MACRO POROSITY FORMATION: A STUDY IN HIGH PRESSURE DIE CASTING

David Blondheim Jr. and Alex Monroe

Mercury Marine – Mercury Castings, A Division of Brunswick, Inc., Fond du Lac, WI, USA

David Blondheim Jr.

Colorado State University, Fort Collins, CO, USA

Alex Monroe
Michigan Technological University, Houghton, MI, USA

Copyright Ó 2021 American Foundry Society

https://doi.org/10.1007/s40962-021-00602-x

Abstract
Porosity formation in high pressure die casting (HPDC) The general region where macro porosity forms is pre-
impacts mechanical properties and casting quality. Much dictable with simulation, but its actual size and distribution
is published regarding micro porosity and its impact on of the voids are random. These results challenge the
mechanical properties, but there is limited research on the industry accepted practices for inspections and process
actual formation of macro porosity. In production appli- controls. This also underscores the importance of up-front
cations, macro porosity plays a critically important role in design for manufacturability to avoid macro porosity-re-
casting quality and acceptance by the customer. This paper lated quality issues.
argues that the most useful definition of macro porosity is
the limits of visual detectability. With this definition, it will Keywords: high pressure die casting, porosity, casting
be shown macro porosity presents stochastically within a defects, defect classification, macro porosity, micro
controlled HPDC process. This means macro porosity has porosity
a random probability distribution or pattern that should be
analyzed statistically and cannot be predicted precisely.

Introduction macro porosity.6,7 The cause and size provide one high-
level classification system for porosity, although more
Porosity is considered a defect, and it is inherent to alu- detailed defect classification systems exist.2,8,9
minum high pressure die casting (HPDC) due to the nature
of metal solidification.1 Research completed by the North Porosity is well documented within the HPDC indus-
American Die Casting Association (NADCA) shows that try.1,2,6,10,11 Shrink porosity is caused by volumetric con-
porosity concerns are one of the leading contributors to traction of metals during solidification. Shrink porosity is
scrap costs and the biggest problem within die casting.2 often irregular, with ragged shapes. Gas porosity occurs
Approximately 30% of the industry has identified when a gas concentration within the liquid metal is higher
addressing porosity as a top concern.3 There is high than the solubility of the metal. Gas porosity is typically
motivation to improve porosity scrap given the die casting spherical in shape. The mechanisms causing these two
industry has $8 billion annual sales in North America4 and types of internal voids can also interact during the solidi-
suffers from an 8% median scrap rate.5 Porosity defects can fication of a casting. This interaction is termed gas assisted
be described by the primary cause: shrink porosity and gas shrink. This porosity has shape characteristics of both
porosity and classified by the size: micro porosity and shrink and gas porosity.

The size of porosity is another distinguishing feature of the

porosity formed in castings. Shrink porosity and gas
Received: 28 January 2021 / Accepted: 19 March 2021

