You are on page 1of 16

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/338467860

Towards Functionally Graded Sand Molds for Metal Casting: Engineering


Thermo-mechanical Properties Using 3D Sand Printing

Article in JOM: the journal of the Minerals, Metals & Materials Society · January 2020
DOI: 10.1007/s11837-019-03975-x

CITATIONS READS

15 808

3 authors, including:

Daniel Martinez Guha P. Manogharan

7 PUBLICATIONS 25 CITATIONS
Pennsylvania State University
94 PUBLICATIONS 2,456 CITATIONS
SEE PROFILE
SEE PROFILE

All content following this page was uploaded by Guha P. Manogharan on 09 January 2020.

The user has requested enhancement of the downloaded file.


JOM
https://doi.org/10.1007/s11837-019-03975-x
 2020 The Minerals, Metals & Materials Society

SOLID FREEFORM FABRICATION 2019

Towards Functionally Graded Sand Molds for Metal Casting:


Engineering Thermo-mechanical Properties Using 3D Sand
Printing

D. MARTINEZ ,1 C. BATE,1 and G. MANOGHARAN 1,2

1.—Department of Mechanical Engineering, Pennsylvania State University, University Park,


PA 16802, USA. 2.—e-mail: gum53@psu.edu

Growing applications of additive manufacturing (AM) have now been adopted


for metal castings via indirect hybrid AM, i.e., 3D sand printing (3DSP). A
comprehensive study on the thermo-physical properties of 3DSP molds based
on 3DSP processing conditions and their effects on materials and mechanical
properties of resulting aluminum castings is presented. The effect of furan
binder content (i.e., 1–3%) on as-printed 3DSP molds is evaluated to deter-
mine changes in dimensional accuracy, density (helium pycnometry), actual
binder content (loss on ignition) and mechanical strength (three-point bending
and tensile testing) and thermal properties (transient plane source technique
and casting runs). Pore network characterization via mercury intrusion
porosimetry does not reveal any significant differences in the morphology of
pore structures due to varying binder concentrations. However, bulk perme-
ability is reduced with increasing binder content. Thermal analysis shows a
reduction in conductivity and heat capacity due to binder degradation. Find-
ings from this study will enable the optimum selection of binder content for
functional grading of 3DSP molds/cores for adequate degassing and mechan-
ical strength, which is ready for adoption by both foundries and 3DSP service
providers.

deposition of an acid catalyst mix (e.g., furan) on a


INTRODUCTION AND BACKGROUND
layer of silica (e.g., foundry sand) based on layer-by-
As automotive, energy and aviation industries layer CAD information until the final mold/core is
continue to increase the demand for high-perfor- printed. Once molds have been completely 3D sand
mance aluminum production parts, the market for printed, they are de-powdered and assembled for
aluminum castings is expected to increase at a pouring like in traditional sand casting. Similar to
compounded annual growth rate (CARG) of 9.1% to most other AM processes, unbound material (i.e.,
USD 86.5 billion by the year 2025.1 As this demand foundry sand) is retrieved, sieved and reused in
for high-quality castings continues to grow, additive subsequent 3DSP builds.
manufacturing (AM) via 3D sand printing (3DSP, a Advancements in 3DSP have led to the develop-
binder jetting AM technique) has been proven to ment of ‘‘rapid casting technology’’,10–12 where the
offer a novel approach to tooling-less manufactur- process part of designing the final casting to pouring
ing2,3 of complex castings4,5 with increased mechan- can be compressed using digital design, simulation
ical properties6,7 and lower lead times.2,8 3DSP and printing. To achieve seamless integration and
offers both technologic and economic advantages to adoption of 3DSP in traditional foundry operations,
the aluminum casting industry compared with it is critical to understand the mechanical and
traditional sand casting. Another advantage of thermal characteristics of 3DSP molds and cores.
3DSP is the large array of alloys qualified for sand These critical characteristics have been the focus of
casting compared with direct metal AM processes several studies ranging from mechanical strength
such as laser-powder bed fusion and directed energy and dimensional accuracy10,13,14 to porosity and
deposition.9 The basic principle of 3DSP is selective permeability.13,15 However, there is a lack of
Martinez, Bate, and Manogharan

