The Pennsylvania State University

The Graduate School

FUNDAMENTAL STUDY ON 3D SAND PRINTED MOLDS: METAL FLOW AND

THERMAL PROPERTIES

A Thesis in

Mechanical Engineering

by

Casey Bate

© 2019 Casey Bate

Submitted in Partial Fulfillment

of the Requirements

for the Degree of

Master of Science

May 2019

ii

The thesis of Casey Bate was reviewed and approved* by the following:

Guha Manogharan

Assistant Professor of Mechanical Engineering

Thesis Advisor

Tim Simpson

Paul Morrow Professor of Engineering Design and Manufacturing

Director, Additive Manufacturing & Design Graduate Program

Co-Director, Penn State CIMP-3D

Dan Haworth

Professor of Mechanical Engineering

Head of Department of Mechanical Engineering

*Signatures are on file in the Graduate School

iii

ABSTRACT

Fueled by the growing popularity of 3D sand printing (3DSP) molds for metal casting,

this thesis took-on two research opportunities for the modeling and characterization of these

materials. The first proposed a novel experimental method using succinonitrile (SCN) for

modeling casting flows. Understanding the metal flow in sand-molds is critical to eliminate

casting defects due to turbulent filling. While numerical methods have been applied to simulate

this phenomenon, harsh foundry environments and expensive x-ray equipment have limited the

experimentation to accurately visualize metal flow in sand molds. This thesis used flow

simulation and experiments using both water and SCN to identify the critical dimensionless

numbers to perform accurate metal flow analog testing. Experimental results show that SCN flow

testing was more accurate in recreating the flow profile of molten aluminum, thus validating its

utility as a metal analog for metal flow research. These findings can be used in future metal flow

analysis such as integrated filling-feeding-solidification studies. Secondly, thermal-physical

properties of 3DSP molds were investigated for the accurate solidification modeling of these

molds. The variable of binder concentration was added to this investigation to analyze effects on

casting solidification. A range of 3D printed molding materials was tested for binder content,

adherence to tolerance, density, specific heat, thermal conductivity, casting solidification time,

and coefficient of heat accumulation (bf). A solidification study of 99.9% pure aluminum showed

binder ratio had no effect on casting solidification time or heat transfer rates in the mold. A bf

value of 739.1 W-s0.5/m2-K was calculated from this experiment. A median density of 2.600

g/cm3 was measured ranging +/- 1% with binder content. Chvorinov’s rule and numerical

simulation were used to predict a thermal conductivity value between 0.29 W/m-K and 0.42

W/m-K. These findings can be used for more accurate representation of 3DSP molds in

solidification simulation analysis.

iv

TABLE OF CONTENTS

LIST OF FIGURES ................................................................................................................. vi

LIST OF TABLES ................................................................................................................... viii

ACKNOWLEDGEMENTS ..................................................................................................... ix

Chapter 1 Introduction ............................................................................................................ 1

1.1 Sand Casting – An Overview ..................................................................................... 1 1.1.1 Solidification Analysis and Riser Design ........................................................ 3 1.1.2 Gating Design .................................................................................................. 6 1.1.3 Pattern and core building ................................................................................. 10 1.1.4 Limitations of Traditional Sand Casting ......................................................... 12 1.1.5 Common Sand Casting Defects ....................................................................... 13

1.2 Introduction of Additive Manufacturing in Casting Through 3D Sand Printing ....... 14 1.2.1 Additive Manufacturing – An Overview ......................................................... 15 1.2.2 Introduction of 3D Sand Printing .................................................................... 16 1.2.3 Advantages and Limitations of 3DSP ............................................................. 17

1.3 Chapter Summary and Thesis Overview .................................................................... 19

Chapter 2 A Novel Approach to Visualization of Metal Flow in Sand-Casting ..................... 20

2.1 Introduction ................................................................................................................ 20 2.2 Materials and Methods ............................................................................................... 27 2.3 Results and Discussion ............................................................................................... 36

2.3.1 Computer Simulation Results .......................................................................... 36 2.3.2 Importance of Dimensionless Number Similarity in water Testing ................ 38 2.3.3 Succinonitrile Comparison to Water and Aluminum ...................................... 42

Conclusions ...................................................................................................................... 46

Chapter 3 Fundamental Study on 3D Sand Printed Molds: Thermal Properties .................... 50

3.1 Introduction ................................................................................................................ 50 3.2 Materials and Methods ............................................................................................... 55

3.2.1 Tolerance Measurements ................................................................................. 56 3.2.2 Loss on Ignition Testing .................................................................................. 57 3.2.3 Density Testing ............................................................................................... 59 3.2.4 Cooling Curve Analysis .................................................................................. 60 3.2.5 Specific Heat Testing ...................................................................................... 62

3.3 Results ........................................................................................................................ 64 3.3.1 Loss on Ignition Results .................................................................................. 64 3.3.2 Tolerance Results ............................................................................................ 64 3.3.3 Density Results ................................................................................................ 67 3.3.4 Specific Heat Results ...................................................................................... 68 3.3.5 Cooling Curve Results .................................................................................... 70 3.3.6 3D Printed Sand Mold Solidification Modeling .............................................. 73

v

3.4 Conclusion ................................................................................................................. 75

Chapter 4 Conclusion .............................................................................................................. 77

4.1 Conclusions from Succinonitirle Flow Testing .......................................................... 77 4.1.2 Future Work Regarding Succinonitrile Flow Testing ..................................... 79

4.2 Conclusions on 3DSP Molds Thermal Properties ...................................................... 80 4.2.1 Future Work Regarding 3D Sand Printing Thermal Properties ...................... 81

References.……………………………………………………………………………………83

vi

LIST OF FIGURES

Figure 1-1: Traditional green sand-casting process……………………………….……….2

Figure 1-2: Left) Solidification time for hammer casting. Right) Solidification for hammer

casting including gating system………………………………………….………………...4

Figure 1-3: Examples of good and bad part arrangement for proper solidification………..5

Figure 1-4: Properly functioning riser displaying shrinkage………...………………..…...6

Figure 1-5: Example of full gating system for metal casting…...…………..……………...6

Figure 1-6: Left) Traditional pour basin. Right) Offset pour basin [1]…………………….7

Figure 1-7: Depiction of melt flow in traditional sprue [1]………………………………..8

Figure 1-8: Proposed sprue redesigns [2]……………………………………………….....8

Figure 1-9: Examples of gating types. A) top. B) side. C) Bottom……………………….10

Figure 1-10: Example of a pattern plate. Left) Drag. Right) Cope…………………….....11

Figure 1-11: Example of a finished sand mold. Left) Drag. Right) Cope………………...11

Figure 1-12: Examples of straight and irregular parting lines…..…..…………………….12

Figure 2-1: Casting geometry of 1995 experimental study, units in mm………………….22

Figure 2-2: Filling images from 1995 study. Each column shows 1 of 3 tests. Letters

correspond to the time the image was taken after filling began a) 0.24s b) 0.5s c) 0.74s

d)1.0s e)1.24s f) 1.5s g) 1.74s h) 2.0s [3]…………………………………………………23

Figure 2-3: CAD design of acrylic mold parts……………………………………………34

Figure 2-4: Example of assembled acrylic mold (from earlier test)………………………34

Figure 2-5: Simulation results of SCN, aluminum, and water at a) 0.7s b) 0.9 s c) 1s d)

1.2s……………………………………………………………………………………….37

Figure 2-6: Expected aluminum fluid properties………………………………………....38

Figure 2-7: Water testing results for Fr matched (right) vs Fr unmatched (left) a) 0.5

seconds after pulling plug b) 0.74 s c) 1 s d) 1.24 s……………………………………......40

Figure 2-8: Water testing results for Re matched (right) vs Re unmatched (left) a) 0.5

seconds after pulling plug b) 0.74 s c) 1 s d) 1.24 s……………………………………......42

Figure 2-9: Succinonitrile testing results a) 0.5 seconds after pulling plug b) 0.74 s c) 1 s

d) 1.24 s…………………………………………………………………………………..44

Figure 2-10: Moody diagram [68] 1) Water Re 28,000 in acrylic 2) SCN Re 6,800 in

acrylic 3) Relative roughness of substance at Re 28,000 correlated to pressure drop

incurred by SCN…………………………………………………………………………45

Figure 3-1: LOI crucibles in oven………………………………………………………...58

Figure 3-2: Pour cup dimensions.……………………………………………….………..60

Figure 3-3: Pour cup experimental set-up………………………………………………...60

vii

Figure 3-4: Cooling curve experimental set-up………………………………………......61

Figure 3-5: Cooling cup after pouring exhibiting shrinkage……………………………...62

Figure 3-6: TA Instruments DSC Q2000.……………………………...………………....63

Figure 3-7: Graph of length deviation from nominal value versus pulse value…………...65

Figure 3-8: Graph of diameter deviation from nominal value versus pulse value………...65

Figure 3-9: Defective pulse 5 pour cups………………………………………………….66

Figure 3-10: Graph of deviation from nominal value versus pulse value for pour cups…..66

Figure 3-11: Graph of printed sample density versus binder content………………..……67

Figure 3-12: Graph of heat flow versus temperature for blank, sapphire, and pulse ……...69

Figure 3-13: Graph of specific heat versus temperature for pulse 2 sample………………69

Figure 3-14: Temperature vs. time from pour graph for thermocouples in pulse 4 mold…70

Figure 3-15: Graph of temperature vs. time at casting center for pulse values 1.5 – 4.........71

Figure 3-16: Plot of temperature vs time from pour for thermocouples placed in the mold

at 5mm and 20 mm from the casting for pulse values 1.5 - 4……………………………...72

Figure 3-17: Plot of weight percentage and heat flow versus temperature………………..73

viii

LIST OF TABLES

Table 2-1: Properties of succinonitrile……………………….……………………..………....…….26

Table 2-2: Properties of Aluminum………………………………………….….………..27

Table 2-3: Matching Reynold’s number values for SCN and aluminum……..….……….28

Table 2-4: Matching Reynold’s number values for water and aluminum……...…………29

Table 2-5: Possible properties of a 1 cubic meter sand mold casting……….…..………...30

Table 2-6: Solidification times for given mold materials and superheats………..……….32

Table 2-7: Experimental set-up assumptions……………………………….……….……33

Table 2-8: Testing initial conditions……………………………………….……….…….35

Table 2-9: Froude’s number match vs unmatched…………………………………..……39

Table 2-10: Reynold’s number match vs unmatched……………..………………..……..41

Table 2-11: Succinonitrile filling results……………………..………….……….………43

Table 3-1: Simulation results for thermal property study……………………….….…….52

Table 3-2: ANOVA results for thermal property study…………………………..………53

Table 3-3: 3D sand printed samples……………………………………………..………..56

Table 3-4: Loss on ignition results………………………………………………………..64

Table 3-5: Analysis of variance results for cylindrical geometries…………………..…...65

Table 3-6: Calculated density values compared to measured values…………………..…68

Table 3-7: Properties used when calculating mold constant……………………………...73

Table 3-8: Comparison of solidification times predicted by Chvorinov’s rule and numerical

simulation………………………………………………………………………………...75

ix

ACKNOWLEDGEMENTS

The author would like to acknowledge and dedicate this thesis to the following group of

people. First my family for always supporting me on a daily basis and helping me with any and

all needs. Next, my friend’s group from home who have also been there for me through the ups

and downs of graduate school and who I know will be life-long friends. My numerous friends I

have made at Penn State through various mediums with who I have spent so much of this time.

The technicians who have helped me with these projects including those at the FAME Lab, MCL,

and Reber basement. The students and lab mates who have helped me with these projects

including the students of the SHAPE lab and Hickner group. And finally, sour cream and onion

Pringles because if you actually managed to read this you deserve something to laugh at

Chapter 1

Introduction

Metal casting, a practice over 5,000-years-old, plays a role in 90% of all manufactured

goods. In the United State alone, metal casting is a $33 billion dollar industry [4] that supplies

industries that include oil and gas, aerospace, and automotive [5,6]. In terms of global production,

the U.S. ranked third behind China and India in 2016 having produced 11.5 million tonnage in

castings across 1,950 facilities [4]. As with all forms of manufacturing, the metal casting

industries has continued to seek higher quality products while minimizing scrap and process lag.

The recent development of three-dimensional (3D) sand printers has opened a new frontier for

metal casting, one that allows for new innovation to both product and process. Prompted by this

drive for innovation, the studies conducted in this thesis have sought to improve the quality of

metal casting through innovative experimental approaches.

1.1 Sand Casting – An Overview

Metal casting can be conducted in many forms including green sand, lost foam, high

pressure die casting, and investment casting [7]. Of these processes, green sand has remained the

most popular accounting for over 70% of all metal castings [8]. The process of metal casting

consists of eight main steps from part concept to finished product. These steps include the

creation of the theoretical part in computer aided design (CAD) software, solidification analysis,

gating and riser design, pattern building, core making and introduction, mold making, pouring

(casting), and post processing (Figure 1-1).

2

Figure 1-1: Traditional green sand-casting process.

In a simplified scenario, a customer, perhaps a large automotive manufacturer, designs a

part to use in their product. The customer contracts a foundry who takes on the CAD file. The

foundry conducts solidification analysis either through one of many commercially available

simulation software or through their own intuition. The foundry may ask that alterations be made

to the part to increase the chances of casting success. The solidification analysis also notifies the

foundry of where shrinkage is likely to occur. Given the shrinkage knowledge and the customer

specified tolerances, the foundry will place risers as necessary and develop pattern specifications

to minimize the shrinkage effects. The gating system is designed to maximize the number of parts

that can be produced in each mold while ensuring the mold cavity will be completely filled. This

system is also developed through modern simulation tools or foundry intuition. Once the gating

and riser system has been designed, a pattern plate will be commissioned. This plate will be

manufactured by a pattern shop which may be in house at the foundry or its own third-party

entity. Once complete, the plate is sent to the foundry where it will be used to impress the design

3

of the gating system, risers, and customer part in blocks of sand. This is called the “molding

process”. This negative space leaves a void in the sand which will be filled with molten metal

during pouring to produce the final part.

In many complex parts, cores are needed to produce voids in the desired part. In the case

of a pipe, a core is needed to produce the inner hollow of the pipe, otherwise the part would be a

solid cylinder. Due to limitations in the molding process, cores are manufactured out of sand as

their own piece. They are inserted after the molding process between the two sand blocks that

make up each mold. There may be numerous cores in a single mold depending on the number of

parts and complexity. After pouring occurs, the metal is allowed to cool in the mold until it

solidifies. The entire mold with cores and solid metal in place is then “shook-out”. In this process,

the molds are tumbled and vibrated until the sand breaks away from the solid metal parts. The

excess metal that filled the gating system may also break away from the finished part. From there,

the three groups are sorted accordingly. A percentage of the sand may be recycled for use in new

molds; however, it is not possible to make a new mold out of purely recycled sand. The gating

system is moved to a scrap collection pile where it may too be melted again and recast. The cast

parts are sent to post processing where excess metal is machined to meet the file customer

specific tolerances. This concludes the sand-casting process. The more detailed aspects of

solidification analysis, gating design, and pattern making are discussed in greater detail in the

following sub-sections.

1.1.1 Solidification Analysis and Riser Design

Solidification analysis is conducted to analysis the casting for short pours, shrinkage and

hot spots. Many experienced metal casters identify these concerns simply from looking at part

geometry; however, most foundries use some form of commercially available numerical

4

simulation tool. Of these some of the most currently most popular software packages include

MAGMA [9], ProCAST [10], SOLIDCast [11], and Flow3D Cast [12]. These packages use a

form of finite element analysis (FEA) to predict the way heat will flow out of the molten metal.

This is done by breaking down imported .stl files into a mesh. The mesh is a grid of volumes,

either in cubic or triangular form based on what type of FEA is being used, that collectively make

up the volume of the casting and the mold. The program then uses heat transfer physics to predict

the rate of heat transfer out of each cube (or triangular) face into the adjacent cubes around it, and

eventually into the mold volume. This process is done for every single volume in the mesh and

then progressed a single time step all the while calculating the temperature of each cube. Once

each volume making up the casting has reached solidification temperature, the simulation ends.

The data can then be reviewed to check for proper solidification (Figure 1-2).