International Journal of Metalcasting

porosity can form in macro and micro size voids. Micro Background on Macro Porosity
porosity is often discussed as voids that form within the
mushy interdendritic liquid.12 Macro porosity is typically The transition dimension between micro porosity and
discussed within the industry as visible porosity that is macro porosity has ebbed and flowed between different
compared to quality specifications. Macro porosity may authors. Huang and Conley21 said micro porosity was from
cause rejects within castings with a radiographic (X-ray) 0.5 to 1.0 mm, and macro porosity was from 1 to 10 mm.
inspection or visual inspection after secondary processing, Lee et al.7 called voids that have a maximum dimension
such as machining.13 A clear definition between micro and larger than 5 mm as macro porosity. A different approach
macro porosity is lacking and will be discussed. used by Liang et al.22 is to call macro porosity anything
that can be visibly seen with less than 5X magnification but
Micro porosity has been well researched. The stochastic or did not provide a dimension. Zhang et al.23 used any void
random nature of micro porosity formation was studied by less than 300 lm as the transition from macro to micro
Atwood.12 Lee et al.7 reviewed five different models porosity. Consistency in a definition and the reason for the
involving complex simulation of micro porosity and selected value is lacking from literature. The exact transi-
microstructure with random nucleation. Several groups of tion dimensions between macro and micro porosity may
authors have studied micro porosity formation and distri- continue to be challenged; however, definitions should be
bution in castings with micro computed tomography (CT) supported by reasoning. Having a practical definition is
equipment.14–16 Cao et al.17 and Niu et al.18 studied needed to provide clarity and consistency within industrial
porosity formation on the micro scale with vacuum-as- research.
sisted high pressure die casting. In addition to vacuum, Niu
et al.18 work also studied other injection parameters to see Given the significant impact that macro porosity has on
impact on the mechanical properties with micro porosity. HPDC foundries, a well-conceived definition is needed.
Zhang et al.19 studied micro porosity formation and The variation seen in the literature is because scope of the
mechanical properties of both entrapped air and shrinkage. definition was too small. Three factors should be consid-
All this work has provided useful insights into how micro ered when defining macro porosity: metallurgical forma-
porosity forms and the impact it has. tion mechanism, functional part requirements, and ability
to inspect the defect. After considering these factors, the
Research on macro porosity is heavily tied to publications limit of detectability in inspection is the best transition
on simulations, and it is generally not studied mechanisti- between macro and micro porosity for HPDC.
cally like micro porosity. Overall, there is a correlation
between porosity results from simulations and where it is
found in a casting.20 Macro porosity can be formed by both Metallurgical Formation Mechanism
shrink and gas porosity. Shrink porosity can be reduced by
improving feeding paths in both casting and tool design. From a metallurgical perspective, the size where micro
Gas porosity is affected by the pour rate, slow shot accel- porosity transitions to macro porosity is based on the sec-
eration, venting, and die lubrication improvements.2 Since ondary dendrite arm spacing (SDAS). This relationship is
these features of die and process design are well known, due to the different nature of the feeding flow of the bulk
there has not been much research on the formation of liquid versus the feeding in the mushy region between the
macro porosity in production environments. This is an dendrite arms. In the bulk liquid, the dendrite arms are not
oversight because porosity is still a major cause of scrap.5 present and therefore cannot affect the maximum size of
Understanding the formation of macro porosity can create a the porosity. This unconstrained macro porosity is formed
better understanding of inspection and classification of when the bulk feeding flow is cut off. Dendrites restrict the
casting defects. feeding flow and constrain the maximum size of the
porosity, creating micro porosity.24,25
The goal of this work is three-fold. First, a definition will
be provided to help distinguish the difference between Micro porosity size can vary significantly in different
micro porosity and macro porosity. This definition will be casting processes because SDAS is not a constant value for
based on an industry perspective within HPDC. Second, the these processes. This is because SDAS is proportional to
stochastic or random nature of macro porosity will be the cubed root of the cooling rate.24,26,27 Sand casting and
reviewed with an industrial case study. Simulation software permanent mold castings have a wide range of cooling
can predict a zone where porosity is likely to form, but it rates, making a universal definition of interdentric micro
falls short of showing the random formation of voids porosity in those processes difficult. HPDC cooling rates
produced within that zone. Finally, a review will highlight are different because they are consistently high with an
that from first principles macro porosity should be expected observed SDAS in the range of 10–40 lm.28 The small
to form randomly in unfed liquid pockets. The end results value means detecting true micro porosity voids in HPDC
from this work should challenge the accepted analysis of is often not industrially practical. This suggests that the
macro porosity in HPDC. definition of macro porosity for HPDC should be driven by

International Journal of Metalcasting

the functional requirements or readily available inspection human vision can generally resolve a visual target that
methods. represents 5 minutes of an arc. This 5 minutes of an arc is
typically referred to as 20/20 vision.32 The equivalent
visual angle of a piece of porosity on a casting is deter-
Functional Porosity Requirements mined based on defect size and surface distance as shown
in Eqn. 1. Conversion from degree to minutes of an arc is
Functional casting porosity requirements influence a needed to compare to a Snellen vision chart. Table 1
working definition for macro porosity. After castings are includes a series of calculated equivalent visual angles for
machined, they are inspected for macro porosity on different defect sizes and distances.
machined surfaces. Acceptance criteria for given surfaces
depend on the required function of the casting.13 Thresh- Defect size
olds are sometimes determined by past practices for a given Tanh ¼ ;
Distance from eye
original equipment manufacturer (OEM) and recommen-   Eqn: 1
Defect size
dations provided by vendors of assembled components h ¼ arctan
Distance from eye
such as piston rings, O-rings, and gaskets. In other cases,
thresholds are set by failure points identified in finite ele- It has been found that human ability to focus on near
ment analysis (FEA) for the product. These porosity stan- objects deteriorates with age. Typically, those in their 20s
dards are treated as confidential trade secrets by OEMs; are able to focus down to 150 mm from their eyes. This
therefore, typical thresholds for specific applications are erodes to 250 mm minimum focal distance as they age.33
not published. Given that 5 mins of an arc is 20/20 vision,32 selecting
250 lm threshold between micro and macro porosity
The North American Die Casting Association (NADCA) provides a realistic size that humans can inspect either
has provided an example of four levels of porosity on without magnification in their youth and with minimal
exposed machined surfaces in Die Casting Porosity magnification as they age.
Guidebook1 with a Level 1 porosity being specified as a
maximum size of 0.4 mm. Experience by the authors has Based on the SDAS of HPDC, the functional specifications
shown typically that the tightest specifications are a max- used to inspect castings, and the ability to detect and
imum porosity diameter between 0.4 and 0.5 mm identify porosity with X-ray and visual inspection, a
(400–500 lm). The threshold for macro porosity would threshold of 0.250 mm (250 lm) is a defendable choice for
have to be smaller than this minimum acceptable porosity the threshold between micro and macro porosity in HPDC.
limit to use functional requirements as a meaningful dis- With a size defined and the reasoning supplied, the porosity
tinction between acceptable and rejectable levels of that causes rejects in HPDC are macro porosity. The focus
porosity during inspection. can now shift to how this macro porosity forms in pro-
duction castings and what can be learned to improve the
HDPC process.
Inspection Ability