fundamental knowledge on the effects of the printing parameters remains largely unexplored.
thermo-mechanical properties of 3DSP molds based Binder saturation simultaneously influences multi-
on printing conditions and the resulting material ple mold properties such as dimensional accuracy
properties of resulting metal castings. This original and density. Therefore, a process map that estab-
research is intended to present a comprehensive lishes this complex interaction of mold properties is
examination of 3DSP mold characteristics, includ- required to obtain optimum mold and casting
ing the thermo-mechanical properties of both sand results. To the best of the authors’ knowledge, there
molds and final aluminum castings under the has been no prior reported study that establishes
varying influence of a critical 3DSP parame- this critical process map for varying 3DSP process
ter—binder saturation, i.e., binder content. conditions (e.g., binder saturation).
The final objective in this research is to leverage
the capabilities of layer-by-layer controlled deposi- MATERIALS AND METHODOLOGY
tion of the sand and binder in a 3DSP system to
3DSP samples were fabricated in a Viridis3D
simultaneously control the local distribution of
RAM sand printer (Enviosiontec, Woburn, MA)
mechanical strength (i.e., handling strength), ther-
using silica sand of grain fineness number (GFN)
mal diffusivity (i.e., solidification rate) and gas
65 round or sub-angular with a dry premixed acid
permeability (i.e., outgassing) within different sec-
catalyst and furan binder. Six different binder
tions of the sand molds. This study also aims to
deposition settings were selected to vary the binder
demonstrate the unique opportunities to engineer
content between 1% and 3%. In this study, the levels
both the mold and casting properties that are not
are labeled as B1.5, B2, B2.5, B3, B4 and B5, which
otherwise feasible via traditional molding methods.
refer to the number of printer-head activation
For instance, binder content can be functionally
pulses per second as set by the manufacturer’s
graded across the volume of molds and cores such
instructions. Specifications of the samples are pre-
that higher mechanical strength is achieved near
sented in Table I. Samples of the same binder
the mold walls and higher permeability via net-
content were printed in a single build box.
worked pores is incorporated for outgassing of mold
The large cylinders, pouring cups and 3 PB
gas away from the mold walls. This study presents a
specimens were used to measure dimensional toler-
systematic framework to study the influence of
ance using a digital caliper. The diameter was
binder saturation on thermal properties and poros-
measured multiple times (n = 4) at the midline
ity. In addition, it aims to provide insights into the
section in two orthogonal directions, and results
correlation between the sand-binder content rela-
were averaged to report the overall diameter along
tionship to achieve adequate mold strength, print-
with cylinder height. Loss on ignition (LOI) testing
ing accuracy and casting characteristics. To the best
was conducted as per AFS Standard 5100-00-S24
of the authors’ knowledge, this study is the first
using the 20Ø 9 20 mm samples (12 g) for consis-
reported comprehensive study that reports on the
tency across all binder conditions. Humidity content
unexplored process map to achieve functionally
was determined by recording the mass of the
graded 3DSP sand molds to enable production of
samples before and after heating in an oven for
high-performance casting.
150 min at 107C. Samples were weighed at 0 min,
In traditional sand molds, mold composition has
90 min and 150 min. Lack of change in sample
been modified to obtain different characteristics in
weight between measurements showed that water
the final metal casting.16–18 The capabilities of
content was completely evaporated. Subsequently,
3DSP have been explored by redesigning the mold
burn-out of binder content was performed by heat-
to accommodate novel insulation capabilities.19–21
ing the samples at 986C. Furan resin degrades at
Hackney and Wooldridge22 focused on characteriz-
180C22 and ignites at 595C.25 Samples were
ing an ExOne S-Max system for changes in dimen-
weighed at 60 min, 120 min and 180 min. Lack of
sional accuracy, density, and tensile, compressive
change in sample weight between measurements
and impact strength using manufacturer-recom-
proved that all binders had been burned out from
mended settings. Mitra et al.13 analyzed the effect
the samples. The amount of binder in the samples
of aging, i.e., mold curing temperature and time, on
was quantified based on the difference in weight of
flexural strength (three-point bending) and perme-
the samples after 180 min at 986C and 190 min at
ability in a similar study using the ExOne S-Print
107C.
machine. Mechanical strength and permeability
To characterize pore network’s properties, mer-
were observed to remain approximately constant
cury intrusion porosimetry (MIP) was employed
over time at curing temperatures £ 100C. More
since MIP is a well-established technique used in
recently, they reported that binder saturation has a
the petroleum industry,26–28 construction engineer-
drastic effect on flexural strength, and no significant
ing29–31 and pharmacy32–34 to identify pore sizes
change is observed in permeability when curing
ranging from 3 nm to 360 nm diameter.35
occurs at room temperature.23 However, the possi-
A common experimental approach to determine
bility of altering intrinsic mold properties, such as
thermal properties in molding materials is the one-
permeability, heat capacity and conductivity, along
dimensional steady heat flux setup.36,37 Other
with varying mechanical properties by modifying
Towards Functionally Graded Sand Molds for Metal Casting: Engineering Thermo-mechanical
Properties Using 3D Sand Printing

Table I. 3D sand printed samples: design of experiments


Number of tests
Sample Dimensions (mm) Test per condition (n)

Pour cup 130 9 130 9 105 Experimental solidification time Heat diffusivity 1
Large cylinder 100Ø 9 50 Transient heat plane source 1
MIP cylinder 7Ø 9 20 Mercury intrusion porosimetry (MIP) 1
LOI cylinder 20Ø 9 20 LOI 2
Helium pycnometry 3
Permeability 50.8Ø 9 50.8 AFS permeability 9
Flexural sample 15 9 15 9 150 3-point bending 8
Tensile coupon 44 9 76 9 25 AFS tensile 9

techniques for determining thermal properties Subsequently, these results can be employed in
include the laser flash technique and inverse-heat Chvorinov’s law to determine the mold constant and
method. The former requires a smooth and polished thermal properties of 3DSP sand molds using the
sample such that laser reflection can be measured. following equation:
However, it is not a feasible technique for rough  2
surfaces of sand molds.38 The latter inverse-heat Ts ¼ B
V
ð2Þ
method involves reproducing measured tempera- A
tures in a casting experiment via numerical model-
ing by back-calculating unknown thermal This analytical expression relates the solidifica-
variables.39,40 tion time Ts (s) of a casting with total volume V (m3),
Using the one-dimensional analytical representa- surface area (m2) and mold constant B (s/m2).45–47
tion of the system, a diffusion equation can be The mold constant is related to the thermal prop-
developed with known boundary temperatures,41 erties of the alloy being poured; hence, to isolate the
i.e., recorded temperatures at known locations impact of mold properties, a coefficient of heat
within the mold using thermocouples. The diffusion accumulation bf can be defined as a function of the
equation can be expressed based on an error mold density qm (kg/m3), specific heat cm (J/m3 K)
function as follows: and thermal conductivity km (W/mK) as shown in
  Eq. 3:
Tx;t  Ts x pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
¼ erf pffiffiffiffiffiffiffiffiffiffiffi ð1Þ bf ¼ qm cm km ð3Þ
Tair  Ts 2 hd  t
 2
where thermal diffusivity hd ms of the mold mate- Density measurements were determined using a
rial is calculated knowing the temperature Tx;t ðKÞ Micrometrics AccuPyc II 1340 helium pycnometer.
at a given point x (m) in the mold. Ts and Tair The technique was selected because of the relative
represent the temperatures (K) at the inner surface ease of execution, accurate results based on
of the mold, which is in contact with metal and reported studies and recent application in binder
ambient temperature, respectively. jetting parts to determine skeletal density.48–50 Gas
Another method to measure thermal properties is pycnometry involves the measurement of the gas
the transient plane source (TPS) or ‘‘hot disk’’ volume (e.g., helium, which was used in the study)
technique. Thermal properties of solid material required to fill a vacuum chamber with the test
can be back-calculated by measuring the increase sample. With the known gas volume for a given
in temperature of a double-spiral Wheatstone pressure and temperature, volume of the testing
bridge. The temperature of a heated sensor through chamber and weight of the sample, skeletal density
the electric current42 is used in the conduction can be determined. The accuracy of this technique
equation of a hot disk within an infinitely large depends on the assumption that most voids within
sample to back-calculate the thermal conductivity the sample are interconnected between each other
and diffusivity of the sample. The tested material and with the sample surface.
must be large enough and the temperature sam-
pling time should be short enough to prevent the Mechanical Properties of 3DSP Molds
boundary conditions from influencing the experi- 3DSP samples were tested for mechanical
ment.43 It is also important to have uniform flat strength via tensile and flexural experiments. Ten-
surface contact between the sensor and sample. TPS sile specimens were tested as per AFS 3301-08-S51
has been demonstrated to have an effective tech- in a Simpson Analytics universal sand strength
nique to measure a wide range of thermal conduc- testing machine using the standard 25.4-mm tensile
tivities with an accuracy of ± 5%.44 specimen design (see Table I). These samples were
Martinez, Bate, and Manogharan

