You are on page 1of 20

An Evaluation of Quality Parameters for

High Pressure Die castings

R. Lumley, N. Deeva and M. Gershenzon


CSIRO Future Manufacturing Flagship, Victoria, Australia

Copyright © 2011 American Foundry Society

Abstract
Several techniques for examining casting “quality” as result, but rotary degassing did appear to remove a portion
it relates to high pressure diecast alloy A380 have been of the oxides present in the melt, thereby improving casting
evaluated in the as-cast condition. The roles of three simple quality. It is shown that of the different analyses conducted,
parameters were considered: a) metal velocity at the gate, b) all could differentiate a degree of casting quality, but some
the effect of increased Cu or Zn content, and c) the effect of techniques (i.e., Weibull statistics combined with flow
rotary degassing on a recycled melt. It was shown that tensile curve derivations based on the Ludwik-Holloman equation)
failure in high pressure die casting (HPDC) specimens is are particularly useful. It is proposed that complex strain
influenced by complex defect clusters and the interaction of a localization and failure occurs in HPDC specimens,
variety of casting defects. The two major defect cluster types which results in a proportionately large fraction of defects
identified in the current work were comprised of a dispersed appearing on the fracture surface.
foam-like shrinkage defect, and/or large oxide films present
on the fracture surfaces. The removal of hydrogen had little Keywords: high pressure diecasting, HPDC, casting quality,
effect on average tensile properties which was a surprising A380 aluminium

Measures of Comparative “Quality” of Al Castings another advantage is that an adverse change in quality may
be quickly detected by comparing results against a known
Statistical Techniques baseline value of -3s. This procedure also allows for a rapid
comparison of the same products manufactured by different
One significant advantage of the HPDC manufacturing facilities, or even different machine operators. The disad-
route is the very high production rates with which parts vantage of this technique is that it assumes a normal distribu-
may be produced. With regards to laboratory testing, this tion of data exists, which is often not true for a statistically
means that large numbers of nominally identical samples large number of castings.
can be produced without substantial variations in prop-
erties, composition, microstructure or dimensions. Since Weibull Analysis
HPDC tensile test bars require no machining, and their
gage dimensions are identical within a very small mar- It has been proposed that examination of the Weibull dis-
gin of variation, rapid determination of large numbers of tribution is a better test of quality in aluminium castings1,2
tensile results is possible. This means an examination of than -3s. Weibull statistics were first developed for brittle
several different techniques for casting quality may be ceramic materials, and have the major advantage in that
quantitatively tested. adjustments can be made based on the size and scale of the
product being made. Another benefit is that the Weibull
± 3 Sigma Analysis modulus, “m,” is a useful means by which to compare
components made by different processes. As the value of
It is a common practice for buyers of aluminium castings to m increases, the narrower the range of tensile strengths
specify values of -3s for the products they purchase from (and ductility) which will exist, so reliability is improved.
metalcasting facilities. That is, the lower bounds are speci- This approach has many merits for aluminium high pres-
fied as the mean minus three times the standard deviation sure diecastings. As with other brittle materials, the tensile
of a statistically relevant number of test samples. This tech- strength data of castings in practice exhibits a wide amount
nique has excellent utility in that it is easy to use and in the- of scatter, and the Weibull distribution provides a simple
ory, only 0.27% of products manufactured will fall outside method to determine accurately the probability of failure
the 3s values. (i.e., 0.135% above or below the +3s or -3s at any given stress. A practical method for using Weibull
values respectively) In the context of aluminium castings, analysis is summarized from Davies:3

International Journal of Metalcasting/Summer 2011 37


The Weibull analysis provides a probability of failure at a A theoretical physical background to the Weibull distribu-
given stress, and is often expressed by the empirical rela- tion has also been established,4 related to the probability
tionship: density f(a) of flaw sizes within the material, with f(a) being
approximated by:

Equation 1
Equation 7

Where P is the probability of failure at a stress, s, m is the


Weibull modulus, su is the stress at which P = 0 (and is nor- Where n in Eqn 7 is the rate at which f(a) tends to zero for
mally taken to equal zero), X is the strength limiting dimen- a>>c/n and c is a scaling parameter. Assuming there are a
sion of the material, and so is a normalizing factor, some- large number of randomly oriented flaws, m and n are related
times referred to as the position parameter. Therefore, for through m = 2n - 2. (Note that “n” used in Eqn 7 is the same
specimens of constant geometry, notation as the original reference 4 and is not the same as
the strain hardening exponent, n, defined later for Eqns 13
and 14.) Therefore, it is clear the scatter in strength data and
Equation 2 hence the value of the Weibull modulus, is directly related
to the flaw size distribution and therefore casting quality.
One additional advantage to the use of Weibull modulus is in
the simplest method of obtaining so and m from a series of utilization of the information gained for material at different
data is to rank the s data from smallest to largest and assign scales. For example, Eqn 1 may alternately be written as:5
P values so that:

Equation 3 Equation 8

Where “i” is the rank and N the total number of specimens,


Eqn 2 can then be expressed as: Where; V is the volume of the component. Therefore, for the
same probability of survival (or failure; note that probability
y = A + Bx Equation 4 of survival = 1- probability of failure) the stress at failure
varies with the volume of the component. For example, if su
Where: = 0, and V1 > V2, then:

Equation 9

and
The best estimates of so and m can then be obtained using
the linear least squares method, for example: s1 / s2 = (V2 / V1)l/m Equation 10

So for an equal probability of survival (or failure), a larger
Equation 5 volume of the component will display a lower stress to cause
failure.

Similarly, the stress ratio between bending and tension for


the same probability of survival (or failure) is also given by:
Equation 6
sbend / stensile = {2(m + 1)2}l/m Equation 11

The Influence of Defects on


Where: Mechanical Properties of Aluminium Castings

A range of different defect types exist in aluminium cast-


ings and all affect mechanical properties. Elongation and
More commonly however, it is simplest to plot the xy data tensile strength are particularly sensitive to the presence of
from Eqn 4 and determine the slope of the line which gives any types of casting defect. Surappa et al.6 has shown that
the value of the Weibull modulus, m. The value of position the decrease in elongation to fracture through the presence
parameter, so, is equal to the value at which 37% (1 - 1/e) of of porosity could be related to the projected area of pores
samples survive.2 on the fracture surface. This model was further expanded