Castings are typically inspected in a raw state with X-ray or Experimental Study
computed tomography (CT) equipment or by visual
inspection after secondary machining. From an X-ray or A small housing casting, produced on a 900-ton die cast
CT image standpoint, production intent equipment has a machine out of A362.0 aluminum alloy, as seen in Table 2,
pixel size resolution of 100 lm on castings that can fit up
to 400 mm 9 400 mm image windows.29,30 Specialized Table 1. Equivalent Visual Angle for Different Defect
micro-CT research equipment exists to capture micro Sizes and Distances
porosity resolution of 10–50 lm on parts typically less than
50 mm cube.14,31 This type of equipment has significant Equivalent visual angle (minutes of an Arc)
limitations in a production castings environment based on Casting defect size (mm) Distance from defect (mm)
size and time needed to perform analysis. Therefore, the
150 250 500 600
capable range of micro-CT equipment should not be con-
sidered in defining an industrial application of macro 0.1 2.29 1.38 0.69 0.57
porosity. The focus should be given to capabilities of
0.25 5.73 3.44 1.72 1.43
industrial, production intent X-ray equipment.
0.5 11.46 6.88 3.44 2.86
Visual inspection of castings is the next challenge. Visual 0.75 17.19 10.31 5.16 4.30
acuity defines what the human eye is capable of detecting. 1 22.92 13.75 6.88 5.73
Snellen eye charts, as typically found at optometrists’
offices, were developed based on the studies showing

International Journal of Metalcasting

 Table 2. A362.0 Chemical Composition Limits

Si Fe Cu Mn Ni Zn Ti Sr

10.5–11.5 0.40 0.20 0.25–0.35 0.10 0.10 0.20 0.05–0.07

was selected for this experiment. Previous X-ray audits X-ray images were reviewed following ASTM E505
showed varying levels of porosity within this casting even standard36 to determine porosity cause. Shrink porosity
when process parameters remained consistent. The porosity (ASTM Category C) and gas porosity (ASTM Category A)
was located near a semi-circular feature on the side were seen within the castings. The castings were graded
opposite the gating. The porosity level found in this loca- with a 1 (best), 2 (moderate), and 3 (worst) score. Figure 2
tion of the casting does not impact overall part quality provides examples of all three gradings.
based on product testing.

Once the die was brought up to temperature with the typ- Results and Analysis
ical start-up process, 100 castings were produced at iden-
tical process settings. These settings matched the Critical process parameters were collected for all sample
production settings and had no adjustments made for the castings (n = 99) during the experiment. The data were
entire sample run. The castings were pin-stamped with a within typical production variation for the casting. Mean
serial number in the robot extraction cell. With this serial and 95% confidence intervals for these parameters are seen
number, all process data are traced to the X-ray images. in Table 3. Based on the grading samples on the images,
The castings were completed in just under two hours. there were 9 samples each identified as Grade 1 (good) and
Grade 3 (worst). These parts that represented the extremes
Castings were sawed so the area of interest could be easily of the macro porosity found within the casting were
X-rayed to provide repeatable images and remove back- reviewed in detail.
ground features non-critical to this study. A sample of the
sawed casting is seen in Figure 1. One casting was dam- The best castings showed scattered shrink porosity typi-
aged during the sawing process used to prepare it for the cally 0.2 mm to 0.4 mm thick and up to 1.0 mm long. This
X-ray (sample number 76). Its process information was porosity was scattered throughout the section in review,
removed from all analysis. This sample showed no sig- with a tendency for it to form furthest way from the gating
nificant difference in process values. (part is gated from the left side of the X-ray images). The
nine Grade 1 (best) castings can be seen in Figure 3.
The X-ray equipment used was a Bosello SRE Max with a
225 max KV power rating. Images were registered using The worst castings showed a gas assisted shrink with
open source Fiji imaging software34 and the selective plane rounder, but irregular shaped voids, consistent of gas
illumination microscopy (SPIM) registration plug-in.35 The feeding into a shrink porosity. Gas porosity within the

Figure 1. Sawed sample casting for X-ray.

International Journal of Metalcasting

 Figure 2. Example X-ray gradings 1 through 3.

Table 3. Mean and Confidence Interval of Critical Injection Parameters

Mean 95% confidence interval for mean Units

Cycle time 68.0 67.8–68.2 Second

Average slow shot velocity 14.16 14.155–14.165 inches/second
Calculated start of fast shot 14.27 14.266–14.274 Inches
Average fast shot velocity 141.69 141.61–141.77 Inches/second
Intensification pressure 5124.9 5119.0–5130.8 PSI
Intensification squeeze distance 0.174 0.171–0.177 Inches
Biscuit Size 1.69 1.65–1.73 Inches

worst castings grade was typically 3.0 mm to 4.0 mm in accepted analysis and quality implications of porosity in
diameter. For shrink porosity, the worst castings had HPDC.
0.4–0.6 mm thick and 5.0–10.0 mm long porosity voids.
Overall, the worst of the worst (WOW sample—#43) had This experiment could did not determine if the Grade 1
an approximate 8.0-mm-diameter void in the casting. The castings were more dense (less porous) than the Grade 3
nine worst castings (Grade 3) can be seen in Figure 4. castings. It simply observes that the levels of macro
porosity are higher while no process parameters varied
Shapiro-Wilk normality tests37 showed the process significantly. This leaves open the question of whether the
parameters to be non-normal in the 99 samples. As a result, Grade 1 castings traded macro porosity for micro porosity
the Wilcoxon Signed Rank test38 was performed to com- that could not be detected by X-ray. Further study is
pare the samples between the best and worst groupings. merited and ongoing.
Table 4 contains all the individual recorded data for the
samples. Table 5 shows the p-values calculated with the
Wilcoxon test. None of the critical process parameters Part Geometry Impact on Stochastic Macro
showed to be significant. Porosity Formation