 
printed in the horizontal position with the mechan- @l P r2 þ 4er
ical loading axis along the recoating direction. Nine ¼ ð6Þ
@t 8gL
samples per binder condition were tested to mini-
mize deviation from true mean strength.52 where r is the radius of the cylinder (m), e is the
Flexural tests are generally used to determine friction coefficient between mercury and sand, and g

mold and core properties in both the AFS53 and m2
is the kinematic viscosity s of mercury. Subse-
German54 standards. 3DSP samples were tested for
ultimate flexural strength using three-point bend- quently, three unique quantitative measures of the
ing tests as per the ASTM E29055 standard on porous media can be calculated: (1) pore size
samples of 15 mm 9 15 mm 9 100 mm dimensions distribution, (2) pore surface area per gram of
(see Table I) for eight samples per bending condi- material and (3) average and median pore diameter.
tion. Four samples were tested with the applied
force normal to print direction and four samples Thermal Properties of 3DSP Molds
with the applied force along the print direction. An Cooling curves of 99.9% pure aluminum were
MTS WTest 100 machine with a 100-kN-load cell determined by pouring 3DSP cooling cups for every
was used to determine the ultimate flexural binder condition (n = 5) based on solidification time
strength based on the length between supports L and thermal diffusivity measurements. The latter
(mm), width w (mm) and thickness t (mm) of the was back-calculated from the one-dimensional
specimen to determine the stress curve rflex (MPa) steady-state solution to thermal diffusion equation.
based on recorded force F (kN) from Eq. 4: The highest binder ratio batch (i.e., B5) failed to
3FL consistently print the pouring cups; hence, the
rflex ¼ ð4Þ pouring experiment could not be performed at this
2wt2
binder condition. Pure aluminum was selected to
The span length, loading rate and data acquisition achieve a definite freezing point in the cooling curve
rate were set to 60 mm, 2.5 mm/min and 10 Hz, using an open-top pour cup (25-mm-thick walls to
respectively. hold 80 mm3 of pure aluminum of about 1.4 kg. as
shown in Fig. 1).
Permeability and Mercury Porosimetry Four TL1815 k-type thermocouples rated for
of 3DSP Molds maximum temperature of 800C with an accuracy
of ± 0.1C were used to measure the temperature at
Permeability tests were performed in accordance
four unique locations within the mold. All probes
to AFS 5224-13-S standards.53 Nine cylindrical
were placed at identical distance of 40 mm from the
specimens per binder condition with dimensions of
top of the cup as shown in Fig. 1a: (A) middle of the
Ø50.8 mm 9 50.8 mm were tested with the direc-
casting volume, (B) 5 mm from the mold wall in
tion of flow normal to the print direction. The AFS
contact with the metal, (C) 5 mm from the mold wall
permeability number was obtained using a Simpson
in contact with ambient air and (D) 5 mm from the
Analytics Digital Absolute Permmeter using Eq. 5,
ambient air within the mold wall (see Fig. 1b).
where Q is the volume of air of the perimeter
Temperature was recorded using a data acquisition
pneumatic cylinder (m3); h and A are the height (m)
unit from National Instruments cDAQ-9171 and NI-
and cross-section (m2) of the 3DSP sample, respec-
9212 through a LabView script at a sampling rate of
tively; P is the differential pressure (Pa); and t is the
10 Hz. The melt temperature prior to pouring was
time for the complete flow of air across the cylinder
measured using an OMEGA OS204 infrared ther-
(s).
mometer and was found to be 765 ± 4C in the
Qh ladle. Data collection was prior to melt pour and
P¼ ð5Þ continued through the liquid cooling phase, liquid–
APt
solid transition and solid cooling to room
MIP tests were performed on one Ø7 mm 9 20 temperature.
mm sample for each binder condition. This tech- Results obtained via analytical one-dimensional
nique involves measurement of volumes of liquid solution to the conduction equation are compared
mercury that intrude into the sample at precisely with the transient plane source (TPS) technique,
controlled pressures.35 The tests were performed in which was performed in a Kapton 8563 F1 hot-disk
a Micromeritics AutoPore V system, and MicroAc- sensor (/19.6 mm) with an output power of 90 mW
tive Software (Norcross, GA) was used for data placed between two 3DSP samples. The samples
processing. By applying the Washburn model,56 the were measured every 80 s while heating the sample
pore structure can be analyzed as a conglomeration to 50C, 100C and 250C to observe variations in
of cylinders of length L (m), which is progressively thermal properties (see Fig. 8). Five measurements
filled with increasing driving pressure P (Pa) using were performed per binder condition at each tem-
the following rule: perature level. The samples were heated at a rate of
Towards Functionally Graded Sand Molds for Metal Casting: Engineering Thermo-mechanical
Properties Using 3D Sand Printing

Fig. 1. (a) 3DSP pour cup dimensions in mm and (b) experimental setup.

