Optimization of Casting Process Design with SOLIDCast and HyperOpt

1
Optimization of Casting Process Design with
SOLIDCast and HyperOpt
2
What is OPTICast?
OPTICast is a new software program from Finite
Solutions, Inc., developers of the SOLIDCast
Solidification Modeling System (formerly sold as
AFSolid 2000). OPTICast works in combination
with SOLIDCast to automatically modify a casting
process design until an optimum condition is
reached. This allows the design engineer to
develop and submit an initial process design,
then let the computer find a final design which
maximizes both quality and yield.
3
What makes OPTICast work?
OPTICast uses the HyperOpt system from Altair
Engineering, Inc. to find the optimum
design. HyperOpt is a world-class optimization
engine which is used by the major automobile
companies, among many others. As an example,
HyperOpt was recently applied to a sheet metal
stamping to reduce the initial blank size by 9
and saving the manufacturer over 40,000 in
material savings. Now, for the first time,
HyperOpt has been applied to the casting process.
4
What is Optimization?
Optimization is a mathematical method for finding
the best solution to a given problem. Optimizati
on allows us to automate the search for a design
solution, freeing the engineers time to work on
other issues, and providing a more thorough and
repeatable design process.
5
So, how does Optimization work?
Optimization requires that you first develop an
initial design. For example, this would
typically be an initial design for gating and
risering the casting. Based on this design, you
then need to select three elements