The results of the experiment have shown that macro Predicting the morphology of macro porosity is not a useful
porosity formation is random when industry accepted exercise. Macro porosity is easy to detect with visual
control parameters are held constant in a production envi- inspection on machined surfaces or X-rays. Feeding-based
ronment. The size and distribution of the voids varied rules have been successfully developed to manage porosity in
significantly throughout the casting run, even though no sand and permanent mold castings. Macro porosity is reduced
process settings were changed. When comparing the best when an adequately large volume of liquid metal is available
and worst samples, the process parameters showed no to replenish the volumetric contraction of the solidifying
statistical difference. These parameters remain important to metal. The feeding volume must be connected by a liquid path
the process, but they do not fully explain the randomness throughout the solidification of the casting’s cross sections
associated with the porosity formation. As will be dis- being fed. From Chorvinov’s rule, it can be surmised that
cussed in the next section, the general location of the thicker sections require more solidification time.25
porosity formation remained predictable, but the actual
macro porosity formation was random. Production con- Computer simulations of these feeding rules are effective
trolled processes produce randomly formed porosity. This for identifying where porosity can form, but they fail to
random macro porosity should challenge many of the properly predict the morphology of the macro porosity.

International Journal of Metalcasting

 Figure 3. 9 best samples (Grade 1).

MAGMA was used to compare with the experimental but gate locations are determined by die design and gate
results.39 Six warmup cycles and one production cycle removal considerations. A temperature gradient from the
were calculated. Figure 5 shows the predicted porosity thick sections of the casting to the gate cannot be ensured.
zone (a) and the pore volume fraction (b) for the studied Resulting unfed pockets of liquid create the macro porosity
casting in the area that was X-rayed. The porosity zone is common to HPDC. Randomly sized and shaped macro
the predicted hot spot that is the extent of the unfed liquid porosity voids should form in these trapped pockets, but
pocket. Porosity volume fraction is concentrated into two also the size and shape of the trapped pockets are random
voids that approximate the center of the porosity observed because the solidification during filling and time that
via X-ray. More porosity is predicted on the right-hand side pressure is applied to the liquid metal after filling is
of the image like the experimental results. uncertain. Small extension or reduction of feeding will
make significant changes to the total porosity due to the
Prediction of the macro pore morphology should be nearly rapid solidification from the HPDC process.
impossible. This is because the pressure drop required to
homogeneously nucleate porosity is in the giga-Pascal Figure 6 plots the liquid volume in the sample area, as seen
range. Instead, pores require a heterogenous nucleation site in Figures 1 and 5, over the entire solidification time of the
such as an oxide bifilm or pre-existing trapped gas pores.20 casting. Feeding is predicted to be cut off to the sample
Filtering, degassing, venting/vacuum, and good furnace area at 9 seconds after filling, leaving 18.4 cm3 of
maintenance practices can reduce the number of remaining liquid aluminum. 1.9% of the sample area will
heterogenous nucleation sites, but they cannot be elimi- contain pores with the assumption that the aluminum will
nated. Their occurrence will also be stochastic by nature. have approximately 6% shrinkage. Increasing or decreas-
Since the nucleation sites are random, the macro pores also ing the feeding time by 0.25 seconds changes the unfed
must occur randomly. The simulation software can predict liquid volume by ± 3.5%. These small variations in feed-
these porosity zones but is incapable of showing the ran- ing time can come from a host of uncertainties in the
domness of the size and shape of the macro porosity. process. For example, cold flakes blocking feeding through
the gate, spray variation, initial metal temperature, and
Additionally, HPDC feeds exclusively through the gating other variables can reasonably be assumed to affect the
system into the part. Pressure is applied to enhance feeding, feeding time by 0.25 seconds or more.

International Journal of Metalcasting

 Figure 4. 9 worst samples (Grade 3).

Recommendations to the DFM activities on new casting designs. The risks

associated with porosity results should not simply be
The stochastic nature of macro porosity formation over the written off to something the process could remove, par-
course of a normal production run and the inability to ticularly if inadequate sample sizes are reviewed and
predict it precisely should challenge accepted practices tested.
within the industry. Cross-functional review of risks asso-
ciated with the random porosity formation must be con- The design validation process should also be reviewed
sidered and identified throughout the casting life cycle. based on this random formation of macro porosity. Larger
Critical areas that should be rethought include the design of lots of testing samples should be a focus during the initial
castings, inspection of castings, and casting process production validation process phase. Functional testing
optimization. should look at both worst case macro porosity formation
and other random samples with castings produced with
normal production settings. This will require large X-ray
Design of Castings studies to help review the amount of macro porosity for-
mation that exist and select the castings that should be
Castings free from macro porosity can only exist in a functionally tested to ensure product performance.
combined effort between casting geometry design and
manufacturing process. Manufacturing process will not
consistently overcome random macro porosity formation in Inspection of Castings
castings designed with a high probability of porosity. A
truly collaborative effort in design for manufacturing Because macro porosity forms randomly, variation should
(DFM) and finite element analysis (FEA) is required for be expected during a typical production run. This should
success. Casting simulation software is accurate at pre- cause one to question traditional audit and inspection
dicting location of macro porosity. However, the random processes in X-ray and at secondary machining. Better
nature of its formation may not guarantee a casting with the audit sizes can be determined by understanding the prob-
worst macro porosity condition is evaluated during func- ability of uncovering a worst-case macro porosity situation.
tional testing. Casting simulation should be a cornerstone

