FINAL REPORT

INVESTMENT CASTING SHELL CRACKING

By
Von Richards

January 2000-May 2003.

University of Missouri, Rolla

Award Number: DE-FC07-00ID13845

Disclaimer: Any Opinions, findings and conclusions or recommendations are those

of the author and do not necessarily reflect the views of the Department of Energy.

Acknowledgement: This material is based upon work supported by the U.S.

Department of Energy under Award No. DE-FC07-00ID13845.
Primary Contractor: Tri-State University, Angola Indiana
Subcontractor: University of Missouri-Rolla, Rolla, Missouri

Other Partners: Steel Founders’ Society of America

Mercury Marine Propeller Division
Conbraco Industries
Precision Metalsmiths
Hitchiner Manufacturing

Contacts: Principal Investigator

Von L. Richards
University of Missouri – Rolla
Department of Metallurgical Engineering
1870 Miner Circle
Rolla, MO 65409-0001
573-341-4730
vonlr@[Link]

Project Team: DOE-HQ contact: Harvey Wang

Industry Contact: Raymond Monroe, Steel Founders’
Society of America
Project mentor: Bob Trimberger, DOE Idaho
Contract Specialist: Marshall Garr, DOE Idaho

2
 Final Technical Report

INVESTMENT CASTING SHELL CRACKING

Executive Summary

This project made a significant contribution to the understanding of the investment

casting shell cracking problem. The effects of wax properties on the occurrence of shell
cracking were demonstrated and can be measured. The properties measured include
coefficient of thermal expansion, heating rate and crystallinity of the structure.

The important features of production molds and materials properties have been indicated
by case study analysis and fractography of low strength test bars. It was found that stress
risers in shell cavity design were important and that typical critical flaws were either
oversize particles or large pores just behind the prime coat. It was also found that the true
effect of fugitive polymer fibers was not permeability increase, but rather a toughening
mechanism due to crack deflection.

I. Comparison of Goals and Accomplishments

Goal Accomplishments
Upgrade in-process testing  AN ANOVA study was completed to show the
benefits of four point bend as opposed to three point
bend testing and to show the effects sample
preparation technique on shell material testing.
 Demonstrated that cristallinity and relaxation effects
occur in pattern wax and influence the stress applied
to shells in the autoclave process.
 Developed a non-contact method to measure wax
expansion that can be done with feedback hotplate,
thermocouple readout and computer interfaced
digital camera.
Reduce shell cracking  Demonstrated the effect of wax reclamation process
scrap on thermal expansion behavior of pattern wax. This
correlated with increased shell cracking scrap at the
participating foundry.
Evaluation of shell  Evaluated three-point and four-point flexure tests
material bench tests and sample preparation techniques for shell strength
samples
 Measured toughness, elastic modulus, and
coefficient of thermal expansion for shell materials.
Evaluate correlation of  Demonstrated that toughness and not permeability
shell process parameters, improvement was the effect that improved autoclave

3
bench test, and shell performance when fugitive polymer fibers were
manufacturing added to the shell formulations.
performance  Demonstrated that low strength shells were limited
by build flaws. Most common flaws were:
1. -large particles at the interface between prime
coat and first back-up coat
2. -delamination or pores at the interface between
prime coat and first back-up layer.

II. Introduction:
Shell cracking is an age-old problem in investment casting. The objective of this
program is to develop an understanding of the problem of shell cracking in the
investment casting manufacturing environment and developing laboratory tests for mold
materials that can be used as control tests to characterize shells in the production
environment. This final report reviews five key areas in the investment-casting program:
four-point bend testing, wax thermal expansion, sonic modulus testing, fracture toughness
testing and permeability of the shells. Each of these key tasks is presented as separate
sections; the student participants under the supervision of P.I [Link] Richards performed
these tests.
The conceptual framework for these tasks is based on the concepts of elastic
stress analysis and fracture. Essentially, the wax expansion during mold de-waxing
imposes on stress on shell since thermal expansion of shell is lower than that of wax. The
stress on the shell depends on the elastic response superposition required to constrain the
free strain of the wax. This results in a high stress at the intersections of long surfaces of
the shell with the ends other surfaces. The tensile properties of the ceramic bars depend
on the size and geometry of the existing flaws. As a consequence we need to characterize
the elastic response of the shell, the strength distribution of the shell (flaw population)
and fracture toughness of shell material. The thermal expansion of shell is also a factor to
be considered in rigorous elastic modeling of shell-wax interaction.

4
III. FOUR-POINT BEND TESTING TO CHARACTERIZE THE STRENGTH OF
CERAMIC MOLD SHELLS

Dr. Von L. Richards

Student Participants: Craig Erford, James Sturgeon, Teck-Seng”Chris”Tan, Deepak R,
Ginger Conin, Phil Jackson and Sony Mascreen
III.(i) Introduction:
The tensile properties of the ceramic bars depends on the size and geometry of the
existing flaws, so there is a considerable scatter in the values of strength determined by a
tensile, bending or fatigue test. Ceramic parts produced with the same materials and using
similar methods fail at very different loads. In order to design structural parts using
ceramics, the probability that a flaw present will cause failure to occur at any given stress
must be known. The Weibull distribution and Weibull modulus provides one statistical
approach to designing with ceramics.
Samples from all of the participating foundries were tested. The three-point bend
test is the prevalent method of determining the strengths of shells in industry but the
four-point bend test is thought to give more accurate results than the three-point method
and was used in all strength measurements. A four-point bend method has been
developed for testing the properties of ceramic mold shells. The four-point method
subjects the tensile portion of the sample to a uniform stress over a specified region. By
comparison, the stress varies over the test length of three-point modulus of rupture test,
reaching the calculated value at only one point. A fully equalized fixture was built and
tested following the concepts of error reduction in ceramic testing proposed by F.I.
Baratta, (Baratta,1982) for technical ceramics. Sample preparation is critical for these
tests. Therefore a wax pattern was developed to minimize the preparation damage and
stress concentrations on the tensile side of the test bar. Although shell cracking usually
occurs during de-waxing, the flaw population of the ceramic is preserved through firing

5
III.(ii) Designing the Test Mold:
In developing the testing procedure we looked at some case studies of shell
cracking at participant foundries. It appears that most frequently the cracking began as a
tensile fracture of the hot face (or primary coat) surface in a highly stressed region.
Therefore we concluded that the tensile strength of the face side of the shell might be
most important.
The test article was designed taking into account some of the sources of error in flexure
testing of ceramics discussed by Baratta (Baratta, 1982). In that reference Baratta takes
into account the following sources of error:
Nonlinear stress distribution
Initial curvature of the specimen
Anticlastic curvature
Large deflections
Wedging stress
Tangency point shift
Neglect of corner radii effects on area moment of inertia
Contact stress.
As with any design, there were some compromises, particularly in making the
test article producible by the participating casters and in attempting to utilize equipment
which would already be available in steel foundries having other molding processes
included in their suite of manufacturing processes. The effect of non-linearity of stress
distribution was limited by use of L/d ratios of nine-to-ten depending on the participant’s
process. L and d are defined in figure 1b.
Initial specimen curvature was typically of the order of radius to thickness greater
than 100, so this effect is negligible. The effect of anticlastic curvature should be
negligible since the range of width to thickness for the bars was four to six. The effects of

6
large deflections were limited by the relatively low strength of the bars being tested, less
than 0.5 percent at the L/d ratio used.
The error due to wedging stress was considered negligible again due to the small
deflection before fracture even though the a/d ratios ranged between 1.2 and 1.4. “a” is
the distance from the loading point to the support point, and “a” and “d” are defined in
figure 1b.
Test bar mold was designed with about a 0.16-in radius while typical specimen
thickness was .25 inches to 0.4 inches. This would lead to about eight percent systematic
error in the specimen strength reported. However, the radius was included based on input
from the participating casters that a tighter radius would lead to build-up of the face coat
in the edges of the tensile face of the test bar. Our test design included using the tensile
surface of the flexure sample and the edges of that surface generated by the shell build-up
process. The existing methods at some of the participants involved either building-up a
sheet of shell material on a flat paddle and releasing it by impact, then testing the material
that does not fracture, or building-up a shell on a flat wax article and sanding or sawing
the edges. In the former case we felt we would have proof tested and eliminated lower
strength material containing critical flaws causing fracture at low stress. In the latter case
we attempted this method, but found that the degree of “proof testing” which occurred
in the sanding or sawing step depended on the manual skills of the technician and that we
were inducing fine cracks and flaws due to particle “pull-out” on the edge of the tensile
side of the test article.
Contact stress was one area in which we violated the recommended criterion due
to equipment space limitations, with a contact radius of only about .25 inches where 1.5
inches would have been calculated from the recommendations given by Baratta. The only
ameliorating factor here is that our breaking stress to elastic modulus ratio was probably
lower than the 10 -3 value used by Baratta.
We initially specified groups of fifteen to twenty samples based on the frequently cited
Weibull modulus of about twenty in the investment casting literature. As will become
apparent later, this sample size was too small because the Weibull modulus of production
materials is below this by a statistically significant amount. Having a fully articulated
fixture compensated the effects of specimen twist.

