Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                

Bearing Design

Download as pdf or txt
Download as pdf or txt
You are on page 1of 6

Available online at www.sciencedirect.

com

ScienceDirect
Procedia CIRP 12 (2013) 276 – 281

8th CIRP Conference on Intelligent Computention in Manufacturing Engineering

Optimization of bearing lengths in aluminum extrusion dies


R. Mayavarama*, U. Sajjaa, C. Seclib, S. Niranjanc
a
Altair Engineering, 7800 Shoal Creek Blvd, Ste 200N, Austin, TX, 78757, USA
b
Altair Engineering, Srl, Via Livorno,60 c/0 Environment Park 10144 Torino, Italy
c
Altair Engineering Pvt. Ltd., Prestige Tech Park, Bangalore 560103, India
* Corresponding author. Tel.: +1512-467-0618; fax: +1-512-467-1382. E-mail address: msr@altair.com.

Abstract

Bearing design plays a key role in producing high quality extruded products. Balancing the flow by adjusting the bearing lengths to
eliminate profile distortion is a common practice in the industry. The bearing correction, however, is accomplished by costly trial
and error approaches largely relying on the expertise of die designers. In this paper we present a numerical algorithm to optimize
the bearing lengths that produce uniform velocity at the die exit. This is based on a finite element model to solve material flow
during extrusion. The solution approach involves iteratively computing velocity, temperature, and strain fields during extrusion and
updating the bearing lengths until balanced flow is achieved. Different aspects of bearing design and robustness of this
optimization technique are illustrated through a few examples.
© 2012
© 2013 The
The Authors.
Authors. Published
Published by
by Elsevier B.V. Open
Elsevier B.V. access under CC BY-NC-ND license.
Selection and/or peer-review under responsibility of Professor Roberto Teti
Selection and peer review under responsibility of Professor Roberto Teti
Keywords: Extrusion; Simulation; Optimization; Bearing.

1. Introduction existing dies. This trial and error correction process is


time consuming and ineffective. This calls attention to
Extrusion of metals is a commercially important numerical simulation and optimization of the bearing
process used for mass production of parts. Aluminum design.
extrusions are increasingly used in a wide variety of Modeling of extrusion process provides a complete
applications in aerospace, automotive, railway, medical, understanding of the material flow and heat transfer
electronics, consumer products and nuclear industries inside the die [1-4]. Numerical simulations together with
due to their high strength to weight ratio, formability and optimization algorithms offer advantages in designing
excellent thermal properties. In extrusion process, metal the bearing and effectively reduce the number of die
is subjected to compressive load and undergoes plastic trials. Reddy et al. [4] presents a response surface based
deformation as it is forced through the die to form the optimization approach for determining bearing lengths.
profile shape. Quality of the extruded shape depends on This study is performed with HyperXtrude®, a
many factors including flow stress, friction at the tool commercial finite element software for extrusion
walls, temperature increase due to stress work, and die simulation. This solver is coupled with an external
geometry. Of all these interrelated factors, friction in optimizer (HyperStudy®). Even though it is successful,
bearing region plays the most critical role in achieving this is a computationally expensive approach requiring
good product quality. To ensure the integrity of the several initial runs to build the response surface. The
profile shape and avoid distortions, the extruded material optimizer uses a general purpose algorithm that does not
should exit the die with uniform velocity. With the consider the specific aspects of extrusion die design and
advent of new alloys and demand for tighter tolerances, the nonlinearity of the problem.
designing a bearing becomes difficult especially for In the current work, a new optimization approach is
complex profiles with large thickness variations. presented for the determination of bearing lengths that
Adjustment of bearing lengths in different regions of the ensure the uniform flow of the material at the die exit.
die by machining is often the recourse for correcting This approach considers the nonlinearity of the flow and

2212-8271 © 2013 The Authors. Published by Elsevier B.V. Open access under CC BY-NC-ND license.
Selection and peer review under responsibility of Professor Roberto Teti
doi:10.1016/j.procir.2013.09.048
R. Mayavaram et al. / Procedia CIRP 12 (2013) 276 – 281 277

uses techniques specific to metal extrusion in adjusting 2.3. Bearing design and optimization
bearing lengths. This algorithm is implemented inside
HyperXtrude solver to avoid inefficiencies of external Bearing is the region of the extruded profile that
coupling. Several simulations have been performed to comes in contact with the die surface and it is defined
demonstrate the robustness of this approach. using a three-dimensional curve on the die surface.
Length of the bearing curve from the die face determines
2. Theory the contact area at any given point (Fig. 1). This length
varies along the profile cross-section. Bearing length is
2.1. Governing equations specified at a few points along the curve and interpolated
for other points. The points at which the length is
Metal extrusion process is modeled as an specified are .
incompressible deformation of a viscoplastic material in
an Arbitrary Lagrangian Eulerian (ALE) framework.
The material flow and heat transfer are governed by
conservation laws of mass, momentum, and energy.

v =0 (1)

v ) v = (2)

v ) T= q (3)