38 International Journal of Metalcasting/Summer 2011


by Cáceres and Selling7 who showed that similar behavior elongation, or the true strain at the onset of necking. This
existed irrespective of the type of defect or numbers of de- means that a defect-free casting will continue to elongate un-
fects, whether it was porosity, dross, oxides or artificially til the maximum stress is achieved and necking begins, lead-
introduced holes. Bulk volumetric porosity had almost no ing to failure. The presence of defects in the material means
correlation to tensile properties, whereas the number or that void growth and crack propagation occurs at levels of
proportion of defects present on the fracture surface was strength and ductility where e < n, these levels being pro-
directly related to the tensile behavior. Cáceres and Sell- portionate to the fraction of defects present in the material.
ing7 proposed a model based on the relationship between a
cross sectional area not containing a defect, Ao, and the cross As an alternate to use of the Ludwik-Holloman equation,
sectional area containing a defect, Ai, such that Ai=Ao(1-f), the Kocks-Mecking Model and the derivation of the latter
where f is the area fraction covered by the defect. In this to give the Voce equation are based on a first principles ap-
case, load equilibrium is maintained if: proach to strain hardening, and represent an excellent al-
ternative.11 Whereas, the Ludwik-Holloman equation may
si(1-f)Aoe-ei =shAoe-eh Equation 12 underestimate experimental results, the Kocks-Mecking
Model may have the opposite effect and be less relevant to
Where si and ei are the true stress and strain inside the de- low levels of strain.11 As may be appreciated, scatter exists
fect, and sh and eh are the true stress and strain outside the in nominally equivalent material, which suggests that indi-
defect. ces based on both models represent the data reasonably well
in application.
If the material follows the well-known relationship (Lud-
wik-Holloman equation): One limitation to the use of the quality index is the fact
that, in practice, it is often only used with a small amount
s = Ken Equation 13 of experimental data, or used with average (mean) results
obtained from a similarly small number of samples. The
Where s is true stress, e is true plastic strain, K is a constant, combination of these two factors may invalidate conclusions
known as the strength coefficient and n is the strain harden- regarding component quality, since the scatter in tensile data
ing exponent; is directly related to the flaw size distribution and hence
the quality of the cast material. Although average values of
then combining Eqns 12 and 13 leads to: elongation and tensile strength are often presented in litera-
ture, what are clearly of great importance are the lowest val-
(1-f)e-ei ein = e-eh ehn Equation 14 ues of mechanical properties as these provide the limits for
safe component operation. These lower limits are important
This equation relates the strain inside the defect to the strain because it is these values around which components will (or
outside the defect. As may be appreciated, when solving should) actually be designed.
Eqn 14, as the defect area present on the fracture surface f,
increases, for a fixed value of strain outside the defect (eh), Experimental Methods
the strain inside the defect containing region (ei ) increases
more rapidly. HPDC alloy specimens for tensile testing were produced us-
ing a Toshiba cold chamber die-casting machine with a 250
Quality Index for Al Castings ton locking force, a shot sleeve with an internal diameter of
50 mm and a stroke of 280 mm. The die provided two cylin-
Quality index (Q) was originally derived by Drouzy et al.8 to drical tensile specimens and one flat tensile specimen from
describe the presence, or absence, of casting defects in per- each shot, and these all conformed to specification AS1391.
manent mold and sand cast Al-Si-Mg alloys such as A356 The cylindrical tensile test bars used for the current tensile
and A357. It was first developed as an empirical formula: tests had a total length of 100 mm with a central parallel
gauge section 33 mm long and a diameter of 5.55 ±0.1 mm.
Q = UTS+150logEf Equation 15 The flat tensile specimen on the runner was not used in the
current testing. The first 10 shots in each run were discarded.
Where Q and UTS are in MPa and Ef is the percent elonga- At least 100 tensile samples (50 shots) were then cast in each
tion at fracture (i.e., elastic + plastic strain). Later, Cáce- batch, from which samples for tensile testing were randomly
res9,10 showed that for any aluminium alloy, derivation of selected. Tensile testing was conducted following standard
the flow curve described by the Ludwik-Holloman equation procedures, at a strain rate of 5 mm/min.
(Eqn 13) could provide the basis for the assignment of qual-
ity indices based on proportions of what would constitute a Degassing was done with a Reading Foundry Products Mod-
defect-free casting, where the true strain e = n. Although the el RE 2/4 rotary degassing unit, operating at 300 RPM with
Ludwik-Holloman equation is an empirical relationship, n a flow rate of high purity argon of 7L/minute for 20 minutes
can be shown experimentally to be equivalent to the uniform (The furnace held approximately 140kg of metal). Reduced

International Journal of Metalcasting/Summer 2011 39


pressure testing was conducted during selected experiments Values of the mean (m), one standard deviation (s), and
using a stainless steel cup with a Dynarad reduced pressure m-3s for the tensile data of each of the two conditions are
test unit, where a vacuum of -100 kPa was applied. Follow- shown in Table 2. As a particularly simple technique for ex-
ing these procedures, density was calculated using the Ar- amining the comparative quality of the two sets of castings,
chimedes technique. this technique shows an excellent representation, since only
0.135% of actual data should fall below m-3s. (Note that m
Scanning electron microscopy (SEM) was conducted using ±1s represents 68.2% of the data, m ±2s represents 95.4% of
either a Leica S440 SEM microscope, or alternately, an FEI the data, and m ±3s represents 99.7% of the data). As shown
Quanta 400FEG environmental SEM. For select cases, Im- in Table 2, values of m -3s for the 82 m/s samples were sub-
age Pro Plus analysis software was used with color segmen- stantially better than for the 26 m/s samples. In this regard
tation to approximate the area of the defect on the fracture the values of m -3s for the 82 m/s condition are actually
surface. reasonably close to the mean values for the 26 m/s condi-
tion. Here it is important to note that these tensile property
Results and Discussion values are only valid for the specific geometry tested, and
importantly, are not necessarily comparable with castings of
Statistical Analysis other geometries or gating systems.

Initial experiments were conducted to determine if mean- Representative samples having close to the mean values of
ingful results related to casting quality could be derived for elongation and tensile strength were chosen for further study.
HPDC components. As a baseline measure of performance, Figure 1 shows fracture surfaces of an as-cast sample pro-
in previous work on HPDC’s12,13 it has been demonstrated that duced with a melt velocity at the gate of 26 m/s. The fracture
some differences exist in the tensile properties derived from surface is shown as a secondary SEM image in Fig. 1(a) and
as-cast samples produced with a velocity at the gate of 26 m/s, as an equivalent Backscattered Electron (BSE) image in Fig.
compared to those produced with higher velocities such as 82 1(b). Figure 1(c) and (d) are higher magnification images of
m/s. The difference in properties was later shown to become the central region of the fracture surface, also in BSE mode.
more significant in heat treated conditions,14 where, as would Casual examination of Fig. 1(a) suggests that no large pores
be expected, a higher quality as-cast part will result in a higher or defects were present on the fracture surface. However, the
quality heat treated one. Data for an A380 composition (Alloy BSE image in Fig. 1(b) reveals a defect that occupies a very
1 from Table 1) were generated using 25 individual tensile large proportion of the fracture surface (the defect was found
tests for each of the two melt velocities. to have an area of approximately 6.1 mm2 using Image Pro