7
III.(iii) Experimental Procedure:
The ceramic test bars used in this procedure were made with the tensile side
against the wax. Each participating foundry was given the test bar mold. The foundries
injected the mold with their wax (Figure 2). The mold then went through each foundry’s
slurry process. In some cases the areas marked by the arrows were wiped wet, but in
other cases they were inaccessible during coating so they were sawn off afterwards. The
tensile bar has a flattened “U” shape to it. Both unfired and fired bars were to be tested by
the four-point bend method. For each test, about fifteen to twenty bars would be prepared.
The test bars would be cut on the compression side (rough side) (Figure 3). Nothing was
done to the tensile (smooth) side (Figure 4). Using the diamond saw, the bar was cut on
the compression side to remove the legs of the “U”. A 260-grit diamond-polishing wheel
was used to obtain the desired flat surface. A very flat surface was desirable. This meant
sanding some of the bars with 220-grit sandpaper. The sandpaper was placed on glass to
ensure a flat sanding. Once the specimens were ready, they were put in the oven at 150°C
for three hours. The three hours in the oven was only for the previously fired bars.
(Unfired bars had a different procedure, which is mentioned below.)
When the samples were cool enough, they were weighed. The length and
thickness were measured in millimeters and recorded. They were then broken on the
four-point bend apparatus (Figure 5). The load to break the bar was recorded in Newtons.
If any bars broke at or outside the four-point bend points, the data was considered suspect
and left out of the statistical analysis.
Unfired bars were also tested using the same preparation as the fired bars. About
15 bars were used per trial. The difference begins after being cut with the saw. Once the
unfired bars were cut, they were not sanded. This was because the unfired bars cut much
more easily. They gave a very flat surface to work with. Length and thickness in
millimeters was once again recorded. Due to the bars being wet for this test, the weight
was not recorded. The bars were placed in 95°C water for one hour. The bars were then
tested in the 4-point bend apparatus. To do this, the hot bar was extracted using tongs. It
was then placed in the 4-point bend apparatus. The load to break the bar was recorded in
Newtons. Testing of unfired test bars helps us simulate the actual conditions of the

8
ceramic material in an autoclave. Heating the test bar in water to boiling temperature
melts the wax on the surface of the bar (the same happens in an autoclave). This test
helps us draw conclusions on the effect of polymers on the strength of shells.
III.(iv) Results:
The results in three areas will be covered:
1. Comparison of cut coupons to the mold coupons from the design discussed
above
2. Four-point bend as a measure of process drift over time
3. Three- point bend versus four-point bend
Comparison of cut bar versus the proposed test article design is shown in table I. The cut
coupon has a lower Weibull modulus but a higher mean strength. Considering a normal
distribution the cut bar shows a higher standard deviation of strength values.
The comparison of three-point bend results to four point bend results at two foundries is
summarized in table II. Although the statistical significance may be questionable, the
trend toward giving a higher nominal strength value with three-point bend is evident.
The process drift over time at one of the participating foundries is shown as a Weibull
plot for three different sampling times in figure 6. These are four-point results using the
proposed test bar design. The Weibull modulus was greatly different at each of these
samplings: 4.8, 2.6, and 11.7. Weibull Modulus for each of the participating foundries
was calculated. The graphs for each of the participating foundries are as shown in
Figures7 to 9. Figure 10 shows the comparison of weibull plot with fibrous shell samples
from participant B.

III.(v) Discussion of Results:

The cut coupon has a higher mean strength than the molded bar from (the designed test
bar) and a broader distribution of test values. During sanding to obtain straight edges on
the test bars cut from flat coupons, several of the bars broke. Thus the weaker bars from
the sample population were eliminated, giving a false high mean strength. At the same
time, the sawing and sanding introduced flaws in the edge of the test pieces on the tensile
side, which superimposed another flaw distribution in addition to the flaws introduced in
the shell-making process. This would give rise to the broadened breaking stress

9
distribution compared to the proposed test article design. The molded bars did have
clean-up diamond saw cutting and sanding on them to generate a uniform and measurable
thickness, but this was all done on the compression side of the flexure bar, while the
tension side and its edges were generated by the wax pattern. The explanation for the
trend to higher mean strengths in the three point bend test as compared to the four point
bend test is not statistically significant, but does seem logical considering that a smaller
region of the sample is exposed to the maximum tensile stress in the three point bend test
than in the four point test. Thus, the probability of encountering a flaw from the large size
tail of the population distribution is lessened when three-point bend is used.
Figure 10 shows the comparison of weibull moduli of non-fibrous polymer shell sample
to the fibrous shell sample. The weibull modulus is slightly higher for the fibrous samples
(5.27) when compared with non-fibrous samples from the same foundry (4.71).
The Weibull moduli are lower than commonly reported from samples prepared by impact
removal from a waxed paddle. The Weibull moduli are lower than the value initially used
to select the sample size, which leads to some additional inaccuracy. As indicated by
figure 11 (from Baratta), at the 90% confidence level the characteristic strength should be
determined within plus or minus ten percent, but the Weibull modulus will only be known
within plus or minus thirty percent at these levels with only twenty to twenty two
samples. However, the process drift is clearly bigger than the error bar on the
measurement, so that the characteristic strength and Weibull modulus do represent real
changes in the process.

III.(vi) Conclusions and Recommendations:

The conclusions we can draw from this study to date are:
1. The molded test bar is more representative of the shell building process than
a sanded or sawed bar prepared from a flat sample
2. The four-point bend method should be more representative of the weakest
link in the shell structure, but this has not been proven to a statistically
significant extent.
3. Weibull moduli of test lots can represent the variation of shell strength due
to process variations.

10
Recommendation:
Use a test bar that has the tension surface and the edges of the tension surface generated
by the prime coat and shell build, rather than prime coat with the edges generated by
abrasion or sawing.

III.(v) References
1. Francis I. Baratta, Requirements for Flexure testing of Brittle Materials, AMMRC
TR 82-20, Army Materials and Mechanics Research Center, Watertown, MA,
1982.

11
Figure 1 a. Stress Distribution in Three Point Bend Testing

Figure 1b. Stress Distribution in Four Point Bend Testing

12
Figure 2: Test Bar Mold, the Injected Wax Portion to be Dipped in Slurry. Arrows
Indicate Optional Wet Strike-Off Flats.

Figure 3: Compression Side of Test Bar Showing Sawn and Sanded Areas

13
 Figure 4: Tensile Side of Test Bar

Figure 5: Four-Point Bend Apparatus

14
 Table I. Comparison of Cut Coupons to Designed Test Bar (Molded)

Table II. Comparison of Three-Point Bend to Four Point Bend

Figure 6: Process Variation over Time at One Participating Foundry (Note that
some of this was due to intentional changes by the foundry.)

15
 Unfired Shell-Foundry C.

Date of Receipt:06/05/01
Date of Test: 10/15/01
Weibull Modulus 6.98
Regression Line:
-0.02939+6.98492 ln (strength)
1
0
lnp

-1
-2
-3

-0.4 -0.3 -0.2 -0.1 0.0 0.1

lns

Figure 7: Plot of Cumulative Probability Versus Fracture Strength

16
 Fired Shell-Foundry C
Firing Temperature: 1300F
Molds are dark in appearance
Date of Receipt: 06/06/01
Date of Test: 12/06/01
Weibull Modulus= 1.559
1
0
lnp

-1
-2
-3

-0.4 -0.3 -0.2 -0.1 0.0 0.1

lns

Figure 8: Plot of Cumulative Probability Versus Fracture Strength

17
 Fired Shell- Foundry A
Date of Receipt:06/05/01
Date of Test: 10/15/01
Weibull Modulus: 5.1598
Regression line:
1.1289+5.1598ln(strength)

R-squared= .9502
1
0
lnp

-1
-2

-0.8 -0.6 -0.4 -0.2

lns

Figure 9: Plot of Cumulative Probability Versus Fracture Strength

18
 Unfired Shell- Foundry A
Date of Receipt:06/05/0
Date of Test: 12/06/01
Weibull Modulus: 6.019

Regression Line: 20886+6.01975ln(strength)

R-squared: 0.9779
1
0
lnp

-1
-2

-0.4 -0.3 -0.2 -0.1 0.0 0.1 0.2

lns

Figure 10: Plot of Cumulative Probability Versus Fracture Strength

19
 3

y = 4.7084x -
1 3.4057

y = 1.5598x +
0
ln(ln(1/(1-

0.6437
- -2 - -1 - 0 0.5 1 1.5
2.5 1.5 0.5
-1
p)))

y = 5.2661x -
-2 4.4953
y = 7.0449x +
0.8347
-3

-4
ln(Strength
)

Foundry "B" Fiber

Foundry "C" Foundry Foundry "B" Containing Samples
"A"

Figure 10: Comparison of Weibull Plots of Participating Foundries

20
Figure 11A . Error in Weibull Parameter Determination based on
Number of Samples (from Baratta, 1982)

Figure 11B . Error in Characteristic Strength Value based on

Weibull Slope and Sample Size (Baratta, 1982)

21
 IV. FRACTURE TOUGHNESS OF INVESTMENT CASTING SHELLS USING
FRACTOGRAPHY

Von L. Richards, Ph.D.

University of Missouri at Rolla
And
Student Research Participant; Sony Mascreen, James Sturgeon

IV.(i) Background
The purpose of fractography is to analyze the fracture features and relate these
features to find the causes of fracture. While conducting the four-point bend test on
ceramic mold shell samples, used in investment casting, we found some samples had
exceedingly low strengths that looked like outliers in the Weibull analysis. Our
participating foundries were interested in examining these low strength ceramic shell
samples. Fractographic analysis was used to find the cause of failure of these samples.
Results showed shell build flaws were responsible for some of the failures; therefore we
decided to find fracture toughness of these broken samples using fractography. This
method is an unconventional way of finding the fracture toughness. The determination of
plane-strain fracture toughness, KIc, is important in understanding the material

22
characteristics of investment casting shells. Once the value of K Ic is calculated we can
more thoroughly understand the fracture mechanics of the shell itself. Using microscopic
analysis we are able to find the critical flaw that induced the fracture, and the fracture
toughness can be calculated by using Griffith’s crack theory.
Additionally, we explored artificially induced flaws using a knoop indenter in
conjunction with the four-point bend test to get fracture toughness. Here a crack is
induced in the sample using a knoop indenter. The breaking strength is obtained by the
four-point bend test and the growth and depth of crack is observed using fractography.
The two methods are compared in the paper.

IV.(ii). Introduction
The study of fracture mechanics and fracture toughness is an important aspect in
describing a material’s resistance to fracture. This differentiates between the material’s
intrinsic resistances to crack extension, which can be considered more or less severe
depending on parameters like processing, handling etc. We know that in brittle fracture,
presence of a crack-like defect leads to unstable rapid failure when the stresses at the
crack tip will exceed a critical value. This critical value called KIC, the plane strain
fracture toughness, is the property of a material that gives its inherent resistance of that
material to failure in the presence of a crack-like defect. The relation connecting the
fracture toughness and the crack length is given by (Ref 1)

KIc=Y ( a) ½……………………………………..(1)
Where: Y is a geometric constant
 is the stress applied to cause crack propagation
a is the length of an edge crack
The tensile stress  is obtained from the four-point bend testing of the molded test
bars. The procedure for this is given in detail in the next section. The geometric factor Y
is dependent on the specimen and crack geometry. For a semi elliptical surface crack with
major plane of crack perpendicular to the applied tensile stress, the stress intensity factor
is given by (Ref 2)
KI2=1.21 a  2/Q………………………………..(2)

23
 Where: Q is the flaw shape parameter.
Figure-1 shows flaw shape parameter Q, plotted against the crack-shape ratio
a/2c. Here “c” is the width of the crack and 0 is the yield strength. In brittle materials
yield strength is equal to the tensile strength, therefore /0 =1. Using these we get the
value of crack shape parameter and hence a more accurate fracture toughness value can
be calculated.
This paper is a continuation of work done earlier in Tri-State University by the
first author. This method used the Knoop hardness tester in an attempt to create a flaw
that could be measured after fracture. The different indentation loads used are 44.15N,
58.86N, and 68.67N, the procedure for which is given later.
Fractography, a very useful and effective technique, has an important role in
understanding the fracture strength and behavior of ceramics. We use fractography to
identify the critical flaw that induces the fracture. Here we are looking into the inherent
flaw in the samples that we have broken using four point bend apparatus to get the
strength of the samples.
The eventual purpose of the work reported in this paper is to determine whether
the reproducible features of the microstructure of shell material can be used to adjust the
toughness. It is a further objective to determine if that toughness is less dependent on the
random flaws than is the measured strength. Varying the toughness could then be used to
make the shells more robust against the random flaws induced during shell build.