International Journal of Metalcasting

 Table 4. Injection Parameter Data for Best and Worst Samples

# Grade Cycle Average slow Calc start Average fast shot Intens rise Intens Intens squeeze Biscuit
time (s) shot velocity (in/s) fast shot (in) velocity (in/s) time (ms) pressure distance (in) size (in)
(psi)

6 1 67.0 14.12 14.30 142.18 69 5096.5 0.157 1.89

13 1 65.8 14.15 14.30 141.34 74 5094 0.167 2.00
24 1 69.3 14.14 14.27 141.65 75 5108.7 0.177 1.79
29 1 67.7 14.18 14.28 141.84 70 5108.7 0.177 1.41
49 1 68.3 14.14 14.28 141.78 71 5123.3 0.157 1.35
60 1 69.0 14.17 14.26 141.87 67 5145.3 0.187 1.67
70 1 68.3 14.15 14.28 141.93 71 5125.8 0.187 1.47
74 1 68.7 14.15 14.27 141.93 72 5164.8 0.177 1.56
79 1 67.6 14.16 14.26 141.58 70 5150.2 0.157 1.59
15 3 67.6 14.13 14.30 142.12 74 5072 0.167 1.91
22 3 67.7 14.15 14.28 141.58 73 5172.2 0.167 1.21
28 3 69.5 14.14 14.29 141.39 70 5116 0.167 1.72
38 3 68.4 14.15 14.28 142.12 73 5101.3 0.167 1.72
40 3 67.8 14.15 14.28 142.12 73 5098.9 0.187 1.71
43 3 66.8 14.17 14.28 142.06 74 5130.6 0.157 1.04
52 3 68.4 14.18 14.27 141.5 71 5223.4 0.128 0.92
67 3 68.5 14.20 14.24 141.14 67 5147.7 0.177 1.76
84 3 68.3 14.21 14.23 142.15 68 5172.2 0.177 1.59

Table 5. Wilcoxon Signed Rank Test p values

Cycle Average slow shot Calc start Average fast shot Intens rise Intens Intens squeeze Biscuit
time velocity fast shot velocity time Pressure distance size

p value 0.496 0.183 0.233 0.910 0.536 1.0 0.611 0.652

With an exception of some structural automotive parts, in- chance none of the castings contain that defect. If the
line X-ray equipment is not typically found in most defect rate and the process remain constant and the
industrial die casting foundries. The cost of equipment and inspection has experience seeing high probability of all
time to process when compared to the risks of porosity good castings, what happens once that inspector finds a
typically makes this an uneconomical process. Instead, defective part? It is slightly better than a coin flip to have a
audit X-rays are performed on randomly chosen samples defect occur every two production weeks, based on a three-
during the production run. The number of samples and a shift operation. The warning flags in the foundry are sent
defect rate caused by worst-case macro porosity can be out, and the troubleshooting begins. Another random
used to understand the probability of selecting a sample lot sample selected could then show all good parts, and the
and seeing no defects. The binomial probability function is investigation of ‘‘what changed in the process’’ may occur,
used to find this probability. Table 6 shows the probability wasting resources because the process has not changed, and
of finding no defects for different defect rates and sample the sampling was just poor.
sizes.
Worse yet, someone may take the failed inspection results
Understanding this probability should influence that how and change the process to try to ‘‘improve it.’’ Now the
troubleshooting is completed within the foundry. As an actual defect rate shifts from 2.5% to some other unknown
example, if a casting truly has a 2.5% scrap rate based on value. This can lead to a spiral of process changes over the
the random macro porosity and the X-ray technician ran- production life of a part, with limited knowledge of what
domly pulls 3 castings for inspection, there is 92.7% scrap rates are occurring.

International Journal of Metalcasting

 Figure 5. Simulated results of predicted porosity zone.

6.0% By having a good working knowledge of actual scrap rates

for given castings based on machining feedback, one can
50 make an educated decision on how to approach the
5.0%
Approximate Porosity (%)

defective audit sample. The Bayes Success-Run Theo-

Liquid Volume (cm3)

40 4.0% rem can be used to help determine a lot size given the
historical defect rate and the confidence the manufacturing
30 desires.40
3.0%

20 2.0%
Casting Process Optimization
10 1.0% The randomness of the macro porosity should cause the
industry to review the techniques used for process opti-
0 0.0% mization. An appropriate sample size is needed for any
0 5 10 15 20 25 optimization process to ensure the worst-case scenarios are
Feeding Time aer Filling (s) detected in the macro porosity formation. Also, additional
process monitoring of the HPDC equipment, beyond the
Figure 6. Volume of liquid in the sample area as
injection system, is needed.
predicted by MAGMA.