Figure 1-2: Left) Solidification time for hammer casting. Right) Solidification for hammer

casting including gating system.

Preferably, solidification occurs unidirectionally in the opposite direction in which the

metal filling occurred. Ideally, the part solidifies from the farthest point working back toward the

gate, with the gate solidifying after the part. The reason for this is due to shrinkage in the metal

volume when changing from a liquid to a solid. This shrinkage takes place over the entire part

and is accounted for by making the pattern slightly larger than the desired final part. However,

5

when the metal solidifies, it sucks in liquid metal from the surrounding area. This causes a

problem if there is not ample molten metal near the solidifying area. If a hot spot occurs (an

isolated area of the part which solidifies slower than surrounding area), then that spot will be

depleted of its molten volume, leaving a void in the part. Hot spots and shrinkage mean that the

layout of the part in relation to the runner in critical. Typically, it is ideal to place the bulkiest end

of the part at the gate entrance so that it will have the longest access to the inflowing metal. The

thinner parts placed farther away will solidify sooner, drawing metal from the bulky area, which

will in turn draw metal from the gating system. If larger areas are separated from the melt flow by

a smaller area, the large area will be cut off from the melt supply during solidification, leading to

shrinkage (Figure 1-3).

Figure 1-3: Examples of good and bad part arrangement for proper solidification.

Despite best efforts, it is not always possible to arrange parts in a way that produces

optimal solidification. For these cases, risers are used. Risers are added volumes of metal that are

not critical to the final part. During solidification, they act as metal reservoirs for critical volumes

to siphon molten metal from (Figure 1-4). It is unimportant if the riser incurs shrinkage as it is

discarded during post processing. The important aspect is that the riser is large enough to supply

enough metal to the part.

6

Figure 1-4: Properly functioning riser displaying shrinkage.

1.1.2 Gating Design

The gating system is the mechanism by which the molten metal reaches the casting

cavity. This system of channels consists of the pour basin, the sprue, runners, and gates (Figure 1-

5). Each of these components has its own set of design rules; however, they all work to transfer

the metal to the casting cavity in a manner that minimalizes the chances of turbulence, slag, and

premature solidification.

Figure 1-5: Example of full gating system for metal casting.

7

The pour basin (Figure 1-6) is the first part of the gating system to contact the melt.

Traditionally, this opening has been little more than a wide funnel carved into the sand with its

sole purpose being to ensure the pourer has a wide opening. Occasionally, ceramic filters may be

used at the base of the pouring basin to catch slag and other unwanted particles in the melt.

However, in the past twenty years, the off-set pour basin design has gained popularity [13].

Figure 1-6: Left) Traditional pour basin. Right) Offset pour basin [1].

The offset basin operates by having a well adjacent to the sprue opening. Rather than

pouring directly through the basin into the sprue, the melt is poured in the well and rises to form a

pressure head in the pour basin. As the melt fills the basin, it flows from the well to the sprue

opening. This flow design provides a steady, uninterrupted flow into the sprue, and reduces the

velocity (as well as turbulence) of the melt by changing the direction of the vertical momentum

during pouring.

The next gating component is the sprue. The sprue, like the pour basin, has traditionally

been simple. Its design has historically been a vertical tapered cone, just like the pour basin;

however, much narrower. The narrowing taper of the sprue is critical for having good melt

characteristics. As the melt falls through the sprue, it naturally narrows as low-pressure air forms

in the gaps between the melt and sprue walls Figure (1-7).

8

Figure 1-7: Depiction of melt flow in traditional sprue [1].

This lack of containment on the melt allows for air entrainment and folding of the melt.

Both of these are large producers of casting defects. It is necessary that this same idea is

preserved throughout the gating system to avoid these defects. The base of the sprue may be met

with a well that is similar to the concept of the offset pour basin. Like the offset basin, the well

aims to slow the vertical momentum of the melt while also allowing for slag and sand caught in

the melt to settle to the bottom of the well. The sprue-well junction is often the location of a great

deal of splashing which is also not ideal for the melt. Splashing allows for folding and isolated

concentrations of melt to break from the bulk flow and prematurely solidify producing slag. Like

pour basins, sprue innovative sprue design has also been explored in recent years (Figure 1-8).

Figure 1-8: Proposed sprue redesigns [2].

9

These new sprue designs aim to reduce turbulence in the melt flow by gradually

transitioning the vertical momentum of the melt to horizontal. By doing this, these sprues are also

able to eliminate the right-angle transition from sprue to runner found in traditional designs which

further eliminates causes of turbulence. Hyperbolic and parabolic sprue designs are possible to

create with intricate pattern designs. Further complex designs like the spiral sprue are only

possible with the ability of 3D sand printing. Studies have shown that the increased complexity of

gating design possible with 3DSP can reduce turbulence compared to designs possible with

patterns [2].

Runners carry the melt horizontally from the sprue exit to various parts of the mold.

These are often the longest part of the gating systems, but these are dependent on the size of the

mold and the overall layout of the design within the mold. Runners have not been subject to as

intense review as sprue and pour basins. It is often best to employ Bernoulli’s theorem when

designing runners, particularly when using multiple gates [14,15]. In this case, the cross-sectional

area of the runner needs to be decreased between the first and second gate to ensure equal

pressure distribution across both gates. Various adaptations such as vents and extensions can be

applied to runners to further improve casting quality. Extensions can help control the slag and

entrainment often found at the leading each of the melt by allowing this edge to flow past the

gate. Similar to a well, the dirty melt is caught by the extension allowing the cleaner melt to enter

the casting cavity [1,15]. Vents can be used to control pressure in the channel. These vertical

passages are open to atmospheric pressure outside the mold. This allows a channel for trapped air

inside the gating system to be pushed out during filling. After filling with melt, vents also act as

undersized risers which can be used strategically to aid in solidification [5].

The gates are the final parts of the gating system. These are the cannels which carry the

melt from the runner to the casting cavity. Similar to runners, gates have also not seen a great deal

of complex innovation. Gates should be arranged in a manner that promotes both complete filling

10

of the mold as well as proper solidification. Most crucial to gating design is the position of the

gates relative to the casting cavity. A part may be top gated, bottom gated, or gated at the parting

line (Figure 1-9). Top gated is advised against as this system does not employ any mechanism for

slowing the rate of vertical momentum in the flow. If an incoming flow has too much energy, it

may damage the sand wall of the casting cavity, ruining the shape of the part and entraining sand

in the melt. Bottom gaiting is recommended for the opposite reasons as top gating. The melt has

to fight against gravity to fill the cavity, thus slowing the flow and reducing turbulence. This form

of gating often requires additional space in the mold for the longer system making it less

practical. Side gating at the parting line may be the easiest form of gating to create particularly for

traditional molding using patterns. The flow effects from this system fall between the top and

bottom characteristics [1,15].

Figure 1-9: Examples of gating types. A) top. B) side. C) bottom.

1.1.3 Pattern and core building

The pattern plate holds all the design information for the part, gating, and riser systems

(Figure 1-10). It is the physical product of the design rules discussed in the previous subsections.

The pattern plate is the positive shape of these systems, sized for proper shrinkage, which leaves

the negative impression in a sand block. Every sand mold is created from two blocks of sand, the

11

cope and the drag, which can be viewed as the top and bottom respectively Figure (1-11). The

line at which these blocks come together is called the “parting line”. This serves as the location

for where the pattern is inserted.

Figure 1-10: Example of a pattern plate. Left) Drag. Right) Cope.

Figure 1-11: Example of a finished sand mold. Left) Drag. Right) Cope.

Patterns can be made from a variety of materials including woods, plastics, and metals.

Each one has their advantages and disadvantages mainly concerning cost, tolerances, and

durability [16]. In addition to oversizing part design to compensate for shrinkage, patterns also

need to have draft on the impression surfaces. Draft is necessary to ensure that sand is not pulled

from the block during removal of the pattern. The larger the draft, the better the result for this

12

scenario; however, draft adds excess material to the final part which is removed during post

processing. Foundries may work with customers to strategically redesign draft into a part and

minimize post processing time.

Cores are additional sand pieces inserted into a mold to form complex geometries such as

hollows and overhangs (Figure 1-12). Cores are produced on their own assembly line separate

from the rest of the molding process. Cores are made by blowing or compressing sand into

custom core tooling called “core boxes”. Like pattern plates, the custom nature of these boxes

adds cost to the casting for each core design needed. This can add significant cost to a casting if

multiple unique cores are needed; however, avoiding cores can limit the amount of geometric

freedom a casting can have or add increased post processing time [5].

1.1.4 Limitations of Traditional Sand Casting

Sand casting is limited in geometric complexity due to the need to build around the

parting line and limit the utilization of cores. Parting lines should be picked for parts at a point

that produces a flat, even line across the length of the part.

Figure 1-12: Examples of straight and irregular parting lines.

13

The pattern will only form the exterior impression of the part so no interior detail is

maintained unless created with cores. This has limited the ability to create or cheaply make

castings with overhang geometry, curve back geometries, fins, steps, and holes. This has also

limited performance of both the casting and the part in its desired role. Geometric limitations

impose constraints on gating system design which without may be able to reduce turbulence and

casting defects. The need for draft and limits of thin wall casting may add unnecessary weight to

many casted parts.

1.1.5 Common Sand Casting Defects

Casting defects are vast in number and causes. Defects include but are not limited to

shaping faults arising from pouring, inclusions and sand defects, shrinkage defects, gas defects,

contraction defects following solidification, dimensional errors and compositional errors [14].

The majority of these defects hinder the strength of the casting, leaving the part unusable and

costing foundries money. Defects can be the result of human or simulation error such as in the

cases of not appropriately accounting for shrinkage, short pouring, or improper riser placement.

Preparation of the melt is another area that can directly attribute to defects. The melt should be

degassed, cleared of slag and other impurities. Alloys need to be mixed to the proper quantities to

ensure the desired properties of the casting. Pour temperature needs to such that surface finished

is optimized while avoiding premature solidification. Grain refiners may be added to improve

solidification quality [15].

Perhaps most critical to casting success is the design of the gating system. Proper gating

system design can aid in gas entrainment, solid entrainment, dimensional defects, film

formations, and complete filling. A large defect factor that can be controlled through gating

system design is turbulence. Turbulence is the result of the melt flowing through the mold at

14

higher than critical velocity. Many texts attribute this velocity to be 0.5 m/s [2,14,15]. Avoiding

turbulence is the most difficult casting rule to follow; however, it is necessary to make best

efforts as at least 80% of all casting defects can be attributed to turbulence [15]. Melt turbulence

produces folding and splashing leading to film creation and entrainment of solidified melt

droplets. If the melt has too much momentum, then it may break off sand from the mold,

entraining it in the melt but also hindering the dimensional accuracy of the casting through

damaging the mold. As high turbulence melt folds, it traps air leaving pours and blowouts during

solidification.

Satisfying such a low critical velocity has made turbulence free casting impossible in

gravity fed sand casting [15]. The limitations in the traditional molding process have made

particularly challenging. This area is one where 3DSP can have a direct impact and instant impact

through the ability to develop more geometrically complex gating systems. While these systems

are expected to show improvement in casting quality [2], the critical velocity value will continue

to ensure turbulence to an issue in metal casting.

1.2 Introduction of Additive Manufacturing in Casting Through 3D Sand Printing

Additive manufacturing (AM) has been a growing area of research over the past two

decades [17]. This new idea of building parts from a space of nothing is in direct contrast to years

of subtractive manufacturing philosophy, where parts are carved from a block of bulk material.

AM has grown to include numerous materials including metals, plastics, sands, and bio materials.

It has been adopted by the industrial sectors of aerospace, automotive, medical, and building

construction. Given this, it was only a matter of time before AM reached the metal casting world.

As AM has continued to grow and find the boundaries of its utilization, hybrid AM

systems have begun to appear. Literal “hybrid additive manufacturing” refers to the combination

15

of AM and subtractive manufacturing with-in the same machine. This is done by incorporating

3D printing and a CNC mill inside the same housing with access to the same build plate. Other

hybrid forms of AM may refer to indirect AM, which is this method by which AM has entered

the casting world. Indirect means that that AM plays a role in the overall manufacturing of a part,

but does not directly produce the part. Through 3D sand printing, AM has been able to produce

sand molds for the casting process. These molds employ many of the benefits of AM, but the part

is still formed by pouring liquid metal into a void, the same way all casting is done. This indirect

influence of AM on the casting world has moved slower than other parts of the AM community,

but it continues to gain momentum. Remaining gaps in the 3DSP-casting research have given

motivation to this study to continue to improve the capabilities of this form of manufacturing.

1.2.1 Additive Manufacturing – An Overview

Additive manufacturing (AM), by ISO/ASTM definition, is a process of joining materials

to make parts from 3D model data, usually layer upon layer [10]. Additive manufacturing is a

relatively recent development that is a contrast to the traditional method of subtractive

manufacturing (SM). AM builds a part in material layers as opposed to SM where material is

removed from a cast or forged billet in the form of milling, cutting, grinding, turning, or drilling.

AM has been a rapidly growing area of research during the last twenty years with publications

increasing logarithmically in scale [17]. Researchers have continued to expand upon the

techniques, materials, complexity, quality, size, and cost reduction of AM. These efforts were

initially focused on plastics and metal materials, but have expanded to a wide variety of printable

materials including metals, ceramics, plastics, polymers, sand, electronic materials, and biological

materials [19].

16

Techniques for additive manufacturing include: material extrusion, material jetting, vat

photopolymerization, binder jetting, powder-bed fusion, fused energy deposition, and sheet

lamination [20]. Each of these processes has its strengths and weaknesses relative to one another

in terms of tolerances, material capability, part size, system size, cost, and support material

removal. As a whole, additive manufacturing offers previously unrealizable design freedom, the

ability to optimize parts for strength versus weight, and has the potential to be a more

environmentally sustainable form of manufacturing [21].

1.2.2 Introduction of 3D Sand Printing

3D sand printing makes use of the binder jetting process. This process builds finished

parts by bonding powder together on a layer-by layer basis. The printer will spread a layer of the

build material, in this case sand, across the entire build area. Then, a glue head, in this case using

a furan resin, passes over the sand layer. The glue head deposits resin only in the areas prescribed

by the model. A new layer of sand is then applied and the process is repeated in this fashion.

Catalyst is embedded in the sand that reacts with the resin to bond the sand together.

The first commercial sand printer was released in 2001 by Generis GmbH of Germany.

This printer used the same binder jetting process used by today’s printers. In 2005, Ex One

released its first “S-Print” sand printer at a cost of $500,000 [22]. Today, companies such as Ex

One [23], Voxeljet [24], and Envisiontec [25] produce sand printers in a variety of sizes ranging

from research units to full-scale production.

17

1.2.3 Advantages and Limitations of 3DSP

When discussing the advantages and limitations in additive manufacturing, it is often

done in terms of opportunistic and restrictive elements introduced by AM when compared to

traditional methods of manufacturing [26,27]. Opportunistic elements are those that AM can

provide that are often considered favorable to traditional. When discussing 3D sand printing, the

main opportunistic feature is near limitless geometric freedom. 3DSP eliminates the need for

cores as the parting line requirement no longer needs to be met. Mold makers or mechanical mold

devices no longer need to access the middle of the mold to create the desired geometric features

in the sand. Additionally, the binder jetting method does not require supports like many other

forms of AM; so, there is no penalty for pushing geometric limits. This geometric freedom can

improve the performance and quality of casted parts in a manner that traditional methods

currently cannot. Reduced geometric limits allow for part designers to extended freedom to

design parts that may be stronger or reduced in weight. Geometric freedom in gating design

allows for the creation of innovative designs such as those shown in Figure 1-8. Innovative gating

designs can improve part quality through turbulence and heat transfer control, helping to avoid

many of the casting defects discussed earlier.

3DSP also allows for low quantity castings to be produced cheaper, faster, and with a

higher degree of individual customization. This is possible through the elimination of the pattern

plate. Since the pattern plate is typically the most complex and costly part of the casting process,

it often takes weeks to manufacture the plate. Since 3DSP does not require any components to

produce a complete sand mold outside of the printer, each 3DSP mold is essentially a “one-off”.

Like most AM methods, 3DSP also allows for object embedding. In casting, this may be useful

for strategically placed chills, filters, or more innovative objects such as velocity sensors and

heating elements.