IV.(iii) Procedure
IV.(iii).a. Sample Preparation
The ceramic test bars used in this procedure were made with the tensile side
against the wax. Each participating foundry was given the test bar mold. The foundries
injected the mold with their wax. The mold then went through each foundry’s slurry
process. The tensile bar has a flattened “U” shape to it as shown in figure 2.
The test bar material was produced by participating foundries using the design
described by Richards, et al (Ref 3). Fired bar material was used for the processes
described here. The test bars would be cut on the compression side (rough side). Nothing
was done to the tensile (smooth) side. Using the diamond saw, the bar was cut on the

24
compression side to remove the legs of the “U” until a flat specimen remained. A
260-grit diamond-polishing wheel was used to obtain the desired flat surface. Once the
specimens were ready, they were heated in an oven at 150C for three hours. Length and
thickness were measured in millimeters and recorded. The samples were then broken on
the four-point bend apparatus shown in figure 3.

IV.(iii).b. Four Point Bend Test

A four-point bend method has been developed earlier for testing the
properties of ceramic mold shells (Ref 3). The bend testing method developed was based
on the ASTM standard C1161-94, and the work of Baratta (Ref 4). The four-point method
subjects the tensile portion of the sample to a uniform stress over a specified region. A
fully equalized fixture was built and tested following the concepts of error reduction in
ceramic testing proposed by F.I. Baratta (Ref 4) for technical ceramics. The load to break
the bar was recorded in Newtons. If any bars broke at or outside the four load points, the
data was considered suspect and left out of the statistical analysis. Fractographic studies
were done on the low strength bars obtained from the four-point bend test. Strength of the
molded samples from each foundry was tested. Richards, et al, discussed the analysis of
these test bars in detail (Ref 3). The Weibull plots for the participating foundries are given
in figure 5. These plots show the lots of bars tested contained multiple flaw distributions,
so that fitting a single Weibull modulus to the data would be inappropriate.

IV.(iii).c. Indentation and Four Point Bend Test

The sample, which was to be tested using an induced flaw, was also prepared as
above. A knoop indenter was used to induce the flaw and the indentation was made on the
center of the tensile side of the sample. The long axis of the indenter was kept
perpendicular to the applied tensile force. The loads used for the indentation were
44.15N, 58.86N, and 68.67N. The area that was indented was dyed using a Neopen J.A.P.
- S.H.F. Penetrant, then the samples were dried at 150 0 C for 3 hours. Afterward all the
bars were broken on the four-point bend apparatus. The load to break the bar was
recorded in Newtons. Method for indenting, dying and breaking was modeled closely
after ASTM Standard C1421-99, for the Determination of Fracture Toughness of

25
Advanced Ceramics at Ambient Temperature. This method was selected because of its
similarity to the strength testing that was already underway. In this case also, if any bars
broke at or outside the four load points, the data was considered suspect and left out of
the statistical analysis.

IV.(iii).d. Fractography
The fracture surface was kept intact after the strength testing and observed under
a microscopic camera (Polaroid DMC 2.0) along with a microscopic scale. Transverse
lighting source was used, that is light was incident at an inclined angle to the fracture
surface of sample. A picture of the fracture surface was taken and then the critical flaw
that resulted in fracture was identified as in figure-4. When looking at the surface, the
crack depth is measured from the smooth surface of the bar to an area that was most
clearly connected to the other fracture surface before fracture. The dimensions of the
crack are then measured using Scion Image analyzer. Once the crack dimensions and
strength are determined, fracture toughness is calculated. In the early work on indentation
--induced flaws, the method of measurement was less precise, using a binocular wide
field microscope fitted with an eyepiece camera.

IV.(iv). Results
The fracture toughness (plane strain critical stress intensity) obtained by
fractographic methods from the low strength samples observed in four-point bend testing
of samples from the three foundries tested. This data is shown in table 1A, along with the
flaw shape parameter determined in each case. The date of sample manufacture by the
foundry is noted, because process changes have been instituted at various times by the
participating foundries. At foundry B a comparison test was run between the standard
slurries and a proposed engineering changes using slurry that included fugitive polymer
fibers.
The data obtained in the early work on indentation-induced flaws is also shown in
table 1B. These samples were produced by two of the same foundries from which
samples were taken for the fractographic approach. In fact the foundry C material was

26
form the same production batch for both methods. In both cased the breaking strength
was determined by the four-point bend method described by Richards, et al (Ref 3). .

IV.(v) Discussion of Results

Standard deviation is smaller in the fractographic analysis than in the initial
indentation work reported here. This can be attributed to better precision in image
capturing and analysis than in earlier indentation work and to the use of the flaw shape
parameter in the fractographic work. The earlier indentation work geometric factor Y was
assigned a nominal value of unity that was improved upon in the subsequent fractography
work with the introduction of the flaw shape parameter, Q, (Ref 1). Comparing relations
(1) and (2), we get Y=(1.21/Q)1/2. In the data reported here flaw shape parameter Q ranges
from 0.75 to 2.05, while in the initial work the unity geometric factor corresponded to an
assumed constant flaw shape parameter of Q=1.1.
As shown in table 1A, the mean toughness of fiber-containing samples (after the
fibers were burned out) appears to be higher. To test whether the mean toughness was
equal to that of shells from the same foundry without fiber, a student’s t-test was
performed. This gave a 5% probability that the mean fracture toughness is equal or a 95%
probability that the fiber containing samples have higher toughness. This can be
explained by the theory of crack deflection. When the shells with the polymer fiber are
de-waxed in an autoclave, the high temperature in the autoclave causes the polymers to
melt and thus leaves voids in the shell. These voids act as crack deflectors. When a crack
tries to run through the shell, it hits the void and spreads to a different direction, thus
giving higher fracture toughness values. The student’s t-test assuming unequal variances
was used for comparing the fracture toughness of similar foundries for the indented
samples and the regular mold samples. The probability that the Foundry C values for
different methods are equal is 23%. The probability that the Foundry B values with out
fibers are equal is 27%. We can see that there is a significant but small probability that the
values generated by the two methods for each foundry are equal. Based on such limited
data, they may be in qualitative agreement. The indentation method needs to be refined
by utilizing the improved flaw measurement methods developed in the fractography
study.

27
 The mean of fracture strength and toughness is compared in table 3. Strength is
directly related with the fracture toughness: the higher the strength, the higher the fracture
toughness.
The polymer fiber containing shells have greater toughness. This may contribute
to their improved autoclave performance with respect to shell cracking. The observed
improvements in the autoclave de-waxing of shell containing the polymer fibers had
previously been totally accredited to improved permeability (Ref 5).
The improvement of toughness of fired shell with the addition of the polymer
fiber is probably due to crack deflection. Crack deflection can take place when there are
local areas in a ceramic that have a lower resistance to crack propagation than an average
plane cutting through right angles to the tensile stress (Ref 6). Crack deflection
mechanism is based on the tilting and twisting out of crack planes around the grains. The
stresses acting in inclined planes near a crack tip depends on the angles involved. For
straight cracks that, tilt is about a direction perpendicular to the crack advance through an
angle . The stress intensity factor required for crack propagation is given by
K( ) = K( =0) sec2 ( /2)…………………………………….(3)
Where: K(=0) is the plain strain stress intensity .
Crack propagation is diverted by a layer of porous structure due to the addition of
polymer fiber material; the effective stress intensity is lessened. K(=0) corresponds to
the apparent stress intensity measured by this technique. Thus, for the load orientation
expected in de-waxing, the shell material will behave as though the fracture toughness is
higher. This improvement in the toughness should also be evident in unfired shells once
the fibers are melted or volatilized. However the toughness measurements in the unfired
shells will be more difficult.

IV.(vi) Conclusions
The conclusions we have reached after this study are:
1. The introduction of polymer fibers in the shell, gives the shell a tougher
microstructure.
2. The tougher microstructure thus produced will enhance autoclave performance
relative to shell cracking.

28
 3. The mean strengths of the shells seem to follow fracture toughness.

IV.(vii) References:
1. George E. Dieter, Mechanical Metallurgy, Published By McGraw Hill, Third
Edition, Pages 354-356.
2. G.R Irvin, Analysis of Stresses and Strains Near the End of a Crack Traversing a
Plate, [Link], Vol24, pg.109-114, 1957
3. Von L. Richards and Ginger Connin, Four-Point Bend Testing to Characterize the
Strength of Ceramic Mold Shells, Investment Casting Institute, 49th Annual
Technical Meeting 2001, pg.13.1-13.10.
4. Francis I. Baratta, Requirements for Flexure Testing of Brittle Materials,
AMMRC TR 82-20, Army Materials and Mechanics Research Center, Watertown,
MA, 1982.
5. Dan Duffey, The Wexcoat Binder System, Investment Casting Institute, 49th
Annual Technical Meeting 2001, pg.10.1-10.10.
6. John B. Wachtman, Mechanical Properties of Ceramics, Wiley-Interscience
Publications 1996,pg. 159-163.

Figure-1. Flaw-Shape Parameter Q for Surface and Internal Elliptical Cracks (Ref 1)

29
Figure-2. Tensile Side of the Molded Test Bar

Figure-3. Four-Point Bend Apparatus.