Table 6. Sample Binomial Probabilities

Binomial probability of different defect rates and sample sizes

Defect rate Number of defects found Randomly sampled lot size
3 5 10 15 30 50 100

1% 0 97.0% 95.1% 90.4% 86.0% 74.0% 60.5% 36.6%

2.5% 0 92.7% 88.1% 77.6% 68.4% 46.8% 28.2% 8.0%
5% 0 85.7% 77.4% 59.9% 46.3% 21.5% 7.7% 0.6%
10% 0 72.9% 59.0% 34.9% 20.6% 4.2% 0.5% 0.0%

International Journal of Metalcasting

Optimization processes are based on feedback from process control. The conducted experiment shows castings
inspection results. With HPDC, that feedback is based on produced with statistically similar injection parameters,
inspection results in either the raw casting state with X-ray cycle time, and biscuit size can produce significantly dif-
images or in a final processed state with a visual inspection ferent levels of porosity as seen in the Grade 1 and Grade 3
for porosity. As discussed previously, lot sizes need to be examples. This work shows the traditional parameters are
reviewed and appropriately selected based on historical not fully capable of reducing the randomness that can exist
scrap rates. Without this, samples may not capture the in the HPDC process. Additional research is needed to
potential worst-case stochastic formation of the macro understand if other process parameters such as variability
porosity. This sample size is also pertinent when setting up in metal cleanliness, furnace temperature, die temperature
an optimization design of experiments (DOE). Without the before die close, spraying, die thermal management, or
correct sample size, the variation that exists due to the others could reduce the predicted porosity zone and
random formation could be missed. therefore region where stochastic porosity forms.

Injection parameters and metal holding temperature are The next area to research is the density in the predicted
typically the focus of in optimization publications in porosity zones between castings. Research has shown that
HPDC.41–44 This approach is logical given the history and density is not a good predictor of mechanical proper-
commercially available products for the industry to capture ties.47,48 The difference levels of void space visible in the
this data. Questions should arise on the focus and priority Grade 1 versus Grade 3 X-rays lead to questions regarding
of these process inputs when significant macro porosity the density of the predicted porosity zones. Are the den-
variation occurs. It is clear from simulation that injection sities of these grade differences the same with different
parameters can influence predicted zones. This study shows distributions of size of the macro and micro porosity? This
the random formation of porosity in these zones. The is a useful question to have answered as its impact on
injection parameters should remain controlled and moni- quality inspection results (acceptable versus scrap casting)
tored; however, data collection on additional process and perceived mechanical properties could be misleading
monitoring systems should be prioritized45 to potentially to the industry.
further reduce this predicted porosity zone. These addi-
tional systems could include die temperature,46 vacuum,18 From this work, we know there is randomness in the size
and equipment cycle time and performance. Optimization and shape of macro porosity. This randomness influences
of these additional parameters could reduce the predicted classification of defects and process optimization decisions.
porosity zone. Therefore, the macro porosity that forms in Misclassification of macro porosity can lead to poor pre-
that zone will also be reduced. dictions of quality when supervised machine learning
algorithms are used. In this case, two significantly different
outputs on X-ray images are produced from nearly identi-
Conclusions cal inputs. It becomes impossible for machine learning to
find a pattern in what fundamentally becomes noise in the
High pressure die casting suffers from a porosity problem. ‘‘results’’ created by random macro porosity formation.
By better understanding how macro porosity forms, Furthermore, sample sizes for optimization studies must be
improvements can be made within the industry. Like micro carefully planned based on these random macro porosity
porosity, macro porosity is randomly formed in HPDC. formations. Small sample sizes will have a higher proba-
Casting design, inspection, and optimization processes are bility that the true worst-case macro porosity formation is
all affected by macro porosity. This random macro porosity not seen, thereby providing misleading optimization guid-
formation has been shown by an industrial case study and ance. These areas are worthy of additional research.
fundamental theory.
In conclusion, the stochastic nature of macro porosity
Simulation software uses casting geometry and tooling formation within the prediction porosity zone should
design to predict porosity zones but cannot predict the challenge the industry to further research HPDC process in
actual random size and distribution of those voids in the production environments. By researching these topics fur-
zone. Understanding this stochastic formation should ther, the industry will be better positioned to help improve
challenge previously accepted practices and improve the overall HPDC casting design and manufacturing of parts.
comprehension and classification of macro porosity defects
in die casting. Specifically, this provides the industry with
three areas of further study.
Funding
The first area of study is within the industry’s process
Research was sponsored by Mercury Marine – Mercury
control approach. Injection parameters are the main focus
Castings, a division of Brunswick, Inc.
within the industry and academic research for HPDC

