International Conference on Aluminium (ALUCAST 2000)

COMPUTER-AIDED DESIGN AND ANALYSIS

FOR ZERO DEFECTS
By
B. RAVI
Associate Professor
Indian Institute of Technology, Bombay

ABSTRACT For a given component, die development

comprises broadly three tasks: design,
Inexplicable patterns of internal defects in manufacturing and tryouts. Die design
diecast parts are often because of includes cavity (parting, cores, draft,
borderline optimization of gating and allowances), feeding and gating systems
feeding systems. In other words, a (feeders, feed-aids, sprue, runners, gates,
particular die design is too sensitive to overflows, vents), cooling system and
process parameters such as melt mechanical system (ejection, guiding,
temperature or filling time. A sufficiently housings). Die manufacturing includes
robust die design can be determined process planning (NC tool path generation
through extensive (but expensive) die and verification), electrode manufacturing
tryouts. Computer simulation programs (for EDM process, if used), machining, die
can replace shop floor trials, but most of finishing (grinding, polishing, etc.),
the available programs are quite difficult- dimensional inspection (conventional or
to-use and computation intensive. To CMM-based) and die assembly.
overcome these limitations, we have
developed a hybrid approach combining Die tryout is usually the most cumbersome
automated design, intelligent analysis and and challenging task, involving several
castability health-checks. The design iterations of casting, inspection, fault
program includes encoded experience to analysis and modification of feeding
give a good first gating and feeding design and/or gating system. Each trial may take
and creates their solid models. Intelligent 1-2 weeks, and typically 2-4 trials may be
analysis combines physics-based required. Some complex castings (such as
simulation with practical experience to an engine block) may require 8-10 trials or
quickly predict internal defects in a more. Even then, the yield may be poor or
casting. The health is assessed and the die may continue to produce occasional
quantified by a set of castability criteria. defective parts.
With this approach, it is possible to
complete several virtual tryouts and finalize
a sufficiently robust die design in a single
day. The system has been successfully
used for gravity die cast parts and is being
extended to pressure diecasting. It is easily
customized for any new metal or process.

Keywords: CAD/CAM, Casting Design, Die Fig.1 Over design, borderline design and
Casting, Simulation. sufficiently robust design of feeder.

In some cases, the gating and feeding

INTRODUCTION design is just correct to get a good casting
under normal and controlled conditions.
Tooling development is not only the most Such a design is sensitive to changes in
expensive activity in die casting process, process variables. This means that even
accounting for over 70% of the lead-time to minor and expected variation in metal
the first article of approval [1], but also composition or pouring temperature
significantly influences part quality. Often, produce sudden and often unexpected
die development is a bottleneck in new increase in the level of defects. We refer to
product development. this as ‘borderline optimization’.
 International Conference on Aluminium (ALUCAST 2000)

To achieve the goal of zero defects, the Several researchers and diecasting
casting engineer should tilt the balance by engineers have encoded their die design
slightly over designing the casting, so it is experience in terms of equations, tables
sufficiently robust against expected process and graphs, which are available in
variations. However, determining the technical literature and handbooks [2,3,4].
optimal value of each design parameter These include equations for determining
itself requires a series of trials (especially the ideal filling time (as a function of
for new castings) at normal operating casting thickness, metal fluidity, die
conditions. Determining the sufficiently temperature, etc.), pressure head
robust value for each design parameter will (depending on material and application)
require even more trials (at expected limits and ejection force (with respect to cast
of operating conditions). The material, metal, surface area and length of core).
energy and labor costs for so many trials This knowledge can be incorporated in
may not be economically justifiable. computer programs, useful for preliminary
die design calculations. These can also be
The preferred way to reduce the time and interfaced with solid modeling programs to
cost of shop floor trials is by virtual semi-automatically generate 3D CAD
tryouts, by creating and simulating a 3D models of die elements. The models can be
model of the casting, described next. exported to NC tool path generation
software for die manufacturing. One such
VIRTUAL DIE TRYOUTS software, Diedifice, for pressure diecasting
is being developed by Neilsoft Ltd., Pune.
The 3D CAD models of the part and die can
be created using solid modeling programs
(AutoCAD, CATIA, Cimatron, I-DEAS, Pro-
Engineer, SolidWorks, Unigraphics, etc.). A
combination of operations such as sweep
(linear and rotational) and Boolean (add,
subtract, intersect) are used. This enables
better visualization, property computation,
rapid modification and compact archival.
Also, the 3D models can be transferred
(through standard DXF, IGES, STEP and
STL formats) and used for other activities,
including stress analysis, cavity shape
modeling, NC tool path generation,
automated inspection (by comparing CMM
data with original model) and process Fig.3 Die assembly design and modeling
simulation. Parametric and features-based
modelers (Pro-Engineer and SolidWorks) Simulation technology has emerged as a
enable modeling in terms of manufacturing boon to diecasting engineers to perform
features (hole, boss, rib, etc.). virtual trials, predict casting defects and
improve the die design without pouring a
single shot. This not only eliminates the
cost of die modification and material/
energy costs, but also provides a better
insight into the process and enables
exploring more alternative solutions.