Table 1. Alloys Examined

Table 2. Data for Alloy 1 Samples Produced at 26m/s or 82 m/s

40 International Journal of Metalcasting/Summer 2011


Plus software). Higher magnification imaging (Fig. 1[c & d]) The Weibull distribution was used to further analyze the
showed that this is a region where dispersed shrinkage po- data presented in Table 2. In Fig. 3, the Weibull distribu-
rosity was present, because there was evidence of an almost tions for tensile strength (UTS) and elongation at failure (Ef)
foam-like, random dendritic structure. There was also a small are shown and the differences between specimens produced
oxide film present in the exact centre of Fig. 1(d) which helps at the two melt velocities are significant. For example, Fig.
highlight the wide range of features that may be present in a 3(b) shows that at a stress level of 320 MPa, the probability
complex defect cluster, formed during diecasting. of failure in the 26 m/s samples is 95%, whereas for those
produced at 82 m/s, the probability of failure is only 1.1%.
Another important result was that the tensile properties of Conversely, 1.1% probability of failure in the 26 m/s sam-
the actual specimen photographed in Fig. 1 were very close ples occurs at close to 260 MPa, or a difference of 60 MPa.
to the statistical mean (Table 2). An implication of this ob- As shown by Fig. 3(a), the Weibull modulus was 27.3 for
servation is that either a) the properties of specimens are not the samples produced at 26 m/s, and 42.3 for samples pro-
particularly affected by the presence of such large defects, or duced at 82 m/s. This increased Weibull modulus for the
b) all of the 25 specimens tested were of equivalently (low) higher velocity material indicates that since the spread of
quality. Since images of the fracture surface of an average data for the tensile strength is reduced, the flaw size distribu-
specimen produced at a higher melt velocity of 82 m/s (Fig. tion must also be reduced. (It is also important to note that
2) show less evidence of large defects, and the specimen has such a lower value of Weibull modulus as seen in the 26 m/s
higher tensile properties, it may be assumed that explanation samples will produce a more pronounced size and bend ef-
a) is likely to be incorrect. As a result of the microstructural fect, as described by Eqns 8-11). As may be appreciated, the
differences between Figs. 1 and 2, it is a reasonable assump- raw data of tensile results may also be used to examine the
tion that strain localization during plastic strain is far more spread of elongation at failure, Ef. Although it is not as usual
severe in the sample produced at 26 m/s and that this differ- to examine the strain term Ef using the Weibull distribution,
ence will arise primarily from the defect distribution which the elongation to fracture is a more sensitive indicator of
is observed on the fracture surface. casting quality2 so both tensile strength and elongation are

Figure 1. Fracture surface of an average sample produced at 26 m/s at different magnifications.

International Journal of Metalcasting/Summer 2011 41


parameters requiring analysis. In Fig. 3(c), this data is plot- It therefore seems clear that this technique provides a ver-
ted as a distribution of Ef, and in Fig. 3(d) as a probability satile representation of the quantifiable differences between
of failure for a level of elongation, P(Ef). Here the stress term the two casting conditions. Furthermore, this approach al-
from Eqns 2-6 is replaced by the strain term, Ef. lows for a robust evaluation for what may be considered to
be “safe” from a design perspective for multiple parameters.1
Here it is very important to point out that the Weibull modu- In contrast to the values of m-3s presented in Table 2, which
lus of Ef is actually less for the 82 m/s melt velocity than for are applicable only to the specific geometry of the casting
the 26 m/s condition, irrespective of the fact that this trend tested, the results from the Weibull analysis are able to be
is opposite to that observed for the tensile strength. These scaled to cope with different specimen volumes, geometries
differences are evident in Fig. 3(d), and the consequences of and loading conditions (Eqns 8-11).
this observation are that even though the values of elongation
are greater in all samples tested in the 82 m/s condition, the Use of the Ludwik-Holloman Equation
distribution of values is broader, and it may then follow that
the distribution of flaw sizes was increased simply because Further investigation of quality in the A380 high pressure di-
a greater proportion of samples are of higher relative qual- ecastings was conducted using the techniques developed by
ity. As may be appreciated, tensile strength does not need to Cáceres,9,10 based on the Ludwik-Holloman equation (Eqn
change as much to result in a large change in elongation, as 13). Values of K and n defined by Eqn 13 were determined
tensile strain increases. In these cases therefore, the Weibull for each of the two data sets, (i.e., the 25 samples for each
modulus as well as the values of the position parameters (so of the two melt velocities tested). Values of n were deter-
and Efo) are also very important measures of quality. (Posi- mined by first converting all 25 of the 0.2% proof stress and
tion parameter for this alloy is discussed in more detail later tensile strength data points to their true stress values, and all
in the section on the effects of alloy chemistry). elongation data points to true strain values. All of the results

Figure 2. Fracture surface of an average sample produced at 82 m/s at different magnifications (see text for details). Note
that the size of the primary composite defect present on the fracture surface is significantly reduced when compared
to the sample produced at 26 m/s.

42 International Journal of Metalcasting/Summer 2011


for both melt velocities were used to form portions of the From these values of n and K, a quality chart based around
average true stress–true strain curve for the alloy. Values of the model flow curve representative of Alloy 1 was derived,
n were then determined from the slope of the log-log plot and is shown in Fig. 5. In Fig. 5 it can be seen that for the
of the line of best fit for each data set; that is, a) 26 m/s, samples produced at 26 m/s, the comparative iso-q values as
b) 82 m/s, and c) both velocities combined. At 26 m/s, the defined by Cáceres9,10 are between 0.05 and 0.1, and for the
strain hardening exponent, n, was 0.274, and the strength samples produced at 82 m/s, the iso-q values are between
coefficient K was 906. At 82 m/s, n was 0.261, and K was ~0.1 and 0.17. Here it is important to note that the lowest
872. When both data sets were combined n was 0.267 and values for the 82 m/s data set were still greater than the high-
K was 888. All analyses showed reasonable agreement. To est value for the 26 m/s data set.
validate the technique, a comparison was made between the
model true stress–true strain flow curve generated from the Effects of Alloy Chemistry
combined average values of n (0.267) and K (888) shown in
Fig. 4, with the same curve determined experimentally for a Two additional A380 alloys were tested (Alloys 2 and 3 from
single specimen diecast at a velocity at the gate of 82 m/s. Table 1). Alloy 2 had a higher level of Cu compared to Alloy
All true stress and true strain data for both velocities are also 1, and Alloy 3 had a higher level of Zn. These alloys were
plotted. The results therefore confirm that this method pro- tested to determine if there were any quantifiable differences
vides a valid means of determining useful values for n and in quality which could be observed because of the composi-
K, for the alloy A380. An added advantage is the ability of tional changes. Raising Cu level within an A380 alloy may
the model to produce values consistent with those resulting produce higher levels of strength, but also greater quantities
from a statistically large batch of tensile tests, in this case a of hard intermetallic domains in the solidified eutectic that
total of 50 tests. adversely influence ductility. The susceptibility to hot tear-

(a) (b)

(c) (d)

Figure 3. Weibull distributions of Alloy 1 produced at 26 or 82 m/s. Views (a) and (c) show the distribution and values
of Weibull modulus, m for the UTS and elongation at failure, Ef; (b) and (d) show the probability of failure at stress or
elongation levels for the two conditions tested.