18

Restrictive elements are those limitations specific to a given AM process. AM printers

are limited by the size of the build volume, and 3DSP is no exception. The largest commercially

available printers feature build volumes of just over 300 ft3 and maximum build heights of around

3 ft. This places limits on the maximum size of a mold or how many molds may be completed in

a single print. Meanwhile, pattern plates (once finished) and modern molding machines are

capable of producing molds as quickly as under a minute. For each part produced, the pattern cost

is decreased while a printed part remains a fixed cost regardless of the number produced.

Additionally, 3DSP as a binder jetting process, has the issue of needing to clean

unbonded sand for the finished mold. This issue has proven to be a hurdle effecting printing time

and the “print bed to use” nature of 3DSP. To clean parts, operators often have to vacuum and

hand brush loose sand from the printed parts. Molds often have to be divided into multiple parts

to allow proper cleaning. Failure to properly clean a mold of loose sand or failure to assemble the

mold properly will cause defects in the casting, thus creating new opportunity for defects.

Finally, high purchasing costs, operating costs, and large installation volumes can be

prohibited to smaller manufactures looking to invest in 3DSP. These elements also propose room

for a gap to form between larger and smaller manufacturers based on having the resources to

invest in this technology. The current state of 3DSP lends itself to be most useful to small

quantity, high complexity castings [17]. Continued development of 3DSP and AM as a whole will

continue to amplify the opportunistic abilities of 3DSP while minimizing the restrictive elements;

however, the themes for both sides highlighted in this section will remain as a whole barring a

significant technological leap in this area.

19

1.3 Chapter Summary and Thesis Overview

This chapter provided an overview for the traditional sand casting process. Key steps

included in the process are solidification analysis, gating system designs, and pattern and mold

building. Each step is critical to manufacturing a successful casting and avoid defects brought-on

by part shrinkage, short pours, and melt turbulence. The Additive manufacturing process of 3D

sand printing (3DSP) was introduce including its current advantages and limitations with-in metal

casting. 3DSP in capable of producing a higher level of geometric complexity for both gating

systems and the actual casting. It does so thorough the use of the binder jetting process which

eliminates the needs for patterns and cores. Parts of high quantity and low geometric complexity

remain better suited for the traditional sand casting process as 3DSP is limited by production

speed.

The remainder of this thesis discusses two research studies derived from the capabilities

offered by additive manufacturing in the metal casting industry. The first study, presented in

chapter two, introduces a new methodology for experimentally visualizing casting flow. This

study was motivated by the need for an inexpensive test to verify innovative, geometrically

complex gating systems possible with 3D sand printing. The second study, presented in chapter 3,

quantified thermal properties of 3D printed sand molds while simultaneously investigating the

effects of binder content on these properties. A conclusion of these studies is presented in chapter

four highlighting the study results, future work, and limitations.

Chapter 2

A Novel Approach to Visualization of Metal Flow in Sand-Casting

In this chapter, a novel approach for casting flow visualization is proposed using

succinonitrile (SCN). Understanding metal flows in sand-molds is critical to eliminate castings

defects due to turbulent filling. While numerical methods have been applied to simulate this

phenomenon for multiple decades, harsh foundry environments and expensive x-ray equipment

have limited the experimentation to accurately visualize metal flow in sand molds. In this study, a

novel approach to solve this challenge is proposed using succinonitrile as a metal analog. SCN

has a long history in solidification research due to its BCC crystal structure and dendrite-like

solidification property, but this is the first reported study on its use for melt flow studies. This

paper used flow simulation and experiments using both water and SCN to identify the critical

dimensionless numbers needed for accurate metal flow analog testing. Froude’s number and wall

roughness were identified as critical variables. Experimental results show that SCN flow testing

was more accurate in recreating the flow profile of molten aluminum, thus validating its utility as

a metal analog for metal flow research. Findings from this study can be used in future metal flow

analysis such as: runner, in-gate and integrated filling-feeding-solidification studies.

2.1 Introduction

Metal casting is the oldest known manufacturing process (> 5,000 years), and plays a role

in 90% of all manufactured goods [4]. In particular, 80% of castings are produced via the

traditional sand-casting method [28]. Traditional sand-casting process involves mold fabrication

(e.g. no-bake green molds) using a pattern to produce mold components (e.g. cores, cope, drag

21

and cheeks) that are assembled at the parting line. In addition to generating the mold cavity for

part geometry, the pattern plate also develops the geometry of gating system, i.e. channels for

metal flow into the mold cavity. The gating system in sand-casting includes: pouring basin, sprue,

runners, gates and risers. Several studies in traditional sand-casting have highlighted the

importance of optimal design of gating systems to minimize casting defects [29,30].

One of the inherent challenges in traditional sand-casting to produce defect-free castings

is minimizing/eliminating turbulence in the melt flow such that critical velocity (<0.5m/s) is not

violated at the ingate [15,31]. Turbulence during pouring can produce air entrainment, splashing,

and film formation all of which hinder the microstructure of the casting and reduce part strength

[32,33]. Turbulence can be minimized through the proper design and analysis of gating systems

[15]. Recent innovations in 3D sand-printing have enabled non-conventional gating system

designs as the geometric limitations of traditional mold fabrication have been greatly expanded

[2]. Such 3DSP-centric gating designs can significantly reduce overall casting defects by as high

as 99.5%, oxide inclusions by 35% and improve mechanical strength of metal casted parts by

8.4% which would both positively save production costs to foundries from scraps and improve

part quality for end-applications.

Experimental analysis of liquid metal flow has been a major challenge in casting

research. This can be attributed to the pouring conditions in castings (e.g. opaque sand-molds,

high temperature, outgassing) and hence do not provide easy access to collecting qualitative or

quantitative data which can be compared to a wealth of simulation studies [34–36] for validation.

Qualitative flow field data can only be collected through the use of expensive X-ray equipment

because of the opacity of the mold, whether it be sand or of the permanent variety [3,37]. Several

studies have reported attempts to collect temperature and deduced velocity data via in-contact

thermal measurement sensors has also proven to have limited success due to dynamic changes in

temperatures and conductivity of the melt [38,39].

22

In 1995, a benchmark study was conducted for the accurate characterization of liquid

metal flows which was motivated by the rapid growth in research efforts on numerical modeling

where a sand mold for 10mm x 200mm x 100mm (x,y,z) aluminum plate was fabricated. As

shown in Figure 2-1, a bottom gating system with a runner of 240 mm length and sprue height of

410 mm from entrance to base of the well was fabricated with an offset pour basin that featured a

removable plug.

Figure 2-1: Casting geometry of 1995 experimental study [3], units in mm.

After a predetermined melt height was reached in the pouring basin, the plug was

removed to eliminate the effects of initial velocity during pouring. A 2.2 kg charge of 99.99%

pure aluminum was poured into the basin at 720°C (approximately 700°C at the sprue entry). The

mold filling (n=3) was observed in an x-ray machine with a sampling frequency of recorded at 50

Hz. The images (see Figure 2-2) captured in this study are the most prevalently employed

benchmark for sand mold filling to exist [40–42]. This can be largely attributed to the prohibitive

23

cost of performing high temperature metal pouring inside an expensive x-ray equipment and

corresponding safety concerns.

Figure 2-2: Filling images from 1995 study. Each column shows 1 of 3 tests. Letters

correspond to the time the image was taken after filling began a) 0.24s b) 0.5s c) 0.74s d)1.0s

e)1.24s f) 1.5s g) 1.74s h) 2.0s [3].

The need for melt flow experiments is derived from the need to evaluate and verify newer

numerical models for metal flow. Water analogs have been largely used in the casting community

as a cheaper alternative to studying liquid metal flow. Thomas et al. [43,44] used water to study

liquid steel flow in continuous casting settings. In two studies, water was used to verify the

accuracy of a numerical model, while also drawing insights from the fluid flow to indirectly

develop the boundary conditions of the model which was being developed for liquid metal

[43,44]. Cleary et al. [45] used water to simulate molten aluminum flow through die cavities

using a water model as the control test for a novel smoothed-particle hydrodynamic (SPH) model

24

and commercially available casting flow simulation tools. The study found that flow simulations

had resemblance to water tests. Renukananda et al. [46] also used water to examine mold filling

in a horizontal multi-gate system and showed that while water had a different gate velocity and

flowrate than the comparison metal, the relationship between these properties and gate location

followed similar trends Additionally, the amount of material deposited at each gate was similar

across all tests.

The reliance on water tests as a means of numerical confirmation for liquid metals is an

approximate ‘similarity testing’ as water has very different thermal-fluid properties when

compared to liquid metal. Water also does not solidify at room temperature unlike molten liquid

metal which leads to limitations in its utility as an evaluation tool that simultaneously simulates

flow and solidification. Several studies [47,48] have been conducted to determine factors that are

critical to establishing metal-water analogs. These studies have primarily focused on the

dimensionless numbers Froude number (Fr) (Equation 2.1) and Reynolds number (Re) (Equation

2.2).

𝐹𝑟 = 𝑉2

𝑔𝐷ℎ (2.1)

𝑅𝑒 = 𝜌𝑉𝐷ℎ

𝜇 (2.2)

Froude number is a ratio of inertial to gravity forces acting on a fluid while Reynolds

number is a ratio of inertial to viscous forces. Sahai et al. [47] showed that matching Reynolds

number in reduced scale water models for continuous casting tundishes is important in achieving

reliable metal-water analogy. Froude’s number was determined to have no effect on these systems

[47]. Another study claimed that despite a difference of about 18% between the kinematic

viscosities of water and steel, little effect was seen when comparing the flow patterns of the two

materials in the nozzle condition [43]. Another study [48] matched Fr and Re in addition to

25

Weber number to simulate air entrainment in plugging steel flows through water models. Despite

these efforts, it was still suspected that air entrainment would be higher in actual steel processing

based on observations in water tests. In summary, there are major unresolved issues in using

water analog tests for melt flow analysis and there is a motivation to identify alternative materials

for casting visualization.

Succinonitrile - C2H4(CN)2 has a low entropy of fusion and is a single plastic from -35°C

to 62°C melting temperature [49]. It has been the focus of decades of solidification research

popularized by Glickman et al. in 1976 [50] with earlier studies into molecular and vibrational

modes conducted in the mid 1950’s by Janz and Fitzgerald [51,52]. It’s properties as a “plastic

crystal” are very useful in the field of crystal growth science focused on dendritic solidification.

Dendrite solidification is the crystal forming process in metals such as nickel, copper, gold, silver,

aluminum, zinc, lead, tin, and indium [53–55]. Plastic crystals such as SCN are a class of

molecular solids (both organic and inorganic) which melt with a relatively small entropy change

and are hence considered as analogs to simple metals for solidification studies. The rotary

motions of these molecules are preserved when a molecule transforms from liquid to solid phase.

Plastic crystals typically have a wide liquid range when compared to most substances which melt

closer to ambient temperature. Additionally, the transparency of plastic crystals makes it suitable

for a wide variety of optical techniques for accurate morphological and kinetic measurements

[50].

The pioneering study by Glickman [50] drew numerous conclusions about the physics of

dendrite modeling while expanding knowledge on physical properties of SCN. Subsequent

studies continued to employ SCN for dendrite formation studies [56–60]. Another study [57]

explored the addition of argon gas and acetone to SCN , and acetone with SCN in 1988. Acetone

was of particular interest to form an SCN alloy that preserved the linear solid-liquidus line

observed in the SCN phase diagram. Succinonitrile has continued to be popular in solidification

26

research as researchers continue to focus on more specific areas of solidification and grain

refinement [58–60].

Another study [61] employed light scattering spectrometry to measure the viscosity and

surface tension of liquid SCN, two properties which are vital to understanding the flow and heat

transfer rate that were not previously well explored. Surface tension (±2%) and viscosity

measurements (±10%) for six different temperatures ranging 60°C to 110°C were recorded for

pure SCN and can be correlated to temperature as shown in (Equations 2.3 and 2.4) [61].

Surface tension (mPa-s) = 43.14 – 0.0823T (2.3)

Viscosity (mN/m) = 4.11 – 0.0263T (2.4)

Where T is temperature in °C. The density of succinonitrile as a function of temperature (°C) is

shown in Equation (5) [62]. Additional thermal and physical properties of succinonitrile are listed

in Table 2-1.

Density (g/cm3) = 1,000(1.0334-(0.000781)T) (2.5)

Table 2-1: Properties of succinonitrile.

Symbol Property Value Reference

W Molecular weight 80.092 g/mol [50]

ΔVm Molar volume change on melting 3.71 cm3 [50]

ρs Density of solid 1,016 kg/m3 [50]

ρl Density of liquid 970 kg/m3 [50]

Tm Melting point 331.24 K, 58.09°C [50]

Tb Boiling point 538.7 K, 265.55°C [63]

L Latent heat of fusion 46,238.7 J/kg [50]

Cp Heat capacity of liquid 1998.23 J/kg-K [50]

Ks Thermal conductivity of solid 0.224 W/m-K [50]

Kl Thermal conductivity of liquid 0.223 W/m-K [50]

The aim of this thesis is to evaluate the suitability of succinonitrile (SCN) as an

alternative to water as a liquid metal flow analog. If successful, SCN flow tests could not only

accurately mimic metal flow but properties of SCN as a plastic crystal and dendrite solidification

27

formation at room temperature could enable novel flow-solidification visualization framework. In

other words, the success of SCN as a means of mimicking metal flow will further bridge the gap

between metal flow and solidification models which would positively impact related

experimental efforts and lead to more accurate numerical solvers for potentially integrated flow-

solidification models. The testing methods detailed in this paper also provide a roadmap for

validation of innovative gating geometries.

2.2 Materials and Methods

In this study, a systematic methodology to achieve similarity values in critical fluid flow

parameters (Re, Fe) and solidification parameters were developed for the proposed flow material

(liquid SCN) for targeted metal flow (Aluminum – Table 2-2 [3]).

Table 2-2: Properties of Aluminum.

Symbol Property Value

ρl Density of liquid 2,373 kg/m3

Tm Melting point 640°C

Tp Pouring Temperature 700°C

μ Dynamic viscosity 0.00125 Pa-s

L Latent heat of fusion 398,000 J/kg

Cp Heat capacity of liquid 1,888 J/kg-K

Hydraulic diameter of a fully filled rectangular channel can be found using Equation 6.

Dh = 2ab

a+b (2.6)

Based on Equation 2.6, a rectangular channel of 19.2mm x 15mm cross-section [3] will

result in a hydraulic diameter of 0.01684 meters. With an average head height of 40 mm during

pouring in the pouring basin, a modified version of Bernoulli’s theorem (see Equation 2.7) found

an initial velocity of 0.886 m/s was likely to occur immediately after the plug had been removed.

28

𝑉𝑖𝑛𝑖𝑡𝑎𝑙 = √2𝑔ℎ (2.7)

Subsequently, Re number of 28,325 was determined for the aluminum at the entrance of

the sprue (Equation 2.2):

𝑅𝑒𝐴𝑙𝑢𝑚𝑖𝑛𝑢𝑚 = 2373 ∗ 0.886 ∗ 0.016842

0.00125= 28,325

There are three different approaches to achieve a desired Re (i.e. 28,325) for any material

(i.e. SCN) by varying the: (1) pour temperature to correspondingly vary the kinematic viscosity of

the fluid, or 2) pour velocity, or 3) hydraulic diameter of the channel opening.

In this case, the kinematic viscosity of aluminum was 5.268E-7 m^2/s. Equations 3 and 4

result in a temperature similarity value of 137.73°C for SCN which violates the physical

properties of SCN. However, at a pouring temperature of 75°C for SCN, a velocity similarity was

found to be 3.969 m/s, about 4.17 times that of aluminum. Similarly, a hydraulic diameter

similarity was found to be 0.07017 meters, again 4.17 times that of aluminum.

Table 2-3: Similarity values for Reynolds number - SCN and aluminum.

