Using simulation for permanent mold casting optimization.

Inside This Story:

* When Teksid Alumino de Mexico found porosity problems in its low pressure cast aluminum bedplate, the firm used computer simulation to find the cause of the defect.

* Once the defect was identified, Teksid Alumino made tooling changes to reduce the reject rate on the cast component.

While the low pressure casting process is a well-established route for manufacturing large safety-critical nonferrous automotive components, feeding-related issues in castings can often result in porosity problems after machining. In order to solve these problems, foundries must maintain stringent control of mold temperatures to promote directional solidification. The best way to solve this problem--casting process optimization software.

Teksid Alumino de Mexico, S.A. de C.V., Frontera, Coahuila, Mexico, experienced this problem firsthand. While the two most common parts made via the low pressure process are wheels and cylinder heads, the process has recently been successfully applied to the manufacture of bedplates. However, in producing its bedplate casting, Teksid found a significant porosity problem located between the ingate and one of the bulkheads in the bedplate.

This article discusses how Teksid Alumino used simulation tools to analyze and product the cause of the porosity outbreak found in its bedplate casting, and the subsequent tooling changes made to implement a quick solution on the shop floor.

Understanding the Problem

In manufacturing the bedplate casting (cast from 319 aluminum) at Teksid, several low pressure permanent mold machines are used. During the process, the furnace temperature was maintained at 1328F (720C). At the end of the cycle, the average temperature of the top mold was 180F (100C) lower than the bottom mold.

The geometry of the bedplate along with the gating system is shown in Fig. 1. The size, location and control of the cooling lines remain confidential to the company's process operational strategies and am not discussed here. The control of the cooling lines resulted in favorable thermal profiles within the die. Some of the strategies that were used to control the mold temperature to promote directional solidification toward the ingates are that:

[FIGURE 1 OMITTED]

* cooling lines are controlled by rime and switched "on" and "off" during the cycle;

* cooling lines are controlled by thermocouples placed at strategic locations within the die;

* rectangular air pockets are machined into the die in strategic locations to act as insulators, delaying solidification in certain areas of the casting (mold geometry prevented placement of air pockets in certain areas of the die);

* air is blown to enhance forced cooling and coating was applied to the die during the cycle.

Porosity was found after machining the interface between the ingate and the bulkhead in Fig. 1. The appearance suggested that the porosity was of the shrinkage variety. Since there were 10 ingates connecting the bulkhead, the grinding operation involved sawing off and machining the ingate stubs connecting the bulkheads of the bedplate.

The majority of the porosity outbreaks were consistently found after machining in the bulkhead section five region of the casting. The engineers discussed several hypotheses, and the scenarios shown in Table 1 were considered worthy of serious investigation,

During the simulation, the filling of the mold was controlled using a pressure versus time curve that was typically used in production of the bedplates. Pressurization was applied as soon as filling was completed, and that pressure was maintained on the cavity until the gates were solidified.

The feeding algorithm included in the simulation can force-feed areas in the casting that become deficient in metal during solidification, based on a preset value. For example, a feeding of 30% (termed feeding effectivity) allows feeding of the metal during solidification until the control volume reaches a fraction solid of 30%.

Even though the software database recommends a feeding effectivity value for a particular alloy (30% for 319 aluminum), any simulation strategy must include simulations using the recommended value and a value incrementally lower than the recommended value. The later will predict somewhat lower feeding situations, a strategy that may be used in seeking out additional porosity-prone areas in the casting that might otherwise be borderline cases.

Testing the Hypotheses

In total, three simulations were carried out to test the hypotheses listed in Table 1.

* Stimulation #1--As received geometry simulated with all shop floor based process parameters and a feeding effectivity value of 30%. All hypotheses were tested using Simulation #1;

* Simulation #2--As received geometry stimulated with all shop floor based process parameters identical to Simulation #1, but with a feeding effectivity value of 20%. Hypothesis #4 was retested to verify if results predicted by this simulation were any different from those predicted in Simulation #1;

* Simulation #3--Modifications to the ingate geometry simulated with all shop floor based process parameters identical to Simulation #1. A feeding effectivity value of 20% was used in the simulation. Hypothesis #4 was further evaluated to verify if the casting problem was solved.

Even though there was a strong bias from the onset that the porosity in bulkhead #5 was solidification related, both filling and solidification results tht were obtained in Simulation #1 were analyzed in detail. To ascertain if the cavity filling (turbulence) played any role in the porosity problem, filling pattern and retained air pressure in the cavity were analyzed.

As a general rule, local flow velocities during filling must be less than 0.5 m/s to prevent turbulence. Not only were the velocity profiles well below this level, but them was no plausible evidence from filling results that the porosity was in any way filling related.

The first issue in the analysis of solidification results was to make sure that the intensification was maintained until the gates were fully solidified. The pressure was maintained for 110 sec after the end of filling, meaning that the solidification result files after 153 sec (including 43 sec of filling time) should show fully frozen ingates. Figure 2 shows that the outside bulkheads (#1, #5, #6 and #10) solidify earlier than the inner bulkheads, and bulkhead #5 completely solidified at 144 sec. This indicated that lack of pressurization did not cause the porosity, eliminating hypothesis #1.