International Journal of Metalcasting/Summer 2011 43


ing may also change. Similarly, increased levels of Zn may size, shape and type, the tensile properties of the three sam-
increase porosity in the casting due to its high vapour pres- ples were virtually identical. In some cases, the defects were
sure above 420C (788F).15 Zn has also been reported to sta- more homogeneously distributed than others (i.e., compare
bilize needles or platelets of the Beta phase Al5FeSi,16 which Figs. 7 and 8). However, their actual effects on tensile prop-
is well known to cause embrittlement within cast aluminium erties were nevertheless similar.
alloys. Alternately, either element may enhance castability
as well as increasing strength, so that relative quality may Weibull distributions were then used to examine the accumu-
instead improve. lated data in greater detail. As before, the samples produced at
26 m/s were inferior to those produced at 82 m/s. The Weibull
A further 25 tensile tests were conducted for samples pro- modulus and so values as they relate to tensile strength for Al-
duced at 26 m/s and 82 m/s for each of Alloys 2 and 3. Val- loys 1-3 are presented in Table 5(a). The Weibull modulus and
ues of the mean (m), one standard deviation (s), and m-3s for Efo values as they relate to elongation at failure for Alloys 1-3
the tensile data of each of the two conditions are shown in are presented in Table 5(b). Respective distributions are pre-
Tables 3 and 4. Examinations were also made of the fracture sented in Figure 9 (for tensile strength only). For specimens
surfaces of specimens of Alloy 2 that had been produced at produced with a melt velocity at the gate of 26 m/s, it is clear
a velocity of 26 m/s, for three samples which had similar from Table 5(a) and Fig. 9(a) that Alloy 1 is inferior to Alloy
tensile properties. Images are shown in Figs. 6-8 and some 2 whereas Alloy 3 falls between the other two. This primar-
large, characteristic defects, with different origins are ob- ily reflects differences in one or both of position parameter and
served. It was also noted that, although the defects ranged in Weibull modulus. Table 5(a) shows this to also be the case for

Table 3. Data for Alloy 2 Samples Produced at 26m/s or 82 m/s

Figure 5. Quality chart for Alloy 1 at two different melt


Figure 4. Experimental data for samples produced velocities, describing the experimental data using the
at 26m/s and 82 m/s overlaid with one experimental flow curve derived from the Ludwik-Holloman equation,
true-stress-true strain curve, as well as the model true after the method developed by Cáceres.9,10 The data for
stress-true strain curve derived using the Ludwik- the 26 m/s condition all falls between the q = 0.05 to
Holloman equation. Values of the strain hardening q = 0.1 iso-quality points, whereas that for the 82 m/s
exponent, n, and strength coefficient, K, derived are condition nearly all falls between the q = 0.1 and q = 0.2
shown within the plot. iso-quality points.

44 International Journal of Metalcasting/Summer 2011


samples produced at 82 m/s, although the relative differences Therefore although all alloys clearly benefit from the higher
in position parameter are much less such that the curves nearly melt velocity at the gate, Alloy 3, which contained higher Zn
overlap (Fig. 9[b]). One feature that is interesting to note is the levels, benefits most from this change. Additionally, despite
differences in Weibull modulus for tensile strength between the the almost identical results of m for Alloys 1 and 2 at 26 m/s,
two different melt velocities, for the various alloys. In Alloy 3, the differences between these two when produced at 82 m/s are
m is more than doubled when going from the lower melt veloc- significant. Differences are also pronounced in comparisons of
ity to the higher one, whereas for Alloy 1, the increase is less. the Weibull distribution for elongation at the different melt ve-

Table 4. Data for Alloy 3 Samples Produced at 26m/s or 82 m/s

Figure 6. Fracture surface of a sample of Alloy 2 showing a combination of an oxide film plus a large foam-like shrinkage
defect. Views (a) secondary electron mode, (b), backscattered electron (BSE) mode show the oxide and defect cluster,
(c & d)) BSE at higher magnification. Note also in (d) there are significant quantities of Fe-bearing particles present on
the fracture surface, the white phase (0.2% proof stress, 184 MPa, UTS 296 MPa, 1.7% Ef).

International Journal of Metalcasting/Summer 2011 45


locities. Alloy 2 displays the same general trends as Alloy 1, also provided in Fig. 3 [c & d]). For Alloy 3 both the Weibull
shown by Table 5(b), where the Weibull modulus is lower and modulus and the position parameter (Efo) are greater for the
the position parameter (Efo) is greater, (results for Alloy 1 are samples produced at 82 m/s.

Table 5a. Weibull Modulus and Position Parameter, so for Tensile Stress

Table 5b. Weibull Modulus and Position Parameter, Efo for Elongation at Failure

Figure 7. Fracture surface of a sample of Alloy 2 showing a large shrinkage foam-like defect cluster on the fracture
surface. Views (a) secondary electron mode, (b), backscattered electron (BSE) mode showing the extensive defect
cluster, (c & d) BSE at higher magnification. (0.2% proof stress 180 MPa, UTS 296 MPa, 1.8% Ef)

46 International Journal of Metalcasting/Summer 2011


The conclusions that can be reached from this are (a), Alloy and (b), both Alloys 2 and 3 display superior quality (higher
3 displays a greater flaw size distribution (i.e., lower Weibull Weibull modulus) to Alloy 1 when cast at 82 m/s. As a result,
modulus) than Alloys 1 and 2 when produced at 26 m/s; it may be concluded that the higher alloying contents do have

Figure 8. A sample of Alloy 2 showing several large oxide defects (in the mm range), on the fracture surface. Views
(a) secondary electron mode, (b), backscattered electron (BSE) mode showing the oxide films, (c & d)) BSE at higher
magnification. Note in (d), the material still shows a small amount of the foam-like shrinkage defect cluster in the centre
of the image and hard, Fe-bearing particles showing white contrast. (0.2% proof stress, 182 MPa, UTS 302 MPa, 1.9% Ef.)

(a) (b)

Figure 9. Two Weibull distributions for the three alloys at (a) 26 m/s, or (b) 82 m/s. Note the differences which exist at
26 m/s are reduced at 82 m/s.

International Journal of Metalcasting/Summer 2011 47


an effect on improving castability, but this is most important 480 MPa, and engineering strain of almost 30% (assuming of
at the higher melt velocity. At the lower velocity, increased course that flow instability did not arise leading to the onset
Cu level has had little effect on Weibull modulus, but higher of early necking).
Zn content has reduced this value and hence has reduced the
relative quality because the flaw size distribution is greater. The results and spread of data shown by the flow curve of
Fig. 11 have additional meaning. By solving Eqn 14 for the
Despite the differences due to composition which could be value of strain hardening exponent n, a relationship between
identified using the Weibull statistics in Tables 5 (a & b), the strain outside of the defect (i.e., true plastic strain), and
such variations in chemistry would of course be considered the strain inside the defect, may be related to the fraction of
to be normal during production, where the metal composi- features which have an equivalent effect to defects present on
tion is expected to vary continually within the specification. the fracture surface. That is, it must be taken into account that
Values of K and n were again determined for the data sets defects spread throughout the microstructure are numerous in
using the procedures outlined in the previous section. Table type and their interactions may be extraordinarily complex.
6 shows all of the strain hardening exponents (n) and values
of the strength coefficient (K) recorded for each of the three
alloys, at both melt velocities.

