1

Numerical Shape Optimisation Numerical Shape Optimisation

in Blow Moulding in Blow Moulding
Hans Groot Hans Groot
2
Overview Overview
1. 1. Blow molding Blow molding
2. 2. Inverse Problem Inverse Problem
3. 3. Optimization Method Optimization Method
4. 4. Application to Glass Blowing Application to Glass Blowing
5. 5. Conclusions & future work Conclusions & future work
3
Inverse Problem Glass Blowing Conclusions Blow Molding Optimization Method
Blow Molding Blow Molding
glass bottles/jars
plastic/rubber
containers
mould
pre-form
container
4
Example: Jar Example: Jar
Inverse Problem Glass Blowing Conclusions Blow Molding Optimization Method
5
Problem Problem
Forward problem
Inverse problem
pre-form container
Blow Molding Glass Blowing Conclusions Inverse Problem Optimization Method
6
Forward Problem Forward Problem

m
Surfaces
1
and
2
given
Surface
m
fixed (mould wall)
Surface
i
unknown
Forward problem
Blow Molding Glass Blowing Conclusions Inverse Problem Optimization Method
7

1
Inverse Problem Inverse Problem
Surfaces
i
and
m
given
Either
1
or
2
unknown
Inverse problem

m
Blow Molding Glass Blowing Conclusions Inverse Problem Optimization Method
8
Construction of Pre Construction of Pre--Form by Pressing Form by Pressing

2
Blow Molding Glass Blowing Conclusions Inverse Problem Optimization Method
9
Optimization Optimization
Find pre-form for approximate
container with minimal distance
from model container
mould wall
model container
approximate container
Optimisation Method Glass Blowing Blow Molding Inverse Problem Conclusions
10
Optimization Optimization
mould wall
model container
approximate container
Minimize objective function
2
2
2
i
d d d
+
* ! ! +

i
d
Optimisation Method Glass Blowing Blow Molding Inverse Problem Conclusions
11
Computation of Objective Function Computation of Objective Function
Objective Function:
Composite Gaussian quadrature:
m+1 control points (

) m intervals
n weights w
i
per interval (

)
2
i
d d
+
* ! +

2 2
i
d ( )
m n
i nj i
j i
d w d
+
+
+ !

x
Optimisation Method Glass Blowing Blow Molding Inverse Problem Conclusions
12
Parameterization of Pre Parameterization of Pre--Form Form
P
1
P
5
P
4
P
3
P
2
P
0
O
R,
1. Describe surface by
parametric curve
e.g. spline, Bezier curve
2. Define parameters as radii
of control points:
3. Optimization problem:
Find p as to minimize
1 2 5
P P P
( , ,..., ) R R R ! p
) ( p *
Optimisation Method Glass Blowing Blow Molding Inverse Problem Conclusions
13
iterative method to minimize objective function
J: Jacobian matrix
: Levenberg-Marquardt parameter
H: Hessian of penalty functions:

i
!w
i
/c
i
, w
i
: weight, c
i
>0: geometric constraint
g: gradient of penalty functions
Ap: parameter increment
d: distance between containers
Modified Levenberg-Marquardt Method
)
T T
i i i i i i i i
+ + A ! + J J I H p J d g
Optimisation Method Glass Blowing Blow Molding Inverse Problem Conclusions
14
Function Evaluations per Iteration
Distance function d:
o one function evaluation
Jacobian matrix:
1. Finite difference approximation:
o p function evaluations (p: number of parameters)
2. Broydens method:
o no function evaluations, but less accurate
function evaluation = solve forward problem
Optimisation Method Glass Blowing Blow Molding Inverse Problem Conclusions
15
Neglect mass flow in azimuthal
direction (u
f
0)
Given R
1
(f), determine R
2
(f)
Volume conservation:
R(f) radius of interface
Approximation for Initial Approximation for Initial Guess Guess
streamlines
3 3 3 3
2 1 m
i
( ) ( ) ( ) ( )
R R R R
N N N N ! +
f
r
Optimisation Method Glass Blowing Blow Molding Inverse Problem Conclusions
16
Initial Initial Guess Guess
approximate
inverse problem
initial guess
of pre-form
model
container
17
Glass Blowing Glass Blowing
Blow Molding Inverse Problem Conclusions Glass Blowing Optimization Method
18
Forward Problem Forward Problem
1)Flow of glass and air
Stokes flow problem
2)Energy exchange in glass and air
Convection diffusion
problem
3)Evolution of glass-air interfaces
Convection problem
Blow Molding Inverse Problem Conclusions Glass Blowing Optimization Method
19
Level Set Method Level Set Method
glass
air
air
> 0
< 0 < 0
= 0
motivation:
fixed finite element mesh
topological changes are
naturally dealt with
interfaces implicitly defined
level sets maintained as signed
distances
Blow Molding Inverse Problem Conclusions Glass Blowing Optimization Method
20
Computer Simulation Model Computer Simulation Model
Finite element method
One fixed mesh for
entire flow domain
2D axi-symmetric
At equipment
boundaries:
no-slip of glass
air is allowed to flow
out
Blow Molding Inverse Problem Conclusions Glass Blowing Optimization Method
21
Comparison Approximation with Comparison Approximation with
Simulation Model Simulation Model
forward problem
pre-form
container
simulation
approximation (u
f
0)
22
Optimization of Pre-Form
inverse problem
initial guess
23
Optimization of Pre-Form
initial guess
inverse problem
24
Optimization of Pre-Form
optimal pre-form
inverse problem
25
Signed Distance between
Approximate and Model Container
26
Summary
Shape optimization method for pre-
form in blow molding
describe either pre-form surface by
parametric curve
minimize distance from approximate
container to model container
find optimal radii of control points
use approximation for initial guess
Application to glass blowing
average distance < 1% of radius
mold
Blow Molding Inverse Problem Glass Blowing Conclusions Optimization Method
27
Short Term Plans
Extend simulation model
improve switch free-stress to no-slip
boundary conditions
one level set problem vs. two level set
problems
Well-posedness of inverse problem
Sensitivity analysis of inverse problem
Blow Molding Inverse Problem Glass Blowing Conclusions Optimization Method
28
Parison Optimization for Ellipse
model container optimal container initial guess
29
Blow Molding Blow Molding
mould
ring
parison
container
e.g. glass bottles/jars
30
Approximation Approximation
Initial guess
pre-form
model
container
31
Incompressible medium:
R(f) radius of interface G
Simple example axial symmetry:
If R
1
is known, R
2
is uniquely determined and vice
versa
1 1
2 2
3 3 3 3
2 1 m
i
0 0
( ) ( ) sin( )d ( ) ( ) sin( )d ) ) ( (
R R R R
!