30
 Figure-4. Fractography of Broken Bar, Oval Indicating the Critical Flaw

2
y = 4.7084x - 3.4057
1
ln(ln(1/(1-p)))

y = 1.5598x + 0.6437
0
-2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5
-1
y = 5.2661x - 4.4953
-2
y = 7.0449x + 0.8347
-3

-4

ln(Strength)

Foundry "C"
Foundry "A"
Foundry "B"
Foundry "B" Fiber Containing Samples

Figure–5. Comparison of Weibull Plots of Participating Foundries

31
 Table-1A. Plane Strain Critical Stress Intensity as Determined by Fractography
Foundry- Flaw Fracture
Received Depth of Width of Shape Breaking Toughness Standard
Date Crack(a) Crack 2c a/2c Parameter Strength (Kic) Average Kic Deviation
Millimeter Meter Millimeter Q MPa Mpa(m)^0.5 Mpa(m)^0.5
Participant A Samples Fired Without Fugitive Fibers
1.25 0.00125 9.43 0.132556 0.75 0.883 0.070266
A 0.9265 0.00093 4.83 0.191822 1.1 0.736 0.041635
6/5/01 0.4186 0.00042 1.79 0.233855 1.18 1.023 0.037557 0.050608 0.0146384
1.566 0.00157 6.87 0.227948 1.18 0.746 0.052973
Participant C Samples Fired Without Fugitive Fibers
1.35 0.00135 5.18 0.260618 1.3 0.566 0.035552
C 1.43 0.00143 3.69 0.387534 1.75 1.118 0.062294 0.039233 0.0158326
2/9/01 1.9 0.0019 4.74 0.400844 1.8 0.515 0.032614
1.49 0.00149 4.59 0.324619 1.425 0.42 0.026472
Participant B Samples Fired Without Fugitive Fibers.
1.9671 0.00197 3.9 0.504385 2.25 1.1435 0.065953
1.344 0.00134 3.0281 0.443843 1.9 2.819 0.145924
B 1.0188 0.00272 3.3727 0.302073 1.35 1.6063 0.140541
04/17/02 0.8913 0.00089 2.3756 0.375189 1.7 1.7928 0.079958 0.103483 0.015901
1.4219 0.00142 3.6344 0.391234 1.75 1.1559 0.06418
1.895 0.0019 3.915 0.484036 2.05 2.6617 0.157949
1.1756 0.00118 3.4855 0.337283 1.55 1.2993 0.069878
Participant B-Fired Samples Containing Fugitive Polymer Fibers.
1.46 0.00146 5.3 0.275472 1.35 3.219 0.206342
1.44 0.00144 4.25 0.338824 1.55 2.063 0.122567
B 1.51 0.00151 3.29 0.458967 2 1.442 0.077232
2/19/02 1.6 0.0016 4.57 0.350109 1.65 1.974 0.119818 0.151627 0.047311
2.2807 0.00228 5.023 0.454051 1.95 2.286626 0.15243
1.0813 0.00108 3.8031 0.284321 1.35 3.045326 0.167995
2.1881 0.00219 6.079 0.359944 1.65 2.036294 0.144541

32
 Table-1B. Fracture Toughness by Indentation Method (Fired and Without Fugitive Fibers)
Flaw Fracture
Depth of Width of Shape Breaking Toughness Average Standard
Foundry Crack(a) Crack 2c a/2c Parameter Strength (Kic) Kic Deviation
0.9823 0.000982 3.1293 0.31 1.41 2.4857 0.127885
B 0.9333 0.000933 3.1293 0.30 1.4 1.1406 0.057403 0.080061 0.033666

3/15/01 0.707 0.000707 3.1007 0.23 1.18 1.568 0.074812

0.8389 0.000839 3.4465 0.24 1.2 1.8929 0.097555

C 0.9006 0.000901 2.2373 0.40 1.85 2.2306 0.095931

2/9/01 0.6102 0.00061 2.6788 0.23 1.18 1.2418 0.055043 0.058065 0.03645
0.4262 0.000426 1.8361 0.23 1.18 0.6268 0.023219

Table- 2. Toughness Comparison of Samples with Fugitive Fibers and Samples without Fugitive Fibers
from Foundry B

Fibrous Sample Non -Fibrous

2/19/02 Sample-- 4/17/02
Average
Toughness 0.151627 0.103483348
Standard
Deviation 0.047311151 0.015900956
Sample Size
N 8 7

Table-3. Toughness and Strength Comparison of Participating Foundries

Foundry Mean Strength Mean K1c

A 0.834 0.0506
B 1.889 0.1034
C 0.5674 0.039233
B Containing
Fugitive Fibers 2.165 0.1516

33
 V. THERMAL EXPANSION OF INVESTMENT CASTING PATTERN WAX.
By
Student Participants; Sony Mascreen, Brian Sutherland, Phil Jackson, Adrienne Orf
And
[Link] L. Richards, Ph.D.
University of Missouri at Rolla

V.(i) Introduction:
The purpose for testing the thermal expansion of pattern wax is due to the
effect of the free strain developed during autoclave de-waxing. Typically
investment-casting shell cracking may occur at combinations of high load and high stress
intensity. The wax expansion during mold de-waxing imposes stress on shell since
thermal expansion of shell is lower than that of wax. The expanding wax induces stress
onto the inner surface of shell, which sometimes is the cause of shell cracking. It is a
widely held paradigm that autoclaving causes shell cracking. Although knowledgeable
ceramic engineers working in the investment casting industry will admit that the flaw
population introduced in shell production affects the resistance to cracking, the effects
often appear during autoclave de-waxing. Therefore, it is also reasonable that the
mechanical properties of the pattern wax will affect performance during the de-wax
cycle.

34
 Our fixture is made of aluminum, which surrounds the wax on all the sides but
one, which enables an excellent heat transfer. There is a glass slide on the side of wax that
isn’t covered by aluminum, which allows us to observe the change in length of wax.
There is a rectangular aluminum plug also, as seen in figure 4, which keeps the flow of
wax uniform when it reaches liquid state.
The thermal expansion of investment casting pattern wax has been measured by a
non-contact method from room temperature up through the melting temperature. The
effects of the following variables on thermal expansion have been examined: section
thickness, orientation with respect to injection flow, injection pressure, heating rate, and
recycling procedure. It is clear that heating rate affects the temperature at which one
observes the onset of relaxation of the non-crystalline portion of the structure. A higher
heating rate raises the relaxation temperature and the thermal expansion that occurs
before the structural relaxation. Thinner sections can cool more rapidly and affect the
degree of crystallinity and free volume of the non-crystalline portion of the wax structure.
Recycling procedures tend to affect the volume change upon melting the crystalline
portion of the wax structure. This is an abrupt volume increase and can apply significant
stress to the shell during autoclaving. Filled and unfilled waxes from participant
foundries were examined as well as virgin waxes suggested by vendors.

V.(iii) Procedure
The waxes used for these experiments are obtained from the participating
foundries. These foundries inject wax into the wax step block pattern to provide samples.
The wax is injected at three different injection pressures; (1) the normal injection pressure
of the foundry, (2) 100 psi higher than normal and (3) 100 psi lower than normal. Vendor
waxes are injected either by the vendor or by a participating foundry. Figures 1 and 2
show the wax step block pattern. The steps are 0.25 inches, 0.375 inches and 0.5 inches
thick. The injection direction is parallel to the steps.
V.(iii).a. Machining The Wax
The first step in preparing a wax specimen for experimentation is to machine a
small sample strip from the wax step mold. This is done by using a vertical band saw, a
vertical milling machine and final sanding with 400 grit sand paper. The wax strip is

35
machined to give a good fit in the test cell channel. To avoid annealing the wax, it is
refrigerated between each of the machining steps to insure that premature heating does
not occur. Once the wax is completely machined, the test cell itself must be prepared.
V.(iii).b. Test Cell
The test cell is made of aluminum with a cover glass and a slider to follow the
wax. The first step in the preparation of a test cell is simply cleaning it by using a light
dishwashing soap. It is then rinsed and dried thoroughly. A small layer of vacuum grease
is applied to the inner-faces of the aluminum top, bottom, and slider. The slider is
lubricated with vacuum grease so that it provides a non-contact measurement method. A
small amount of no-stick solution is placed inside the channel where the wax will be
tested. The wax strip is inserted into the test cell channel. Once all of these preparations
have been made, the cell is bolted together and put back into the refrigerator. The
assembled and disassembled test cell is shown in figures 3 and 4.
V.(iii).c. Setting Up The Experimental Apparatus
In order to get a highly accurate measurement of the expanding wax, experimental
apparatus is set up underneath a microscopic camera (Polaroid DMC 2.0). The test cell is
partially submerged in a water bath that sits on top of a hot plate. A feedback
thermocouple is positioned in the water bath. This way, the water temperature can be
controlled by the hot plate. Additional thermocouples are located inside the aluminum
body of the test cell. These temperatures are then read using two-port digital
thermocouple readout. Figures 5 and 6 show the test cell set-up.
V.(iii).d. Procedure For Length Measurement
A series of pictures of the wax is taken at different temperatures with a
microscopic camera. Scion Image analysis software is used to measure the wax lengths
from the pictures. Earlier measurements in this study used a video coordinate measuring
machine to measure the length at the same time increments as currently used with the
Polaroid DMC in recent measurements.
V.(iii).e. Procedure For Slow Heating Process
When the slow heating process is being used, the sample is started below room
temperature. A length measurement is taken, and then the test cell is allowed to
equilibrate with the room temperature by adding water. Once room temperature is

36
reached, another length measurement is taken. From this point on, the test cell is heated
at a rate of 3C with length measurements being taken between each 3C increment
(roughly seven to nine minutes between increments to get an equilibrium temperature).
The entire process lasts until the wax reaches a temperature of about 60 to 65C. At this
temperature, the wax is mostly liquid. The experiment is terminated when the liquid wax
starts to leak past the slider. A typical measurement cycle requires two and one-half
hours.
V.(iii).F. Procedure For Accelerated Heating Process
Similar to the slow heating process, the sample is started below room temperature.
An initial length measurement is taken, and then the water is applied to the test cell and is
allowed to equilibrate to room temperature. From this point, the hot plate is set at 300C,
and measurements are made at three-degree increments. The entire set of measurements
takes about eight minutes.