The accumulated true stress and true strain data is presented


in Fig. 10, which shows all of the combined tensile test re-
sults plotted together with the model flow curve representing
all 150 tensile tests of the A380 alloy samples.

Figure 11 shows the corresponding quality chart describ-


ing all three alloys, at both melt velocities together with the
model flow curve describing the data, based on an n value of
0.259 and a K value of 883 (Table 6). Since all three alloys are
within the same compositional specification and the values
are similar, it is not necessary to plot lines of constant quality
(iso-q lines) on the diagram. Points corresponding to values
of q of between 0.01 and 1 are marked. As well as describ-
Figure 10. Experimental true stress–true strain data for
ing the behavior of the three different A380 alloys, the curves
samples of Alloys 1-3 produced at 26m/s and 82 m/s
of Fig. 11 also accommodate reasonably well the combined overlaid with the model true stress-true strain curve
results from other specimen geometries, literature values for derived using the Ludwik-Holloman equation. Values
A380 alloy test bars, and samples machined from castings,17-23 of the strength coefficient, K, and the strain hardening
where these are converted to true-stress and true-strain values. exponent, n, derived to generate the model flow curve
Therefore, it may be suggested that Fig. 11 could permit ten- are shown within the plot.
sile properties from any casting made from A380
alloy to be evaluated against this common baseline.
As will be appreciated, this plot is valid for A380
alloy with 0.2% proof stresses between 160 and
190 MPa, and lower values of 0.2% proof stress and
tensile strength resulting from a coarser Al grain,
or Si particle size, causes the strength coefficient
K to be reduced and hence the curve may be offset
downwards. Similarly, increased Mg content above
0.2% for other 380 type alloys (i.e., C380) causes
the strength coefficient to be raised and the curve
is then offset upwards. What is also interesting to
consider from these derivations of n and K, is their
relative magnitude, and therefore, the scope for im-
provement in the alloy. In particular, it is interesting
to note that, A380 alloy is reaching less than 20% of
its maximum potential, and that if straining was not Figure 11. Quality indices (q) plotted along the flow curve determined
limited by defects (i.e., a defect free casting) neck- for the combined results for three A380 alloys at two different
ing would only begin at a true stress value of 622 melt velocities. Area fraction equivalent defects are plotted on the
MPa and a true strain value (where e = n) of ~0.26. second Y axis, determined by solving Eqn 14 for the value of strain
This would correspond to an engineering stress of hardening exponent, n, shown. Symbols are the same as for Fig. 10.

48 International Journal of Metalcasting/Summer 2011


In Fig. 11, the equivalent fraction of defects (generated making Alloys 1-3. It was expected that the use of recycled
by solving Eqn 14 for a strain hardening exponent of n = material containing larger proportions of oxides and other
0.259) may be examined against the spread of data pres- defects would simply produce a greater spread in the re-
ent. For the samples produced at 26 m/s, the proportional sults and would therefore produce a good comparison. Two
spread of equivalent defects present on the fracture sur- batches of samples were prepared. In the first batch of sam-
face ranges from around 0.32 to 0.44. For the samples ples, the recycled material was cast with no special prepara-
produced at 82 m/s, the equivalent values are 0.22 to 0.34. tion. Once molten and the temperature stable, the dross was
It is important to note that these numbers generated us- skimmed from the surface prior to casting. A reduced pres-
ing Eqn 14 appear to be very large since it is most com- sure test sample taken just before casting commenced (but
mon to have less than 2% average porosity by volume in after skimming) is shown in Fig. 12(a) and it will be seen
a HPDC. However, as noted by Cáceres and Selling,7 the that, as expected, the material contained substantial amounts
volume fraction of defects present in a casting is a poor of hydrogen. It is also reasonable to assume, oxides are pres-
representation of the tensile fracture behaviour. In addi- ent in suspension. Following preparation of the first set of
tion, as pointed out by Sigworth,2 the defects present on samples, the melt was treated with the rotary degassing unit
the fracture surface may be of the order of three to twenty operating at 300 RPM with high purity argon flowing at 7
times the volume fraction of porosity present in the alloy. litres/min. for a total of 20 minutes, and a reduced pressure
Examination of Figs. 1, 2, 6, 7, 8 agree with this state- test specimen was again taken before casting. The results
ment and suggest that a very large proportion of the frac- are shown in Fig. 12(b), and the absence of pores confirms
ture surface may be part of a composite defect structure. that the melt had been successfully degassed. In addition,
In addition to porosity, oxides and Fe bearing particles, it would be reasonable to assume, a percentage of the ox-
other microstructural features such as Cu bearing inter- ides present in the melt were also eliminated.1,25 Gallo25 has
metallics, coarse silicon plates, heterogeneous grain sizes shown that by melt degassing alone without the use of flux,
arising from pre-solidified grains24 and weaker than aver- around 60% of oxide films are removed from the melt, along
age aluminium grains, are also likely to be contributing to with up to 80% of the spinel-like oxide inclusions that are
the defect equivalence, where complex flow localization present. One interesting observation in this regard appears
is considered. In addition, the skin of an A380 HPDC may in Fig. 12(c), which shows the results of a reduced pres-
have tensile properties higher than the average,* meaning sure test specimen taken after holding the molten metal for
the weaker, interior material, occupies a relatively large 16 hours after the test of Fig. 12(b). The sample presented
proportion of the cross sectional area and therefore will in Fig. 12(c) has almost the same measured density as Fig.
also play an important role in strain localization leading 12(a) due to the re-hydrogenation of the molten aluminium,
to failure. (*Note: Preliminary experiments on a limited but the appearance of the porosity present is somewhat dif-
number of samples [three] of Alloy 1 produced at 26 m/s, ferent to that observed in the lower section of Fig. 12(a) and
where the centre was carefully machined out of the cast- may have resulted from oxides in the melt. Irrespective of
ing leaving only the skin layer, suggest the strength prop- the reasons for this curious difference, both hydrogen and
erties are approximately 20% higher than the average, oxides are well known to adversely influence casting quality
for the test bars used in the current work.) In the current and it would be quite reasonable to suggest samples made
samples, the “skin” layer is 400-500μm thick, and there- from the material shown in Fig. 12(b) should be of superior
fore accounts for approximately 1/3 of the cross sectional quality to those made from the material shown in Fig. 12(a).
area. Therefore, although the numbers generated by solv-
ing Eqn 14 for the equivalent defect contents are high, Twenty-five tensile tests were conducted with material
they do not appear to be unrealistic in light of the actual cast both before and after the rotary degassing, therefore
defect clusters observed on the fracture surfaces. representative of the metal obtained from the reduced
pressure test samples shown in Figs. 12(a & b). Values
Effect of Melt Degassing of the mean (m), one standard deviation (s), and m -3s