Pour Material Aluminum SCN SCN SCN

Liquid Density 2,373 kg/m3 955.3 kg/m3 975 kg/m3 975 kg/m3

Dynamic Viscosity 0.00125 Pa-s 0.00051 Pa-s 0.00214 Pa-s 0.00214 Pa-s

Velocity 0.886 m/s 0.886 m/s 3.969 m/s 0.886 m/s

Hydraulic diameter 0.01684 m 0.01684 m 0.01684 m 0.07017 m

Temperature 700°C 137.73°C 75°C 75°C

Re Number 28,325 28,325 28,325 28,325

When the methodology to identify similarity values was repeated for aluminum and

water, it was found that that water flow at a temperature at 53.1 °C would result in Reynolds

number similar to molten aluminum for the same volumetric flow conditions as highlighted in

Table 2-4.

29

Table 2-4: Matching Reynold’s number values for water and aluminum.

Pour Material Aluminum Water Water

Liquid Density 2,373 kg/m3 997.05 kg/m3 986.61 kg/m3

Dynamic Viscosity 0.00125 Pa-s 0.00089 Pa-s 0.00052 Pa-s

Velocity 0.886 m/s 0.886 m/s 0.886 m/s

Hydraulic diameter 0.01684 m 0.01684 m 0.01684 m

Temperature 700°C 25°C 53.1°C

Re Number 28,325 16,732 28,319

As seen in Equation 2.1, Froude’s number is comprised of velocity, gravity and hydraulic

diameter. As this study desired to keep a uniform geometry across all tests, Fr similarity was

solely a product of matching initial velocity. Using the modified Bernoulli’s formula found in

Equation 2.7, it was determined that initial velocity could be controlled by fluid head height in the

pour basin. If two fluid were held at the same head height prior to pulling the basin plug, then

their Froude’s numbers would match. This concept was used when testing the importance of Fr

similarity during testing.

It is well established that many issues in metal casting stem from premature

solidification. Succinonitrile’s ability to melt at low temperatures and solidify at room

temperature may be able to offer new insights to this issue as an experimental tool. Chvorinov’s

rule [64] is a formula for relating solidification time to mold parameters, geometric parameters,

and thermal parameters of the melt material (see Equation 2.8).

𝑇𝑖𝑚𝑒𝑆𝑜𝑙𝑖𝑑𝑖𝑓𝑖𝑐𝑎𝑡𝑖𝑜𝑛 = [𝜌𝑙𝐿

𝑇𝑚−𝑇0]

2[

𝜋

4𝑘𝑚𝜌𝑚𝑐𝑚] [1 + (

𝑐∆𝑇𝑠

𝐿)

2] [

𝑉

𝐴]

2 (2.8)

km = Thermal conductivity of mold (W/m-K)

ρm = Density of mold material (kg/m3)

cm = Specific heat of mold (J/kg-K)

V = Volume of casting (m3)

A = Surface area of casting (m2)

30

ρl = Liquid density of pour material (kg/m3)

L = Latent heat of pour material (J/kg)

c = specific heat of pour material (J/kg-K)

Tm = Melting temperature of pour material (K)

T0 = Ambient Temperature (K)

ΔTs = Superheat, temperature at which material is pour minus melting temperature (K)

Consider that a desired object to be cast is in the shape of a cube. The cube has sides that

are 0.1 meters in length. Therefore, the cube has a volume of 0.001 m3 and a surface area of 0.06

m2. The cube is to be casted using a traditional green sand mold. The properties of this sand mold

can be found in Table 2-5.

Table 2-5: Possible properties of a 1 cubic meter sand mold casting.

Property Value Reference

Sand Thermal Conductivity 0.59 W/m-K [65]

Sand Density 1,522 kg/m3 [65]

Sand Specific Heat 1,075 J/kg-K [66]

Ambient Temperature 298K, 25°C

If these mold conditions and the shape of the cast object are maintained, then the

solidification time of the casting becomes solely dependent on the thermal properties of the pour

material. This means that aluminum poured at 700 °C would result in a solidification time of 518

seconds using Chvorinov’s rule and the mold conditions previously described (aluminum density

= 2,373kg/m3, latent heat = 398,000 (J/kg), specific heat = 1,888 J/kg-K, melt temperature = 660

°C). If the casting substance was changed from aluminum to succinonitrile, then it would be

found that an identical solidification time of 518 seconds would be obtained for a superheat of

31

9.5°C (meaning pour temperature of 67.5°C) considering the properties of SCN found in Table 2-

1 and the mold properties in Table 2-5. This ability to match solidification time of aluminum for a

small super heat has great potential for SCN as an experimental casting research tool. The

solidification research of SCN has already been discussed; however, it has not been found where

SCN has been used as a method of studying gating systems in metal castings. The ability to match

solidification time can accurately represent the problem of premature solidification in gating

systems. This allows for the experimental testing of innovative gating designs, specifically those

of the thin-walled verity. It should be noted that Chvorinov’s rule is designed to quantize

conductive heat transfer of a stationary fluid through a mold after filling has been completed.

Chvorinov’s rule does not consider convective heat transfer, which plays a large role in the

premature solidification problem, or radiative heat transfer.

The optimal case for SCN as a metal analog would be to provide the ability to bring

casting research out of the foundry and into a laboratory environment. This would lead to lower

costs, smaller equipment, and possibilities to visualize and quantify flow parameters. The

simplest way to accomplish this goal would be to substitute the sand mold with a common,

inexpensive transparent material. Acrylic plastic is easy to obtain, machine, and assemble as well

as transparent and relatively cheap. It has already been used in casting experiments most notably

in 2016 [46]. Unfortunately, the thermal characteristics of acrylic do not yield themselves well to

the previous solidification time example. Acrylic has the values of 0.21, 1,200, and 1,500 for

thermal conductivity (W/m-K), density (kg/m3), and specific heat (J/kg-C) respectively. For the

same geometry with SCN, the solidification time would jump to 1,322.7 seconds, more than twice

that of the aluminum in the sand mold. A realistic super-heat could not be found for SCN and

acrylic using Equation 2.8.

Glass is a transparent material that would allow for the matching of solidification time of

aluminum in a sand mold; however, it is neither cheap nor easy to build custom, intricate designs.

32

Glass has the values of 0.75 W/m-K, 2457.6 kg/m3, and 834.6 J/kg-C for thermal conductivity,

density, and specific heat respectively. This would result in a SCN solidification time of 517.9

seconds for a super-heat of 21.4°C (pour temp of 79.4°C) (see Table 2-6). While higher than the

sand mold, this pour temp is still well below the 266°C boiling temperature of SCN. Other mold

materials may also work, but finding one that allows for the proper matching of solidification

time and transparency will be difficult if possible, at all.

Table 2-6: Solidification times for given mold materials and superheats.

Pour material Aluminum SCN SCN SCN

Mold Material Sand Sand Acrylic Glass

Super Heat 40°C 9.44°C 9.44°C 21.44°C

Solidification Time 517.9 s 517.9 s 1322.7 517.9 s

Computer simulations were conducted to analyze the effect that Reynold’s number had

on the flow profile for the three substances. Aluminum at 700°C was compared to water at 25°C,

and SCN for the velocity and hydraulic dimeter matched conditions. The mold geometry from the

1995 study was recreated in CAD software, the dimensions of which can be seen in Figure 2-1.

The geometry was imported to the commercial software Flow3D – Cast [12]. Simulations were

conducted for all materials flowing through a sand mold with identical characteristics.

Additional limitations were imposed for the experimental portion of this testing. These

limitations and assumptions are as follows:

33

Table 2-7: Experimental set-up assumptions.

Limitation Explanation

A constant head was not

maintained in the pour

basin.

The 1995 experimental study maintained a constant head height

of 40 mm in the pour basis throughout the mold filling. This

study filled the pour basin to a desire head height, ceased

pouring, and pull a plug to release the flow into the mold. Effects

this may have had on velocity were neglected.

Mold permeability was not

matched.

Mold permeability is a necessity to eliminate back pressure in

the mold due to trapped air. Back pressure impedes the flow of

the melt if not eliminated. The 1995 study used 60 AFS-grade

silica sand bonded with 1.2 wt.% phenolic urethane resin. A vent

of random size was added to the test mold in this study to

alleviate the back pressure rather than an exact permeability

match.

Wall roughness was not

matched.

Wall roughness plays a role in flow velocity and profile. The

roughness values on the 1995 sand mold and that of the acrylic

mold in this study differ. No efforts were taken to correct this

difference and roughness was ignored.

Reynold’s similarity for

SCN and Aluminum could

not be satisfied.

Issues velocity and diameter derived similarity were discussed in

“section 2.1”. The necessary temperature of 138°C to thermally

match exceeded the 77°C temperature rating or the acrylic.

Severe cracking was seen in attempts to reach higher

temperatures. Therefore, SCN was tested at 75°C.

An acrylic mold was made to mimic the 1995 geometry found in Figure 2-1. A CNC mill

was used to cut the pattern into 0.5-inch-thick acrylic sheets. The casting geometry was parted at

the middle of the sprue so that a maximum depth of 7.5 mm was cut into each acrylic sheet. The

majority of the geometry fit into a 12 inch by 12-inch acrylic sheet. A 6 inch by 6-inch piece was

used for the top of the sprue. A 1.3 mm deep shelf was cut around the edge of the casting

geometry on one half of the mold so that a rubber gasket could be applied.

34

Figure 2-3: CAD design of mold parts.

A 1/16th inch thick piece of rubber was cut into strips and glued along the non-recessed

edge of the gating geometry. Clear silicon was also applied around the edges of the embedded

geometry to form a liquid seal. One of the two 6-inch pieces was fixed to one of the 12-inch

pieces with acrylic cement. The remaining two acrylic pieces were fastened to the glued pieces

with three M8 bolts and ten 8-32 bolts. The fastening of the bolts compressed the rubber/silicon

seal to form a liquid-tight seal.

Figure 2-4: Example of assembled acrylic mold (from earlier test).

35

Notches were removed from the top of the sprue. A rectangular slot was cut into the

bottom of a 1-quart food-grade container. The container was placed on top of the mold so that the

top of the sprue came through the slot. The base of the container was coated with silicon to form a

seal with the top of the mold. This container served as the pour basin. A rubber plug was cut to

match the rectangular shape of the sprue. A screw was inserted into the plug and a string was tied

around the screw. This mechanism was used to control when fluid would enter the mold. Masking

tape was place along the sprue and runner edges and marked every 0.5 inches so that velocity data

could be gathered from the mold.

Pour testing was conducted in a fume hood due to the health hazards imposed by SCN.

Succinonitrile is Category 2 skin irritant and a Category 2A eye irritant. It is also a Category 4

acute toxin (oral) and a Category 3 specific organ toxin (respiratory) [67]. For these reasons, the

Personal Protective Equipment (PPE) for this work included EN 166 safety glasses, Nitrile rubber

gloves, type P95 respirators, and a lab coat. A Casio EXILIM highspeed camera was used to

record video of the mold at 300 fps. Green food dye was added to the pouring substances to

improve viewing. The initial conditions for each test are shown in Table 2-8.

Table 2-8: Testing initial conditions.

Test Substance Head Height Temperature Initial Re #

Fr matched Water 40 mm 53°C 28,316

Fr unmatched Water 80 mm 34°C 28,575

Re matched Water 40 mm 53°C 28,316

Re unmatched Water 40 mm 22°C 15,616

SCN 1 Succinonitrile 40 mm 75°C 6,804

SCN 2 Succinonitrile 40 mm 75°C 6,804

SCN 3 Succinonitrile 40 mm 75°C 6,804

A sand bath connected to a JKEM Geminin temperature controller was used to heat the

substances to the desired temperatures. For the lower temperature water tests, warm tap was used.

36

Each substance was poured into the pour basin until the specified head height was reached. The

rubber plug was then pulled, allowing the substance to enter the mold.

2.3 Results and Discussion

2.3.1 Computer Simulation Results

It was found that speeding up the volumetric flow rate of the SCN via greater initial

velocity or larger hydraulic diameter produced different results from aluminum. This was due to

the law of continuity which states that mass flow rate of any substance must be conserved. By

setting SCN to a higher initial velocity to match the Re number of aluminum, the mold filled

quicker and more violently than the aluminum simulation. Enlarging the hydraulic diameter as

described in Table 2-3 also produced this result. Despite the initial velocity matching that of the

aluminum simulation, the larger volume of mass in the sprue meant the mass flow rate matched

that of the speed-up test. Therefore, the SCN flow behaved in the same erratic manner. These

same findings support the need for Froude number similarity in casting modeling.

These findings revealed the need to match the mass flow rates for each substance and the

limitations involved with attempting match Reynold’s number through flowrate in casting

situations. A comparison simulation was run for aluminum at 700°C, water at 25°C, and SCN at

75°C using a consistent flow rate and the geometry in Figure 2-1. The results of this simulation

(Figure 2-5) showed that the water and Al molds filled in 2.4 seconds and the SCN filled in 2.5.

The simulations show similar flow structures at the same times across the three substances. These

structures resembled those shown in Figure 2-2.

37

(a)

(b)

(c)

(d)

Figure 2-5: Simulation results of SCN, aluminum, and water at a) 0.7s b) 0.9 s c) 1s d)

1.2s.

38

2.3.2 Importance of Dimensionless Number Similarity in water Testing

Using Bernoulli’s Equation, the initial velocity and the velocity at the base of the sprue

were estimated for the 1995 study [3]. These values were found to be 0.886 m/s and 3.687 m/s

respectively. This means an average velocity of 2.287 m/s over the length of the sprue.

Figure 2-6: Expected aluminum fluid properties.

The camera from the study was unable to capture the entire mold in a single frame. From

the images in Figure 2-2, it appears that the sprue filled in the first 0.24 seconds and the mold was

fully filled in about 2 seconds.

2.3.2.1 Importance of Froude’s number

The time to fill, average sprue velocity, and average runner velocity was calculated for

each test. Due to limitations imposed by the testing procedure, the molds were not completely

39

filled by the time the pour basin was emptied. Therefore, “filled” was defined as the time at

which the fluid level rose above the 12-inch mark across the width of the plate. This left room for

quantitate error as the fluid level did rise uniformly across the plate width. Additionally,

limitations in the camera field of vision meant that the entire mold could not be viewed in one

shot. Therefore, time calculations were taken from when the fluid reach the 7-inch mark in the

sprue. The average sprue velocity was calculated from the time it took the fluid to travel from the

7-inch mark to the 14-inch mark in the sprue. Similarly, the average runner velocity was

calculated from the time it took the fluid to travel from the 2 to the 6-inch mark on the runner.

These velocities denote purely vertical velocity in the sprue, and horizontal velocity in the runner.

These quantitative results showed that the higher value Froude’ number test (unmatched)

exhibited higher velocities and a shorter fill time than the lower Froude’s number test (Table 2-9).

These results agree with the computer model.

Table 2-9: Froude’s number match vs unmatched.

Test Time to fill

(seconds)

Average Sprue

Velocity (m/s)

Average Runner

Velocity (m/s)

Fr matched 1.690 1.976 1.404

Fr mismatched 1.513 2.143 2.425

Qualitative analysis was able to offer greater insight on the fluid profile comparison for

each test, as well as if either test matched the 1995 aluminum images [3]. These images (Figure

2-7) show that the higher Fr value test appeared to carry more kinetic energy throughout the

filling. As both tests featured a similar Reynold’s numbers, it was hypothesized that the higher

viscosity would dampen the increased energy, thus preserving the fluid profile. That was not the

case.

40

(a)

(b)

(c)

(d)

Figure 2-7: Water testing results for Fr matched (right) vs Fr unmatched (left) a) 0.5

seconds after pulling plug, b) 0.74 s c) 1 s, d) 1.24 s.

Neither test perfectly matched the results of the 1995 study [3]. This is most evident at

the 0.5 second mark. In the aluminum test, the melt traversed the base of the entire runner, and

did not explode into the plate area until after rebounding off the end of the runner. For both water

tests, the water turbulently filled the runner as it traversed from left to right. The water entered the

plate area simultaneously as it did this. Better agreement between the two studies was seen for the

41

following three-time frames; however, the water appears both water tests appear to lead the

aluminum test in filling rate.

2.3.2.2 Importance of Reynold’s number

Higher levels of agreement were seen between the water tests at different Reynold’s

numbers (see Table 2-10). The tests differed in filling time by 0.05 seconds and featured identical

average sprue velocities. These results align with the computer model and the predictions of

Bernoulli’s Equation.

Table 2-10: Reynold’s number match vs unmatched.