V.(iv) Results:
As a result of performing thermal expansion tests on pattern waxes, we were able
to better examine many of the thermal properties for these waxes. We used several
different experimentation methods to help broaden the amount of information we could
obtain for a wax. Some of the different experimental approaches included looking at
mold injection pressures, injection flow rates, water quenched waxes, recycled versus
virgin waxes, wax thickness, heating rates, and wax orientation. The results for each of
these tests will be discussed in this section.
V.(iv).a. Wax Thickness
One of the first tests that we performed on a pattern wax was designed to
determine the effect of wax thickness. To perform this test, we designed a step block
mold that would be used for the duration of the program. This block consisted of three
different thicknesses: .25 in, .375 in, and .5 in. All three different thicknesses were tested
using the same experimental setup and procedure. Once all the data was collected, the
resulting graphs were compared. From these graphs, we were able to determine that the
thicker sections expand at a slightly smaller rate than the thinner sections (figures 7 and
8). Also, the majority of the expansion difference between the two wax thicknesses was

37
present where all of the crystalline material transformed into liquid. This can possibly be
related to the pressure at which the wax was injected. This causes the amount of
crystalline material in the wax to increase.
V.(iv).b. Recycled Wax vs. Virgin Wax
Another test that was performed on the waxes was developed to determine how
recycled waxes compare with virgin waxes. We used the same thickness in our study to
eliminate the cooling rate factor. Figure 9 showed that recycling of the cerita 29-51 had
the same effect as reported by for participant “A” recycled versus virgin wax – a greater
crystalline expansion portion of the curve with there recycled wax. To be more precise, it
should be noted that this effect was less significant with the Cerita 29-51 and the
participant “C” recycling system than was indicated in earlier work for foundry “A”.
It is very important to note that different recycling procedures will yield different
properties within the wax (figure 10). Recycled wax #1 and #2 were prepared using two
different reclaim methods. As is evident from the graph, recycled wax #1 has a great deal
of crystalline expansion present while the other recycled wax shows more expansion
while the wax is still solid. Although neither recycling process was able to perform as a
virgin wax, the second recycled wax was very similar in the amount of crystalline
expansion.

V.(iv).c. Injection Pressures

To conduct this experiment, we used a wax injected at three different pressures
from the same participant. The three pressure levels were one the normal injection
pressure of the participating foundry and other two were 100 psi lower and 100 psi higher
to the normal pressure. Again, they were tested using the same setup and procedure. The
same thickness was again used to avoid the cooling rate factor. Figure 11 and 12
illustrates the effect of injection pressure. Assuming that we did not overlook any
entrapped gas bubbles, the high injection pressure may have resolved several crystalline
melting events due to blended molecular weights. Also, the main crystalline melting
event is produces a more significant volume increase when higher injection pressures are
used. The higher the pressure, the more crystalline melting expansion can be observed.

38
V.(vi).d .Vendor Waxes
As our study with waxes continued, a few participants asked that we examine some
commercial waxes to help reference our data. The main thrust of the work done during
this year in the wax expansion area was to examine the wax used by Sabau, et al (Sabau,
2001) at Oak Ridge National Laboratory to help provide a reference benchmark for the
participating foundry waxes that have been examined in this program. Coincidentally,
Participant C was using the same wax, Cerita 29-51, but they used reclaimed wax for
patterns as well as sprue and had a very thoroughly controlled staged reclamation system.
This provided an opportunity to compare reclaimed Cerita 29-51 to as-received or
“virgin” wax. The vendor and participant C injected these waxes. The effects of pattern
section thickness and injection pressure were also examined.
The participant foundries recommended examining some commercial waxes to provide a
reference for the database. Remet, and Kindt Collins offered their best “starter wax”.
“Starter wax” is defined as their recommendation to a caster starting an investment
casting process. Cerita 29-51 was used to reference the Oak Ridge study, (Sabau et al
2000). Figure 13shows comparison of all the vendor and participant waxes to date. This
gives participating foundries a means to relate their wax to the vendor waxes and other
research.

V.(iv).e. Injection Flow Rate

The thermal expansion of wax injected at different flow rates were tested.
The wax we used for this was from participant D. The different flow rates used by the
participants were
1. Low flow rate (Flow rate of 2 on a scale of 1- 10)
2. Control flow rate (Flow rate of 3 on a scale of 1- 10)
3. High flow rate (Flow rate of 7 on a scale of 1-10)
We also compared the thermal expansion of these waxes with the water-quenched
wax injected at a control flow rate. These waxes are water quenched in 109 degree F
water for 1 hour).

39
Figure 14 shows the comparison of different injection rates of participant D wax. The low
flow rate wax showed more crystalline expansion when compared to the control flow rate
and water quenched wax injected at the control flow rate.
Water quenching the wax injected at the control flow rate didn’t show any difference in
the linear thermal expansion of the wax when compared with the wax injected at the
control flow rate. Higher injection shear rate of wax doesn’t seem to increase the
formation of crystalline components in the wax. The formation of the crystalline material
is dependent upon the injection pressure only. We were unsuccessful in getting repeated
thermal expansion results from the high flow rate waxes. This was due to the numerous
air bubbles in the wax step block and this in turn gave erroneous results.

V.(iv).f. Wax Orientation

We did complete one study that involved testing waxes that were parallel and
perpendicular to the wax injection direction. They were the same thickness and were
tested using the same procedure. There was very little difference between the two as far
as expansion was concerned. The results for this test can be seen in figure 17.

V.(iv).g. Heating Rates

Another experiment that was used to test the pattern wax involved changing the
rate at which it was heated. For our testing, we have used two different heating rates.
The first rate is a very slow heating process that takes the wax from roughly 15C to
70C in about 2.5 hours. With this process, we can clearly see any critical temperatures
such as the glass transition temperature. We can also observe when the crystalline
material transforms into a liquid substance. It was important to apply this type of heating
rate so that we would be able to get a very close idea of the characteristics of the wax.
The second heating process took the wax from 15C to 70C in nearly 8 minutes. We
used this rapid heating rate to try to replicate an actual autoclave process. Again, we are
able to see critical temperatures and phases, but this time we are observing how the wax
would act in an actual manufacturing environment. Once we had collected an adequate

40
amount of data for each heating rate, we made a comparison between the two (figure18).
As one can see, the amount of crystalline material that goes into liquid is very similar
between the two heating rates, but the temperature at which the amount of volume
relaxation occurs is different. The onset of volume relaxation occurs at a higher
temperature with faster heating.

Figure 1: Top view of wax step block

41
 Figure 2: Side view of wax step block

Figure 3: Top view of disassembled test cell

42
Figure 4: Top view of the assembled test cell

Figure 5: Complete Experimental Apparatus

Figure 6: Water Bath and Hot Plate Setup

43
 Fractional Volume vs. Temperature

0.016

0.014

0.012

0.01
Fractional Volume

0.008

0.006

0.004

0.002

0
0 10 20 30 40 50 60 70
Temperature °C)
(
400 psi 3/ 8" 400 psi 1/ 4"

Figure7: Comparison of Participant C Wax of Different Thickness

Fractional Volume vs. Temperature

0.016

0.014

0.012
Fractional Volume

0.01

0.008

0.006

0.004

0.002

0
0 10 20 30 40 50 60 70 80
Temperature ( °C)

1/ 4 inch sample 3/ 8 inch sample 1/ 2 inch sample

Figure 8: Comparison between Different Thicknesses of Water-Quenched Wax

44
 Fractional Volume vs. Temperature

0.016

0.014

0.012

0.01
Fractional Volume

0.008

0.006

0.004

0.002

0
0 10 20 30 40 50 60
Temperature ( °C)

Participant C 1/ 4 inch 400PSI Wax OakRidge (cerita 29-51) 1/ 4 inch 400 PSI
Wax

Figure 9: Comparison of Cerita 29-51 Virgin Wax to Recycled Wax (Cerita 29-51 from participant C)

Participant A Wax
1/4 inch recycled #1vs. 1/4 inch virgin

0.024

0.022

0.02
Fractional Volume

0.018

0.016

0.014

0.012

0.01

0.008

0.006

0.004

0.002

0
0 10 20 30 40 50 60 70 80

Temperature (?C)

1/ 4 i nch recycl ed #1 1/ 4 i nch vi rgi n 1/ 4 i nch recycl ed #2

Figure 10: Comparison Recycling Procedures of Participant A Wax with Virgin Wax

45
 Fractional Volume vs. Temperature

0.016

0.014

0.012

0.01
Fractional Volume

0.008

0.006

0.004

0.002

0
0 10 20 30 40 50 60
Temperature ( °C)

Parti ci pant C 400PSI Parti ci pant C 200 PSI

Wax
Figure 11: Comparison of Different Injection Pressures of Participant C ¼ inch Wax

Figure12: Comparison of Different Injection Pressures of Participant A Wax

46
 ?

Figure 13: Comparison of Different Vendor Waxes to Participant Waxes

Fractional Volume vs. Temperature

Participant D Wax 1/ 4 inch Thickness
0.03

0.025

0.02
Fractional Volume

0.015

0.01

0.005

0
0 10 20 30 40 50 60 70 80
Temperature (°C)

Control Fl ow Rate Low Fl ow Rate WaterQuenched Sampl e

Figure14: Comparison of Different Injection Flow Rates

47
 Figure 15: Comparison of Water Quenched Wax with Low Flow Rate Wax

Figure 16: Comparison between Different Thicknesses of Water-Quenched Wax

48
 ?

Figure17 Comparison of Thermal Expansion of Wax with Different Orientation

Figure18 Comparison of Thermal Expansion of Wax with Different Heating Rates

49
 VI. Sonic Test For The Determination Of Young’s Modulus
By
Teck-Sang “Chris” Tan (Student Participant)
And
Dr. Von L. Richards.
VI.(i) Introduction:
Sonic testing is a nondestructive testing used to determine the elastic modulus or
Young’s modulus. The reason for using this test is to obtain a guideline for four-point
bend test that determines the stress of the ceramic shells of participating members.
During the coordination meeting in January 2001, a question was raised about the
apparent inverse correlation between strength and elastic modulus in the early data. We
have explored this relationship on two batches of ceramic bars from two foundries, with
concurrent strength testing on the same bars. These bars are thinner than the normal test
bar we have been using, which leads to a little more scatter in the strength result. These
foundries will be named as foundry M and foundry H. Research was carried out to choose
the best way doing the test and after referring to some sources, a test using speed of
sound was carried out. Referring to ASTM C1259-98 (Standard Test Method for

50
Dynamic Young’s Modulus, Shear Young’s Modulus, and Poisson’s Ratio for Advanced
Ceramics by Impulse Excitation of Vibration).