To further test these proce-


Table 6. Individual and Combined Strain
dures and to examine one
Hardening Exponents, n and Strength Coefficients, K, for Alloys 1 to 3
potential way to improve
casting quality, an addition-
al set of experiments were
conducted utilizing the prin-
ciples outlined above. Alloy
4 from Table 1 was manu-
factured at a melt veloc-
ity of 82 m/s from recycled
runners, biscuits, and reject
samples of A380 used for

International Journal of Metalcasting/Summer 2011 49


for the tensile data of each of the two test conditions are or no benefit in rotary degassing of the molten metal. This
shown in Table 7. It is important to note the averages would appear to be a very surprising result and not con-
of each set of tensile data were nearly identical; the un- sistent with what is known about the advantages of degas-
treated condition in Table 7 corresponded to the metal in sing molten metal. One important observation was that
Fig. 12(a), and the treated condition corresponded to the there was no increase in the upper bounds of the data set
rotary degassed metal shown in Figure 12(b). The data corresponding to the treated (degassed) metal. This tends
presented overlaid with the model true-stress–true strain to suggest that hydrogen did not have an adverse effect in
flow curve is presented in Fig. 13, and the corresponding the current sample geometry, since if it did, it would be
quality chart is presented in Fig. 14. expected that an overall improvement would be observed
in the treated metal, similar to the differences seen be-
On first examination, considering only the mean values tween melt velocities of 26 and 82 m/s. It may however
presented in Table 7, it would seem that there was little be likely that, in thicker sections, hydrogen would have a

Table 7. Data for Alloy 4 Samples Produced from Recycled Metal,


either a) Untreated, or b) Treated by Degassing with High Purity Argon

Figure 12. Reduced pressure tests for (a), as-melted alloy, untreated, r = 2.39g/cm3. (b), rotary degassed alloy, r = 2.74g/
cm3. (c), as (b) +16h hold before taking sample; r = 2.5g/cm3. Sample diameter was 5cm at the base.

50 International Journal of Metalcasting/Summer 2011


far greater effect as it would segregate to the centre of the failure in Fig. 15(b). Note that the Weibull plot for Ef was
casting, thereby contributing to shrinkage porosity. What similar in form to that for the tensile strength; for the un-
is most likely, however, is simply that hydrogen porosity treated material m was 5.6 and the position parameter Efo
is, at least over this specific range of tensile results and was 3.83. For the treated material, m was 7.9 and Efo was
elongations, a defect of lesser importance. That is, either 3.89. From the engineering tensile stress data for these sets
much larger defects contribute to the failures, or the effect of tests, the Weibull modulus, m, was 25 with a so value
of hydrogen on the defect clusters present, is negligible. of 329 MPa for the untreated material. The value of m was
In summary, at this level of quality, hydrogen has little or 44.7 for the treated alloy, with a so value of 332 MPa. For
no effect on the measured mechanical properties. Examina- both stress and elongation data, the position parameter is
tion of the values of m -3s (for the engineering stress and almost the same, but the values of Weibull modulus are
strain), and those shown on the flow curve of Fig. 13, (for different, which suggests a change in flaw size distribution
the true stress and strain) suggests there may be a change between the two conditions. These differences in Weibull
in the relative spread or distribution of results. It may also modulus are actually similar to those between the melt ve-
be expected that the value of Weibull modulus, m, will also locities for Alloy 1, shown in Table 5a.
show differences because this value is a direct indicator
of the flaw size distribution. Weibull plots for the tensile During the tensile testing to generate the results used in
strength are presented in Fig. 15(a), and for elongation to Figs. 13-15, samples displaying oxide defects on the frac-

Figure 13. True stress-true strain data overlaid on the Figure 14. True stress-true strain plots for untreated and
model flow curve derived from the Ludwik-Holloman treated samples that correspond to Fig. 13.
equation, showing the relative spread of data between
the materials which was a) recycled and untreated, or b)
subsequently treated by degassing with high purity Ar.

(a) (b)

Figure 15. Weibull plots for Alloy 4 with either no melt treatment, or with a melt treatment, where the procedure conducted
was rotary degassing with high purity Ar. Plot (a) is for the tensile strength, and (b) is for the elongation at failure, Ef.

International Journal of Metalcasting/Summer 2011 51


ture surface that were visible to the naked eye were routinely use is the subject of much debate,11 but it is important to
segregated for further examination. Three samples from the note each technique has widely varying complexities. The
untreated batch of material were placed aside, whereas only techniques used in this current work all provide excellent
one showed this type of defect in the treated material. After information regarding casting quality and all can clearly
testing was complete, and the data analyzed, it became clear identify what constitutes a “better” casting. The methods
that these four samples also represented the lowest four data developed by Cáceres9,10 may be considered to be robust in
points in the combined set of test results, falling within the that they can represent different casting geometries from
range of q = 0.05 to q = 0.1 in Fig. 14. different sources, for a single alloy specification, and allow
them to be compared against a unified measure of casting
Figure 16 shows secondary electron images of the frac- quality. An even more powerful tool for analysis of as-cast
ture surfaces of the four segregated samples in (a, c, e high pressure diecastings is to use these accumulated re-
and g). Corresponding backscattered electron images are sults in combination with a derivation of the Weibull mod-
provided in (b, d, f, and h). For reference, the specimens ulus of both tensile strength and elongation.
with the fracture surface shown in Fig. 16 (g & h) demon-
strated the lowest properties (159 MPa 0.2% proof stress, Whichever technique that is used, most crucial is the
283 MPa UTS and 2.1% Ef), with all three of the other ability to establish the baseline performance with high
samples showing almost identical properties to each other reliability. Once the baseline condition and importantly,
(157-165 MPa 0.2% proof stress, 304-306 MPa UTS and the spread of data, are known, changes in quality and the
2.7-2.8% Ef). Interesting comparisons may be made be- reasons for these changes are relatively simple to detect.
tween the fracture surfaces. Figure 16 (a & b) represented Secondly, samples generated during the evolution of the
the one sample showing a visible oxide defect where the baseline data, provide a wealth of reference information.
molten metal was treated by degassing. In (c & d), close In the current work, it has been demonstrated that casting
examination revealed there were two oxide defects pres- quality may be improved by increasing the melt velocity
ent on the fracture surface (arrowed). A small amount of at the gate, which has the effect of decreasing the size of
the foam-type dispersed shrinkage porosity mentioned defect clusters appearing on the fracture surfaces. Cast-
earlier was also able to be detected from the BSE im- ing quality is also influenced by alloy composition, but
age. Alternately, in (e & f), only a small oxide defect was this effect would appear to be complex, and further in-
present but a larger amount of the foam-type shrinkage vestigation regarding the actual basis of the differences is
porosity was evident as the darker region in the centre of warranted. Rotary degassing to remove hydrogen and ox-
Figure 16(f). In (g & h), it is important to note that, not ides from the melt does not influence quality in the same
only is the amount of the foamy shrinkage defect present way as melt velocity at the gate, but it does decrease the
relatively significant, the size of the oxide defect on the spread of results generated. In this instance, casting qual-
fracture surface was also larger than in all other exam- ity would appear to be improved through the removal of
ples. This sample corresponded to the lowest mechanical a proportion of oxides from the melt, which has an effect
properties derived from the combined set of tests. on the value of Weibull modulus. As may be appreciated,
it is impossible to completely eliminate oxides from the
What is most significant is that the four lowest values of HPDC process as many will originate simply due to the
tensile properties within the combined data sets all dis- extreme turbulence that exists during casting. However,
played the presence of one or more oxide films present on eliminating large oxides already present in the melt which
the fracture surface. This would suggest that the presence arise from the use of recycled metal would appear clearly
or absence of these films is critical to the properties and to be advantageous.
relative quality of HPDC components. In this instance, it
may be concluded that rotary degassing clearly had very Finally, it is important to note that defects with substantially
little effect on average tensile properties, but did have an different size, shape and distribution may have similar or
effect on the spread of results. Quality was improved in equivalent effects on the tensile failure. In HPDC defects
the treated material since the Weibull modulus was in- are usually present in complex clusters where many differ-
creased from 25 to 44.7. This increase corresponded to ent features have an inter-relationship, and, as a result, such
the proportion of samples displaying large, visible oxide defect clusters may behave as singular defects with a sub-
defects on the fracture surface being reduced by melt stantially different size.
degassing, and as a result, the quality of the final cast
product was improved. Acknowledgements