International Journal of Metalcasting

Availability of data and material 15. M. Weidt, R.A. Hardin, C. Garb, J. Rosc, R. Brunner,
C. Beckermann, Prediction of porosity characteristics
Not applicable. of aluminium castings based on X-ray CT measure-
ments. Int. J. Cast Met. Res. (2018). https://doi.org/10.
Conflict of interests Research was completed as part of require-
ments for an author’s PhD degree. 1080/13640461.2018.1467105
16. C. Gu, Y. Lu, A.A. Luo, Three-dimensional visual-
Code availability Not applicable. ization and quantification of microporosity in alu-
minum castings by X-ray micro-computed
tomography. J. Mater. Sci. Technol. 65, 99–107
REFERENCES (2021). https://doi.org/10.1016/j.jmst.2020.03.088
17. H. Cao, M. Hao, C. Shen, P. Liang, The influence of
1. J. Brevick, Die Casting Porosity Guidebook. North different vacuum degree on the porosity and
American Die Casting Association, 2008. mechanical properties of aluminum die casting.
2. W.G. Walkington, Die Casting Defects: Causes and Vacuum 146, 278–281 (2017). https://doi.org/10.
Solutions (North American Die Casting Association, 1016/j.vacuum.2017.09.048
Rosemont, IL, 1997). 18. X.P. Niu, B.H. Hu, I. Pinwill, H. Li, Vacuum assisted
3. D. Twarog, ‘‘State of the Industry 2012.’’ North
high pressure die casting of aluminium alloys.
American Die Casting Association, 2012, [Online].
J. Mater. Process. Technol. 105(1–2), 119–127 (2000).
https://www.diecasting.org/archive/dce/212online2.
https://doi.org/10.1016/S0924-0136(00)00545-8
pdf.
4. J. Folk, ‘‘U.S. Aluminum Casting Industry - 2019,’’ 19. Y. Zhang, E. Lordan, K. Dou, S. Wang, Z. Fan,
Die Casting Engineer, vol. July 2019, 2019. Influence of porosity characteristics on the variability
5. S. Midson, ‘‘Report on the 2014 Die Casting Bench- in mechanical properties of high pressure die casting
marking Survey Part 2 of 3: Operations,’’ in Report on (HPDC) AlSi7MgMn alloys. J. Manuf. Process. 56,
the 2014 Die Casting Benchmarking Survey, North 500–509 (2020). https://doi.org/10.1016/j.jmapro.
American Die Casting Association, 2014. 2020.04.071
6. S. Viswanathan et al., Eds., ‘‘Shrinkage Porosity and 20. J. A. Dantzig and M. Rappaz, Solidification, 1 st.
Gas Porosity,’’ In: Casting, ASM International, 2008, EPFL Press, 2009
pp. 370–374. https://doi.org/10.31399/asm.hb.v15. 21. J. Huang, J.G. Conley, Modeling of microporosity
a0005222 evolution during solidification processes, in Review of
7. P.D. Lee, A. Chirazi, D. See, Modeling microporosity progress in quantitative nondestructive evaluation. ed.
in aluminum–silicon alloys: a review. J. Light Met. by D.O. Thompson, D.E. Chimenti, (Springer, US,
1(1), 15–30 (2001). https://doi.org/10.1016/S1471- 1998), pp. 1839–1846
5317(00)00003-1 22. T. Liang, C. Mobley, N. Tsumagari, ‘‘The Effects of
8. J. Campbell, Castings, 2nd edn. (Butterworth-Heine- Shot Delay Time on the Microstructures and
mann, Oxford, 2003). Mechanical Properties of a Die Cast Aluminum
9. E. Fiorese, F. Bonollo, G. Timelli, L. Arnberg, E. Alloy’’, Presented at the Die Casting Toward The
Gariboldi, New classification of defects and imper- Future (Rosemont, IL, 2002). https://www.diecasting.
fections for aluminum alloy castings. Int. J. Met. 9(1), org/archive/transactions/T02-053.pdf
55–66 (2015). https://doi.org/10.1007/BF03355602 23. B. Zhang, S.L. Cockcroft, D.M. Maijer, J.D. Zhu, A.B.
10. H.H. Doehler, Die Casting (McGraw-Hill Book Phillion, Casting defects in low-pressure die-cast
Company, New York, 1951). aluminum alloy wheels. JOM 57(11), 36–43 (2005).
11. NADCA Product Specification Standards for Die https://doi.org/10.1007/s11837-005-0025-1
Casting, 10th Edition. Arlington Heights, IL: North 24. K.D. Carlson, C. Beckermann, Prediction of shrinkage
American Die Casting Association, 2018. pore volume fraction using a dimensionless Niyama
12. R. Atwood, ‘‘A Combined Cellular Automata and criterion. Metall. Mater. Trans. A 40(1), 163–175
Diffusion Model for the Prediction of Porosity (2009). https://doi.org/10.1007/s11661-008-9715-y
Formation During Solidification,’’ University of Lon- 25. G.K. Sigworth, Shrinkage, feeding and riser design.
don, 2001. AFS Trans. 14(002), 25–36 (2014)
13. Product Design for Die Casting E-606, Sixth Edition., 26. M. Shabani, A. Mazahery, Prediction of mechanical
vol. E-606. North American Die Casting Association, properties of cast A356 alloy as a function of
2009. microstructure and cooling rate. Arch. Metall. Mater.
14. I. Brill, B. Kappes, and S. Midson, An Initial (2011). https://doi.org/10.2478/v10172-011-0073-1
Evaluation of CT Scanning for Measuring and Char- 27. M. Easton, C. Davidson, D. St John, Effect of alloy
acterizing Porosity in Aluminum Die Castings, Indi- composition on the dendrite arm spacing of multi-
anapolis, IN, 2018, vol. T18-083. http://www. component aluminum alloys. Metall. Mater. Trans. A
diecasting.org/archive/transactions/T18-083.pdf 41(6), 1528–1538 (2010). https://doi.org/10.1007/
s11661-010-0183-9