0.42C per minute with a rest time of 60 min at each Photonics (Hangzhou, Zheijiang). Postprocessing
temperature level to reach steady state. This test and analysis were performed using open-source
was carried out for four binder levels, B1.5, B2, B2.5 ImageJ.
and B5. Due to setup error, results for the B2.5 level
were only obtained during the heating phase. RESULTS AND DISCUSSION
Loss on Ignition (LOI) of 3DSP Molds
Mechanical Testing of Aluminum Castings
Loss on ignition (LOI) testing reveals significant
Flexural bending tests were performed as per
deviation from expected values provided by the
ASTM E290 standard55 on six aluminum samples
equipment manufacturer, specifically at lower bin-
per binder condition. The samples of dimensions
der content levels (see Fig. 2a). Intermediate pulse
20 mm 9 70 mm 9 4.5 mm were machined from
levels result in binder contents that are closer to
the center of the castings to eliminate edge effects
measured values, with an error near 10%. Binder
(see Fig. 9). The span length, loading rate and data
conditions of B2.5, B3 and B4 show comparatively
acquisition rate were set to 60 mm, 2.5 mm/min and
small changes in binder content, which were also
10 Hz, respectively.
observed in dimensional tolerances (see Fig. 3) and
mechanical testing (see Fig. 4). Actual binder con-
Microstructure of Aluminum Castings
tent ranges between 1.49% and 2.74%. Water
Grain microstructure was evaluated from content remains < 0.2% in all binder conditions
12 mm 9 12 mm 9 6 mm specimens machined (see Fig. 2b). The experimental values represent the
from the bottom corner of the castings to include average of two LOI tests, except for B5 level, for
edge effects of solidification sand molds, which could which only one result was obtained because of
be correlated to thermal properties of the 3DSP measurement error in one of the trials. The values
molds (see Fig. 9). The surface nearest to the bottom of binder content as expected by the manufacturer
surface was ground and polished down to 1 lm are supposed to follow a nearly linear relation (R2
using a diamond polishing suspension. To observe value of 0.9696) with the number of pulses applied
the grain structure, samples were etched for 6– to the binder head.
8 min using Keller’s reagent of the following chem- Using helium pycnometry (n = 3), a nearly linear
ical composition: 95 mL H2O (water), 2.5 mL HNO3 reduction in density with an increase in binder
(nitric acid), 1.5 mL HCl (hydrochloric acid) and concentration is observed, following the expression:
1 mL HF (hydrofluoric acid).57 Images were cap-
tured with an Olympus SZH10 Research Stereo q ¼ 3:4222x þ 2:6659 ð7Þ
microscope with an Omax A3530U 3.2MP digital where x is the concentration of the binder in the
camera using Toupview software from ToupTek sample (wt.%) with an R2 value of 0.9931.This
Martinez, Bate, and Manogharan

Fig. 2. (a) Binder and (b) water content within the samples tested via AFS loss on ignition test and (c) density measurements obtained via helium
pycnometry and calculated using a linear relation between silica sand and furan binder.

Fig. 3. Average deviation from the nominal dimensions (in mm) along and across the printing direction for (a) flexural samples, (b) pouring cups
and (c) cylinders. The print orientation is presented for every geometry condition.
Towards Functionally Graded Sand Molds for Metal Casting: Engineering Thermo-mechanical
Properties Using 3D Sand Printing

Fig. 4. (a) Ultimate tensile strength and sample weight and (b) ultimate flexural strength in two orientations with respect to the print direction.

linear relationship suggests that the arrangement Dimensional Accuracy of 3DSP Molds
of silica particles in each sample is not altered by
The flexural bending samples, pouring cups and
sand-binder interactions. The densities of loose sand
large cylinders are measured and analyzed for
and liquid furan binder were determined as 2.651 g/
dimensional accuracy after 3DSP. These samples
cm3 and 1.142 g/cm3, respectively. A linear equation
were measured several days after printing; hence, it
to predict density values as a function of binder
is assumed that shrinkage due to binder cross-
condition (wt.%) can be derived:
linking will have reached steady-state conditions. It
qprintedsmaple ¼ xqsand þ yqBinder ð8Þ must be noted that the rate of shrinkage during the
initial few hours after printing (when it is challeng-
Bias towards overestimating the actual density ing to handle the molds) has been demonstrated to
with an increase in binder content is observed when change based on different binder and catalyst
using this linear equation (see Fig. 2c). This could concentrations.59 However, this study is not rele-
be attributed to the accumulation of acid catalysts vant to practice because of the need for higher green
with relatively lower weight,13 which is not consid- strength to both handle and transport 3DSP molds.
ered in this study. Since binder concentration An increase in binder content reduces printing
> 3 wt.% results in inconsistent print quality and accuracy, consistent with reported studies,60,61 and
higher costs58 to print molds, this analytical approx- is attributed to binder bleeding, which is the
imation of mold density is considered accurate for phenomenon of capillary flow of binder beyond the
normal printing conditions with acceptable accuracy deposited zone. The flow of binder out of the
(% error < 1.5%). immediate droplet area will agglomerate additional
Martinez, Bate, and Manogharan