Summary and Conclusions The authors would like to thank Andy Yob and Gary Sav-
age for assistance with casting, reduced pressure testing and
Here it is important to comment on the different measures rotary degassing. The authors would also like to thank Drs.
of what constitutes casting “quality,” and the various ways Carlos Cáceres, Geoffrey Sigworth and Prof. Ian Polmear
in which it can be evaluated. What is the best method to for their valuable comments on this manuscript.

52 International Journal of Metalcasting/Summer 2011


Figure 16. Fracture surfaces of the four samples containing naked eye visible defects on the fracture surfaces. Views (a, c,
e and g) are secondary electron images, and (b, d, f and h) are the corresponding BSE (Back Scattered Electron) images.

International Journal of Metalcasting/Summer 2011 53


References 14. Lumley, R.N., O’Donnell, R.G., Gunasegaram,
D.R., Givord, M., “Heat Treatment of High Pressure
1. Campbell, J., “Castings: The New Metallurgy of Cast Diecasting Alloys,” International Patent Application
Metals,” 2nd ed., Elsevier, Oxford, UK (2004). PCT/2005/001909, WO2006/066314.
2. Sigworth, G.K., “Quality Issues in Aluminium 15. Lumley, R.N., and Schaffer, G.B., “Anomalous Pore
Castings,” Fundamentals of Aluminium Metallurgy, Morphologies in Liquid Phase Sintered Al-Zn Alloys,”
Chapter 7, R.N. Lumley, ed., Woodhead Publishing, Met. and Mat. Trans. A, 30A, (6), pp.1682-1685
Cambridge, UK, pp. 155-182 (2010). (1999).
3. Davies, I.J., “Best Estimate of Weibull Modulus 16. Zak, H., and Tonn, B., “Melt Treatment, Grain
Obtained Using Linear Least Squares Analysis: An Refinement and Modification 1: Effect of Alloying
Improved Empirical Correction Factor,” J. Mat. Sci., Elements on Iron-Containing Intermetallics in Al-Si-
39 (4) pp. 1441-1444 (2004). Mg-Cu 319 and 380 Alloys,” Conf. Proc. ICAA11,
4. Jayatilaka, A., De, S., and Trustrum, K., “Statistical Aachen, Germany, pp. 279-284 (2008).
Approach to Brittle Fracture,” J. Mat. Sci., 12, 17. Wang, L., Makhlouf, M., and Apelian, D., “The Effect
pp.1426-1430 (1977). of A380 Alloy Chemistry on its Microstructure and
5. Hertzberg, R.W., “Deformation and Fracture Mechanical Properties,” AFS Trans. vol. 103, pp. 675-
Mechanics of Engineering Materials,” 4th ed., John 681 (1995) (and references 1-5 therein)
Wiley and Sons Inc., p.271 (1995). 18. Lumley, R.N. and Griffiths, J.R., “Fatigue and
6. Surappa, M.K., Blank, E.W. and Jaquet, J.C., “Effect Fracture Resistance of Heat Treated Aluminium High
of Macro Porosity on the Strength and Ductility Pressure Die-castings,” Conf. Proc. 112th Metalcasting
of Cast Al-7Si-0.3Mg Alloy,” Scripta Metall, 20, Congress, NADCA, paper T08-042 (2008).
pp.1281-1286 (1986). 19. NADCA Product Specification Standards for Die
7. Cáceres, C.H., and Selling, B.I., “Casting Defects and Castings 6th ed., NADCA publication 402, Standard
the Tensile Properties of an Al-Si-Mg Alloy,” Mat. Sci. A-3-2-06, NADCA, Wheeling, Illinois. (2006)
Eng. A, A220, pp.109-116 (1996) 20. MatWeb Material Property Data, Online Materials
8. Drouzy, M., Jacob, S., Richard, M., “Interpretation of Information Resource,
Tensile Results by Means of a Quality Index,” AFS Int. http://www.matweb.com/search/DataSheet.aspx?Mat
Cast Met. J., vol. 5, pp. 43-50 (1980). GUID=5f92a8f7d6ad416c8ce9398cae14a363&ckck=1
9. Cáceres, C.H., “A Rationale for the Quality Index of Accessed Feb. 2010.
Al-Si-Mg Casting Alloys,” Int. J. Cast Met. Res., vol. 21. Kaufman, J.G., “Properties of Aluminum Alloys,
10, pp.293-299 (1998). Tensile, Creep, and Fatigue Data at High and Low
10. Cáceres, C.H., “A Phenomenological Approach to the Temperatures,” The Aluminum Association, ASM
Quality Index,” Int. J. Cast Met. Res., vol. 12, pp.367- International, Ohio, USA, pp. 266-267 (2002).
375 (2000). 22. Alcan Primary Foundry Alloys, Alcan Primary Metal
11. Tiryakioğlu, M., Campbell, J., and Alexopoulos, N.D., Group, Alcan Aluminum Corp. Ingot Sales, p.64
“Quality Indices for Aluminum Alloy Castings: A (2002).
Critical Review,” Met. And Mat. Trans. B, 40B, pp. 23. Makhlouf, M.M., Apelian, D., Wang, L. Microstructure
802-811 (2009). and Properties of Aluminum Die Casting Alloys,
12. Gunasegaram, D.R., Finnin, B.R., Polivka, F.B., NADCA Publication, #215, p.71 (2007)
“Effect of Flow Velocity on the Properties of High 24. Murray, M.T., “High Pressure Diecasting of
Pressure Diecast Al-Si Alloy,” Mat. Forum, vol.29, Aluminium and it’s Alloys,” Fundamentals of
pp.190-195 (2005). Aluminium Metallurgy, Chapter 9, R.N. Lumley, ed.,
13. Gunasegaram, D.R., Finnin, B.R., Polivka, F.B., “Melt Woodhead Publishing, Cambridge, UK., p. 217-259.
Flow Velocity in High Pressure Die Casting: It’s Effect (2010).
on Microstructure and Mechanical Properties in an 25. Gallo, R., “Cleaner Aluminium Melts in Foundries: A
Al–Si Alloy,” Mat. Sci. Tech., vol. 23, #7., pp.847-856 Critical Review and Update,” AFS Trans., vol. 116,
(2007). pp.195-220 (2008).