VI.(ii) Results:
Referring to Table III-1 (Foundry M), Table III-3 (Foundry M) and Figure III-1,
there is a noticeable relationship of the elastic modulus to the strength of the bars. By
taking the result obtained from the four-point bend test, the strength of two test bars (1_1
and 2_4) is almost the same (at about 10.92 MPa), but there is a big difference in their
elastic modulus values from the sonic test (about 4 GPa). Then, referring to the result
obtained from bar 3_3, this bar has the highest strength (16MPa), but it has about the
median elastic modulus value. When we examine the test results obtained from foundry
H (Please refer to Table III-2, Table III-5 and Figure III-2), the elastic modulus and the
strength of the test bars have narrow distributions (mean for the elastic modulus is 7.20
GPa and the standard deviation is 1.4 GPa; mean for the strength is 3.1MPa and standard
deviation is 0.45MPa). Why is there such a great difference between these two batches of
bars from these two foundries? Pore size distributions were found to be substantially
different. The bars from Foundry M have a broader distribution of pore sizes. One might
suggest that the pore size controls strength, while the pore volume fraction controls the
elastic modulus of the bars. For example, as shown in Figure III-3a, Figure III-3b and
Figure III-3c, it is found that the structure of the bar that has a more regular structure
gave the highest value compared to the other two bars. However, the bars from Foundry
H have a more regular and dense microstructure compared to Foundry M.
The result, which was plotted as strength versus the Young’s Modulus has a
specific relationship for the bars, tested for Foundry M, as shown in Figure 2.
Mathematically speaking, the curve was a second power polynomial curve. The student
participant was asked to explain the phenomenon, and he found that the relationship is
due to the pore volume and size distribution in the ceramic bars. This may follow the
relationship derived by J.K. MacKenzie, E=E (1-1.9P+0. 9P2) (P-pore volume fraction,
0

E -matrix elastic modulus, and E- elastic modulus of sample) to describe the change of
0

elastic modulus due to the addition of low modulus material as second phase.
Specifically, we add pore spaces that have approximately zero bulk modulus values

51
(Kingery, p.775.) If the strength is proportionate to non-pore volume, while modulus
involves a second power term, a second power relationship of strength to modulus is
reasonable.
The modulus results suggest predicting the strength of the bars that are from the
same foundry before testing them with the four-point bend test. But, this does not
conclude that the modulus test is a reliable predictor of strength.

Table 1 -- Foundry M Results from Sonic Modulus Test

Calculation for Young's Modulus for test bar

Date: 2/26/01
Foundry name: M
Received: 02-09-01

b(mm) L(mm) t(mm) f (Hz) T1 m (g) E (GPa) L/t density Type

21.42 130.16 6.17 1888 1.015 132 198.6778 21.1 7.6967a-36
25.593 100.73 5.03 4960 1.016 49.9 369.8704 20.01 3.8456alumina
20.18 130.32 6.39 1856 1.016 45.2 62.92835 20.39 2.6897aluminum
19 98.01 4.5 896 1.014 12.5 5.236655 21.78 1.49171_1
19 98.01 4.5 864 1.014 12.5 4.869287 21.78 1.49171_1
19 98.01 4.5 896 1.014 12.5 5.236655 21.78 1.49171_1
19.13 98.85 4 800 1.011 11.2 5.410177 24.71 1.48071_3
19.13 98.85 4 800 1.011 11.2 5.410177 24.71 1.48071_3
19.13 98.85 4 800 1.011 11.2 5.410177 24.71 1.48071_3
19.1 98.4 4.8 768 1.016 13.2 3.375962 20.5 1.46323_5
19.1 98.4 4.8 896 1.016 13.2 4.595059 20.5 1.46323_5
19.1 98.4 4.8 768 1.016 13.2 3.375962 20.5 1.46323_5
19.1 98.05 4.85 704 1.016 13.2 2.721859 20.22 1.45333_2
19.1 98.05 4.85 768 1.016 13.2 3.239237 20.22 1.45333_2
19.1 98.05 4.85 736 1.016 13.2 2.974924 20.22 1.45333_2
19.18 97.8 4.89 672 1.016 13 2.355783 20 1.41733_1
19.18 97.8 4.89 672 1.016 13 2.355783 20 1.41733_1
19.18 97.8 4.89 640 1.016 13 2.136765 20 1.41733_1

52
 19.12 98.38 4.6 928 1.014 12.3 5.203416 21.39 1.42152_3
19.12 98.38 4.6 896 1.014 12.3 4.850746 21.39 1.42152_3
19.12 98.38 4.6 928 1.014 12.3 5.203416 21.39 1.42152_3
19.1 98.47 4.85 768 1.016 14.1 3.504276 20.3 1.54583_3
19.1 98.47 4.85 640 1.016 14.1 2.433525 20.3 1.54583_3
19.1 98.47 4.85 896 1.016 14.1 4.769709 20.3 1.54583_3
19 98.92 4.23 864 1.012 11.5 5.535052 23.39 1.44652_2
19 98.92 4.23 736 1.012 11.5 4.01652 23.39 1.44652_2
19 98.92 4.23 832 1.012 11.5 5.132641 23.39 1.44652_2
19.07 98.55 4.8 608 1.016 12.9 2.080391 20.53 1.432_4
19.07 98.55 4.8 576 1.016 12.9 1.867165 20.53 1.432_4
19.07 98.55 4.8 576 1.016 12.9 1.867165 20.53 1.432_4
21.42 130.16 6.17 1888 1.015 132 198.6778 21.1 7.6967a-36
25.593 100.73 5.03 4960 1.016 49.9 369.8704 20.01 3.8456alumina
20.18 130.32 6.39 1856 1.016 45.2 62.92835 20.39 2.6897aluminum

Table 2 -- Foundry H Results for Sonically Determined Modulus

Foundry:H
RECEIVED: 3-15-01
TEST DATE:03/30/01

b(mm) L(mm) t(mm) f (Hz) T1 m (g) E (GPa) L/t density Type

21.5 130.2 6.13 1888 1.01 133 202.42 21.2 7.742 a-36
25.6 100.6 5.03 4864 1.02 49.9 354.41 20 3.859 alumina
20.1 130.3 6.35 1856 1.02 45.4 64.586 20.5 2.725 aluminum
18.9 94.2 3.74 864 1.01 12.7 7.665 25.2 1.907 H1
18.9 94.2 3.74 832 1.01 12.7 7.1077 25.2 1.907 H1
18.9 94.2 3.74 800 1.01 12.7 6.5715 25.2 1.907 H1
19.1 100.5 4.1 864 1.01 15 8.2837 24.5 1.911 H5
19.1 100.5 4.1 992 1.01 15 10.92 24.5 1.911 H5
19.1 100.5 4.1 800 1.01 15 7.1019 24.5 1.911 H5
19 89.13 3.6 992 1.01 11.5 8.6475 24.8 1.886 H6
19 89.13 3.6 704 1.01 11.5 4.3553 24.8 1.886 H6
19 89.13 3.6 832 1.01 11.5 6.083 24.8 1.886 H6
18.9 97.59 4.5 928 1.01 16 7.1253 21.7 1.925 H9
18.9 97.59 4.5 928 1.01 16 7.1253 21.7 1.925 H9
18.9 97.59 4.5 992 1.01 16 8.142 21.7 1.925 H9
18.9 95.74 3.9 960 1.01 14.5 10.035 24.5 2.06 H13
18.9 95.74 3.9 832 1.01 14.5 7.5375 24.5 2.06 H13
18.9 95.74 3.9 896 1.01 14.5 8.7418 24.5 2.06 H13
19.1 100.2 4.5 960 1.01 15.2 7.7841 22.3 1.768 H14
19.1 100.2 4.5 960 1.01 15.2 7.7841 22.3 1.768 H14

53
19.1 100.2 4.5 864 1.01 15.2 6.3051 22.3 1.768 H14
19 97.02 4 672 1.01 13.8 4.4794 24.3 1.872 H15
19 97.02 4 608 1.01 13.8 3.6668 24.3 1.872 H15
19 97.02 4 608 1.01 13.8 3.6668 24.3 1.872 H15
19 100.3 4.02 832 1.01 14.8 8.0342 25 1.937 H16
19 100.3 4.02 864 1.01 14.8 8.6641 25 1.937 H16
19 100.3 4.02 800 1.01 14.8 7.4281 25 1.937 H16
18.9 96.31 3.95 800 1.01 14.1 6.6162 24.4 1.959 H17
18.9 96.31 3.95 800 1.01 14.1 6.6162 24.4 1.959 H17
18.9 96.31 3.95 768 1.01 14.1 6.0974 24.4 1.959 H17
19 96.16 3.85 864 1.01 13.2 7.7291 25 1.877 H18
19 96.16 3.85 768 1.01 13.2 6.107 25 1.877 H18
19 96.16 3.85 960 1.01 13.2 9.5421 25 1.877 H18
21.5 130.2 6.13 1888 1.01 133 202.42 21.2 7.742 a-36
25.6 100.6 5.03 4864 1.02 49.9 354.41 20 3.859 alumina
20.1 130.3 6.35 1856 1.02 45.4 64.586 20.5 2.725 aluminum

Table 3. Foundry M Results for Young’s Modulus and Strength

Sample Strength (Pascals) Elastic Modulus(GPa)

1_1 10916179.34 5.114198755
2_2 12354209.3 4.894737722
2_3 13502565.49 5.085859171
2_4 10924663.52 1.938240207
3_1 11774092.37 2.341366459
3_2 14690179.11 2.978673175
3_3 16195233.54 3.569169751
3_5 13634380.45 3.782327428

Table 4 -- Foundry H Results for Strength and Young’s Modulus

Elastic Modulus
Sample (GPa) Strength (MPa)
H1 7.114754 3.80108
H5 8.768534 2.713753

H6 6.361932 3.112619
H9 7.464192 3.694802
H13 8.771506 2.635552
H14 7.291106 3.378315
H15 3.937699 2.988031
H16 8.042135 3.372148
H17 6.44325 2.56
H18 7.792731 2.749888

54
 1.80E+07

1.60E+07

1.40E+07

1.20E+07
Strength (Pa)

1.00E+07

8.00E+06

6.00E+06

4.00E+06

2.00E+06

0.00E+00
0 1E+09 2E+09 3E+09 4E+09 5E+09 6E+09
Elastic Modulus (Pa)