sand grains, which results in printed parts with printing plan. Ultimate flexural strengths are lower
higher than nominal dimensions. Another study than the values reported in the literature for other
reports an opposite effect on binder jetting of finer 3DSP systems using the furan binder and employ-
materials.62 ing the ExOne13,60 and Voxeljet14 systems. Tensile
The deviation from nominal dimension is smaller strength is also shown to be lower than flexural
across the print direction in all three samples with strength, coinciding with other reported values for
the exception of the distance between the interior 3DSP systems.14 It must be noted that exact
walls of the pouring cup where maximum deviation comparison with other 3DSP systems is difficult
from the nominal is at the highest binder concen- since in some cases the layer thickness and sand
tration level (> 3 mm). However, this is explained grain size are not reported. Table II presents a
by the fact that the measurement is composed of two summary of the reported values for ultimate flexu-
different printed sections (i.e., binder bleeding in ral strengths in cold box, heat-cured and 3DSP
the two walls of the cup). However, deviation of wall cases. The 3DSP system reported in this study has
thickness is similar to trends in other samples. At lower strength than that in previously reported
the lowest binder level, the deviation is around studies.66
0.5 mm across the print direction. At the interme-
diate levels (i.e., B2, B2.5, B3 and B4) deviation is Permeability and Mercury Porosimetry
between 0.5 and 1 mm, which follows a similar
MIP results are expressed as pore volume distri-
trend as the reported LOI results.
bution per diameter. Translating these results to
Similarly, an increase in dimensional deviation is
pore count per diameter is difficult because of the
observed with an increase in binder concentration
assumption about pore geometry.69 B2 and B4
along the print direction. However, the deviation is
samples show pore sizes of larger median (67 lm
lower across all samples compared with measure-
and 64 lm, respectively) and mean (62 lm and
ments along the print direction, which can be
59 lm, respectively) based on the Washburn model
attributed to the compressive effect that the recoat-
(see Fig. 5). These samples also display broader
ing rake has been demonstrated to exert on the
distributions in both cumulative area and incre-
printing bed.60 This compression could be accentu-
mental intrusion. Calculations of mean pore size
ating the binder bleeding in the direction perpen-
based on the assumption of straight cylindrical
dicular to the printing plane, and further research
pores of circular cross-section range between 57 lm
is required to verify this hypothesis. In the case of
and 62 lm. No distinctive correlation can be drawn
deviation in the depth of the pouring cup (see
between pore size and binder content. Samples used
Fig. 3b), an opposite trend is observed, which could
for the MIP test were destroyed at final pressure
be attributed to the retrieval and cleaning of loose
because of permanent deformation.
sands from the build box using compressed air. This
Permeability measurements validate the hypoth-
final cleaning step is believed to have affected the
esis that increasing difficulty in degassing will occur
printed depth of the cups.
during casting with an increase in binder concen-
tration (see Fig. 6a). This could be attributed to the
Mold Mechanical Properties
reduction in pore space and/or higher tortuosity of
Ultimate tensile strength 3DSP samples have a the pore network. MIP results suggest no significant
non-linear relation to the binder content. Optimum difference in the pore network but do not provide
strength is obtained at the B2.5 level across the build detailed information on the geometric properties of
orientation (see Fig. 4). At the highest binder the pores to understand the reduction in perme-
content level, tensile strength increases again. The ability. Future tests will include micro-computed
mean strength regardless of binder content for this tomography to analyze the exact geometric charac-
system is lower than the reported values for the teristics of the pore networks.
tensile strength of chemically bonded sand using Figure 6b shows a very nearly linear relation
furan binder in a ZCorp machine.63 This could be between mold density and permeability results,
attributed to the coarser grain size distribution used which is not seen as clearly when analyzing it as a
in this study, which has been identified as an function of binder concentration. The relation
important factor influencing tensile strength.64,65 between permeability and density is direct, which
As expected, the fracture was brittle in all cases. might seem counterintuitive but is due to the fact
Ultimate flexural strength shows a similar trend that the furan binder is less dense than the silica
to that of ultimate tensile strength, with increasing sand; therefore, as binder content increases, the
strength as the binder concentration reaches the B3 total density decreases.
level, a drop in B4 and a final increase with the
maximum binder concentration level. Flexural Mold Thermal Properties
strengths along and across build orientations show
Mold and casting temperatures for each 3DSP
very similar results; however, at intermediate
sample were measured during and after pouring.
binder levels, a lower maximum strength can be
Figure 7a shows representative temperature curves
observed when the force is applied parallel to the
at the four locations in the mold during pouring-
Towards Functionally Graded Sand Molds for Metal Casting: Engineering Thermo-mechanical
Properties Using 3D Sand Printing

Table II. Comparison of ultimate flexural strength for different binder types
UFS
System Sand type Binder (MPa) Ref.

Cold box (phenolic urethane) Round silica GFN 55.56 1.25 wt.% phenolic urethane 1.8 67
Cold box (phenolic) H32 silica 1.2–1.9 wt.% phenolic resin 0.38a 68
Warm box (furanic) H32 silica Between 1.2 wt.% and 1.9 wt.% furan resin 0.72a 68
Warm box (urea-furanic) H32 silica Urea-furanic resin 0.44a 68
Voxeljet VX 1000 – Furan resin 1 wt.% 1.3 14
ExOne SPrint Round silica, GFN 97 Furan resin 1–3 wt.% 3.2a 13
Viridis 3D RAM Round silica, GFN 65 Furan resin 1–3 wt.% 0.89 This study
a
These values were shown to change with curing time; maximum flexural strength at room temperature curing is reported in this table.