Test Time to fill

(seconds)

Average Sprue

Velocity (m/s)

Average Runner

Velocity (m/s)

Re matched 1.690 1.976 1.404

Re mismatched 1.743 1.976 1.482

The qualitative analysis also shows good agreement between the two tests despite the

roughly 13,000 difference in Reynold’s number (Figure 2-8). As in the Froude test, both water

tests had the same disagreements with the aluminum test. Reynold’s number is thought to be less

significant at values greater than 105. All tests in this thesis are below that value; however, all

water tests have been subjected to the same mold and same mold roughness. Given this, a

difference of 13,000 in the Reynold’s number will have minimal effect for these tests. The sand

mold has a higher roughness value than the acrylic and therefore has a larger impact on the flow

despite matching the Reynold’s number. This, along with back pressure consideration, may

explain the differences in the flow profile between the two substances.

42

(a)

(b)

(c)

(d)

Figure 2-8: Water testing results for Re matched (right) vs Re unmatched (left) a) 0.5

seconds after pulling plug, b) 0.74 s c) 1 s, d) 1.24 s.

2.3.3 Succinonitrile Comparison to Water and Aluminum

The succinonitrile had a slower filling time than both the water and the aluminum tests

despite having a faster sprue and runner velocity than the water (Table 2-11). The SCN tests had a

43

Reynold’s number less than a quarter of the other tests in this study due to the thermal limitation

of the acrylic mold. Similarly, the SCN had a dynamic viscosity that was two to four times that of

the water (depending on the water temperature). The higher viscosity and lower Reynold’s

number likely contributed to the SCN filling the plate area in a slower, more laminar fashion.

Table 2-11: Succinonitrile filling results.

Test Time to fill

(seconds)

Average Sprue

Velocity (m/s)

Average Runner

Velocity (m/s)

SCN 1 2.037 2.134 1.778

SCN 2 2.107 2.134 2.319

SCN 3 2.353 2.134 1.524

Qualitative results of SCN (Figure 2-9) showed higher agreement with the aluminum test

than any of the water tests. Again focusing on the 0.5 second frame, it was seen that the SCN

traveled across the runner in the same manner as the aluminum, and entered the plate area only

after rebounding off the end of the runner. The manner in which the SCN entered the plate

showed similar shape as the aluminum (see Figure 2-2), but was not as distinct in height or shape.

44

(a)

(b)

(c)

(d)

Figure 2-9: Succinonitrile testing results a) 0.5 seconds after pulling plug, b) 0.74 s c) 1 s,

d) 1.24 s.

Given these results, further insight was needed as to how the greatest similarity occurred

between the aluminum and SCN tests despite the significant gap in Reynold’s number. Insights

were obtained through the Moody diagram (Figure 2-10), which relates Darcy-Weisbach friction

factor to Reynold’s number. The aluminum was calculated to have a Re value of about 28,000,

45

therefore, the water tests were matched to that value. Both tests occurred at the same x-position

on the Moody diagram, however, the relative roughness values differed. The water and SCN tests

occurred in polished machined acrylic, which was assumed to be “smooth”. The aluminum tests

occurred in a sand mold which had a higher relative roughness. This difference in relative

roughness created a larger pressure drop in the aluminum, altering its flow geometry. “Smooth”

walls express a near constant liner decrease in pressure drop as Re values increase. SCN tests

occurred at a lower Re value which exhibited a larger pressure drop. The pressure drop during

SCN testing correlated to a relative roughness value roughly 6 times larger than that of the water.

From this, it was concluded that pressure drop due to friction is vital to producing flow similarity

between two substances.

Figure 2-10: Moody diagram [68] 1) Water Re 28,000 in acrylic 2) SCN Re 6,800 in

acrylic 3) Relative roughness of substance at Re 28,000 correlated to pressure drop incurred by

SCN.

46

Conclusions

This study examined the need for dimensionless number similarity in water testing for

metal casting purposes, water testing’s ability to recreate a previously recorded aluminum pour,

and succinonitrile’s ability to recreate the same aluminum pour. All of these tests were conducted

ignoring wall roughness and permeability effects. The following conclusions were deduced:

• Froude’s number similarity is imported for mold filling testing in order to

preserve the fill rate and energy with which the fluid enters the mold across all

tests.

• Reynold’s number was not found to have a direct result on the fluid profile.

Rather, pressure drop as a function of wall roughness ands a function of

Reynold’s number is speculatedargued to have a more significant effect.

Reynold’s number similarity may play a larger role at lower values than those

depicted in this study.

• Water was able to roughly mimic the aluminum test, but an exact match was not

able to be produced regardless of Reynold’s and Froude’s number similarity.

• Succinonitrile was able to mimic the aluminum test better than the water tests

conducted in this study. This was despite the SCN having a Reynold’s number

less than 25% of the aluminum. It is speculatedargued that the lower Reynold’s

number at which the SCN tests occurred wasproduced a better representation of

the aluminum given the higher wall roughness of the sand mold compared to the

acrylic.

This study proposed an argument for succinonitrile as a better metal analog than water.

The results of this study found that SCN was able to mimic the flow pattern of aluminum as well,

if not better, than water. The testing process used in this study consisted of forming a mold made

47

of acrylic plastic, and pouring liquid SCN through the mold. The filling profile was filmed with a

high-speed camera. The advantages of this form of testing are as follows:

• The ability to cheaply visualize experimental casting flow.

• The ability to visualize experimental casting research in a safe lab setting as

opposed to a foundry setting.

• The ability to rapidly and cheaply create and test various gating geometries

• The ability to reuse the test mold multiple times

• The ability to quantify flow velocity through high-speed video imaging

This study proved that testing in this manner was not only possible, but was able to

produce beneficial results. Despite this, there are still flaws with this form of testing:

• The health hazards of associated with succinonitrile means it is best practice to

handle inside a fume hood.

• The limited thermal capacity of the acrylic prohibited the SCN from being heated

to desired viscosity. Attempts to use higher temperatures lead to severe cracking

of the mold.

• The material properties of the acrylic lead to a long solidification time for the

SCN (~40 minutes). This is frustrating for running multiple tests in succession.

Additionally, these properties limit the ability to match solidification time to that

found in sand casting.

• Cleaning the mold between tests is difficult. If opened prior to complete

solidification, the SCN will leave larger amounts of residue.

• Creating proper alignment and sealing of the mold can be difficult.

• Creating molds with more 3D or thicker features can add cost due to the cost of

thicker acrylic

48

The ultimate goal of succinonitrile testing is to allow for advancements in three areas:

flow visualization and quantification, advanced gating geometries, and experimental flow and

solidification merging. The limiting factor to these advancements is the acrylic mold. As shown

in this study, the acrylic did not allow for the desired thermal and solidification conditions of

SCN to be met. A material is needed to that allows for visual transparency, but is also able to

meet the thermal needs of this testing. Incorporating flow sensors into the mold may provide a

way to circumvent the need for optical transparency, and provide velocity values of the melt flow.

Experimentally measured velocity values would provide a reference for which numerical solvers

could be compared.

Wall roughness was shown to be a significant factor in matching the sand mold flow to

the acrylic mold flow. Wall roughness similarity may be met through the addition of baffle

geometries on the acrylic mold wall. These baffles could be correlated to the roughness value a

sand mold. Further work would be needed to properly design the baffles, and it is unknown if this

system would come as a detriment to visual transparency. Another key mold characteristic,

permeability, allows for trapped air to escape the sand mold and reduce back pressure against the

incoming melt flow. Permeability similarity between the sand and acrylic molds may be met

through a vent placed in the top of the acrylic mold. The vent size would need to be correlated to

the sand permeability.

The ability to 3D print gating geometries via stereolithography printing (a vat

photopolymerization process) is hypothesized to offer potential for studying advanced 3DSP

gating geometries. Current limitations exist when it comes to the thermal aspects of the print

material, which are similar to those of acrylic. Many stereolithography printers also feature small

build volumes (roughly 5 in. by 5 in. by 5in.) that limits the ability and time needed to print full-

sized gating systems. Therefore, the priority for continuing this form of SCN testing should be to

49

perfect the practice in reference to traditional green sand molds. Once successful, 3D printed

elements may be incorporated.

In the next chapter, Chapter 3, the research focus changes from melt flow in 3DSP molds

to 3DSP properties affecting casting solidification. The thermal properties are comprised of

density, specific heat, and thermal conductivity of the molding material. These three properties

are needed for analytical and numerical prediction of casting solidification time. The study

presented in chapter 3 attempted to quantify these values for 3D printed sand molds as no values

had been published. An additional focus of this study analyzed the effects of binder content on

mold thermal properties. Results from this study can be found at the end of chapter 3.

Chapter 3

Fundamental Study on 3D Sand Printed Molds: Thermal Properties

3.1 Introduction

The rate of heat transfer between a metal and the mold is an important factor in

determining the metallurgic structure of a casting during solidification. Many common casting

alloys are sensitive to this cooling rate [69], therefore, accurate representation of this heat transfer

is critical to modeling casting success. The rate at which heat moves move from molten metal to

the molding material is largely determined by the thermal properties of the metal and the mold.

The growing popularity of 3D sand printing (3DSP) has created need for numerical simulation

devices to expand their databases. Most commercial numerical solvers provide easy to access

properties for a range of permanent molds, various types of sand molds, and investment molds.

The properties of 3D printed sand molds need to be added to this group; however, the manner in

which a 3DSP mold is printed may cause variation of these properties. Specifically, the main

ingredients of a printed sand mold, silica sand and furan resin, can vary in proportion. The effect

this variation in ratio has on the solidification time of a casting has not been explored.

Chvorinov’s rule (Equation 3.2) is a common analytical approach to determining the

solidification time of a casting [14,64,70]. This expression relates the solidification time of a

casting to the ratio of the casting volume to surface area. The mold constant (b), expressed in

units of s/m2, accounts for thermal properties of the mold and the pour material. Measuring the

solidification time of a simple casting geometry can provide insight to the mold thermal

properties in the form of mold constant.

51

𝑇𝑠 = 𝑏 (𝑉

𝐴)

2 (3.1)

If properties of the pour material and pour temperature are known, the mold constant can

be reduced to the coefficient of heat accumulation, bf (Equation 3.3). This term isolates the

density, thermal conductivity, and specific heat of the molding material. These three mold

properties are significant to quantifying heat transfer between mold and melt. Therefore, inputs

for these properties are needed for most commercial simulation devices.

𝑏𝑚 = √𝜌𝑚𝑐𝑚𝑘𝑚 (3.2)

The thermal properties of 3D printed sand molds are hypothesized to differ from those of

traditional green sand molds. Additionally, these values are hypothesized to vary as the ratio of

sand to binder changes. The manufacturing process of a 3DSP mold is unique to other sand molds

used in the casting industry. Traditional molding methods involve compacting sand around a

pattern. 3DSP molds, by definition of an additive manufacturing process [18], are assembled in a

layer by layer fashion. The specific process, binder jetting, utilizes a mechanical arm to spread a

layer of sand on a print table. Next, a binder head traverses the sand layer, depositing binder only

in areas desired to be bonded. An acid catalyst in the sand reacts with the liquid binder which

solidifies the sand and binder. This process is repeated in the vertical direction until the desired

height of the mold is reached. It is unknown how this manufacturing process impacts the thermal

properties of the mold compared to traditional methods of mold making.

The binder jetting process offers manufacturers some degree of customization in the form

of sand layer height and amount of binder applied. These factors were hypothesized to impact the

properties of the mold. To gain insight on how variation in mold thermal properties effect casting

solidification time, a 33 ANOVA study was conducted in the simulation program SOLIDCast [11]

(Table 3-1). This study simulated a 100mm cube of 308 aluminum in a green sand mold with

52

100mm walls. The study varied the three physical mold properties on three levels: 1 – 25% below

the default value, 2 – the default value, 3 – 25% above the default value.

Table 3-1: Simulation results for thermal property study.

Thermal

Conductivity (K)

Density (ρ) Specific Heat (Cp) Solidification Time in

Minutes

1 1 1 38.91

2 1 1 29.38

3 1 1 23.63

1 2 1 33.06

2 2 1 25.00

3 2 1 20.14

1 3 1 28.77

2 3 1 21.78

3 3 1 17.56

1 1 2 33.10

2 1 2 25.00

3 1 2 20.14

1 2 2 27.67

2 2 2 20.97

3 2 2 16.90

1 3 2 23.87

2 3 2 18.08

3 3 2 14.60

1 1 3 28.83

2 1 3 21.83

3 1 3 17.60

1 2 3 23.90

2 2 3 18.13

3 2 3 14.64

1 3 3 20.50

2 3 3 15.55

3 3 3 12.58

The solidification time results ranged from 12.58 minutes at the all high setting to 38.91

minutes at the all low setting. This range in solidification time would affect gating design,

strength of the casting, surface finish, and shake-out time. A general liner model was produced.

This model found that all contributing mold factors were of equal significance with a p-value of

53

zero (Table 3-2). The results of this ANOVA study showed that improper representation of mold

properties produces significant variation in numerical solidification time prediction. Therefore, it

was necessary to gather data for a variety of 3D printed sand molds to ensure accurate modeling

in the foundry industry.

Table 3-2: ANOVA results for thermal property study.

Many studies have attempted to measure thermal properties in molding materials,

including resin bonded sand, with varying degrees of success [69–75]. One of the main reasons

for the difficulty of this measurement is binder decomposition which occurs at ~ 500 °C, well

below the pour temperature of nearly all casting alloys. Many measurement techniques are

available for determining the thermal properties of materials. The laser flash technique is popular

for measuring thermal diffusivity, but it was determined that this technique was not suitable for

testing bonded sand samples due to the preparation requirements. Laser flash requires a smooth,

polished sample from which the laser can be reflected. The sample would also need to be coated

with conductive material [76]. Additionally, this study would provide a value for thermal

diffusivity which is a product of thermal conductivity, density, and specific heat. Accurate values

for density and specific heat would need to be known to extract the value for thermal

conductivity.

Next, a one-dimensional steady heat flux set-up was considered. Zych et al. [69] used this

method to quantify thermal conductivity vs temperature in a verity of resin bonded mold

Source DF Adj SS Adj MS F-Value P-Value

K 2 576.24 288.121 256.74 0.00

p 2 237.86 118.929 105.98 0.00

Cp 2 234.64 117.319 104.54 0.00

Error 20 22.44 1.122

Total 26 1071.18

54

materials. This set-up was unable to overcome the binder burnout issue and values for thermal

conductivity were only reported for 150 °C to 450 °C. A new testing set-up was created that

provided opportunity to calculate thermal conductivity through unsteady heat transfer analysis.

This set-up used thermal couples to capture temperatures for the molding sand during the pouring

of grey cat iron at 1400 °C. The equation to categorize thermal conductivity through this form of

analysis is given by Zych el at as:

𝐾 = (𝐴 + 𝐵 ∗ 𝐶) ∗ 𝑚 (3.3)

where m is the cooling rate, C is the heat storage capacity of the tested material and A and B are

constants categorizing the testing set-up. Each of these symbols is a representation of a

mathematically complex structure. Through this analysis, thermal conductivity values of 1.0 to

1.4 W/m-C were reported for reclaimed silica sand with up to 1.0% furan binder over the range of

0 to 1,000 °C.

Finally, the inverse approach to finding thermal conductivity was also considered. The

inverse method uses numerical simulation to recreate an experimental set-up. The experiment is

often the pouring of a casting with embedded thermocouples. Using know properties of the

experiment, a small group of unknown variables can be manipulated until agreement is found

between numerical and experimental results. This process requires a simulations software capable

of accurately recreating the real-world experiment on a detailed level. MAGMAsoft [9] has proven

to be successful in past studies[72,73].

This thesis chose to pursue thermal properties for 3D printed sand molds at 6 different

binder levels. The effects of this variation were recorded in the forms of mold tolerance, density,

specific heat, casting solidification time, and heat transfer through the mold. Thermal properties

were determined through a combination of mathematical equations and individual tests which

included the use of a helium pycnometer, thermal DSC, and a foundry pour test of 99.9 % pure

55

aluminum. Detailed descriptions of how these tests were conducted in are presented in section 3.2

while results and discussion are presented in section 3.3.