Figure 1 -- Foundry M: Strength vs. Elastic Modulus

55
 4.00E+06

3.50E+06

3.00E+06
Strength (Pa)

2.50E+06

2.00E+06

1.50E+06

1.00E+06

5.00E+05

0.00E+00
0 2E+09 4E+09 6E+09 8E+09 1E+10
Young's Modulus(Pa)

Figure 2 Foundry H: Strength vs. Young's Modulus

Figure 3a -Foundry M 1_1 bar. (Scale increments are 1 mm)

56
Figure 3b -- Foundry M 3_3 bar (scale increments are 1 mm)

Figure 3c -- Foundry M 2_4 bar (scale increments are 1 mm)

57
 VII. STATISTICAL ANALYSIS OF THREE-POINT AND FOUR-POINT BEND
TESTING FOR THE CHARACTERIZATION OF CERAMIC MOLD SHELLS

Student Participants; Sony Mascreen, Phil Jackson

And
[Link] L. Richards, Ph.D.
University of Missouri at Rolla

VII.(i) Introduction:
The objective of our program was to examine measurements that could be used
to characterize variability in the process that relates to the frequency of shell cracking.
The four-point bend method puts more sample volume of material under uniform stress.
This facilitates finding the critical flaw that may be randomly placed in the sample as a
consequence of process variability. In the earlier section (section II) it was explained that
the molded test bar is more representative of the shell building process than a sanded or
sawed bar prepared from a flat sample and the four-point bend method should be more
representative of the weakest link in the shell structure. But it was not been proven to a
statistically significant extent. A two-way ANOVA with replication (which looks into the
possibility of interaction) is used to compare the two bend tests and the two sample types
to reach a statistical conclusion. The two factors used here are, one is the specimen
condition and the second is the type of bend test. The two types of specimen conditions
are molded test specimen and sawed test specimen. The preparation for which will be
explained later in this section.

VII.(ii) Sample Preparation:

The ceramic test bars used in this procedure were made with the tensile side
against the wax. Figure 1 shows the wax mold for the test specimen. The mold then went
through the shell build process used by Pittsburg State University. The tensile bar thus
produced has a flattened “U” shape to it as shown in figure 2. The samples for these tests
were made in Pittsburg State University. Two types of sample where used from the same
shell. One is the molded sample and the other sawed sample.

58
VII.(ii).a. Molded Samples:
The bars were produced by Pittsburg State University using the mold designed for
producing four point bend test bars. Fired bar material was used for the processes
described here. The test bars would be cut on the compression side (rough side). Nothing
was done to the tensile (smooth) side. Using the diamond saw, the bar was cut on the
compression side to remove the legs of the “U” until a flat specimen remained. A
260-grit diamond-polishing wheel was used to obtain the desired flat surface. Once the
specimens were ready, they were heated in an oven at 150C for three hours to dry them.
Length and thickness were measured in millimeters and recorded. The samples were then
broken on the bending apparatus shown in figure 3.
VII.(ii).b. Sawed Samples:
The sawed samples are cut out from the back shell that we have after cutting the
molded samples. The dimensions of the sawed sample are kept as close to that of the
molded sample so that there is no any statistical errors due to dimensional discrepancies.
Also each sawed sample is identified with the molded sample in terms of the similar shell
from which both were sawed. Fired bar material was used for the processes described
here. The tensile sides of the bar were affected when we cut the specimen. Using the
diamond saw, the bar was cut on the compression side to get the rough shape. A 260-grit
diamond-polishing wheel was used to obtain the desired dimensions. Once the specimens
were ready, they were heated in an oven at 150C for three hours to dry them. Length and
thickness were measured in millimeters and recorded.

VII.(iii) Experimental Procedure:

The samples are carefully sorted out and grouped together in batches of molded
and sawed samples. The four point and three point bend tests were conducted. The
breaking strength of the samples was calculated in MPa. An analysis of variance
(ANOVA) was done for two-factor replication. The two factors taken into account are the
bend testing and the specimen conditions.

VII.(iv) ANOVA Two-Factor with Replication:

An ANOVA analysis for two-factor replication was done on the data obtained.
The two-way analysis of variance with replication deals with the case where two different

59
experimental factors have been studied simultaneously. For each factor, it tests whether
the differences resulting from the various levels are statistically significant. It also tests
for any interaction between the two factors - i.e. does some combination of treatments
produce a special effect that could not be accounted for as the simple sum of the two
factors acting separately.

VII.(v)Results:
The statistical analyses of the tests done are show below. Table 1 shows ANOVA table for
two-factor replication. These results are got using Stat Graphics, a statistical data analysis
tool. Figures 6 and 7 show the scatter plot of the two factors i.e. Test type and specimen
condition. Figure 8 and 9 shows the mean and 95 % confidence intervals of the two
factors explored and figure 10 shows the interaction between them and the 95%
confidence interval.

VII.(vi) Discussion of Results:

The ANOVA for two-factor replication done on the three-point and four-point
bend test with different specimen conditions are shown in table 1. The sample size used is
128 samples each for both the tests and a total of 256 samples were used as for this
analysis. The ANOVA table decomposes the variability of result into contributions due to
various factors. The P-values test the statistical significance of each of the factors. The
first factor chosen was the test type. Since the P value is less than 0.05, the test type
(4-point and 3-point) has a significant effect on result at the 95% confidence level. The
second factor was the specimen condition, molded or sawed specimen, for this factor the
P value is less than 0.05, and so the specimen condition has a significant effect on the
result at 95% confidence level. Now we see the interaction between the two factors. The
P values for the interaction of two factors is less than 0.05, hence the two factors show a
significant interaction on the result at 95% confidence interval.
Figure 3 and 4 shows the scatter plot of the test types and the specimen conditions
respectively. Figure 3 shows that the three-point bend test is more scattered than the
four-point bend test. The scatter plot suggests that four-point bend test is more
appropriate to estimate the mean strength of a population because the variation of the

60
strength in the sample population is low whereas three-point bend test shows a large
variation of strength, which in turn will affect the population mean of the sample. Figure
4 shows the scatter plot for the specimen conditions that is the molded and sawed
specimens. The scatter plot doesn’t show a big scatter in either of the specimen condition
other than a single flyer in the sawed sample. One of the important things to keep in mind
while examining a scatter plot is that it is just gives a basic idea of the scattering of
results. One cannot come to any statistical conclusions just based on scatter plot. Figures
5 and 6 shows the further exploration of results in terms of mean and the 95% confidence
intervals. Figure 5 shows that the three-point bend test gives a higher mean for the same
samples. A logical explanation for this trend to higher mean strengths in the three point
bend test as compared to the four point bend test is that a smaller region of the sample is
exposed to the maximum tensile stress in the three point bend test than in the four point
test. Thus, the probability of encountering a flaw from the large size tail of the population
distribution is lessened when three-point bend is used. The comparison of mean and 95%
confidence intervals of molded and sawed samples are shown figure 7. Here the molded
samples show a lower mean strength and the sawed samples seem to have higher mean
strength. This trend can be attributed to the breaking of the weak bars during sanding
operation on the sawed samples to obtain straight edges on the test bars from flat
coupons. Thus the weaker bars from the sample population were eliminated, giving a
false high mean strength. At the same time, the sawing and sanding introduced flaws in
the edge of the test pieces on the tensile side, which superimposed another flaw
distribution in addition to the flaws introduced in the shell-making process. This would
give rise to the broadened breaking stress distribution compared to the proposed test
article design. The molded bars did have clean-up diamond saw cutting and sanding on
them to generate a uniform and measurable thickness, but this was all done on the
compression side of the flexure bar, while the tension side and its edges were generated
by the wax pattern.
The interaction between the two factors and the 95% confidence intervals of each
factors are shown in figure 8. Here one can see that there is a significant interaction
between the sample preparation and the bend tests. The decrease in the mean strength of
the molded specimen from the three-point to the four-point bend test was higher (1.0

61
Mpa) when compared to sawed specimen (0.4 Mpa). This again emphasizes the fact that
the sawed test specimen doesn’t reveal the actual flaw population of the ceramic samples
used. The sawing and sanding introduced flaws in the edge of the test pieces on the
tensile side, which superimposed another flaw distribution in addition to the flaws
introduced in the shell-making process. This gives a false higher mean strength in both
the four-point and three-point testing method. The reason that there is a significant
interaction can be attributed to the fact that three-point bend test gives a higher mean
strength for the molded bar where as four-point bend test shows much lower strength.
Since the tensile side of the bar was kept intact during machining and sanding of the test
bar, the original flaw population is preserved. On a three-point bend test, only smaller
portion of sample is exposed to maximum tensile strength, hiding the flaws outside that
region, this gives a false higher mean strength for the samples.

VII.(vii) Conclusions and Recommendations

The conclusion we have reached after these statistical analyses are:
1. The molded test bar is more of a representative of the shell building process than
a sanded or sawed bar prepared from a flat sample.
2. It is statistically proven that the four-point bend method should be more
representative of the weakest link in the shell structure.
3. The mean strength of the population is dependent upon both the factors i.e. the
test type and the sample condition on a 95% confidence level.
Recommendation:
To know the actual strength and affect of all the flaw population it is
recommended to use a molded sample on a four-point bend test for the bend tests.

VII.(viii) Reference:
1. Richards, Von L. and Connin, Ginger., “Four-Point Bend Testing to Characterize
the Strength of Ceramic Mold Shells”, Investment Casting Institute, 49th Annual
Technical Meeting 2001, Paper No. 13, pg.13.1-13.10.
2. Baratta, Francis I., “Requirements for Flexure testing of Brittle Materials”,
AMMRC TR 82-20, Army Materials and Mechanics Research Center, Watertown,
MA, 1982.