Fig. 5. (a) Pore size, (b) pore area and (c) incremental volume obtained via mercury intrusion porosimetry (MIP).

filling-solidification. The cooling curve for location A transfer coefficient is associated with binder decom-
corresponds to the thermocouple at the center of the position and water vapor transport across the
casting, which shows the characteristic horizontal mold.36,37 Since temperatures remain < 500C at
line between 80 s and 730 s representing the tran- the locations measured within the wall, binder
sition from liquid to solid in a pure metal.70 All decomposition can be assumed to occur only within
binder conditions showed a similar solidification the 5 mm of the wall nearest to the solidifying
time. The temperature in locations C and D was aluminum.
used to calculate the thermal diffusivity across the TPS results show a clearer effect of increased
thickness of the mold at 5 mm and 20 mm from the thermal diffusivity and conductivity at lower tem-
mold’s internal wall, using Eq. 1. The average peratures because of the increase in binder content
thermal diffusivity calculated at these two points (see Fig. 8), which increases the heat capacity of the
during solidification is shown in Fig. 7b. The change mold. All binder levels show a drastic reduction in
in heat diffusivity as well as corresponding heat thermal conductivity during the cooling cycle. The
Martinez, Bate, and Manogharan

Fig. 6. (a) AFS permeability and weight of standard AFS samples (n = 9) as a function of binder concentration and (b) mold density.

Fig. 7. (a) A representative cooling curve for the four thermocouple positions; (b) thermal diffusivity versus temperature is calculated using the
average temperature measured at locations C and D.

most pronounced effect is observed at the highest which is a pairwise t test used to determine
binder level B5 with a reduction of 36%, which is differences between pairs of samples compared
attributed to both binder degradation at 180C and against each other.71 The lowest binder levels, B1.5
the evaporation effect of humidity within the and B2, show the highest mean strength.
samples.25
Similar heating cycles are reported in the litera- Grain Structure of Aluminum Castings
ture to simulate mold aging during extended stor-
Grain structures observed in cast aluminum
age of molds.13 These results suggest that heat
samples (see Fig. 10) are counted via visual inspec-
capacity and conductivity will be significantly
tion, and percentage areas of casting defects (in-
reduced because of mold aging, and additional
clusions and pores) are determined using ImageJ.
experimentation is required to confirm this effect.
The larger differences in grain number and size
Thermal diffusivity is found to be in the same order
across each face reveal that the analyzed surface
of magnitude based on both casting and TPS runs.
area is insufficient to represent the castings as a
whole; therefore, no conclusions on microstructural
Mechanical Testing of Aluminum Castings
differences are drawn. Grain morphology is
Mechanical testing of castings was performed via expected to be columnar near the mold surface72;
three-point bending tests on six samples at each however, only cross-sectional images are visible.
binder level. Maximum flexural stress is reported The types of defects observed are dispersed micro-
since none of the samples suffered fracture at peak porosity, gas pores and possible bifilm defects. This
stress (see Fig. 9). An ANOVA test was performed, last defect is identifiable as irregular pores with
showing statistically significant differences between visible cracks with entrapped air (see Fig. 10c).73
the binder levels. Subsequently, a Fisher least The defects are common to samples across binder
significant difference test (LSD) was performed, content levels.
Towards Functionally Graded Sand Molds for Metal Casting: Engineering Thermo-mechanical
Properties Using 3D Sand Printing

Fig. 8. (a) Experimental setup for the transient plane source (TPS) technique, (b) thermal conductivity, (c) specific heat and (d) thermal diffusivity
for varying binder contents.

Fig. 9. (a) Mean maximum flexural stress; (b) Fisher pairwise comparison of groups with significantly different means with 95% confidence; (c)
location of extracted samples within the casting and three-point bending.

DISCUSSION analytical calculation of heat accumulation coeffi-


cient bf (see Fig. 11) based on Eq. 3 for 3DSP
Analysis of density measurements obtained via
printed molds of varying binder content. Thermal
helium pycnomentry and thermal properties data obtained at 50C are used in this calculation
obtained using the TPS technique enables
Martinez, Bate, and Manogharan

Fig. 10. (a) Representative micrograph of etched casting, (b) grain morphology and (c) typical casting defects including microporosity and
solidification shrinkage.

Fig. 11. Heat accumulation coefficient (bf) for B1.5, B2, B2.5 and B5 3DSP samples for different binder contents.

since density measurements are obtained at room work will address the temperature effects on mold
temperature conditions. The value of bf at higher density and subsequently on the heat accumulation
temperatures could be lower because of binder coefficient.
degradation and loss of humidity, which could lead This study aimed to quantify the thermal,
to reduced density. mechanical and permeability properties of 3D
The heat accumulation coefficient determined printed sand molds due to varying binder contents
from this study lies within the range of typical in 3DSP. This represents an initial stage in a novel
rammed sand molds compared with the comprehen- approach toward creating functionally graded sand
sive report by Jelinek and Elbel.45 It should be molds by applying Design for AM (DfAM)
noted that our present study is the first to report the approaches. Such a framework to leverage 3DSP
wider chilling effect that can be achieved by mod- through this novel approach is proposed in Fig. 12.
ifying binder deposition during printing. Future Findings on thermo-mechanical properties of 3DSP
Towards Functionally Graded Sand Molds for Metal Casting: Engineering Thermo-mechanical
Properties Using 3D Sand Printing

Fig. 12. Novel framework for the functionally graded 3DSP mold to control thermal, mechanical and degassing properties of sand molds.

