1. INTRODUCTION
In extrusion of solid aluminium sections, metal flow
is controlled primarily by adjusting bearing lengths
around a die orifice, although in some cases die
layout and pockets also play important roles in flow
control. It is well known that metal flows faster in
the billet centre than in the periphery zone due to
the friction between the billet and container wall.
Also, it is easier for aluminium to flow through a
wide part of a die orifice than through a narrow part.
These considerations formed the basis of the
conventional bearing design rules developed since
(1)(2)(3)
. However, as metal always tends
the 1960s
to flow along the path that provides less resistance,
the flow in a particular region depends not only on
its width and distance from the container wall, but
also on the geometry of the surrounding subregions
of the die orifice (4). Because of the complexity and
variety of aluminium sections, interaction of orifice
subregions is practically impossible to be described
in any form of design rules. Furthermore, it is very
difficult to quantify the impact of temperature and
die deflection on flow. Thus, bearing design still
relies on experience and often a trial-and-error
approach is employed when developing a new
design.
In recent years, several research institutes and
universities have initiated studies on the feasibility
of using computer modelling with finite element
(5)
methods for extrusion die design. Herberg et al
analyzed flat-face dies using commercial software.
They demonstrated that 3-D finite element analyses
could provide valuable information for improving die
designs. However, due to the high cost involved in
2. MODELLING METHOD
& = A [sinh (
Steady-state Approach
Extrusion is a non-steady process as the billet
length and temperature conditions continually
change during the extrusion cycle. Ideally, the
process should be simulated using the Lagrangian
description in which the finite element mesh
deforms with the billet. In this way, the thermomechanical history and free surface formation of
the extruded section can be studied directly.
However, this method requires frequent re-meshing
during an analysis due to excessive distortion of
elements in the deformation zones. This not only
severely reduces computation efficiency, but also
adversely affects the robustness of the modelling
process. Therefore, this method is not suitable for
flow simulation of the process involving a large
extrusion ratio.
In contrast, the Eulerian description uses a
fixed mesh that is independent of the deforming
material. This method is ideal for analysing steadystate forming processes. As no re-meshing is
required, large deformation can be simulated with a
relatively low computation cost. For modelling
purposes, extrusion can be regarded as a quasisteady-state process. Except for the early and final
stages of the cycle, the deformation zone in the
billet is largely stable, although variation in
temperature may change its shape to some extent.
From the die design viewpoint, a steady-state flow
analysis is adequate and justifiable, as the die has
to be designed based on the flow pattern at a
particular stage of the extrusion process.
Accordingly, only steady-state flow analysis using
the Eulerian description is considered in the
present work.
&ij
(2)
(3)
(4)
(5)
Friction Conditions
ij = 2
)] n exp [(Q/(RT)]
(1)
2
Vs
m tg 1( Vo )( Vs
)
Vs
3
(5)
&
Iterative Algorithm
The non-linear equation (4) is solved by means
of iteration from an initial estimate. Very often, the
initial approximation is far from the true solution and
the difficulty in convergence occurs. For problems
involving a large number of degrees of freedom, a
slow convergent rate greatly increases the
computation cost, making numerical analysis
virtually impractical. In the last decade, a persistent
effort has been made to find an efficient and robust
iterative algorithm applicable to the rigid
viscoplastic finite element method (16).
This
resulted in the development of a novel iterative
algorithm (17) which has been successfully applied
to our die design simulations. According to our
experience, usually a solution can be obtained
within fifteen iterations. No divergent problems
have ever been encountered since the
implementation of this algorithm.
(6)
4. APPLICATIONS
Dogbone Section
Akeret and Strehmel (19) conducted extrusion
trials on a dogbone section which consisted of two
round bars, 10 and 14 mm in diameter respectively,
and a connecting web of 2 mm in thickness (Figure
1). This section was selected as the first test case
for the present study to check out whether the
simulation could correctly predict the width effect.
The extrusion conditions were assumed as follows:
Container diameter 150 mm;
Ram speed 5 mm/s;
Billet and die temperature 450 C;
Container temperature 400 C.
2
10
14
60
2.0
1.8
1.6
1.4
Relative flow rate
1.2
calculated
averge
1.0
0.8
0.6
0.4
0.2
0.0
0
10
20
30
40
50
60
70
80
Relative
flow rate
empirical
simulation
10 bar
1.21
10
web
0.36
14 bar
1.53
11
13
Wide Section
1.4
1.2
1.0
Container
184 mm
0.8
0.6
0.4
0.2
e
b
c
a
d
154.4
Figure 5:
f
g
0.0
3.2
20
40
60
d
80
100
e
120
140
160
180
200
220
10
9
8
4. DISCUSION
7
6
5
4
by simulation
by die designer
2
1
0
0
b
20
40
c
60
80
100
e
120
f
140
160
180
200
g
220
Finite element meshing is the most timeconsuming part of the simulation process. In our
method, a 3-D mesh is obtained by extruding 2-D
elements along the extrusion direction. It means
only two-dimensional discretion of the die face is
required for modelling of flat-face dies. Excluding
bearing details from the model further simplifies the
process of mesh generation. Nowadays most
commercial CAD systems and finite element
software packages are equipped with efficient twodimensional mesh generator. Thus, finite element
meshing should not constitute a major barrier to
numerical simulation of flat-face dies. According to
our experience, provided appropriate computer
hardware and software are used, model
preparation for a flat-face die with moderate
complexity should take no more than 3 to 4 hours.
Computation Time
Table 2 provides the CPU time used for the
present simulations on a Silicon Graphics (Indigo)
workstation. It shows that the current modelling
technique has reduced the CPU time required for a
detailed simulation to a level acceptable for
practical applications.
Dies
No. of nodes
No. of elements
No. of iteration
CPU time (minute)
Table 2
Dogbone
3510
2904
14
67
Wide Section
3090
2420
11
44
5: REFERENCES
1. Baker, B, Aluminium Extrusion Die Design,
Kaiser Aluminium & Chemical Sales Inc., 1963.
5: CONCLUSIONS
The present study demonstrates that it is
feasible to use finite element simulations to design
bearings for flat-face dies. For two dies modelled,
the simulation correctly predicted the trends in flow
rate distribution at the die entry. The bearing
lengths determined from the simulation generally
agreed with the experimental observations and
actual bearing assignment.
Simplification of the simulation process has
been achieved by implementing a series of novel
modelling techniques.
This has resulted in a
significant reduction in total computation cost with
minimal compromise in the quality of the analysis.
Thus it is now possible to apply finite element
simulations to studies of specific dies, especially
those that require stringent control of metal flow.
In our approach, the input from the die designer
still plays an important role in bearing length
assignment.
Due to the cost constraint, it is
impractical to include all process variables into a
finite element model.
Some tasks, such as
compensation for die deflection and bearing
smoothing, can be easily performed by a skilled die
designer, but can pose a challenge to a completely
automatic process.
Thus, computer aided die
design with finite element methods should be
regarded as a supplementary technique rather than
a replacement for the experience of a die designer.