International Journal of Metalcasting

28. J. Cho, C. Kim, The relationship between dendrite arm 115–125 (1980). https://doi.org/10.1016/0143-
spacing and cooling rate of Al-Si casting alloys in 8174(80)90003-7
high pressure die casting. Int. Metalcasting 8(1), 41. V.D. Tsoukalas, Optimization of porosity formation in
49–55 (2014). https://doi.org/10.1007/BF03355571 AlSi9Cu3 pressure die castings using genetic algo-
29. ‘‘SRE MAX,’’ Bosello High Technology , a ZEISS rithm analysis. Mater. Des. 29(10), 2027–2033 (2008).
company. https://bosello.eu/products/sre-max/. Acces https://doi.org/10.1016/j.matdes.2008.04.016
sed 29 Dec 2020 42. S.G. Lee, A.M. Gokhale, Formation of gas induced
30. ‘‘Phoenix Vtomex C | 3D CT Scanner (Mini Focus),’’ shrinkage porosity in Mg-alloy high-pressure die-
Waygate Technologies. https://www.bakerhughesds. castings. Scr. Mater. 55(4), 387–390 (2006). https://
com/industrial-x-ray-ct-scanners/phoenix-vtomex-c-ct doi.org/10.1016/j.scriptamat.2006.04.040
. Accessed 29 Dec 2020 43. Q.-C. Hsu, A.T. Do, Minimum porosity formation in
31. ‘‘Xradia 610 & 620 Versa.’’ https://www.zeiss.com/ pressure die casting by taguchi method. Math. Probl.
microscopy/us/products/x-ray-microscopy/zeiss-xra Eng. 2013, 1–9 (2013). https://doi.org/10.1155/2013/
dia-610-and-620-versa.html. Accessed 29 Dec 2020 920865
32. T.J. Schorn, Improving the Effectiveness of Visual 44. J. Zheng, Q. Wang, P. Zhao, C. Wu, Optimization of
Inspection (American Foundry Society, Schaumburg, high-pressure die-casting process parameters using
IL USA, 2018). artificial neural network. Int. J. Adv. Manuf. Technol.
33. J.F. Koretz, G.H. Handelman, How the human eye 44(7–8), 667–674 (2009). https://doi.org/10.1007/
focuses. Sci. Am. 259(1), 92–99 (1988). https://doi. s00170-008-1886-6
org/10.1038/scientificamerican0788-92 45. F. Bonollo, N. Gramegna, G. Timelli, High-pressure
34. J. Schindelin et al., Fiji: an open-source platform for die-casting: contradictions and challenges. JOM 67(5),
biological-image analysis. Nat. Methods 9(7), 901–908 (2015). https://doi.org/10.1007/s11837-015-
676–682 (2012). https://doi.org/10.1038/nmeth.2019 1333-8
35. S. Preibisch, S. Saalfeld, J. Schindelin, P. Tomancak, 46. D. Blondheim Jr., ‘‘Unsupervised Machine Learning
Software for bead-based registration of selective plane and Statistical Anomaly Detection Applied to Thermal
illumination microscopy data. Nat. Methods 7(6), Images’’, Presented at the 2018 NADCA Congress and
418–419 (2010). https://doi.org/10.1038/nmeth0610- Exposition (Indianapolis, IN, 2018). http://www.die
418 casting.org/archive/transactions/T18-071.pdf
36. ASTM E505–15, Standard reference radiographs for 47. C.H. Cáceres, B.I. Selling, Casting defects and the
inspection of aluminum and magnesium die castings tensile properties of an AlSiMg alloy. Mater. Sci. Eng.
E505–15. ASTM International (2015). https://doi.org/ A 220(1–2), 109–116 (1996). https://doi.org/10.1016/
10.1520/E0505-15 S0921-5093(96)10433-0
37. S.S. Shapiro, M.B. Wilk, An analysis of variance test 48. R. Lumley, N. Deeva, M. Gershenzon, An evaluation
for normality (complete samples). Biometrika 52, of quality parameters for high pressure die castings.
591–611 (1965) Int. J. Met. 5(3), 37–56 (2011). https://doi.org/10.
38. F. Wilcoxon, Individual comparisons by ranking 1007/BF03355517
methods. Biom. Bull. 1(6), 80–83 (1945)
39. MAGMAsoft. Kackerstrasse 11, 52072 Aachen, Ger-
Publisher’s Note Springer Nature remains neutral with
many: MAGMA Gmbh, 2019
regard to jurisdictional claims in published maps and
40. M.R. Brand, An examination of certain Bayesian
institutional affiliations.
methods used in reliability analysis. Reliab. Eng. 1(2),

International Journal of Metalcasting