properties from this study and future efforts serve highest binder concentration, 2.76% (i.e., B5), the
as inputs for the design-simulation–optimization maximum tensile strength was observed. Flexu-
3DSP casting framework. ral strength followed a similar trend.
This unique approach will enable production of  Permeability decreased with higher binder con-
3DSP molds with optimized mechanical strength, tent. The pore mean size is experimentally
point-by-point control of permeability across the determined to range between 56 lm and 62 lm,
mold and venting designs. In addition, it can be calculated using mercury intrusion porosimetry.
extended to accurately simulate filling and solidifi-  Density decreased with higher binder content,
cation, which will lead to superior design of sand ± 1% from the median value of 2.6 g/cm3 based
molds (e.g., ingate location and number, runner on helium pycnometry. This suggests that
profiles, riser volumes, etc.) and/or create direc- reordering of sand grains does not occur as
tional solidification within the casting. binder content changes: therefore, density can be
predicted accurately by knowing the binder
CONCLUSION content in the mold.
 Thermal conductivity, diffusivity and heat
In summary, this study reports on comprehensive
capacity increase with increasing binder content
experimental testing of 3DSP samples at six differ-
at room temperature. Binder degradation affects
ent binder concentrations ranging from 1.49% to
thermal properties, particularly, at higher bin-
2.76%, which rendered the following results:
der levels.
 Variation in the mechanical strength of 3DSP  A solidification analysis of 99.9% pure aluminum
molds ranges between 25% and 50% based on showed that binder content had no significant
actual binder content compared with the ex- effect on casting solidification time. Grain mor-
pected values presented by the manufacturer. phology was shown to be similar in all cases. A
The actual binder content was lower than the slight increase in the flexural strength of the
expected value because of the higher concentra- aluminum casting was observed at lower binder
tion at the lower three pulse values and lower levels.
concentration at the higher three pulse values.
 Increasing binder content had a negative effect ACKNOWLEDGEMENTS
on part tolerance.
The authors thank Dr. Robert Voigt for his expert
 Tensile strength increased with binder content
advice and Chris Anderson and Travis Richner for
up to a binder content of 1.95 wt.%. At the
Martinez, Bate, and Manogharan