54 International Journal of Metalcasting/Summer 2011


Technical Review & Discussion tained would be to have much higher cooling rates than are
achieved in diecasting, and even then, it might be expected
An Evaluation of Quality Parameters for that the hydrogen will still precipitate out of solid solution
High Pressure Diecastings as porosity. Any hydrogen porosity present and oxide films
R. Lumley, N. Deeva, M. Gershenzon; must contribute to the equivalent defect fraction appear-
CSIRO Future Manufacturing Flagship, Victoria, Australia ing on the fracture surface, but it is the largest oxides that
are most visible and appear in conjunction with the lowest
Reviewer: As important to die casting results as gate veloc- values of tensile properties (Figure 16). For the readers in-
ity, is terminal pressure and achieved final cavity pressure. terest, an excellent representation of how rotary degassing
These values are not given in the paper, but higher terminal can influence inclusion size and content is provided in ref
pressure likely resulted from the higher gate velocity and 1, page 150. We believe that degassing mainly eliminates a
was the reason for better performance in some experiments. proportion of the oxides that contribute to the equivalent de-
fect fraction present on the fracture surface. In this regard,
Authors: In the current results, we have found that the dif- the Weibull modulus does show a significant change / reduc-
ferences in quality due to gate velocity arise primarily be- tion in the flaw size distribution, since the value of modulus
cause of the dispersed shrinkage porosity observed on the is substantially increased (from a value of 25 to a value of
fracture surfaces of the tensile samples. This is substantially 44.7 for tensile strength).
reduced in the higher velocity material. The question then
becomes “Which part of the shot pressure and velocity pro- Reviewer: 3. The comment that Cu (alloy 2) might cre-
file relates to this microstructural defect?” It is our opinion, ate hard intermetallic domains does not seem reason-
and based on discussions with a colleague who has conduct- able when compared to the Si and beta type crystals al-
ed experiments examining these effects, that increasing melt ways present in 380 alloy HPDC components, but Cu
velocity at the gate primarily influences shear of the metal at does reduce ductility simply by increasing strength.
the point of restriction. From his experience and examina-
tion of shot traces taken under differing conditions, the melt Authors: Whereas the Si and b- Al5FeSi crystals are
velocity only influences the impact (terminal) pressure (i.e. clearly of primary importance in affecting ductility in the
the instant at which the cavity is full), and not the maximum cast material, the solidified Cu bearing phases are also
in-cavity pressure during intensification, which occurs after extensively distributed within the solidified eutectic in the
the cavity is full. as-cast condition. As may be appreciated, there may be
1% Fe by weight in the casting, but 3-4% by weight of
Reviewer: The results regarding hydrogen (degassed vs. Cu. Of this, some is retained in solution in the aluminium
not degassed) should not be a surprise—at HPDC pres- grains during cooling, but much of it actually is observed
sures, hydrogen solubility in the liquid would be higher in to solidify in the eutectic, or subsequently precipitates
the solidifying metal than during gravity casting, and any out of the solid solution as coarse equilibrium Al2Cu par-
hydrogen pores that might have formed would, according to ticles, forming on grain boundaries and other high free
the gas laws, have been compressed to extremely small and energy sites. This intermetallic phase also adversely in-
insignificant size. While it can be argued that degassing also fluences ductility. Increasing the content of any brittle
removed oxides, only removal of large films or dross would particle which resides in the fracture path will result in
be significant because the extreme turbulence encountered lower ductility, whether it is b- Al5FeSi, Si, or any other
as melt is sprayed at high velocity into the die cavity would hard intermetallic phase.
rapidly react with any oxygen-bearing atmosphere or lubri-
cants to create new finely divided oxides. Reviewer: Zn (alloy 3) in 380 at normal and increased
levels does little more then increase the density of the al-
Authors: We essentially agree with the reviewers comment. loy (greater weight of metal in each casting). This was
The solubility of hydrogen in aluminium under pressure shown to be significant in past studies when lower cost
might reasonably be assumed to increase under pressure. 380 having 3% Zn proved to actually be more expen-
Certainly, it has been proven that this occurs with Si, for sive because of the greater weight that was shipped, and
example. However we have not currently seen documented the lower cost 380 did not offset that additional cost.
proof of this effect for hydrogen, and the effect may not be
significant. Irrespective of the solubility during solidifica- Authors: Zn is an interesting element in diecastings and we
tion, even during cooling to room temperature, it follows believe its role requires further investigation; it seems to
that any hydrogen residual in the solid solution is non-equi- impart both advantages and disadvantages. Many diecast-
librium and likely to be forced out of solution, in the same ers believe that at least 0.5% is required (this amount may
way that Cu can precipitate onto grain boundaries as equi- always have been there in the alloy purchased), and higher
librium Al2Cu. That is, the hydrogen must precipitate out Zn alloy (e.g. 2%) is generally cheaper, as mentioned. We
as porosity if it was actually present previously in the solid have observed here that increased Zn content was appar-
solution. The only way in which this hydrogen would be re- ently beneficial to castability, but only at the higher melt

International Journal of Metalcasting/Summer 2011 55


velocity used. At the lower melt velocity, increased Zn con- Authors: The tables contain important data to help clarify
tent led to an increased flaw size distribution and hence re- the shape of the distribution for the results, and the authors
duced quality. There is also some evidence that suggested believe it is instructive to the reader. That is, ±1s describes
high Zn contents correlated historically with porosity out- 68.2% of the data, ±2s describes 95.4% of the data, and
breaks and increased reject rates. We believe if a higher ±3s describes 99.7% of the data.
Zn alloy was proven to be more expensive in production,
both from the perspective of increased density and from Reviewer: Is there a significant difference in Alloy 1 and 2
increased reject rates, then a lower Zn version of the alloy at 26 m/s?
would be preferable. (Note B380 is the low Zn version of
A380 alloy, having <1%Zn). Authors: The Weibull modulus of tensile strength for Alloys
1 and 2 are nearly identical, as stated in the text, but the po-
Reviewer: Table 2 contains a lot of data. The 1 sigma, and 2 sition parameter for the two is different (i.e. the failure prob-
sigma lines could be eliminated. ability curve for Alloy 2 is moved to the right, Figure 9a).

56 International Journal of Metalcasting/Summer 2011

You might also like