 1 Basis Shapes


Basis 1

CDOC

Allowable flash percentage: 5% Three simple basis shapes


Basis 2

All basis shapes give more underfill at the front end Each basis vector contains 648 shape

Basis 3

co-ordinates (216 points)

Resultant Shapes
Basis 1

CDOC

Reduced basis technique decreases the number of design variables to three

Basis 2

Possible preform shapes

Basis 3

Level 1 Results
Level 1 best weights
a1 0 a2 0.76 a3 1.0

CDOC

Level 1 best shape

Forged part with underfill

Rectangular nature of Basis 3 is most crucial Contribution of Basis 2 provides the tapering profile Level 1 best shape becomes the Basis 1 in Level 2

Level 1 Results
Level 1 best shape

CDOC

Forged part showing underfill Performance characteristics


14

0 .0 8

Basis 1

12 10

Basis 2

Basis 3

0 .0 6

Strain Variance

0 .0 5 0 .0 4 0 .0 3 0 .0 2 0 .0 1 0

Flash %

8 6

Basis 2

Level 1 best shape

0 .0 7

Level 1 best shape

4 2

0 1 2

Basis 1

Required flash percentage

Basis shape

Flash percentage

Strain variance

Basis shape

Basis 3

Level 2
Basis 1

CDOC

Four basis shapes in Level 2 Each basis shape has different


Basis 2

cross-section Number of shape co-ordinates are

Basis 3

1125 (375 points)

Basis 4

Basis shapes

Resultant Shapes
Basis 1

CDOC

Four deign variables

Basis 2

Basis 3

Possible preform shapes

Basis 4

Increasing the number of basis shapes also increases the DOE points to 25 from 15

Level 2 Results
Optimum Weights
a1 1.0 a2 0.6 a3 0.76 a4 0

CDOC

Optimum Preform

Final Forged Part

Complete die fill achieved, Flash volume: 5% Contribution of Basis 1 reduces the curvature of Basis 2 Basis 3 Cross-sectional radii reduced Any contribution of Basis 4 increases the strain variance

Level 2 Results
Optimum Preform

CDOC

Final Forged Part

Performance characteristics
12 0 .1 2 10 0.1

Basis 1

Basis 2

Basis 2

Strain Variance

0 .0 8

Flash %

Basis 3

Basis 4

Preform

0 .0 6

0 .0 4

Basis 1

0 .0 2

0 1 2

Basis shape
3

Basis shape

Basis 3

Basis 4
5

Preform

Flash percentage

Strain Variance

Summary
Introduced a novel concept for 3-D preform design
Reduced basis technique

CDOC

Enables design variable definition for complex 3-D components Utilized simple basis shapes in multi-level optimization Knowledge based basis shapes aids faster optimization Optimum preform shapes can be easily manufactured Applicable for both 2-D and 3-D forging processes

Any questions ???

Thank you

Vous aimerez peut-être aussi