helping with casting trials. This work was partially Emphasis on Reduction and Presentation of Experimental
funded by a PA Manufacturing Fellows Grant from Data (Micromeritics Inst Corp, 2001), https://www.microme
ritics.com/pdf/app_articles/mercury_paper.pdf. Accessed 21
the PA Department of Community and Economic May 2019.
Development. 36. P.K. Krajewski, J.S. Suchy, G. Piwowarski, and W.K. Kra-
jewski, Arch. Foundry Eng. 15, 47 (2015).
REFERENCES 37. P.K. Krajewski, G. Piwowarski, and J.S. Suchy, Arch.
1. Grand View Research, Aluminum Casting Market Worth Foundry Eng. 14, 67 (2014).
$86.5 Billion by 2025 (Grand View Research, 2019). https:// 38. Y. Xu and D. Chung, Cem. Concr. Res. 30, 59 (2000).
www.grandviewresearch.com/press-release/global-alumi 39. T.J. Williams, MSc Thesis, University of Iowa, 2016.
num-casting-market. Accessed 21 May 2019. 40. T.J. Williams, R.A. Hardin, and C. Beckermann, in Pro-
2. E.S. Almaghariz, B.P. Conner, L. Lenner, R. Gullapalli, G.P. ceedings of the 68th SFSA Technical and Opererating Con-
Manogharan, B. Lamoncha, and M. Fang, Int. J. Metalcast. ference, Paper No. 4.6 (2014).
10, 240 (2016). 41. M.T. Sun and C.H. Chang, J. Heat Transf. 129, 1119 (2007).
3. D. Strong, I. Sirichakwal, G.P. Manogharan, and T. Wake- 42. R. Coquard, E. Coment, G. Flasquin, and D. Baillis, Int. J.
field, Rapid Prototyp. J. 23, 577 (2017). Therm. Sci. 65, 242 (2013).
4. E.S. Almaghariz, MSc Thesis, Youngstown State University 43. S.E. Gustafsson, Rev. Sci. Instrum. 62, 797 (1991).
(2015). 44. T. Log and S.E. Gustafsson, Fire Mater. 19, 43 (1995).
5. J. Wang, S.R. Sama, and G. Manogharan, Int. J Metalcast. 45. P. Jelinek and T. Elbel, Arch. Foundry Eng. 10, 77 (2010).
13, 2 (2019). 46. N. Chvorinov, Proc. Inst. Br. Foundrym 32, 229 (1938).
6. S.R. Sama, J. Wang, and G. Manogharan, J. Manuf. Process. 47. P. Beeley, Foundry Technology, 2nd ed. (Oxford: Butter-
34, 765 (2018). worth-Heinemann, 2001).
7. S.R. Sama, T. Badamo, P. Lynch, and G. Manogharan, Ad- 48. S. Atre, K. Kate, and J. Porter, in European Congress and
dit. Manuf. 25, 563 (2019). Exhibition on Powder Metallurgy. European PM Conference
8. C. Weller, R. Kleer, and F.T. Piller, Int. J. Prod. Econ. 164, Proceedings (2016), p. 1.
43 (2015). 49. D. Hong, D.T. Chou, O.I. Velikokhatnyi, A. Roy, B. Lee, I.
9. D. Snelling, Q. Li, N. Meisel, C.B. Williams, R.C. Batra, and Swink, I. Issaev, H.A. Kuhn, and P.N. Kumta, Acta Bio-
A.P. Druschitz, Adv. Eng. Mater. 17, 923 (2015). mater. 45, 375 (2016).
10. P. Hackney and R. Wooldridge, Procedia Manuf. 11, 457 50. B. Yun and C. Williams, Rapid Prototyp. J. 21, 177 (2015).
(2017). 51. AFS, Procedure 3301-08-S, Mold Core Test Handbook, 3rd
11. B. Rooks, Assem. Autom. 22, 40 (2002). ed. (Shaumburg-IL: AFS, 2001).
12. A. Bernard, J.C. Delplace, N. Perry, and S. Gabriel, Rapid 52. B. Primkulov, J. Chalaturnyk, R. Chalaturnyk, and G.
Prototyp. J. 9, 327 (2012). Zambrano Narvaez, 3D Print Addit. Manuf. 4, 149 (2017).
13. S. Mitra, A. Rodrı́guez de Castro, and M. El Mansori, Int. J. 53. AFS, Procedure 5224-13-S, Mold Core Test Handbook, 3rd
Adv. Manuf. Technol. 97, 1241 (2018). ed. (Shaumburg-IL: AFS, 2001).
14. K. Nyembwe, M. Mashila, P.J.M. van Tonder, D.J. de Beer, 54. VDG, Biegefestigkeit von warmhärtenden, kunstharzge-
and E. Gonya, S. Afr. J. Ind. Eng. 27, 136 (2016). bundenen feuchten Formstoffen (Düsseldorf, 1999).
15. T. Sivarupan, M. El Mansori, K. Daly, M.N. Mavrogordato, 55. ASTM E290-09, Standard Test Methods for Bend Testing of
and F. Pierron, Rapid Prototyp. J. 25, 404 (2019). Material for Ductility (West Choshohocken: ASTM E290-09,
16. A.V. Adedayo, J. Braz. Soc. Mech. Sci. Eng. 32, 171 (2010). 2009).
17. A.V. Adedayo, B. Aremo, and P. Si, J. Miner. Mater. Char- 56. E.W. Washburn, Phys. Rev. 17, 273 (1921).
act. Eng. 10, 387 (2011). 57. G.F. Vander Voort, Metallography—Principles and Practice
18. C. Saikaew and S. Wiengwiset, Appl. Clay Sci. 67–68, 26 (Materials Park: ASM International, 1999).
(2012). 58. P. Sirivimonpa and N. Osothsilp, Key Eng. Mater. 765, 255
19. C.Y. Deng, J.W. Kang, H.L. Shangguan, T. Huang, X.P. (2018).
Zhang, Y.Y. Hu, and T. Huang, China Foundry 15, 37 59. H. Khandelwal and B. Ravi, Mater. Manuf. Process. 30, 1465
(2018). (2015).
20. C. Deng, J. Kang, H. Shangguan, Y. Hu, T. Huan, and Z. 60. N. Coniglio, T. Sivarupan, and M. El Mansori, Int. J. Adv.
Liu, J. Mater. Process. Technol. 255, 516 (2018). Manuf. Technol. 94, 2175 (2018).
21. H. Shangguan, J. Kang, C. Deng, J. Yi, Y. Hu, and T. 61. K.J. Hodder and R.J. Chalaturnyk, Addit. Manuf. 28, 649
Huang, Int. J. Adv. Manuf. Technol. 96, 2175 (2018). (2019).
22. P.M. Hackney and R. Wooldridge, Rapid Prototyp. J. 23, 7 62. S. Tamas and B. Dermot, in Solid Free Fabrication Sym-
(2017). posium Proceedings (2007), p. 470.
23. S. Mitra, A. Rodrı́guez de Castro, and M. El Mansori, J. 63. D. Snelling, C.B. Williams, and A.P. Druschitz, in Solid Free
Manuf. Process. 42, 202 (2019). Fabrication Symposium Proceedings (2014), p. 197.
24. AFS, Procedure 5100-00-S, Mold Core Test Handbook, 3rd 64. O. Tsumura, H. Narita, D. Tomigashi, M. Okino, and K.
ed. (Shaumburg-IL: AFS, 2001). Miyauchi, Jpn. Foundry Eng. Soc. 81, 77 (2009).
25. A. Gardziella, L. Pilato, and A. Knop, Phenolic 65. H. Khandelwal and B. Ravi, J. Manuf. Process. 22, 127
Resins—Chemistry, Applications, Standardization, Safety (2016).
and Ecology, 2nd ed. (Berlin: Springer, 2000). 66. EnvisionTEC, ‘‘Viridis3D’’ (EnvisionTEC), https://envision
26. I. Chatzis and F.A.L. Dullien, Int. Chem. Eng. 25, 47 (1985). tec.com/3d-printers/robotic-additive-manufacturing/. Ac-
27. H.H. Yuan and B.F. Swanson, SPE Form. Eval. 4, 17 (1989). cessed 24 Sept 2019.
28. M.M. Labani, R. Rezaee, A. Saeedi, and A. Al Hinai, J. Pet. 67. J. Thole and C. Beckermann, Int. J. Metalcast. 4, 7 (2010).
Sci. Eng. 112, 7 (2013). 68. B.J. Stauder, H. Kerber, and P. Schumacher, J. Mater.
29. R.A. Cook and K.C. Hover, Cem. Concr. Res. 29, 933 (1999). Process. Technol. 237, 188 (2016).
30. A.B. Abell, K.L. Willis, and D.A. Lange, J. Colloid Interface 69. K. Meyer and P. Klobes, Fresenius J. Anal. Chem. 363, 174
Sci. 211, 39 (1999). (1999).
31. C. Gallé, Cem. Concr. Res. 31, 1467 (2001). 70. C. Gau and R. Viskanta, J. Heat Transf. 108, 174 (1986).
32. A.B. Selkirk and D. Ganderton, J. Pharm. Pharmacol. 22, 71. L.J. Williams and H. Abdi, Fisher’s least significant differ-
79S (1970). ence (LSD) test.Encyclopedia of Research Design, ed. N.
33. F. Carli and A. Motta, J. Pharm. Sci. 73, 197 (1984). Salkind (Thousand Oaks: Sage, 2010),.
34. A.M. Juppo, Int. J. Pharm. 129, 1 (1996). 72. D. Stefanescu and R. Ruxanda, Solidification structure of
35. P.A. Webb, An Introduction to the Physical Characterization aluminum alloys, in ASM Handbook Volume 9—Metallog-
of Materials by Mercury Intrusion Porosimetry with
Towards Functionally Graded Sand Molds for Metal Casting: Engineering Thermo-mechanical
Properties Using 3D Sand Printing
raphy and Microstructures (Novelty, ASM International, Publisher’s Note Springer Nature remains neutral with re-
2004), p. 107. gard to jurisdictional claims in published maps and institutional
73. D. Stefanescu, Science and Engineering of Casting Solidi- affiliations.
fication (New York: Springer, 2002).

View publication stats

You might also like