An Application of 3-D Solidification Analysis
to Large Complex Castings
Chung-Whee M. Kim
Frank J. Sant
[Presented at the 2nd Pacific Rim International Conference on
Modeling of Casting and Solidification Processes, January, 1995]
Abstract
We have provided a tool to casting engineers for solidification
analysis of complexly shaped parts for use on the foundry floor.
This tool is accurate, timely, flexible, and user friendly.
First, we chose to model the process as a heat conduction problem
with latent heat of fusion. Then, the Finite Element Method is
chosen for a numerical method since it is geometrically flexible.
A unique time integration scheme is used to obtain stable,
accurate, and efficient results.
A new meshing technique is developed specifically to utilize the
geometric flexibility of the Finite Element Method to obtain accurate
results for the casting and die/mold. Also, the meshing technique
provides convenient editing and updating of the three-dimensional
finite element mesh for subsequent analyses, allowing the engineer
to quickly and easily obtain the ideal process.
The new technique, with an efficient solidification analysis program,
allows large complex automobile castings to be analyzed with minimum
requirements of computer resources. A large (725,000) finite
element model of an automobile V6 engine block casting can be
analyzed on mid-range engineering workstations.
1.0 Introduction
Casting solidification analysis has become an important part of
everyday foundry practice. Yet, a large scale three-dimensional
(3-D) solidification analysis, such as an automobile engine block,
is still impractical, mostly because of the geometrical complexity
of the casting.
Applications of FEM to shaped castings have been limited. This is
because most mesh generation methods have been developed for stress
analysis, which requires modeling of the part only. Foundry
applications require inclusion of the mold/die in process modeling.
Therefore, it was necessary to develop an enmeshing technique with
the geometric flexibility of FEM, plus the ability to model both the
part and the mold/die.
2.0 Solidification Modeling
Although there exist different levels of complex solidification
modeling for casting processes, we chose to model the casting
solidification process as a heat conduction problem with latent heat
of fusion. This model is simpler than heat and mass transfer models,
yet more accurate than a pure geometric approach. Thus, one can
write: [1]

3.0 Numerical Method
To solve equation (1) for a complexly shaped casting, one needs to
use a numerical method. The Finite Difference Method (FDM) and the
Finite Element Method (FEM) are both particularly suitable for this
task. A simple but not complete comparison of these methods is
illustrated here. (Note that the Control Volume/Direct Difference/
Finite Volume methods are another possibility, but they may suffer
model construction difficulties similar to those of FEM in practical
applications.)

First, FDM offers much easier formulation than FEM, therefore, an FDM
program can be simpler and require less computer resources. On the
other hand, an FEM mesh has inherent geometric flexibility, unlike an
orthogonal FDM mesh. Thus, FEM is a more suitable method for casting
solidification analysis, due to the geometric complexity of most
castings. However, in practice, FDM programs offer simpler, although
crude, meshing schemes, in whihch the pre-set orthogonal mesh and
casting geometries are overlapped. This may cause difficulties in
sorting out the inside mesh from the remainder. Also, satisfactory
enmeshment of curved surfaces and/or thin walled castings is
inherently very difficult, if not impossible, with FDM programs.
Most FDM programs depend upon the mesh generation technique developed
originally for structural analysis, where only the part needs to be
meshed. For casting solidification analysis, both the part (casting)
and the surrounding material (mold/die) need to be modeled for
accurate results. The existence of the mold/die makes the FEM mesh
creation procedure very difficult with conventional meshing
techniques, in addition to making the total number of elements very
large. Thus, we could summarize that meshing of FDM for casting
solidification is simple, but is limited. Also, meshing for FEM is
flexible, but is difficult. At this point, there are more
commercially available FDM software programs for casting solidification
than FEM based ones.
4.0 Meshing Technique
Geometric flexibility of the FEM is essential to casting engineers to
obtain accurate solidification analysis results. In addition, stress
analysis of parts and/or dies can be performed with FEM with little or
no modification, along with fluid flow analysis.

Therefore, we
decided to use FEM over FDM. As mentioned earlier, one of the biggest
obstacles to use FEM effectively for casting solidification is
enmeshment, becasue traditional meshing techniques have been
developed for stress analysis. To ease this difficulty, we have
developed a new meshing technique.
Our new meshing technique utilizes two-dimensional (2D) meshes to
build a three-dimensioanl (3D) mesh. First, a 2D base mesh is created,
which describes most of the part's complicated geometry. Next, one or
more 2D extension meshes are created. See
Figure 1. The extension mesh(es) 'grab' the base mesh elements,
adn 'grow' them into a 3D mesh. The 3D mesh has a layered structure.
Every 3D element is created by extension of a 2D element in the base
mesh. Each element in the 3D model is related to the elements
topologically 'above' and 'below' it, which are all created from the
same original 2D element. Finally, a batch file is created, whic can be
edited and reach time a change is made to the model. The batch file
is a great benefit of the meshing technique, since it greatly simplifies
editing of the model for subsequent analyses of part an/or die
changes.
This method is particularly suitable to casting solidification, but
it is also useful for fluid flow and stress analyses.
5.0 Analysis
5.1 FEM
After semi-discretization in space using finite elements, equation
(1) becomes

