Shawn Mahaney
Dr. Chung-Whee Kim
EKK, Inc., Walled Lake, Michigan
Foundry process simulation software has become a powerful tool for improving
casting processes. This is especially true for die casting, where tooling
costs are high. Reducing the number of trial-and-error development cycles
is the key. Simulation software can help reduce die development costs up-front
and aid in improvement of existing dies. In this article, we discuss what
is important in simulating die casting solidification and filling processes
and we show a practical example.
Since 1991, EKK has developed and marketed solidification and filling analysis
software for the metal casting industry. EKK also offers consulting services
using these software tools. This software is for all casting processes and
materials, however they are particularly suited to aluminum and magnesium
die casting processes.
Most commercially available solidification and mold (cavity) filling analysis software use either the Finite Element Method (FEM) or the Finite Difference Method (FDM, some vendors call this the Control Volume Method). Although the detailed mathematical discussion is not important to the die casting engineer, it is important to realize the practical differences. FDM uses an orthogonal mesh to represent cavity and die geometry (figure 1a). As you can see, it is very difficult to model a complex casting shape, especially a thin wall casting, with a reasonable number of elements. Even use of millions of elements may not improve the analysis result too much because the interface length on any out-of-plane section is still incorrect. This casting/die interface is one of the most influential variables in metal mold process.
FEM, on the other hand, can use not only a non-orthogonal mesh (figure 1b), but also different types of mesh where needed. In FEM, it is easy to incorporate zero-thickness interface elements, which model the cast metal/die metal contact. Because of this interface element, we can separate the casting and die surface temperatures. This is not the case with FDM. The geometric flexibility of FEM is essential for die casting engineers to obtain accurate analysis results.
In addition, stress analysis of parts and/or dies can be performed with FEM
with little or no modification of the mesh. Successful use of the same FEM
mesh for casting solidification analysis and stress analysis has been reported
[1]. Thus, FEM is better suited to die casting analysis
Historically, filling simulation software has used FDM [2]. With the recent
development of free-surface tracking algorithms for finite element flow solvers,
use of FEM meshes for accurate flow solving has become practical. FEM meshes
are especially attractive for thin-walled die castings. The FDM meshes for
thin-wall shapes require a huge number of cells to adequately mesh the thin
walls and the die blocks uniformly. Also, the geometric error in "stair stepped"
meshes can introduce serious error. There are ways to work around these problems
computationally, but they have a negative impact on efficiency. FEM does
not suffer from these problems at all, as element size and orientation can
be adjusted as necessary. Accurate and stable FEM flow solving is now available
with reasonable resources.
Use of the geometric flexibility of FEM for die casting process simulation
requires a new type of mesh generator (pre-processor). Most mesh generation
software has been developed for stress analysis, which only requires modeling
of the part. Die casting applications, however, require inclusion of the
die information (including water/oil lines, die inserts, etc.) so the ability
to generate a multi-region 3-d mesh is needed. Also, the die casting engineer
needs to test various casting process configurations. Thus, ease of modification
of the mesh is essential. To this end, an unique mesh generator has been
developed [3]. This mesh generator is not fully automated, but has the built-in
flexibility the die casting engineer needs.
Because of time and resource limitations, often one cycle of solidification
analysis is performed under a uniform initial die temperature assumption.
These results, however, sometimes mislead the die casting engineer, because
these results do not reflect the individual mold design and casting parameters,
especially cooling effects. Thus, cyclic analysis, wherein several casting
cycles are simulated in succession, is essential to the simulation of die
casting processes. The computational efficiency of the analysis software
is especially critical for cyclic analysis with limited computer resources
in a reasonable length of time.
Ganton Technologies of Sturtevant, Wisconsin was chosen by New Venture Gear of East Syracuse, New York to produce an automotive die cast aluminum transfer case housing. The process was modeled with foundry process simulation software. At about the same time, early trials produced castings with serious mis-runs and other less-serious thermal problems (figure 2). Several significant die changes were quickly simulated. When the changes were incorporated into the die, castings with greatly improved quality were produced.
Fluid filling and thermal analyses were performed to provide process information
which was used to assist in die and process development. Multiple iterations
of each were performed, reflecting product and process changes, which "greatly
reduced the development and sampling time and cost for the die," according
to Doug Guinn, Manager of Product Design and Analysis at Ganton. This article
documents the results of the analyses and their use in improving the
process.