In the above equations, v is material velocity vector


relative to the grid velocity, is the total stress tensor,
is density, T is temperature, q is heat flux vector, and
is the heat addition due to stress work. The stress tensor
and heat flux vector in the conservations laws require
constitutive models relating them to primary variables. Fig. 1. Bearing profile is shown in grey color. This is the region
between the die face and the bearing curve.
These governing equations are solved using a
commercial finite element solver, HyperXtrude [5]. A reasonably complex profile will have a few
This work is focused on steady state simulation of direct hundred control points and adjusting all of them is
extrusion process. computationally expensive. Most extruded profiles have
repeating patterns and hence, the control points have
2.2. Constitutive model interdependencies. To make use of these patterns,
control points are grouped into optimized points,
The stress tensor, is written as a sum of a spherical dependent points, and fixed points (Fig. 2). Only the
pressure tensor and a viscoplastic deviatoric tensor. optimized points are directly adjusted, dependent points
derive their adjusted lengths from the optimized points,
= -P 1 + (4)
and the fixed points have a constant bearing length.
D v + vT) (5)

In the above questions, is the viscoplastic deviatoric


stress tensor, P is pressure, is viscosity, D is rate of
deformation tensor, and 1 is the identity tensor.
Viscosity in the above equation is computed from the
flow stress of the material, which is a function of strain,
strain rate, and temperature. In this study, Sellars
Tegart [6] and Hensel Spittel [7] models are used to
compute the flow stress.

Fig. 2. Classification of control points: optimized points are red in


color and dependent points are green in color.
278 R. Mayavaram et al. / Procedia CIRP 12 (2013) 276 – 281

The objective of the optimization algorithm is to 3.1. Example 1


obtain a uniform mass flow at the die exit. This
uniformity is measured by comparing the local velocity Fig. 4 shows the first example and this model has
to the extrusion speed. Fig. 3 shows the overall solution three legs of varying widths. Since the legs are of equal
algorithm. In this algorithm, bearing correction is length and arranged radially, the effect of frictional
accomplished by the solver using a comprehensive set of resistance on each of the legs, due to container surface,
flow rules based on friction, material flow imbalance, is more or less the same. In addition, the large central
local die opening thickness, and local/global ratio of opening provides a mechanism for mass flow to be
perimeter to cross-sectional area. These factors are balanced around the center. AA6063 alloy using
discussed in Miles, et al. [8]. When the desired Sellars-Tegart model based on data from [9] with 0.005
uniformity in exit speed is accomplished, the m/s ram speed and 700 K billet preheat is extruded.
optimization iterations end. Analysis is controlled by
specifying the allowed variation in the exit speed and
maximum number of optimization iterations.

Fig. 4. Cross-section of the profile used in example 1 with dimensions


in meters

Fig. 5 shows the nose cone and the exit velocity


distribution in the initial design. Material flow is
significantly slower in the thinnest leg and faster in the
thicker leg. This difference is a function of ram speed;
sometimes extruders take the short cut of pushing the
billet at a lower ram speed to overcome or reduce the
imbalance. However, this improvement comes at the
cost of productivity. Fig. 6 shows the nose cone and the
Fig. 3. Optimization algorithm used to adjust bearing profile balanced exit velocity distribution in the final design.
Initial and final bearing profiles are shown in Fig. 7.

3. Results and Discussion

Three profile shapes demonstrating various aspects of


bearing length optimization are presented in this section.
These three test cases are selected from the standard set
of test shapes in Miles et al. [8] and presented in
increasing order of complexity. Example 3 (heat sink
model) is presented in [8] as the ultimate test of bearing
profile design. These profiles are chosen based on the
cross-sectional shape, circumscribing circle, local width
of the opening, and the distance from the container wall.
All these features play an important role in bearing
design. For all three test shapes, a uniform initial
bearing design is used as the starting point for
optimization. Fig. 5. Nose cone and velocity distribution with initial uniform bearing
profile for example 1
R. Mayavaram et al. / Procedia CIRP 12 (2013) 276 – 281 279

Fig. 8. Cross-section of the profile used in example 2 with dimensions


in meters.

AA6063 billet with a preheat temperature of 700 K is


extruded at a ram speed of 0.005 m/s. Sellars-Tegart
Fig. 6. Nose cone and velocity distribution with optimized bearing model with data from [9] is used as the material model.
profile for example 1 Fig. 9 and Fig. 10 show the nose cone plot and
temperature distribution with the initial and final design.
Variation in the exit speed decreased from 14% to 0.5%
in six optimization iterations. There is less than 2%
variation of temperature in the profile region, which is
acceptable and will result in a good surface quality.

Fig. 9. Nose cone and temperature distribution with initial uniform


Fig. 7. Initial and optimized bearing profile for example 1 bearing profile for example 2

Variation in the exit speed decreased from 25% to 4%


in the six optimization iterations. Even though
theoretically there may be many feasible solutions,
manufacturability of a particular design largely depends
on the choice of control points and to some extent on the
minimum and maximum bounds for bearing length at
each control point.