Note that latent heat of fusion effects can be included in [M]
or {f} depending on the formulation. [2] [3]
5.2 Time Integration Scheme
Equation (2) further reduces to

at each time step, with a time integration scheme.
To reduce equation (2) further, a time integration scheme has to be
applied to equation (2). In the case of a casting solidification
simulation, the solution domain consists of casting and mold, which
have considerably different initial temperatures and material
properties. These severe initial conditions, which are assigned at
the interface(s) of the sub-domains, could cause some oscillations
at early stages even with an unconditionally stable integration
method.[4] However, the fully implicit method is truly stable.
Therefore the fully implicit time integration method seemed to be
preferable for casting solidification analysis. Although, it is
well known that this method can be less accurate than the Crank-
Nicholson method. Therefore, we combined the advantages of the
fully implicit method and the Crank-Nicholson method. That is, we
used the fully implicit method at the early calculation stage, then
change to the modified Crank-Nicholson method, in order to improve
efficiency without losing accuracy.[5]
To solve equation (3), the most general purpose FEM program, originally
developed for stress analysis, uses a kind of direct solver. This
solver is very powerful to solve relatively small systems of linear
equations, in terms of available computer resources, but its memory
requirement grows exponentially with the number of unknowns. In
casting solidification analysis the system of equation (3) becomes large
mainly due to the mold components. The elimination of the mold from
casting solidification analysis has been tried with limited success.[6]
Recently, the iterative solver regained its popularity for large
systems of equations, arising from the FEM, because they generally
require less computer resources compared to direct solvers. In
particular, the conjugate gradient (CG) method became popular for FEM
application.[7]
This technique minimizes the residual, {r}, of equation (3).

In theory, convergence can be achieved at a finite number of iterations.
In real problems, however, convergence is slow or may not be
achieved at all. Therefor the CG technique is used with 'preconditioning'
to accelerate convergence. With the preconditioning matrix [P],
equation (3) becomes

The choice of [P] is very important to obtain results efficiently.
For best results, [P] should be close to the inverse of [A], and it
should be easy to calculate. We adapted the Symmetric Successive
Over-Relaxation (SSOR) method for preconditioning because we did not
need to calculate any new matrix coefficients.[8]
6.0 Application
Solidification and fluid flow analyses were performed on a North
American V6 engine block. The engine block (to be produced in Europe)
is a gravity semi-permanent mold casting with cast-in cast iron
liners. Because of an integrated water crossover and high pressure
oil lines, casting integrity is a major concern. This engine block
will be used in high volumes, so cost and ease of manufacturing were
also major considerations.
6.1 Mesh Generation
Due to the size and complexity of the casting, mesh generation of the
entire process was a challeng. Using the previously discussed mesh
generation techniques, a finite element mesh of the casting, gating,
liners, permanent mold, water lines, and cores was produced. The
finite element mesh took six weeks to build. The total mesh size
was approximately 650,000 elements and nodes. A majority of these
elements are in the mold and cores. Interface elements, which are
automatically created in the analysis software, made the total mesh
size 726,000 elements and 724,000 nodes. Figures
2 and 3
show the casting with the cast in liners.
6.2 Solidification Analysis
The importance of cyclic solidification analysis for permanent molds
has been reported.[9] Therefore, cyclic solidification was performed
on the V6 block. The casting cycle was divided into three segments:
1)Dwell (solidification), 2)Mold Open (convection to air), and 3)
Placement of the cast in liners and cores. Five cycles were simulated
to reach a quasi-steady state of the permanent sections of the die.
This analysis was performed on an HP 735/99MHz computer and took 13
hours to run.
Solidification simulation indicated several areas for potential
porosity. Potential shrinkage porosity can be identified by iso-contour
plots of solidification time. Figure 4
shows the isolated region of liquid metal. Cooling lines and risering
are then modified to ensure directional solidification. Other areas
of potential porosity can be identified by cut sections seen in
figures 5 and
6. These areas will not be identified
without cyclic analysis. These areas are caused by an "over heated"
section of the die.
Modifications of the mold and process were modeled to ensure a
directionally solidified casting. Due to the proprietary nature of this
component, we cannot discuss mold and gating configurations.
7.0 Conclusion
A new 3D enmeshing technique for FEM has been developed for shaped
casting applications. This technique, with an efficient solidification
analysis program, allows large complex castings to be analyzed with
minimum requirements of computer resources.
The new meshing technique has been successfully applied to an
automobile V6 engine block casting. Several changes have been made to
produce quality castings based on these analyses.
8.0 Acknowledgments
Thanks to J. DeLeeuw of EKK, Inc., for review and discussions.