Figure 2 also indicates that the ingate connecting bulkhead #5 solidifies after the bulkhead, which was consistent with solidification time results that showed temperature contours progressing from the lower solidification time values toward the higher solidification time values. Development of such well-banded solidification time contours is highly desired in the low pressure permanent mold process and gives a clear indication of directional solidification progressing from the top mold toward the bottom. The solidification time results clearly indicate chat the ingate does not solidify before the bulkhead, and hypothesis #2 was determined unlikely.

[FIGURE 2 OMITTED]

The simulation result shown in Fig. 3 has greater significance. Even though the evolution of isolated areas with the longest fraction solid time (with an effectivity value of 30%) can be clearly seen deep inside the casting, such isolated areas are not formed at the machining interface between bulkhead #5 and the ingate. Such bottlenecks can be seen in Fig. 3 forming in the lower part of the ingate, and, during casting production, these regions end up porosity prone. However, these areas remain deep inside the casting and are surrounded by enough cast material that the structural integrity of the casting is not compromised.

[FIGURE 3 OMITTED]

The first two hypotheses cannot be completely eliminated unless the secondary dendrite arm spacing (SDAS) results are analyzed, as the SDAS values are good indicators of local solidification times within a casting. The predicted values and the measured values of SDAS are listed in Table 2. The SDAS versus time relationship lot results obtained in Simulation #1 showed excellent correlation with the measured values, allowing the firm co eliminate the first two hypotheses as inapplicable.

Even though there were some indications of thermal conditions (hypothesis #3) that point toward impending casting problems in bulkhead #5, conclusive predictions of porosity were not obtained. The only area that warranted further investigation was the feeding-related issue. Because the borderline feeding problems were not predicted using a feeding factor of 30%, the simulation with a feeding factor of 20% was rerun.

In running Simulation #2, all other process parameters were maintained identical to those in Simulation #1. However, when the simulation was repeated (Fig. 4), it showed an island of high fraction solid time value at the machining interface. This isolated island of high fraction solid time could not be fed as the bottleneck causes porosity. Since no other bulkheads showed this trend, feeding issues were clearly the reason for the porosity defect.

[FIGURE 4 OMITTED]

Tooling Changes

Having identified the problem as feeding related, Teksid Alumino decided to make a local tooling change to the ingate feeding the problematic bulkhead. The design change to the ingate involved decreasing the taper on the ingate feeding bulkhead #5 (Fig. 5).

[FIGURE 5 OMITTED]

The final simulation with the design change was carried out with a feeding effectivity of 20% and process parameters similar to earlier simulations. Following the tooling change, the problems in feeding the bulkhead disappeared. The result was considered a major improvement, and subsequent tooling changes with the modified ingate design change were implemented in the foundry.

Since the implementation of the design change across all the tools, a reduction in reject rates was experienced across the board in the manufacturing facility, resulting in substantial improvements in the bedplate's reject rates.

Table 1. Hypotheses for Possible Causes of Porosity in the Bedplate
Bulkhead.

Hypothesis                           Comment

1. Was pressurization removed        If the pressure on the casting was
   prior to the solidification of    maintained until all of the
   the ingates?                      ingates were fully solidified,
                                     then the porosity is not due to
                                     lack of pressurization. This was
                                     to be tested via simulation.
2. Is the ingate solidifying         If this argument were true, then
   before the bulkhead #5 area       solidification time and dendrite
   (machining interface)?            arm spacing (SDAS) measurements
                                     would shed more light on this
                                     issue. There appeared to be merit
                                     in this argument since tooling-
                                     related issues prevented placement
                                     of air pockets in the vicinity of
                                     the problem area. This hypothesis
                                     was to be tested in the simula-
                                     tion, and SDAS values obtained in
                                     the simulation were to be compared
                                     with actual values obtained in
                                     section five
3. Is the area of porosity out-      Criteria functions suggested in
   break an area that has a          literature can be used to verify
   cooling rate that is favorable    this hypothesis. While this would
   for the formation of gas          only give an indication of the
   porosity?                         formation of porosity, its actual
                                     occurrence would depend on several
                                     parameters such as gas content in
                                     the melt and directional solidifi-
                                     cation.
4. Is early development of inter-    Simulations can be carried out to
   dentric network (dendrite         predict if section five (machining
   coherency) preventing feeding     interface) is an area that lacks
   in section five of the casting?   feeding. This was to be checked by
   If so, could this result in       varying the feeding effectivity in
   porosity formation?               the simulations.

Table 2. Actual SDAS Measurements and Their Comparison with Those
Predicted by Simulation #1.

                              Measured SDAS    Measured SDAS
Location                      Bulkhead #5      Bulkhead #6

Bulkhead interface            39.4 [micro]m    36.4 [micro]m
6.5 mm below the interface    41.7 [micro]m    36.3 [micro]m
9 mm above the interface      41.0 [micro]m    35.9 [micro]m

                              SDAS predicted by
Location                      Simulation #1

Bulkhead interface            Simulation indicates SDAS in
                              the range 39.9-39.4 [micro]m in
                              all regions just above and
                              below the machining
                              interface in bulkhead #5
6.5 mm below the interface
9 mm above the interface

For More Information

"Effect of Pressure on the Feeding Behavior of A356 Alloy in Low Pressure Casting," K. D. Li, M. C. Cheng and E. Chang, 2001 AFS Transactions, Paper No. 01-026.

Winston Sequeira is the applications manager at Magma Foundry Technologies, inc., Arlington Heights, III, Rafael R. Rosales is the process engineering manager for Teksid Alumino de Mexico, S.A. de C.V., Frontera, Coahuila, Mexico.