3.2 Materials and Methods

3D printed sand samples were acquired from Hazleton Casting Company in Hazleton,

Pennsylvania. The samples were made by a Viridis3D sand printer manufactured by Envisiontec

[25]. The 3D printed sand was comprised of silica sand GFN 65 round or subangular, a dry

premixed acid catalyst, and furan binder. As the goal of this project was to analyze heat transfer

effects due to changing binder ratio in the printer, a select group of parts (Table 3-3) was printed

at 6 different binder settings. These settings were categorized by the manufacturer as pulse

settings 1.5, 2, 2.5, 3, 4, and 5. Each pulse setting has an expected percent binder by mass of sand

ratio. These expected values were 1%, 1.2%, 1.8%, 2.1%, 2.5%, and 3% respectively (Table 3-3).

Loss on ignition (LOI) testing was conducted to verify the amount of binder present in each pulse

level.

56

Table 3-3: 3D sand printed samples.

Part Dimensions

(mm)

Build Volume

(in3) Test

Pour Cup 130 x 130 x 105 108.29

Solidification Time

Mold Heat Transfer

Large

Cylinder 100Ø x 50 30.51 N/A

Medium

Cylinder 19Ø x 19 0.42

Loss on Ignition

Density

Small

Cylinder 12.6Ø x 2.5 0.02 Specific Heat

3.2.1 Tolerance Measurements

Adherence to tolerance is a common study for all additive manufacturing processes.

Often these restrictive design studies offer insights into minimum feature size and directional

printing effects on part tolerance. This study was not designed for this analysis, nor is it the first

to investigate tolerance in 3D sand printers [77]. Despite this, the collected data is relevant as the

Viridis3D printer is less common in research than the Exone S-Max [23]. The data correlates

tolerance-to-binder ratio, and the three cylindrical geometries provided opportunity for a size-to-

tolerance correlation.

All measurements were taken with a Husky digital caliper in millimeter units. Loose sand

provided a common source of error as a single sand grain could alter measurement by as much a

0.2 mm. A faro arm would be recommended for a more detailed study. All cylindrical geometries

57

were measured in the same manner. The diameter measurement was taken at the midline of the

part. The part was then rotated 90 degrees and measured again. These diameter measurements

were averaged to represent the overall dimeter of the part. The length of the cylinder was then

measured. The jaws of the calipers were wiped between measurements to reduce the interference

of loose sand. A total of 9 samples were measured at each binder level for the 19 mm X 19 mm

samples and the 12.6 mm X 2.5 mm samples. In addition, 4 samples at each level were measured

for the 100 mm X 50 mm samples. The measured values for all samples were averaged on a per

binder level, resulting in an overall average diameter and length value for each part size per

binder level. This data was viewed both as the average measured value and the average tolerance

value relative to the nominal dimensions presented in Table 3-8. The latter was used in the

software Minitab for an ANalysis of VAriance (ANOVA) study relative to part size, binder level,

and diameter versus length.

The pour cups were also measured in a similar manner; however, these values were not

included in the ANOVA study. The interior width of the cup was measured, rotated 90 degrees,

and measured again. These two values were averaged like the diameter measurements. The

exterior dimensions of the cup were measured in the same way. The interior depth and overall

height of the cups were also measured. A total of 4 cups were measured at each binder level with

the data being compiled such as the cylindrical geometries.

3.2.2 Loss on Ignition Testing

Loss on Ignition (LOI) testing was conducted in ordinance with AFS Standard

5100-00-S [78]. Six ceramic crucibles were collected and weighed on a digital scale.

One 19 mm X 19 mm sample was placed in a crucible, each one being of a different

binder saturation. The crucible-sample combinations were weighed again and recorded.

58

The samples were placed in an oven at 107 °C for two and a half hours (Figure 3-1).

This allowed for the evaporation of moisture in the sample (occurring at 100 °C). The

samples were weighed at the 1.5- and 2.5-hour marks. Lack of change in sample weight

between measurements proved that all moisture had been removed from the samples.

Figure 3-1: LOI crucibles in oven.

The second part of AFS Standard 5100-00-S [78] calls for the sampled to be

returned to the oven at 986 °C. Furan resin evaporates at approximately 500 °C,

therefore, the increased heat was used to remove the binder from the bonded sand

samples. Samples were weighed at the 1, 2, and 3-hour marks. Lack of change in

sample weight between measurements proved that all binder had been removed from

the samples. The amount of binder in the samples was quantified by taken the

difference in weight of the samples after the 3-hour mark at 986 °C and after the 2.5-

hour mark at 107 °C.

59

3.2.3 Density Testing

Density measurements of 3D printed sand samples were taken using a micrometrics

AccuPyc II 1340 helium pycnometer. This device used the gas displacement technique to

accurately measure sample volume. Each sample was sealed in the instrument compartment of the

pycnometer, and helium was used as the displacement medium. The instrument compartment was

of a known, precise volume. Helium, when pumped in, filled all the remaining voids (as small as

one angstrom) in the chamber not taken up by the sand sample. Then, the helium was discharged

into a second chamber of known volume. The computation of sample solid phase volume was

made possible through measuring the pressure of the helium in the second chamber.

Prior to testing, the 19 mm x 19 mm sand samples were shaved with a metal file until

they were small enough to fit in the sample cup. Samples were measured after this alteration with

all samples falling below 16.5 mm in dimeter and 17.5 mm in length. The sample cup was

cleaned and placed on a scale. The scale was tarred and the sample was placed in the cup. This

provided the sample weight. Entering this value into the pycnometer software allowed for the

calculation of the sample density. This process was repeated for three samples at each of the six

tested. In total, eighteen sand samples were measured for density.

Density measurements were also taken for the raw sand and liquid binder components.

Three measurements were taken for each in the same manner as the bonded samples. Between

tests, the liquid furan binder was emptied into a waste container. The sample cup was rinsed with

ethanol, wiped with a paper towel, and air dried with compressed air.

60

3.2.4 Cooling Curve Analysis

Cooling curves of 99.9% pure aluminum were measured in printed cooling cups of

various binder ratios to analyze the effect binder percentage had on solidification time. The

99.9% pure aluminum was chosen as the melt material to produce the cleanest cooling curve as

possible in comparison with more popular aluminum casting alloys. An open top pour cup was

designed to hold an 80mm (~3 lb.) cubic aluminum charge (Figure 3-2). The cup featured 25mm

thickness walls on all vertical sides of the cube as well as the bottom. Four TL1815 k-type

thermocouples rated for a maximum temperature of 800 °C at an accuracy of +/- 0.1 °C were

used to measure temperature at various points in the mold. These probes were 5 mm in diameter

at 100 mm in length. All probes were placed in the same plane at 40mm from the top of the cup.

The probes were placed so that the probe tip measured temperature at the center of the casting, in

the casting at 5 mm from the mold wall, in the mold wall at 5mm from the casting, and in the

mold wall at 5 mm from ambient air (Figure 3-3).

Figure 3-2: Pour cup dimensions.

Figure 3-3: Pour cup experimental set-up.

61

Temperature versus time data was collected in a LabView script via a National

Instruments cDAQ-9171 with a NI-9212 insert. This DAQ was capable of supporting 8

thermocouples at one time allowing for two molds to be tested each pour (Figure 3-4). Samples

were taken at a rate of 10 Hz.

Figure 3-4: Cooling curve experimental set-up.

The temperature of each mold was taken with an OMEGA OS204 infrared thermometer

prior to pouring. Molds were placed in different locations on the sand bed for each pour to reduce

preheating of the molds from earlier tests. 50 lbs. of 99.9% pure aluminum was heated to 815 °C

in an induction furnace. This temperature was verified via a thermo-probe prior to the melt being

emptied into the pour ladle. The melt was tested again once in the ladle. The melt was allowed to

cool to 765 +/- 4 °C and then immediately poured into the mold. Data was collected on two molds

62

at a time; however, the molds were poured individually to verify pour temperature. Data

collection was started prior to the melt being poured and continued through the liquid cooling

phase, liquid-solid transition, and solid cooling. When the temperature reading at thermocouple D

became constant or began to decrease, data collection stopped. A large amount of shrinkage

occurred in each casting (Figure 3-5).

Figure 3-5: Cooling cup after pouring exhibiting shrinkage.

After both castings solidified, six thermocouples were removed and used in the

succeeding rounds of testing. The two probes in the center of the casting were not retrieved do to

being solidified in the castings. These two thermocouples were replaced in the DAQ and the

process was repeated for the remaining molds. Ten molds were poured in total, two for each pulse

value of 1.5, 2, 2.5, 3, and 4. The pulse 5 molds were not used for this study due to poor print

quality.

3.2.5 Specific Heat Testing

Specific heat testing was conducted according to ASTM designation E1269-11 [79]. A

three-run method was used to find the specific heat of a pulse 2 bonded sample. This method

63

compared the heat flow results of the sample to a blank and a sapphire sample. The blank,

sapphire, and sand sample underwent the same conditions inside a TA DSC Q2000 (Figure 3-6)

that included remaining isothermal at 0 °C for fifteen minutes, ramping to 300 °C at a rate of 20

°C/min, and holding isothermal at 300 °C for fifteen minutes.

Figure 3-6: TA Instruments DSC Q2000.

For this type of analysis, the blank should produce a positive or nearly zero constant heat

flow. The sapphire should produce a negative heat flow value. The test sample heat flow (in this

case the sand) should fall between the curves of the blank and the sapphire. The specific heat of a

given test sample can then be found from the equation:

𝐶𝑝𝑠𝑎𝑚𝑝𝑙𝑒 = 𝐶𝑝𝑠𝑎𝑝𝑝ℎ𝑖𝑟𝑒 ∗𝐷𝑠𝑎𝑚𝑝𝑙𝑒𝑀𝑠𝑎𝑝𝑝ℎ𝑖𝑟𝑒

𝐷𝑠𝑎𝑝𝑝ℎ𝑖𝑟𝑒𝑀𝑠𝑎𝑚𝑝𝑙𝑒 (3.4)

where Msample is the mass of the test sample in mg, Msapphire is the mass of the sapphire sample,

Dsample is the difference in heat flow between the sample and the blank in mW, and Dsapphire is the

difference in heat flow between the sapphire sample and the blank.

64

3.3 Results

3.3.1 Loss on Ignition Results

Loss on ignition testing quantified the amount of water, binder, and sand for pulse values

1.5 – 5. These percentages are listed in Table 4-4. It can be seen that none of the resulting binder

values matched the expected values supplied by the manufacturer. The pulse 2.5 sample was

found to have a higher level of binder than the pulse 3 sample. Little difference was found is the

content of binder for pulse values 2-4. Water content was found to be insignificant in each sample

measuring consistently below 0.2%.

Table 3-4: Loss on ignition results.

Pulse Expected

Binder %

% Water % Binder % Sand

1.5 1.0 0.11 1.49 98.40

2 1.2 0.11 1.82 98.08

2.5 1.8 0.14 1.95 97.91

3 2.1 0.11 1.93 97.96

4 2.5 0.11 2.08 97.81

5 3.0 0.17 2.74 97.09

3.3.2 Tolerance Results

For the cylindrical samples, the ANOVA results found binder content to have a significant impact

on part tolerance while size, and diameter vs length did not (Table 3-5). This result was expected

and the only recognizable trend from this data. All length measurements for this set of samples

were found to be greater than the nominal value (Figure 3-7) while the same was true for

diameter measurements except for two of the three pulse 1.5 sets (Figure 3-8).

65

Table 3-5: Analysis of variance results for cylindrical geometries.

Source DF Adj SS Adj MS F-Value P-Value

Size 2 0.3803 0.1901 1.06 0.361

Binder 5 6.5531 1.3106 7.29 0

Direction 1 0.2775 0.2775 1.54 0.225

Error 27 4.8515 0.1797

Total 35 12.0624

Figure 3-7: Graph of length deviation from nominal value versus pulse value.

Figure 3-8: Graph of diameter deviation from nominal value versus pulse value.

66

The pour cups offered some observational insights prior to measurement. Each cup was

supposed to be printed with a 2 mm diameter through hole; however, this geometry was not

produced on any of the cups regardless of binder level. Also, two of the four pulse 5 cups were

scrapped due to pour print quality (Figure 3-9).

Figure 3-9: Defective pulse 5 pour cups.

The measurements of the remaining samples showed that pulse 5 had the worst tolerance

accuracy while 1.5 and 3 had the best among tested values (Figure 3-10). Out of the four

measured dimensional aspects, the depth of the pour cup had near perfect dimensional accuracy

omitting pulse 3 and 5. No other trends could be deduced from this data.

Figure 3-10: Graph of deviation from nominal value versus pulse value for pour cups.

67

3.3.3 Density Results

Density measurements were found to decrease as resin content increased (Figure 3-11).

This result was expected as the sand had a higher density than liquid resin. The range in measured

density was found to be small ranging less than +/- 1% from the median value. A liner trendline

was fitted to the data with an R2 value of 0.9931. This trendline was able to correlate printed

density to resin percentage in the sample using the equation:

𝜌 = −3.4222𝑥 + 2.6659 (3.5)

where x is the concentration of binder in the sample.

Figure 3-11: Graph of printed sample density versus binder content.

Measurements were taken for the raw material inputs resulting in an average density of

2.6510 g/cm3 for sand across three samples and 1.1424 g/cm3 for liquid furan binder. It was

proposed that the density values of the raw material could be used to predict the density of the

printed sample in the form of:

𝜌𝑝𝑟𝑖𝑛𝑡𝑒𝑑 𝑠𝑚𝑎𝑝𝑙𝑒 = 𝑥𝜌𝑠𝑎𝑛𝑑 + 𝑦𝜌𝐵𝑖𝑛𝑑𝑒𝑟 (3.6)

68

While error percentages were below 2%, an exact match between the measured printed

density values and those formulated from the raw materials and LOI results were unable to match

(Table 3-6). Matching values would have required the resin content to be as much as doubled in

some cases, or for the resin density to be less which for some calculations meant negative.

Table 3-6: Calculated density values compared to measured values.

X Sand Y Furan Calculated Density

Measured Density % error

0.985 2.6510 0.0149 1.1424 2.6285 2.6145 0.53

0.982 2.6510 0.0182 1.1424 2.6236 2.6051 0.71

0.981 2.6510 0.0195 1.1424 2.6216 2.6005 0.81

0.981 2.6510 0.0193 1.1424 2.6219 2.5981 0.91

0.979 2.6510 0.0208 1.1424 2.6196 2.5946 0.97

0.973 2.6510 0.0274 1.1424 2.6096 2.5721 1.46

Factors not considered by this model include the amount of catalyst in the sand and the

effects of the chemical reaction between the liquid binder and catalyst. The former could be found

through LOI testing of the printer sand; however, it is assumed this value is low. Density testing

of the sand post LOI could also give insight to the individual densities of the sand and catalyst

components.

3.3.4 Specific Heat Results

Heat flow and temperature signals were extracted from the DSC software for the blank,

sapphire, and a pulse 2 sand sample. These signals can be viewed in Figure 3-12. The mass of

sapphire used in testing was 61.3 mg, and the mass of pulse 2 sample used was 22.5 mg. Values

for the specific heat of sapphire (in J/g-K) were provided by ASTM E1269-11[79] in 10 Kelvin

increments. These values were used in conjunction with Equation 3.4 to calculate specific heat

values of the pulse 2 bonded sand sample over the range of 6.85 °C to 296.85 °C (Figure 3-13).

69

Specific heat of the pulse 2 sample increased from 0.71 J/g-C at 16.85 °C to 1.03 J/g-C at 296.85

°C. The graph trend suggests that the specific heat would continue to increase at higher

temperatures.

Figure 3-12: Graph of heat flow versus temperature for blank, sapphire, and pulse 2

sample.

Figure 3-13: Graph of specific heat versus temperature for pulse 2 sample.

70

3.3.5 Cooling Curve Results

Thermocouples were placed at four different locations in the pour cup to record the

temperature versus time data in the casting and the sand mold. An example of these temperature

curves for a single mold is presented in Figure 3-14. The hottest curve data came from

thermocouple couple “A” (see Figure 3-3), which collected data at the center of the casting. This

point was suspected to be the last to solidify in the casting, therefore providing the best cooling

curve representation.

Figure 3-14: Temperature vs. time from pour graph for thermocouples in pulse 4 mold.