Most of the die filling and solidification

simulation programs available today are
based on Finite Element Method [5]. The
main input is a 3D CAD model of the part
and die, created using a solid modeling
program. This model has to be meshed,
that is, broken down into a number of
simple elements (cubes or tetrahedrons). A
Fig.2 Die modeling and visualizing. smaller mesh size gives slower but more
reliable results. Adaptive meshing (finer in
 International Conference on Aluminium (ALUCAST 2000)

critical regions and coarse elsewhere) gives INTELLIGENT DESIGN ASSISTANT

faster results without comprising on
accuracy, but requires expertise in mesh The AutoCAST software provides a single
generation. After meshing, material integrated environment for casting design,
properties (density, thermal conductivity modeling, simulation, analysis and project
and specific heat at different temperatures, data management. Advanced geometric
and latent heat of cast metal) and process reasoning and knowledge-based functions
variables have to be either selected from a have been incorporated in the software,
library or specified interactively. Then the making it work like an intelligent assistant
boundary conditions have to be specified, to casting engineers. The methodology for
in terms of heat transfer rates from casting design (mainly feeding and gating
different regions of the die, which is systems), process simulation, castability
influenced by the properties of the die analysis and optimization is explained here
material and the variable air gap between with the help of a case study.
the part and die, besides process variables.
The product is an Aluminum alloy (LM4)
Simulation programs give accurate results compressor cover casting produced by
if the CAD model, FEM mesh, material permanent mold (gravity die casting)
properties and boundary conditions are process. The largest dimension of the part
correct (otherwise: garbage in, garbage is 186 mm, minimum wall thickness is 4
out). Material property and boundary mm and the weight is 1.2 kg.
conditions data usually have to be
determined and fine-tuned through The part is modeled using AutoCAD 2000
experimentation. This may take several and exported to a file in the standard STL
weeks, which is beyond the scope of most format. The STL file is then imported into
small and medium companies. Second, AutoCAST software. The part model is
these programs require engineers with oriented along a vertical parting line, as
higher academic qualifications, CAD/CAM per the current practice. The program
skills and die casting experience to suggests the mold size and after
conduct the simulation and interpret the confirmation from the user, creates the
results properly. Third, the programs are mold model around the part model. Both
computation intensive, and require models can be visualized (Fig.4).
powerful engineering workstations. Even
then, a single iteration of CAD model
creation, mesh generation, boundary
condition specification, simulation and
visualization can take 2-5 days for a
complex part. Thus it may take several
days to arrive at an optimal die design.