62
 Figure 1. Wax Test Mold

Figure 2. Tensile Side of the Molded Test Bar

63
 Figure 3. Bending Apparatus with a Four Point Fixture

Analysis of Variance for Two Factors with Replication

Source Sum of Squares DF Mean Squares F-Ratio P-Value

MAIN EFFECTS
A: Test Type 34.65 1 34.65 44.22 0.0000
B: Specimen_condition 29.70 1 29.70 37.91 0.0000
INTERACTIONS
AB 5.19 1 5.19 6.63 0.0106
RESIDUAL 197.45 252 0.78

Total (Corrected) 266.99 255

Table 1

64
 10

0
3 Point 4 Point

Figure 4. Scatter Plot for Test Type

10

4
Mpa

0
Molded Sawed

Figure 5. Scatter Plot for Specimen Condition

65
 7
6.8
6.6
6.4
Mpa

6.2
6
5.8
3 Point 4 Point

Figure 6. Mean And 95% Confidence Interval for Different Test Types

7.1
6.9
6.7
6.5
Mpa

6.3
6.1
5.9
Molded Sawed

Figure 7. Mean And 95% Confidence Interval for Different Specimen Condition

66
 Specimen
7.3 condition
Molded
6.9 Sawed

6.5
Mpa

6.1

5.7

5.3
3 Point 4 Point

Figure 8. Interactions and 95% Confidence Interval for Different Test Type

67
 VIII. THERMAL EXPANSION OF SHELL: A STUDY

VIII.(i) Introduction:
Thermal expansion of shell allows calculation of the free strain that can occur in
shell systems. With this data and the properties of the wax, the superposition of strain
method can be used to calculate the stress during critical steps of the investment casting
process. The thermal expansion of the shell not only looks in to the volume expansion of
shell on heat-up, but also the extent of sintering that can occur in the preheat furnace. In
certain cases the stresses can occur as a result of temperature gradient between the inside
and outside of the shell. These stresses can result in metal fining. The thermal expansion
behavior of the shell can affect casting dimensions in another way besides the normal
increase in size with temperature and sintering shrinkage. Shell cracks that occur during
the autoclave or furnace pre-heat operations often result in metal fins on casting. The fins
increase the distance between features perpendicular to the fin. They also cause the shell
wall around the fin to bulge outward for an inch or two. This metal together with the fin
must be removed by surface grinding.

VIII.(ii) Sample Preparation:

The sample for testing the thermal expansion is cut off from a broken test bar
used for 4-point bend test by sawing and grinding. The dimensions are approximated to
length equal to 1 inch, thickness of 0.15 inch and width of 0.2 inch. The sample is ground
to have flat ends, so that it fits readily without any slipping in the dilatometer.

VIII.(iii) Procedure:
The sample dimensions were measured. An Orton’s Dilatometer measured the
expansion of the shell material produced by participating foundries. The expansion
temperature range was set between 50 C and 1000 C and the heating rate was 3C/min to
the peak temperature. The cooling rate was also kept at 3C/min to the lowest temperature.

68
VIII.(iv) Results:
Figure 1 shows that the thermal expansion of participant A and C follows same
pattern. The sintering shrinkage starts around 850C for both participants’ material. No
sintering shrinkage was observed in material from participant B. Figure 2 shows the
standard material at foundry B and a special mix that contained polymer fiber. The plots
of both materials from participant B overlap. This shows that the fugitive polymer fiber in
the specimen seems to have no role in thermal expansion of shell.

VIII.(v) References:
1. Snow, Jerry. D., and David Scott, “Prime Slurries for Investment Casting”,
Investment Casting Institute, 47th Annual Technical Meeting 1999, pg.6.1-6.21.

Shell Expansion

0.4

0.3

0.2
Percent Linear Change

0.1

0
0 200 400 600 800 1000 1200

-0.1

-0.2

-0.3
Temperaure C

Participant C Participant A

Figure 1: Comparison between Shell Expansions of Participant A with Participant C

69
 Shell Expansion

0.8
0.7
Percent Linear Change

0.6
0.5
0.4
0.3
0.2
0.1
0
0 200 400 600 800 1000 1200
Temperature
C
Containing Fugitive Polymer Fiber Contains No Polymer Fibers
WitSamples SampleSampleSamples

Figure 2: Comparison between Fired Shell Expansion of Materials that Contained Fugitive Polymer
Fiber and Materials that did not Contain Fugitive Polymer Fiber from Participant B

70
 IX. PERMEABILITY AND MICROSTRUCTURE ANALYSIS
OF INVESTMENT CASTING SHELLS
By
Phillip Jackson (Student Participant)
University of Missouri – Rolla
And
Dr. Von Richards Ph.D.
University of Missouri – Rolla

IX.(i) Background
The concept behind performing a permeability analysis on investment
casting shells is to determine the ability of the shell to relieve the stress accumulated by
the creation, movement, and trapping of gasses during the auto-clave process. It is
logical that the more permeable shell will provide a greater volume of gas to escape, and
thereby help to reduce the total amount of stress provided against the shell. A secondary
goal that developed in the course of experimentation was to test if the addition of a
polymer fiber to the ceramic slurry increases the permeability of the sample. Previous
work supported that this was true. In the course of work we produced results that
supported the idea that the fibers actually reduced permeability.
For this experiment we used a series of samples created with the test bar mold for fracture
toughness testing. The design of the experiment was to analyze the flow of air through a
core sample of the ceramic material. Several participant foundries produced the shells
used for the experiment. These foundries will be referred to as A, B, C. An analysis of
results was used to compare the participant foundries. Foundry B was used to develop
the work for the results of the secondary goal.

VIII.(ii) Results to Date

As a result of permeability testing we were better able to develop an
understanding as to which methods of shell making are better suited to prevent cracking
in the production process. Most foundries have a preferred selection of materials that are
used to create the ceramic shells used for casting. The experiment designed for
permeability produces a valid measure toward the choice of method for creating shells.

71
IX.(ii).a. Sample Selection and Preparation
Samples were collected from each of three participant foundries. Using the back
plate from the fired shells created for fracture toughness analysis, a 0.5” core samples
were drilled. The drilled samples were labeled in accordance with the foundry and shell.
These samples were processed with a device to measure the pressure drop across the
sample. (Figure 1). Using Darcy’s Law a specific permeability was determined for each
sample and the average was taken for comparison.

IX.(ii).b. Foundry Results

The permeability tests run for the participant foundries resulted in a collaboration
of results. Ten samples from foundry A were tested for permeability. Figure 2 shows
results for specific permeability from these tests ranged from 2.391 E -12 m2 to 4.812 E
–12 m2. The average result for the testing of samples containing polymers was 3.671 E
–12 m2.
Foundry B participated with two types of shells. Ones referred to as NP do not
contain the added polymer fibers in the slurry. Those referred to as A contained the
fibers. Sixteen samples from category A were tested for permeability. (Figure 2) Results
for specific permeability from these tests ranged from 8.053 E –13 m2 to 1.431 E –12
m2. The average result for the testing of samples containing polymers was 1.003 E –12
m2.
Fourteen samples from category NP were tested for permeability with results from
these tests ranged from 7.415 E –12 m2 to 1.387 E - 11 m2. The average result for the
NP test was 1.285 E –11 m2 viewable in figure 2.
Foundry C results were based off of fifteen samples that were tested for
permeability. These results for specific permeability ranged from 8.57 E -13 m2 to
1.67 E –12 m2. The average result for the testing of samples 1.243 E –12 m2 as show in
figure 2.
A comparison was done between each of the three participants. Students T test
was used to calculate the probability that samples from one source would fall into the
specific permeability range of another location. In all cases the T test stated that 99.99%
of the samples from any given foundry would fall in the range of another. A clear

72
distinction can be made of the non-fibrous samples from foundry B. This high
permeability rate suggests that the slurry material and preparation method are preferable
to improve the permeability of shells.

IX.(ii).c. Polymer Fiber versus Non – Fiber samples

A comparison was done between the previously mentioned samples A and NP
from participant foundry B. Figure 3 shows clearly a large difference in permeability
between the sample types. Figure 3 also shows that the measured permeability in all cases
was significantly higher in the non-fibrous samples. This suggests that the NP samples
would perform much greater in the dispersion of gasses. However, in congruence with
fracture toughness (Section III), the results do not suggest an overall increase in
performance of the ceramic shells.

IX. (ii).d. Microstructure

Some research was done with the microstructures of the permeability samples.
The set up included microscope pictures using a light microscope with camera
attachment. The concept was to use point and line count measurements to obtain a
volume to surface area ratios. The initial pictures were taken and the prime coat as the
assumption was made that this coat provided the limiting factor for the permeability of
the sample. Several problems were encountered with this process. Mostly due to the
inability to produce pictures with clear distinctions between mounting material and
ceramic matrix, the results were limited namely to qualitative analysis. No conclusive
evidence has yet been obtained from this micro-picture study. Figures 4 and 5 show
examples of the microstructures.

73
Figure. 1

74
 Foundry A vs. Foundries C and B
3.00E-11

2.50E-11

Foundry C
2.00E-11
Specific Permeability

NP
1.50E-11
A
1.00E-11
Foundry A
5.00E-12

0.00E+00
Figure 2

75
 Specific Permeability of NP samples and
A samples for Foundry B

3.00E-11
Specific Permeability (Units)

2.50E-11

2.00E-11
NP
1.50E-11
A
1.00E-11

5.00E-12

0.00E+00

Figure 3

76
 Figure 4: NP22 at 50X

77

NP22 at 50X
Mounting Material

Prime Coat

13C at 20X
Figure 5

78
 X. List of Papers Published

(1) Four Point Bend Testing to Characterize the Strength of Ceramic

Mold Shells, Von L. Richards and Ginger Connin, Investment
Casting Institute 49th Annual Meeting and Exposition paper
number 13, 2001.
(2) Thermal Expansion of Investment Casting Pattern Wax, Brian V.
Sunderland and Von L. Richards, Investment Casting Institute
49th Annual Meeting and Exposition paper number 2, 2001.
(3) Investment Casting Shell Cracking, Von [Link], Deepak
Rasquinha, and Brian Sunderland, Steel Founder’s Society
Technical and Operating Conference Proceedings November 2001.
(4) Four­point Bend Testing Method More Accurately Characterizes
Strength of Ceramic Mold Shells, Von L. Richards, and Ginger
Connin, INCAST, April 2002, pp. 22­25.
(5) Thermal Expansion of Investment Casting Pattern Wax, Von L.
Richards and Sony Mascreen, AFS Transactions, 2003, paper
number 03­040 accepted for publication.
(6) Fracture Toughness of Investment Casting Shells Using
Fractography, Von L. Richards and Sony Mascreen, Investment
Casting Institute 50th Technical Conference and Expo, paper
number 13, September 2002.
(7) Statistical Analysis of Three­point and Four­Point Bend Testing
for the Characterization of Ceramic Mold Shells, Von L.
Richards, Sony Mascreen, Phillip Jackson, and Thomas Hahn,
Investment Casting Institute 51st Technical Conference and Expo,
Paper Number 8, 2003.

79

13C at 20X