Cooling curves for elemental substances have three unique phases. The first phase has a

negative linear slope representing the cooling of the liquidous material. This is followed by a

linear region of zero slope during which the substance converts from liquid to solid. Finally, a

second negative linear slope region shows the cooling of the solid material. The solidification

time was calculated as the time from pour until the end of the liquid-solid transition phase. The

cooling curves for each pulse value were plotted on the same chart of temperature versus time

(Figure 3-15). Cooling curve analysis showed that binder level had little to no effect on cooling

rate of the casting. The cooling curves for pulse values 2 – 4 were found to be near identical,

71

reporting a solidification time of about 730 seconds. The pulse 1.5 value was found to take about

40 seconds longer in the transition region. The cooling rates prior to the transition region were

found to be identical for all pulse values. Cooling rates after the solidification region were also

shown to match with high agreement.

Figure 3-15: Graph of temperature vs. time at casting center for pulse values 1.5 – 4.

The temperature for the thermocouples placed in the mold at locations “C” and “D” were

also plotted (Figure 3-16). These curves also showed little to no variation across the testing pulse

values. This finding agrees with the solidification curve results in Figure 3-15, and implies that

binder content does not affect heat transfer between the casting and the 3D printed sand mold.

Furthermore, it is implied that the no significant change to the thermal properties of 3D printed

sand molds occurs with variation in binder content.

72

Figure 3-16: Plot of temperature vs time from pour for thermocouples placed in the mold

at 5mm and 20 mm from the casting for pulse values 1.5 - 4.

It can also be seen in Figure 3-16 that neither thermocouple reached 500 °C, the

temperature of binder decomposition [69]. This shows that binder decomposition only occurred

inside of 5mm from the casting. This information could be used to model the mold as two

separate regions, one for the given properties of 3D printed sand, and one for the binder burnout

area. SDT testing may be able to quantify the amount of energy absorbed by binder transition

from solid to gas. A single SDT test was conducted over the span of 0 °C to 1400 °C for a pulse 3

sample (initial sample weight of 21.938 mg). Figure 3-17 shows a decrease in heat flow (W/g)

from 500 °C to 1400 °C (the temperature range at which binder burnout would occur). These

results are scalable to the amount of pulse 3 sand effected by burnout to predict the total amount

of heat absorbed by the evaporation of the binder.

73

Figure 3-17: Plot of weight percentage and heat flow versus temperature

3.3.6 3D Printed Sand Mold Solidification Modeling

The solidification of the 80 mm cubic aluminum castings poured in this study was

modeled using Chvorinov’s rule (Equation 3.1). The cooling curves in Figure 3-15 reported a

solidification time of about 730 seconds for pulse values 2-4. The castings were designed to have

a volume of 5.12E-4 m3 and a surface area of 0.0384 m2. Using these values in Equation 3.1

produced a mold constant (b) of 4,106,250 s/m2. Given the properties of aluminum and the pour

conditions used in this study (Table 3-7), the mold constant from Chvorinov’s rule was able to be

arrange in a manner that produced the coefficient of heat accumulation (bm). This value was

calculated to be 739.1 W-s0.5/m2-K.

Table 3-7: Properties used when calculating mold constant.

Property Value Reference

Aluminum Density 2700 kg/m3 [14]

Aluminum Latent Heat of Fusion 398,000 J/kg [14]

Aluminum Specific Heat 917 J/kg-C [14]

Aluminum Solidification Temperature 673 °C Observed

Mold Initial Temperature 23 °C Observed

Melt Superheat 92 °C Observed

74

Further insights about the thermal conductivity of 3D printed sand molds can be drawn

from the calculated coefficient of heat accumulation and the results for density and specific heat

found in this study. At ~25 °C, the pulse 2 sample was found to have a density value of 2,605

kg/m3 and a specific heat of 720.1 J/kg-C. From bf, the pulse 2 sample had a thermal conductivity

of 0.29 W/m-K.

The program SOLIDCast [11] was used to validate the analytical results. The casting

geometry was recreated in the software. Model inputs were changed to match those found in

Table 3-7 and the mold properties found in this study. The pour time was set to 2 seconds and

ambient temperature set to 23 °C. The external heat transfer coefficient representing natural

convention was left as the default vale of 8.5 W/m2-K. The mold was recreated to replicate that

used in the experimental study with 25 mm walls on all sides omitting the top. A mesh was

generated comprised of 100,000 nodes. For the calculated pulse 2 thermal properties, the

simulation results did not match the 12.167 min solidification time found in the experimental

cooling curves. The thermal conductivity value was iterated in the simulation until a solidification

time of 12.148 minutes was achieved at the casting center for a “k” value of 0.42 W/m-K. The

difference in calculated versus simulation prediction for thermal conductivity yielded a percent

error of 30.95 %. For comparison, the same simulation set-up was ran for the default green sand

thermal properties found in the SOLIDCast [11] database. Simulations were also conducted for a

25% increase and 25% decrease in these values. The 25% decrease failed to converge after a

prediction of over 100 minutes, yielding a percent error of greater than 83%. A percent error was

found to range between 25.86% and 32.92% for the default and 25% increase simulations (Table

3-8). The difference in thermal conductivity found between Chvorinov’s rule and simulation falls

with-in the error range. Further analysis is needed to confirm a thermal conductivity value of 3D

printed sand molds; however, these results suggest the value should lie between 0.29 W/m-K and

0.42 W/m-K.

75

Table 3-8: Comparison of solidification times predicted by Chvorinov’s rule and numerical

simulation.

Test Coefficient of

heat accumulation

(W-s0.5/m2-K)

Chvorinov

Prediction

(Minutes)

Simulation

Result

(Minutes)

Percent

Error

Default

(Green Sand) 982.6 6.884 9.286 25.86

Default +25% 1,375.7 3.512 5.235 32.92

Default -25% 638.2 16.318 >100 >83.68

3.4 Conclusion

A study was conducted to quantify the thermal properties of 3D printed sand molds and

propose an optimal binder content for the printing of such molds. 3D printed sand samples were

acquired in the forms of pour cups and cylindrical geometries each at 6 different pulse values.

Loss on ignition test was conducted to quantify the amount of binder in each pulse value. These

results showed binder content was higher than the manufacturer prediction for the lower three

pulse values and lower than manufacturer prediction for the higher pulse values. A dimensional

accuracy study showed increased binder content had a negative significant effect on part

tolerance. Density values were shown to decrease as binder content increased ranging +/-1% from

the median value of 2.600 g/cm3. The values of unbonded sand and furan binder where also

reported to have density values of 2.6510 g/cm3 and 1.1424 g/cm3 respectively. A solidification

analysis of 99.9% pure aluminum showed that binder content had no significant effect on casting

solidification time or heat transfer in the mold. The solidification time recorded in this test was

used with Chvorinov’s rule to predict a mold constant of 4,106,250 s/m2. Further examination of

this constant produced a coefficient of heat accumulation of 739.1 W-s0.5/m2-K. The specific heat

of a pulse 2 found to be 720.1 J/kg-C. This value, along with the measured density of a pulse 2

sample was used to derive a thermal conductivity value from the coefficient of heat accumulation.

This value was 0.29 W/m-K. The commercial simulation software SOLIDCast [11] was used to

76

recreate the aluminum solidification experiment. Using the measured density and specific heat

values for the pulse 2 mold, the simulation was able to reproduce the experimental solidification

of 12.167 seconds for a thermal conductivity of 0.42 W/m-K. The percent difference between

these values fell with-in the percent error range between Chvorinov’s rule and SOLIDCast [11]

simulation for other mold thermal properties.

Future work can be conducted to validate and extend the insights of this study.

Microstructure and strength analysis of the castings poured in this study will be used to verify the

conclusion that binder content did not have an effect on casting quality. The temperature data

collected from thermocouples may be used with a one-dimensional unsteady heat transfer

analysis to draw further insight to the thermal conductivity of 3D printed molds. This analysis is

limited to the 4 locations at which thermocouples were place to gather temperature data.

Additional tests may be needed if additional temperature locations are required. Similarly,

thermal conductivity may be found through the use of additional commercial simulation software;

however, this is limited by the cost to access these softwares. Finally, further tests are needed to

gather thermal property data for a range of temperatures. This study found density values at ~25

°C and specific heat up to 300 °C. These values will change with increased temperature which are

useful to the improvement of numerical models. Gathering this data will be difficult due to the

needed to hold the sample at a constant temperature in the case of density and the binder burnout

effect occurring at 500 °C Future models may benefit from separately modeling the areas affected

and unaffected by binder burnout. As shown in this study, analysis using a thermal SDT may be

the first step in this process. A recap of these conclusions is presented in chapter 4 along with a

final summary of the work conducted in this thesis.

Chapter 4

Conclusion

This thesis discussed two research studies derived from the capabilities of 3D sand

printing process in the metal casting industry. The first study, presented in chapter two,

introduced a new methodology for experimentally visualizing casting flow. This study was

motivated by the need for an inexpensive test to verify innovative, geometrically complex gating

systems possible with 3D sand printing. The second study, presented in chapter 3, quantified

thermal properties of 3D printed sand molds while simultaneously investigating the effects of

binder content on these properties. As discussed in chapter one, accurate solidification and flow

modeling are imperative to the production of a successful casting. The studies presented in this

thesis targeted these exact areas for 3D sand printing. The findings of this thesis will improve the

modeling capabilities of 3D printed sand molds allowing designers to further leverage

opportunistic design in metal casting. The conclusions and future work for each study are

presented next.

4.1 Conclusions from Succinonitirle Flow Testing

A novel test methodology for experimental flow analysis of metal casting gating systems

was proposed and analyzed through the use of the substance succinonitrile. Through the design of

experiments to validate the use of this substance for this process, it became necessary to analyze

the largely accepted practice of simulating casting flow with water. Two dimensionless fluid

dynamic values, Reynold’s number and Froude’s number, were identified as entities that may or

may not have been important to match across test substances in this area of research. This thesis

78

examined the need for dimensionless number similarity in water testing for metal casting

purposes, water testing’s ability to recreate a previously recorded aluminum pour, and

succinonitrile’s ability to recreate the same aluminum pour. All of these tests were conducted

ignoring wall roughness and permeability effects. Froude’s number similarity was found to be

necessary to preserve the fill rate and energy with which the fluid enters the mold across all tests.

Reynold’s number was not found to have a direct result on the fluid profile. Rather, pressure drop

as a function of wall roughness and Reynold’s number is proposed to have a significant effect.

Experimental results showed water was able to roughly mimic an aluminum pour, but an exact

match was not able to be produced regardless of Reynold’s and Froude’s number similarity.

Succinonitrile produced a more accurate representation of the aluminum pour than the water tests

conducted in this study. This result occurred despite the SCN having a Reynold’s number less

than 25% of the aluminum. Through the Moody diagram, it was seen that the lower Reynold’s

number correlated to a higher pressure drop which may have been a better representation of the

wall roughness found in a sand mold.

The findings of this thesis support the argument for succinonitrile as an equal or better

metal analog than water. The testing process used in this study consisted of forming a mold made

of acrylic plastic, and pouring liquid SCN through the mold. The filling profile was filmed with a

high-speed camera. This testing methodology offers many advantages to metal casting research,

particularly when compared with the alternative of x-ray visualization. The advantages of this

testing method include:

• The ability to cheaply visualize experimental casting flow.

• The ability to visualize experimental casting research in a safe lab setting

• The ability to rapidly and cheaply create and test various gating geometries

• The ability to reuse the test mold multiple times

• The ability to quantify flow velocity through high-speed video imaging

79

This study proved that this manner of testing was not only possible, but was able to produce

beneficial results. Despite this, flaws with this form of testing remain. The health hazards of

associated with succinonitrile (namely being an eye and skin irritant) means it is best practice to

handle inside a fume hood. The limited thermal capacity of the acrylic prohibited the SCN from

being heated to desired temperature for Reynold’s number similarity. Attempts to use higher

temperatures lead to severe cracking of the mold. Furthermore, the physical properties of acrylic

lead to a long solidification time for the SCN (~40 minutes). This was difficult for running

multiple tests in succession in addition to cleaning the mold between tests. If opened prior to

complete solidification, the SCN left large amounts of residue in the mold. This residue hindered

visual transparency and could have potential affects on the flow profile. Finally, the physical

properties of acrylic did not provide for the ability to match solidification time to that found in

sand casting. The acrylic had a significantly longer solidification time which would not be

accurate for predicting the success of thin wall castings of premature solidification in gating

system design.

4.1.2 Future Work Regarding Succinonitrile Flow Testing

The ultimate goal of succinonitrile testing is to allow for advancements in three areas:

flow visualization and quantification, advanced gating geometries, and experimental flow and

solidification merging. The limiting factor to these advancements is the acrylic mold. As shown

in this study, the acrylic did not allow for the desired thermal and solidification conditions of

SCN to be met. A material is needed to that allows for visual transparency, but is also able to

meet the thermal needs of this testing. Incorporating flow sensors into the mold may provide a

way to circumvent the need for optical transparency, and provide velocity values of the melt flow.

80

Experimentally measured velocity values would provide a reference for which numerical solvers

could be compared.

Wall roughness was shown to be a significant factor in matching the sand mold flow to

the acrylic mold flow. Wall roughness similarity may be met through the addition of baffle

geometries on the acrylic mold wall. These baffles could be correlated to the roughness value a

sand mold. Further work would be needed to properly design the baffles, and it is unknown if this

system would come as a detriment to visual transparency. Another key mold characteristic,

permeability, allows for trapped air to escape the sand mold and reduce back pressure against the

incoming melt flow. Permeability similarity between the sand and acrylic molds may be met

through a vent placed in the top of the acrylic mold. The vent size would need to be correlated to

the sand permeability.

The ability to 3D print gating geometries via stereolithography printing (a vat

photopolymerization process) is hypothesized to offer potential for studying advanced 3DSP

gating geometries. Current limitations exist when it comes to the thermal aspects of the print

material, which are similar to those of acrylic. Many stereolithography printers also feature small

build volumes (roughly 5 in. by 5 in. by 5in.) that limits the ability and time needed to print full-

sized gating systems. Therefore, the priority for continuing this form of SCN testing should be to

perfect the practice in reference to traditional green sand molds. Once successful, 3D printed

elements may be incorporated.

4.2 Conclusions on 3DSP Molds Thermal Properties

A study was conducted to quantify the thermal properties of 3D printed sand molds and propose

an optimal binder content for the printing of such molds. 3D printed sand samples were acquired

in the forms of pour cups and cylindrical geometries each at 6 different pulse values. Loss on

81

ignition test was conducted to quantify the amount of binder in each pulse value. These results

showed binder content was higher than the manufacturer prediction for the lower three pulse

values and lower than manufacturer prediction for the higher pulse values. A dimensional

accuracy study showed increased binder content had a negative significant effect on part

tolerance. Density values were shown to decrease as binder content increased ranging +/-1% from

the median value of 2.600 g/cm3. The values of unbonded sand and furan binder where also

reported to have density values of 2.6510 g/cm3 and 1.1424 g/cm3 respectively. A solidification

analysis of 99.9% pure aluminum showed that binder content had no significant effect on casting

solidification time or heat transfer in the mold. The solidification time recorded in this test was

used with Chvorinov’s rule to predict a mold constant of 4,106,250 s/m2. Further examination of

this constant produced a coefficient of heat accumulation of 739.1 W-s0.5/m2-K. The specific heat

of a pulse 2 found to be 720.1 J/kg-C. This value, along with the measured density of a pulse 2

sample was used to derive a thermal conductivity value from the coefficient of heat accumulation.

This value was 0.29 W/m-K. The commercial simulation software SOLIDCast [11] was used to

recreate the aluminum solidification experiment. Using the measured density and specific heat

values for the pulse 2 mold, the simulation was able to reproduce the experimental solidification

of 12.167 seconds for a thermal conductivity of 0.42 W/m-K. The percent difference between

these values fell with-in the percent error range between Chvorinov’s rule and SOLIDCast [11]

simulation for other mold thermal properties.

4.2.1 Future Work Regarding 3D Sand Printing Thermal Properties

Future work for this study aims to continue to provide values for the thermal properties of

3D printed sand mold. Refinement in testing technique with either an SDT or DSC machine

should yield specific heat values for bonded sand samples at temperatures below 450 °C.