To slash the total die development time, it

is important to minimize the time taken for
die design, casting simulation and analysis
of results. This is possible by adapting an
integrated approach for the above three
tasks. An automated die design program,
which incorporates casting domain
knowledge, will quickly give a good first Fig.4 Part modeled, imported and oriented.
design. An intelligent simulation program,
which does not require cumbersome user A preliminary simulation of casting
inputs, will reduce the time for virtual solidification was carried out. The program
trials. Finally, automated interpretation of automatically generates the mesh, sets the
the results will point out directions of boundary conditions (based on metal and
improvement and reduce the number of process), computes the progress of
design iterations. Such an integrated solidification, post processes the results
approach, based on an intelligent casting and displays the location and extent of
design system, called AutoCAST, has been shrinkage porosity (Fig.5). All this took less
developed. This is described next. than 15 minutes on a Pentium computer.
 International Conference on Aluminium (ALUCAST 2000)

The layout is then changed to horizontal.

For this, the part model is turned in the
mold, the parting is set to horizontal and
the mold is remodeled. The feeder location
is specified by selecting the connection
point on the part surface. The program
automatically computes the significant
modulus (ratio of volume to cooing surface
area) around the hot spot in the part, and
designs the feeder dimensions to ensure
that its solidification time is longer than
the hot spot. After user confirmation, the
feeder model is created automatically.
Again, simulation shows that the feeder
alone is unable to eliminate porosity, and a
Fig.5 Simulation predicts porosity chill is also modeled (as described earlier)
at the bottom of the casting.

The gating layout is also created by the

program semi-automatically. The user only
specifies the ingate connection points on
the part surface. The program
automatically suggests the sprue location
and the runner path, which can be
modified by the user if necessary. The ideal
filling time is computed based on the
metal, process, weight, wall thickness and
pouring temperature, for confirmation or
modification by the user. Then the
dimensions of all gating elements are
automatically computed, and a solid model
Fig.6 Chill designed and modeled of the gating system is created.

The solid models of the feeder, chill and

gating system, along with the part model,
are displayed for visual feedback (Fig.8).
The size of feeder and chill are optimized
through several iterations of design-model-
simulate-analyze until simulation predicts
zero porosity defects even for the highest
quality requirements (Fig.9).

Fig.7 Simulation still shows porosity

To increase the heat transfer rate near the

porosity zone, a chill is modeled in the
center (Fig.6). This involves specifying the
chill type, location, shape and size. The
software automatically models the chill and
considers its effect on casting solidification.
Simulation shows that porosity is reduced,
but not completely eliminated (Fig.7). Fig.8 Horizontal layout with feeder, gating
 International Conference on Aluminium (ALUCAST 2000)

LEVERAGING DOMAIN KNOWLEDGE

We are also using AutoCAST as a platform

to capture and leverage specialized domain
knowledge of a particular organization, for
even better and faster decision support.
There are two areas of focus: (1) improved
design calculations to obtain a good first
design of feeding and gating systems, and
(2) more accurate simulation results for
improved matching between predictions
and actual observations.