Model construction and analyses were performed using finite element software
packages commercially available from EKK, Inc. The model includes casting,
runner and gating system with biscuit, overflows, main inserts, operator
and helper slides, main bores cores, and water lines. The model consists
of approximately 360,000 nodes and 393,000 finite elements. Shaded pictures
of the baseline casting and runner/gating system appear in
figure 3a. The model from the improved process
appears in figure 3b.
The die components consist of two upper stationary inserts with two bore cores, one lower ejector insert with three cores, and two slides (operator and helper.) The die components include all water lines as detailed in blueprints, namely one and four water lines in the lower and top upper inserts respectively, nine water lines in the operator slide, eight water lines in the helper slide, nine water lines in the ejector insert, and one fountain in each of the main cores. The latent heat release curve for aluminum 380 was modeled and included in the solidification analysis.
A five part process cycle was modeled, with a total cycle time of 85 seconds. Five cycles were modeled to ensure quasi-steady state values in the die were obtained. A sixth and final solidification cycle was also modeled. The numerical efficiency of current analysis software allows such a simulation to be run over a single workday on a modest workstation or PC.
Two simulated thermocouples were included: one in the top stationary insert,
and one on the casting/die interface in the ejector insert. A typical plot
of these thermocouple traces, as in figure 4,
shows the die approaching steady-state operating temperatures.
The filling models were subsets of the models used for the thermal simulations.
The portions used consisted of the cavity, overflows, runners, and a short
section of the shot chamber. There were a total of approximately 65000 nodes
and 50000 finite elements in the models. The simulations were run on the
full Navier-Stokes equations with a finite element free surface tracking
algorithm. Mold venting and backpressures were not modeled for these analyses,
although this can be done with some increase in analysis time. Typical simulation
run times were around 40 hours on a low end UNIX workstation.
Die temperature pictures for the first analysis, as in figure 5a, show hot and cold spots in the die. Added cooling was recommended where the operator and helper slide mate near the upper two gates. Also, the casting fills with flow across the large cool area, which could cause mis-run problems. It was recommended that two coolers be turned off behind this area.
In the final iteration, a new runner system was implemented. The new runner
between the ejector and side cores adds heat to this area, somewhat relieving
the temperature problem. Figure 5b, however,
still indicates cool spots on the side cores that could overcool the liquid
metal entering the cavity, causing filling defects. These results still suggest
shutting off waterlines in this area.
The simulation of the original process shows the flow around the helper side of the casting outpaces the flow around the operator side, resulting in the last place to fill being not at the side core parting line at the end of the casting, but on the operator side in an area of heavy bosses and no venting potential (figure 6a).
The objective with the subsequent design iterations was to balance the flow
between the operator and helper sides of the casting causing the flow fronts
to meet at the side core parting line at the end of the casting. Both the
second and third designs came closer to the desired filling pattern, but
still filled slightly past the parting line to the operator side. The fourth
design with four gates on a large runner between the ejector and side cores
filled the casting as desired, and included a new overflow between the side
cores to collect metal at the last place to fill and add heat to that end
of the die (figure 6b).
Finite Element mesh generation, solidification analysis, and filling analysis
software are available to die casting engineers. The Finite Element Method
allows accurate and efficient representation of any shape, including thin-walled
die castings. Available software has been applied to an aluminum die cast
transfer case and helped improve casting quality while shortening die development
time.
The authors wish to generously thank Ganton Technologies and New Venture
Gear for making information in this article available. We also wish to especially
thank Doug Guinn of Ganton for working closely with EKK on the analyses and
keeping us informed of later progress on this casting.
1) Volz, Todd J. and Mihara, Takeshi, "Optimizing a Squeeze Cast Wheel Design", Transactions of the 18th International Die Casting Congress and Exposition, pp. 13-15. NADCA paper no. T95-014
2) Hirt, C. W., and Nichols, B. D., "Volume of Fluid (VOF) Method for the Dynamics of Free Boundries," Journal of Computational Physics, 39, 201, 1981.
3) Kim, Chung-Whee, "A Cyclic Analysis of Permanent Mold Casting," Modeling of Casting, Welding, and Advanced Solidification Processes VI, 1993
Figure 1a - Finite Difference Mesh
Figure 1b - Finite Element Mesh
Figure 2 - Casting from Initial Process
Figure 3a - Shaded Finite Element Model
of Original Casting
Figure 3b - Shaded Finite Element Model
of Improved Casting
Figure 4 - Cyclic Thermocouple Traces
Figure 5a - Slide Surface Temperatures Before
Filling, Original Process
Figure 5b - Slide Surface Temperatures Before
Filling, Improved Process
Figure 6a - Filling Pattern, Original
Process
Figure 6b - Filling Pattern, Improved
Process