3.2. Example 2

The second profile studied is shown in Fig. 8. The


profile has several steps with varying local die opening
thickness. Hence, the optimal bearing profile will have
largely varying bearing lengths. Since the flat surface is
shared by all the steps, this will affect the surface quality Fig. 10. Nose cone and temperature distribution with optimized
of the extruded part. bearing profile for example 2
280 R. Mayavaram et al. / Procedia CIRP 12 (2013) 276 – 281

3.3. Example 3 For this example, the Hensel-Spittel constitutive


model is used. This model considers the effect of strain
Fig. 1 and Fig. 2 show the profile shape of the final softening on flow stress. AA6063 billet at a temperature
example used in this study. This is a heat sink model of 703 K is extruded at a ram speed of 0.005 m/s. Fig.
with long fins of varying cross-section and a base with 11 and Fig. 12 compare the nose cone and the velocity
two thin feet. This is a challenging profile for bearing distribution of the initial and final design. Variation of
optimization. Fig. 2 shows the optimized and the exit speed reduced from 50% to less than 2% in 10
dependent points for this profile. Every dependent point iterations.
is related to an optimized point. The selection of
optimized points and the corresponding dependent points
is based on repeating patterns and fabrication
constraints. In addition, it is possible to identify some
points as
change. For each point, minimum and maximum bounds
have to be specified. In this example, the same minimum
and maximum bounds are applied to all the points.
Often, identifying and grouping the control points into
respective categories is the only step that needs to be
done by the designer. The rest of the process is quite
robust and automatic.

Fig. 13. Initial and optimized bearing profiles for example 3

This model has 812,441 elements and 263,583 nodes;


it took 17 hours CPU time to complete the solution.
Since the solver and the optimizer are internally coupled,
every new solution started with the previous solution as
the initial guess. Hence, a typical optimization run took
only 2.5 times the CPU time needed for a standard
steady state run. Fig. 13 shows the comparison between
the initial and final bearing profiles.

Fig. 11. Nose cone and velocity distribution with initial uniform
bearing profile for example 3

Fig. 14. Nose cone and temperature distribution with initial uniform
bearing profile for example 3

With the optimized bearing profile, variation of exit


Fig. 12. Nose cone and velocity distribution with optimized bearing temperature in the extruded profile is less than 5%,
profile for example 3 however, the fins were a few degrees hotter than the
base. This temperature distribution pattern changed
R. Mayavaram et al. / Procedia CIRP 12 (2013) 276 – 281 281

significantly from the initial uniform bearing to the final [2] Zhang, C., Zhao, G., Chen, H., Wang, H., 2011. Numerical
Simulation and Die Optimal Design of a Complex Section Thin-
optimized bearing as shown in Fig. 14 and Fig. 15. In the Walled Aluminum Profile Extrusion, Advanced Materials
optimized bearing the fins are uniformly hotter at the Research,p. 148.
tips than at the base. This is due to the uniformity of the [3] van Rens, BJE, Brekelmans, WAM, Baaijens, FPT, 2002.
Modelling friction near sharp edges using a Eulerian reference
strain rate pattern in the optimized bearing region. frame: application to aluminum extrusion, Int. J. Num. Meth.
Eng., 54:3, p. 453.
[4] Reddy, M., Mayavaram, R., Durocher, D., Carlsson, H.,

International Aluminum Extrusion Technology Seminar 2004, II,


p.171.
[5] HyperXtrude User's Manual, Version 11.0, Altair Engineering,
Inc., Troy, MI, 2011.
[6] Sellars, C.M., Tegart, W.J., 1972. Hot Workability, Int. Metall.
Rev; 1972;17:1, p.1.
[7] Hensel, A., Spittel, T., 1978. Kraft und Arbeitsbedarf bildsamer
Formgebungsverfahren, VEB Deutsche Verlag für
Grundstoffindustrie, Leipzig 1978.
[8] Miles, N., Evans, G., Middleditch, A., 1997. Bearing lengths for
extrusion dies: rationale, current practice and requirements for
automation. J. Mat. Proc. Tech.; 1997; 72, p.162.
[9] Prasad, YNRK, Sasidhara, S., 1997. Hot working guide: a
compendium of processing maps, ASM International, 1997.

Fig. 15. Nose cone and temperature distribution with optimized


bearing profile for example 3

Results shown in all three examples clearly indicate


that the optimized bearing profiles will serve as an
excellent starting point for shop floor try-outs.

4. Conclusions

A new algorithm for the optimization of bearing


lengths based on the thermo-mechanics of the extrusion
process is presented. Results presented clearly show that
the process is robust and efficient. A typical
optimization run took less than three times the CPU time
of a standard steady state run and this is independent of
the number of control points and the problem size.
Examples presented in the paper cover a wide spectrum
of extrusion profiles. The results of this study can be
extended to indirect extrusion and it is a subject matter
of future research.

Acknowledgements

Authors acknowledge the feedback and input


provided by Dr. Mahender Reddy while preparing this
paper and Altair for giving the permission to publish.

References
[1] Bastani, A.F., Aukrust, T., Brandal, S., 2011. Optimization of
-
element simulations, J. Mat. Proc. Tech.; 211, p. 650.

You might also like