To improve the results of design

Fig.9 Optimized, defect-free casting calculations, various equations are being
generalized and their coefficients made
Finally, castability health-checks assess accessible to the user, who can easily
the casting design on a scale of 0-100. A change them. For example, ideal pouring
value of zero implies impossible to cast and time calculation is based on three
100 implies ideal castability. The actual parameters (weight, thickness and fluidity),
values usually lie in-between. The Fettling and six factors, as follows:
criterion, for example, is assessed by
comparing the casting wall thickness POUR_TIME_IDEAL =
POURTIME_FACTOR*FLUIDITY_FACTOR*
(connected portion) to thickness of feeder fluidity/1000* (SIZE_FACTOR +
neck or ingate. Weights are attached to (THICK_FACTOR* thickness/20))*
each criterion, and the weighted sum of power ((WEIGHT_FACTOR * weight ),
assessment of individual criteria indicates WEIGHT_POW_FACTOR)
the overall health of the feeding or gating
design. The weights may be different for The POURTIME_FACTOR is an overall
each project, depending on customer correction factor for ideal filling time and
requirements and production constraints, can be modified if there is a consistent
and can be changed by the user. The difference between suggested and practical
weighted assessment is not only useful as values (always higher or always lower). The
an absolute measure, but also for SIZE_FACTOR accounts for the casting size
comparing the relative health of two and is taken from a table of values against
different casting designs. Table 1 shows casting size ranges. Similarly, the
the assessment of the final layout of the THICK_FACTOR accounts for the average
compressor cover casting. The weighted wall thickness of the casting. A particular
assessments of feeding and gating criteria company may have its own knowledge base
are above average, but can be improved of ideal pouring time for a family of
further by part design for castability. castings (size, shape, wall thickness) of a
particular metal produced by a particular
Table 1. Castability Health-Checks process. This knowledge can be
incorporated in the library of metal-specific
Criterion Weight Assessment process characteristics to match the
suggested ideal pouring time with the
values used in practice.
Feeder Yield 50.0% 63.7%
Feed Efficiency 30.0% 02.6%
The second application is fine-tuning the
Feeder Fettling 20.0% 98.0%
results of solidification simulation for an
Weighted Assessment 52.2% existing or a new combination of metal and
process. This is really useful when
Filling Time 20.0% 62.5% accurate values of the thermo-physical
Choke Velocity 30.0% 33.3% properties (density, thermal conductivity,
Gating Yield 20.0% 54.5% specific heat, etc.) of the casting metal or
Gating Fettling 30.0% 65.3% mold material are not available and the
Weighted Assessment 53.0% boundary conditions (metal-mold interface
air gap thickness, heat transfer coefficient,
 International Conference on Aluminium (ALUCAST 2000)

etc.) are unknown. For example, the CONCLUSION

FR_RANGE_FACTOR interprets the results of
solidification simulation to map the extent To ensure zero defect castings, it is
(spread) of shrinkage porosity from the hot necessary to correctly optimize the casting
spot, depending on the quality level design (feeding and gating). We have shown
desired. This factor can be adjusted to how this can be achieved by combining
match the prediction with observed results. automated design, intelligent simulation
For this purpose, several simulation runs and castability health-checks. By
are carried out, each with a slightly integrating all three in a single
different value of the factor, until the environment, the design-simulate-analyze
matching is perfect. The factor is then cycle time has been reduced to less than
incorporated in the database for automatic one hour. This makes it possible to
application when the same combination of optimize even complex castings in a single
metal and process is used in another day. We have also shown how the system
casting project. can be customized to leverage the
specialized domain knowledge of a
In a recent project, the software was particular company for reuse on new
customized (in a single day) for a high-Mg projects. This will ensure better casting
aluminum alloy ASTM tensile bar casting, design decisions, faster than ever before.
produced by gravity die casting. The
solidification analysis parameters were
fine-tuned for one layout (Fig.10), and then ACKNOWLEDGEMENTS
used to analyze a different layout (Fig.11).
The predictions perfectly matched the The author acknowledges the support of
known results for the second layout. Advanced Reasoning Technologies, New
Mumbai in developing the integrated
approach to die design based on their
AutoCAST software. Thanks are also due to
Mr. Deepak Tanksale of Neilsoft, Pune for
providing the solid models of the crankcase
die cavity and die assembly. The
compressor cover simulation project was
carried out in association with Cummins,
Pune. The customizing project using ASTM
specimens was carried out in association
with Crompton Greaves, Mumbai.

REFERENCES
Fig.9 First layout used for customizing
1. R.C. Creese, “Benchmarking and
Lead Time Reduction,” American
Metalcasting Consortium Project
Report, 1996.
2. American Society of Metals, Casting
Design Handbook, ASM, Metals
Park, Ohio, 1962.
3. Arthur C. Street, The Diecasting
Book, Portcullis Press, London, 2nd
Edition, 1986.
4. Russ Van Ress, Gating Diecasting
Dies, North American Diecasting
Association, Ohio, 1996.
5. R.W. Lewis, H.C. Huang, A.S.
Usmani and M.R. Tadayon,
“Solidification in Casting by Finite
Fig.10 Second layout used for validation Element Method,” Material Science
and Technology, 6(5), 1990.