T T
N N N N N N N N
Initial Guess Initial Guess
3 3 3 3
2 1 m
i
R R R R
!
R(f)
32
Inverse Problem Inverse Problem

1
given (e.g. plunger)

m
,
i
given
determine
2

1
Optimization:
Find pre-form for container with minimal difference in
glass distribution with respect to desired container
33
Inverse Problem

i
and
m
given

1
and
2
unknown
Inverse problem
34
1 1
2 2
3 3 3 3
2 1 m
i
0 0
( ) ( ) sin( )d ( ) ( ) sin( )d ) ) ( (
R R R R
!

T T
N N N N N N N N
Volume Conservation
(incompressibility)
R
1
R
2
R
i
R
m
35
1 1
2 2
3 3 3 3
2 1 m
i
0 0
( ) ( ) sin( )d ( ) ( ) sin( )d ) ) ( (
R R R R
!

T T
N N N N N N N N
Volume Conservation
(incompressibility)
R
m
fixed
R
i
variable
with R
1
and R
2
R
1
, R
2
??
R
i
R
m
R
1
R
2
36
Blow Moulding
preform
container
Forward problem
Inverse problem
37
Hybrid Broyden Method
Optimisation Results Introduction Simulation Model Conclusions
)
)
i i i
i i
i i
i i i
i i i i
i i i
i i
i i
i i
i i
i i i i
r r r
J J
J J
p J r
r p J r
p J r
r r
p p
p p
p p
p p J r
! A
+ !
+ !
+
A A
A A A
+
A A
A A
A A
A A
A A
A A A
+

1
1
1
1 1
with
otherwise ,
: method bad s Broyden'
if ,
: method good s Broyden'








[Martinez, Ochi]
38
Example (p = 13)
Optimisation Results Introduction Simulation Model Conclusions
Method # function evaluations # iterations *
Hybrid Broyden 32 8 1.75
Finite Differences 98 9 1.36
Conclusions:
similar number of iterations
similar objective function value
Finite Differences takes approx. 3 times longer
than Hybrid Broyden
39
Optimal preform
Preform Optimisation for Jar
Model jar
Initial guess
Results Level Set Method Introduction Simulation Model Conclusions
40
Preform Optimisation for Jar
Model jar
Results Level Set Method Introduction Simulation Model Conclusions
Approximate jar
Radius: 1.0
Mean distance: 0.019
Max. distance: 0.104
41
Conclusions
Conclusions Optimisation Introduction Simulation Model Results
Glass Blow Simulation Model
finite element method
level set techniques for interface tracking
2D axi-symmetric problems
Optimisation method for preform in
glass blowing
preform described by parametric curves
control points optimised by nonlinear least
squares
Application to blowing of jar
mean distance < 2% of radius jar
42
Thank you for your attention
43
Comparison Comparison
Inverse problem Forward problem
two unknown intefaces one unknown interface
Inverse problem
Forward problem

m
Inverse problem under-determined or forward
problem over-determined?
44
Inverse Problem
Optimisation Results Introduction Simulation Model Conclusions
preform container
Unknown surfaces
45
Forward Problem
Optimisation Results Introduction Simulation Model Conclusions
preform container
R
m
known
R
i
unknown
46
Incompressible medium:
R(f) radius of interface G
Simple example axial symmetry:
If R
1
is known, R
2
is uniquely determined and vice
versa
1 1
2 2
3 3 3 3
2 1 m
i
0 0
( ) ( ) sin( )d ( ) ( ) sin( )d ) ) ( (
R R R R
!

T T
N N N N N N N N
Volume Conservation Volume Conservation
3 3 3 3
2 1 m
i
R R R R
!
R(f)
47
Incompressible medium:
R(f) radius of interface G
Simple example axial symmetry:
If R
1
is known, R
2
is uniquely determined and vice
versa
1 1
2 2
3 3 3 3
2 1 m
i
0 0
( ) ( ) sin( )d ( ) ( ) sin( )d ) ) ( (
R R R R
!

T T
N N N N N N N N
Volume Conservation Volume Conservation
3 3 3 3
2 1 m
i
R R R R
!
R(f)