• Design Variables
• Constraints
• An Objective Function

6
Design Variables
These are things that are allowed to vary when
the computer is searching for an optimum process
design. A typical example would be the height and
diameter of a riser. Other examples might be
pouring temperature, pouring time or mold preheat
temperature. A feature on the casting (such as a
pad or rib) might also be designated as a design
variable.
7
Constraints
A constraint is some aspect of a design that
determines whether that design is acceptable or
not. For example, porosity level might be a
constraint. Since SOLIDCast can predict the
level of porosity in a casting, you might specify
a certain level as being the maximum allowable.
Any designs which result in higher-level porosity
will be rejected. Process yield might also be a
constraint. If the foundry has a target level
for yield, the system can be instructed to reject
any designs which produce a yield below the
target level.
8
The Objective Function
The Objective Function is the single result which
you are trying to either maximize or
minimize. For example, one possible Objective
Function might be minimization of porosity in the
casting. Another might be maximization of
yield. Still others might be minimization of
total solidification time, maximization of
cooling rate, or minimization of predicted
microporosity.
9
The Optimization Process
Once you have defined your Design Variables,
Constraints and the Objective Function, then
OPTICast takes over. The system begins running
a series of simulations, varying the design with
each simulation, until it is satisfied that the
desired objective has been achieved. At that
point, the optimization is complete and OPTICast
reports to you the combination of design
variables which best satisfies your objective.
10
A Simple Example
Lets look first at a very simple example a
casting with a single top riser.
11
In this example, we have imported a casting model
from CAD, and then created an initial riser
design as a cylindrical top riser within
SOLIDCast.
12
Having selected the casting alloy, the mold
material and the type of riser sleeve, we next
mesh this model. This is a typical step in
SOLIDCast prior to running any simulation.
13
Now, we tell the system that we want to optimize
this casting by selecting Create New Optimization
Project from the Mesh menu.
14
This creates a blank Optimization Project. All
we have to do is fill in the blanks. First, we
select the shapes which comprise the riser, and
designate these as a Design Variable by clicking
on the Add Variable button.
15
The Vertical Scale and the Horizontal Scale of
the riser are now Design Variables. In this
example, we are allowing these to vary up to 1.5
times or as low as 0.5 times our initial design.
16
Now we need to select a Pin Point. This is the
attachment point of the riser to the casting.
This point will remain at a constant position
while the dimensions of the riser are scaled up
or down. To select the Pin Point, we can hide
the casting and click on the bottom center
17
of the riser. This establishes our attachment
point for this geometric feature.
18
For this example, we have only one riser to
design. This means that we have two design
variables the height and the diameter of the
riser.
19
We could also select other items of process data
such as the fill time or the initial temperature
of the casting or mold materials as design
variables. (In this simple case we wont select
these.)
20
Now we may want to specify a constraint. As you
can see, there are numerous items that we could
pick. Here, we have selected Material Density (a
measure of shrinkage porosity) as a constraint,
with a minimum value of 0.994. If we can achieve
this, our casting will be substantially free from
shrinkage.
21
Finally, we select an Objective Function. In
this case, we have elected to maximize the yield.
In effect, were telling the system to find the
smallest riser which produces a sound casting (no
shrinkage).
22
Now that weve set up our optimization run, all
we have to do is select Start Optimization Run
from the menu.
OPTICast will now begin running a series of
simulations, varying the design variables until
the smallest feasible riser is found.
23
Later, when the Optimization Run is complete
24
we can view the results. OPTICast can display
a series of graphs to show us how it arrived at
the final result. To view the graphs, we select
View Graphs from the menu.
25
The first graph shows the value of the Objective
Function for each simulation which was run. In
this case, we started with a yield of about 82
and ended up with a yield of 86.7 after 6
simulations were run.
26
This graph shows the values of the constraint
(Material Density) for each run. The final value
was 0.9999, which indicates a sound casting.
27
Here the system has plotted the values which it
tried for the Vertical Scale (the height) of the
riser. The final value was 80 of the original
value.
28
And a final plot shows the Horizontal Scale
values which were tried. The riser ended up
about 91 of its original diameter.
29
Another way to view results is to select View
Iteration Data from the menu
30
which brings up an Excel spreadsheet that shows
what happened in each successive simulation that
was run by OPTICast.
31
The result?
32
Process yield was increased about 5, and quality
was maximized, in 6 simulations run automatically
by OPTICast.
33
Now, lets look at a more complex example.
34
This is a large steel casting, imported from a
CAD system. Typically, we might first run a
simulation of this casting with no gates or
risers, to see what the natural order of
solidification might be. This helps us decide
where to place gates and risers.
35
We can examine the final temperature distribution
36
the Progressive Solidification
37
or an X-Ray View showing molten metal during
solidification. By looking at these plots, we
decide on an initial design of gates and risers
to produce this casting.
38
Based on this information, we establish a rigging
design for this casting as shown below. The next
step is to create an Optimization Project for
this casting.
39
First, we designate the end riser as Riser 1, and
specify to the system that this is a Design
Variable.
40
Next, we select Riser 2
41
and Riser 3
42
and Riser 4
43
and finally, Riser 5. Each of these risers
will be allowed to independently vary its height
and diameter.
44
We can easily establish the Pin Point (the
contact point) for each riser by hiding the
casting, rotating the view, and then clicking on
the riser contact point to establish the (x,y,z)
coordinates of the Pin Point.
45
Now we specify a Constraint. In this case, we
select Material Density (macroporosity) and set a
minimum value of 0.994.
46
Finally, we select Yield Maximization as the
Objective Function.
47
Selecting Start Optimization Run from the menu
will begin the automatic process of optimizing
this design.
48
When the Optimization Run is complete
49
we can view the results by first plotting the
Objective Function. Here, the yield started at
about 48 and reached about 78 after 100 cycles.
Note how the system initially got the yield up
into the range of 60-70, and then found a way
to increase it to 78.
50
Plotting Material Density shows the soundness of
the casting in each design cycle. The final
value was 0.9954, which was above the specified
constraint.
51
We can view the progress that OPTICast made in
deciding the size of each riser. Here, we have
plotted the progressive changes in the height of
Riser 1.
52
And here are the progressive values showing the
diameter of Riser 1. We can plot the values for
all of the risers this way
53
or we can view a spreadsheet showing all of the
values for all of the designs investigated.
54
The final design is available to load and view as
a model in SOLIDCast. Notice the reduction in
size of each of the risers.
55
The final view shows a plot of Material Density
(shrinkage). With the design as given by
OPTICast, shrinkage is confined to the risers,
and the casting appears sound.
56
The Final Result?
The process yield was increased from 48 to 78,
with all five risers individually designed to
produce a sound casting with no operator
intervention!
57
Some Questions
Why did it take 100 iterations? The number of
runs is roughly the square of the number of
design variables. In this case, we had the
height and diameter of five risers as design
variables, so there were 10 total design
variables. (10x10 100) We could have reduced
the number of design variables to 5 if we had
held the riser height constant and allowed only
the diameters to vary.
58
More Questions
How long did this optimization take? On a 500-MHz
PIII computer, this run took 10 hours (six
minutes per simulation). This time would have
been less than 3 hours on a 1.7-GHz P4
computer. The number of nodes used in each
simulation was about 250,000.
59
More Questions
What processes can I use OPTICast for? Any
process that can be simulated with SOLIDCast can
be optimized with OPTICast. This means that
green sand, chemically-bonded sand, permanent
mold and investment processes, in ferrous and
non-ferrous alloys, can all be optimized.
60
More Questions
Are there any special considerations for
optimizing a casting process design? In general,
you should use the minimum number of nodes
possible in order to reduce processing time.
Also, be aware that OPTICast is changing the
size of features in the model, so if there is a
possibility that some features may overlap, shape
priorities must be set properly. Non-casting
material (gates, risers, feeders) must be
61
Special considerations (Contd) created with
Riser material in the SOLIDCast model. This
means that these components need to be separate
shapes in the model. The number of design
variables should be kept as low as possible to
reduce the number of runs. OPTICast has a
special mode (called a Parameter Study) that
allows you to check how much a specific design
variable influences the outcome so you know
whether its worth designating this as a DV.
62
More Questions
What can I use as design variables? Any geometric
feature (riser, gate, feeder or casting) that is
a separate shape in the SOLIDCast model can be
designated a design variable. Also, the initial
temperature of the casting alloy or any material
can be a DV.
63
More Questions
What can I use as a constraint? You can specify a
minimum value for Yield, Material Density,
Temperature Gradient, Cooling Rate, Niyama
Criterion, and Hot Spot Criterion. You can
specify a maximum value for Solidification Time
(just in the casting, or in casting and risers),
FCC Criterion and Critical Fraction Solid
Time. You can select more than one constraint for
an optimization run.
64
More Questions
What can I use as an objective function? The
objective function can be Yield, Material
Density, Temperature Gradient, Cooling Rate,
Niyama Criterion, Hot Spot Criterion, FCC
Criterion, Solidification Time, or Critical
Fraction Solid Time. You can tell the system to
maximize or minimize the objective function.
Only one objective function can be specified for
each optimization run.
65
What kind of cost reduction is possible with
OPTICast?
A few examples
66
Cost Savings
Example 1 250 pound steel casting 100 castings
produced per month 10 yield improvement Annual
Savings 21,240
67
Cost Savings
Example 2 8 pound permanent-mold aluminum
casting 3600 castings produced per month 5
yield improvement Annual Savings 3,670
68
The next generation of Solidification Modeling
69
Where can I find more information?
Contact Finite Solutions, Inc Dave
Schmidt Phone [phone] Email
FiniteIL_at_aol.com Larry Smiley Phone
513-821-5220 Email LSmiley1_at_aol.com