82

Determining specific heat values will provide values for thermal conductivity through the mold

constant relationship. Furthermore, continued processing of the unsteady heat transfer data

gathered in the molds during pour testing should also offer values for thermal conductivity that

can be compared to the mold constant calculation. Finally, grain structure imaging and strength

testing of the aluminum castings made in this study will be able to confirm the conclusion of this

paper that binder concentration does not affect casting solidification time or part quality.

Future work can be conducted to validate and extend the insights of this study. Microstructure

and strength analysis of the castings poured in this study will be used to verify the conclusion that

binder content did not have an effect on casting quality. The temperature data collected from

thermocouples may be used with a one-dimensional unsteady heat transfer analysis to draw

further insight to the thermal conductivity of 3D printed molds. This analysis is limited to the 4

locations at which thermocouples were place to gather temperature data. Additional tests may be

needed if additional temperature locations are required. Similarly, thermal conductivity may be

found through the use of additional commercial simulation software; however, this is limited by

the cost to access these softwares. Finally, further tests are needed to gather thermal property data

for a range of temperatures. This study found density values at ~25 °C and specific heat up to 300

°C. These values will change with increased temperature which are useful to the improvement of

numerical models. Gathering this data will be difficult due to the needed to hold the sample at a

constant temperature in the case of density and the binder burnout effect occurring at 500 °C

Future models may benefit from separately modeling the areas affected and unaffected by binder

burnout. As shown in this study, analysis using a thermal SDT may be the first step in this

process.

83

References

[1] Sama, S., and Manogharan, G., 2017, “Sand Casting Design Rules.”

[2] Sama, S. R., Wang, J., and Manogharan, G., 2018, “Non-Conventional Mold Design for

Metal Casting Using 3D Sand-Printing,” J. Manuf. Process., 34, pp. 765–775.

[3] Sirrell, B., and Holliday, M., 1996, “Benchmark Testing the Flow and Solidification

Modeling of AI Castings,” (March), pp. 20–23.

[4] “About Metalcasting | American Foundry Society” [Online]. Available:

https://www.afsinc.org/about-metalcasting. [Accessed: 02-Oct-2018].

[5] Almaghariz, E. S., 2015, “Determining When to Use 3D Sand Printing : Quantifying the

Role of Complexity By Eyad S . Almaghariz A Thesis Submitted in Partial Fulfillment of

the Requirements for the Degree of Master of Science in The,” (May).

[6] State, U., Agency, U. . E. P., and Assurance, O. of E. and C., 1997, Profile of Metal

Casting Industry, EPA Office of Compliance, Washington, D.C.

[7] Staff, E., 2006, Guide to Casting and Molding Processes.

[8] Banchhor, R., and Ganguly, S. K., 2014, “OPTIMIZATION IN GREEN SAND

CASTING PROCESS FOR EFFICIENT , ECONOMICAL AND QUALITY CASTING,”

Int. J. Adv. Technol., 5(1), pp. 25–29.

[9] “Startpage MAGMA North America” [Online]. Available:

https://www.magmasoft.com/en/. [Accessed: 20-Mar-2019].

[10] “ProCAST - QuikCAST - Casting Processes Simulation Software | ESI Group” [Online].

Available: https://www.esi-group.com/software-solutions/virtual-

manufacturing/casting/procast-quikcast-casting-processes-simulation-software. [Accessed:

20-Mar-2019].

[11] “Finite Solutions Casting Simulation Software” [Online]. Available:

https://finite.solutions/. [Accessed: 20-Mar-2019].

84

[12] “FLOW-3D CAST | Metal Casting Simulation | CFD Software” [Online]. Available:

https://www.flow3d.com/products/flow-3d-cast/. [Accessed: 20-Mar-2019].

[13] Yang, X., and Campbell, J., 1998, “Liquid Metal Flow in a Pouring Basin Liquid Metal

Flow in a Pouring Basin,” Int. J. Cast Met. Res., 0461(July).

[14] Beeley, P., 2001, Foundry Technology, Butterworth-Heinemann, Woburn.

[15] Campbell, J., 2015, Complete Casting Handbook, Elsevier Ltd.

[16] Luther, N., 1997, “Economics of Good Pattern Equipment.Pdf,” Cast. Source Dir.

[17] Upadhyay, M., Sivarupan, T., and El Mansori, M., 2017, “3D Printing for Rapid Sand

Casting—A Review,” J. Manuf. Process., 29, pp. 211–220.

[18] 52900:2015, A., 2015, “Standard Terminology for Additive Manufacturing – General

Principles – Terminology,” ASTM Int., i, pp. 1–9.

[19] Strong, D., Kay, M., Conner, B., Wakefield, T., and Manogharan, G., 2018, “Hybrid

Manufacturing – Integrating Traditional Manufacturers with Additive Manufacturing

(AM) Supply Chain,” Addit. Manuf., 21(February), pp. 159–173.

[20] Wong, K. V., and Hernandez, A., 2012, “A Review of Additive Manufacturing,” ISRN

Mech. Eng., 2012, pp. 1–10.

[21] Meteyer, S., Xu, X., Perry, N., and Zhao, Y. F., 2014, “Energy and Material Flow

Analysis of Binder-Jetting Additive Manufacturing Processes,” Procedia CIRP, 15, pp.

19–25.

[22] Wohlers, T., and Gornet, T., 2012, “History of Additive Manufacturing Introduction of

Non-SL Systems Introduction of Low-Cost 3D Printers,” Wohlers Rep. 2012, pp. 1–23.

[23] “ExOne Home | ExOne” [Online]. Available: https://www.exone.com/. [Accessed: 20-

Mar-2019].

[24] “VX Sandmold | Voxeljet 3D-Print” [Online]. Available:

https://www.voxeljet.com/branchen/cases/probeform-aus-sand-vx-wuerfel/. [Accessed:

85

20-Mar-2019].

[25] “Viridis3D Archives | EnvisionTEC” [Online]. Available: https://envisiontec.com/3d-

printers/robotic-additive-manufacturing/. [Accessed: 20-Mar-2019].

[26] Kumke, M., Watschke, H., Hartogh, P., Bavendiek, A. K., and Vietor, T., 2018, “Methods

and Tools for Identifying and Leveraging Additive Manufacturing Design Potentials,” Int.

J. Interact. Des. Manuf., 12(2), pp. 481–493.

[27] Floriane, L., Frédéric, S., Gianluca, D. A., and Marc, L. C., 2017, “Enriching Design with

X through Tailored Additive Manufacturing Knowledge: A Methodological Proposal,”

Int. J. Interact. Des. Manuf., 11(2), pp. 279–288.

[28] Dańko, J., Dańko, R., and Holzer, M., 2003, “Reclamation of Used Sands in Foundry

Production,” Metalurgija, 42(3), pp. 173–177.

[29] Campbell, J., 2006, “Entrainment Defects,” Mater. Sci. Technol., 22(2), pp. 127–145.

[30] Cross, M., Pericleous, K., Croft, T. N., McBride, D., Lawrence, J. A., and Williams, A. J.,

2006, “Computational Modeling of Mold Filling and Related Free-Surface Flows in Shape

Casting: An Overview of the Challenges Involved,” Metall. Mater. Trans. B Process

Metall. Mater. Process. Sci., 37(6), pp. 879–885.

[31] Mi, J., Harding, R. A., and Campbell, J., 2004, “Effects of the Entrained Surface Film on

the Reliability of Castings,” Metall. Mater. Trans. A Phys. Metall. Mater. Sci., 35 A(9),

pp. 2893–2902.

[32] Dai, X., Yang, X., Campbell, J., and Wood, J., 2003, “Effects of Runner System Design

on the Mechanical Strength of Al-7Si-Mg Alloy Castings,” Mater. Sci. Eng. A, 354(1–2),

pp. 315–325.

[33] Tiryakioǧlu, M., Campbell, J., and Alexopoulos, N. D., 2009, “Quality Indices for

Aluminum Alloy Castings: A Critical Review,” Metall. Mater. Trans. B Process Metall.

Mater. Process. Sci., 40(6), pp. 802–811.

86

[34] Sutaria, M., and Ravi, B., 2014, “Computation of Casting Solidification Feed-Paths Using

Gradient Vector Method with Various Boundary Conditions,” Int. J. Adv. Manuf.

Technol., 75(1–4), pp. 209–223.

[35] Sutaria, M., Gada, V. H., Sharma, A., and Ravi, B., 2012, “Computation of Feed-Paths for

Casting Solidification Using Level-Set-Method,” J. Mater. Process. Technol., 212(6), pp.

1236–1249.

[36] Lee, P. D., Chirazi, A., and See, D., 2001, “Modeling Microporosity in Aluminum-Silicon

Alloys: A Review,” J. Light Met., 1(1), pp. 15–30.

[37] Kashiwai, S., Ohnaka, I., Kimatsuka, A., Kaneyoshi, T., Ohmichi, T., and Zhu, J., 2005,

“Numerical Simulation and X-Ray Direct Observation of Mould Filling during Vacuum

Suction Casting,” Int. J. Cast Met. Res., 18(3), pp. 144–148.

[38] Di Sabatino, M., Syvertsen, F., Arnberg, L., and Nordmark, A., 2005, “An Improved

Method for Fluidity Measurement by Gravity Casting of Spirals in Sand Moulds,” Int. J.

Cast Met. Res., 18(1), pp. 59–62.

[39] Sulaiman, S., and Hamouda, A. M. S., 2004, “Modelling and Experimental Investigation

of Solidification Process in Sand Casting,” J. Mater. Process. Technol., 155–156(1–3), pp.

1723–1726.

[40] Lewis, R. W., and Ravindran, K., 2000, “Finite Element Simulation of Metal Casting,”

Int. J. Numer. Methods Eng., 47(1-3)(April 1999), pp. 29–59.

[41] Esparza, C. E., Guerrero-Mata, M. P., and Ríos-Mercado, R. Z., 2006, “Optimal Design of

Gating Systems by Gradient Search Methods,” Comput. Mater. Sci., 36(4), pp. 457–467.

[42] Vaghasia, D., 2009, “Gating System Design Optimization for Sand Casting,” (June), p. 79.

[43] Thomas, B. G., Mika, L. J., and Najjar, F. M., 1990, “Simulation of Fluid Flow inside a

Continuous Slab-Casting Machine,” Metall. Trans. B, 21(2), pp. 387–400.

[44] Thomas, B. G., and Najjar, F. M., 1991, “Finite Element Modelling of Turbulent Fluid

87

Flow and Heat Transfer in Continuous Casting,” Appl. Math. Model., 15(5), pp. 226–243.

[45] Cleary, P. W., and Ha, J., 2002, “Three-Dimensional Smoothed Particle Hydrodynamics

Simulation of High Pressure Die Casting of Light Metal Components,” J. Light Met., 2(3

SPEC.), pp. 169–183.

[46] Renukananda, K. H., and Ravi, B., 2016, “Multi-Gate Systems in Casting Process:

Comparative Study of Liquid Metal and Water Flow,” Mater. Manuf. Process., 31(8), pp.

1091–1101.

[47] Sahai, Y., and Emi, T., 1996, “Criteria for Water Modeling of Melt Flow and Inclusion

Removal in Continuous Casting Tundishes.,” ISIJ Int., 36(9), pp. 1166–1173.

[48] Kuyucak, S., 2006, “Sponsored Research : Clean Steel Casting Production - Water

Modeling Studies of Bottom-Pouring Ladle Operations,” Trans. Am. Foundry Soc.,

114(06-), pp. 1–8.

[49] Derollez, P., Lefebvre, J., Descamps, M., Press, W., and Fontaine, H., 1990, “Structure of

Succinonitrile in Its Plastic Phase,” J. Phys. Condens. Matter, 2(33), pp. 6893–6903.

[50] Glicksman, M. E., Schaefer, R. J., and Ayres, J. D., 1976, “Dendritic Growth - a Test of

Theory,” Metall. Trans. A, 7(November), pp. 1747–1759.

[51] Janz, G. J., and Fitzgerald, W. E., 1955, “Infrared Spectrum and Molecular Structure of

Succinonitrile,” J. Chem. Phys., 23(10), pp. 1973–1974.

[52] Fitzgerald, W. E., and Janz, G. J., 1957, “Vibrational Spectra and Molecular Structure of

1,2-Dicyanoethane,” J. Mol. Spectrosc., 1(1–4), pp. 49–60.

[53] Weinberg, F., and Chalmers, B., 1951, “Dendritic Growth In,” Can. J. Phys., 29, p. 382.

[54] Weinberg, F., and Chalmers, B., 1952, “Further Observations on Dendritic Growth in

Metals,” Can. J. Phys., 30, p. 488.

[55] Chalmers, B., 1953, “THE PREPARATION OF SINGLE CRYSTALS AND

BICRYSTALS BY THE CONTROLLED SOLIDIFICATION OF MOLTEN METALS,”

88

Can. J. Phys., 31, p. 132.

[56] Huang, S. C., and Glicksman, M. E., 1981, “Overview 12: Fundamentals of Dendritic

Solidification-I. Steady-State Tip Growth,” Acta Metall., 29(5), pp. 701–715.

[57] Chopra, M. A., Glicksman, M. E., and Singh, N. B., 1988, “Dendritic Solidification in

Binary Alloys,” Metall. Trans. A, Phys. Metall. Mater. Sci., 19 A(12), pp. 3087–3096.

[58] Shen, H. F., and Beckermann, C., 2002, “An Experimental Study of Deformation of a

Columnar Dendritic Mushy Zone Using a Transparent Succinonitrile-Acetone Alloy,”

Metall. Mater. Trans. B Process Metall. Mater. Process. Sci., 33(1), pp. 69–78.

[59] Esaka, H., Wakabayashi, T., Shinozuka, K., and Tamura, M., 2003, “Origin of Equiaxed

Grains and Their Motion in the Liquid Phase,” ISIJ Int., 43(9), pp. 1415–1420.

[60] Khalifa, W., Tsunekawa, Y., and Okumiya, M., 2010, “Ultrasonic Grain Refining Effects

in A356 Al-Si Cast Alloy,” AFS Trans., (April 2010), pp. 1–8.

[61] Tin, P., and de Groh III, H. C., 2004, “Surface Tension and Viscosity of Succinonitrile–

Acetone Alloys Using Surface Light Scattering Spectrometer,” Int. J. Thermophys., 25(4),

pp. 1143–1153.

[62] Ceynar, D. L., and Beckermann, C., 2001, “Measurement of the Density of Succinonitrile

– Acetone Alloys,” 222(September 2000), pp. 380–391.

[63] Petrakis, L., and Rao, A., 1963, “Rotational Transition and Self-Diffusion in

Polycrystalline Succinonitrile,” J. Chem. Phys., 39(7), pp. 1633–1635.

[64] Chvorinov, N., 1940, “Theorie Der Erstarrung von Gussstücken. Giesserei,” Heft, 10.

[65] Park, S. I. I., and Hartley, J. G., 1996, “Measurement of the Effective Thermal

Conductivities of Molding Sands at High Temperatures,” 10(4), pp. 480–488.

[66] Ho, K. A. I., and Pehlke, R. D., 1990, “Metal-Mold Interfacial Heat Transfer,” J. Am.

Ceram. Soc., 73(8), pp. 2316–2322.

[67] Sheet, S. D., 2018, “SIGMA-ALDRICH,” pp. 1–8.

89

[68] Munson, B. R., Young, F. D., and Okiishi, T. H., 2006, Fundamentals of Fluid Mechanics,

John Wiley & Sons.

[69] Zych, J., and Mocek, J., 2015, “Thermal Conductivity of Moulding Sand with Chemical

Binders , Attempts of Its Increasing,” Arch. Metall. Mater., 60(1).

[70] Engineering, F., 2010, “Chvorinov ’ s Rule and Determination of Coefficient of Heat

Accumulation of Moulds with Non-Quartz Base Sands,” 10(4), pp. 77–82.

[71] Hot, K., and Pehlke, R. D., 1990, “Simultaneous Determination of Thermal Conductivity

and Specific Heat for Refractory Materials,” 22(198539), pp. 2316–2322.

[72] Williams, T. J., Hardin, R. A., and Beckermann, C., 2014, “Thermophysical Properties for

ASK Chemical and Exochem Riser Sleeves for Steel Castings,” (4), pp. 1–23.

[73] Williams, T. J., Hardin, R. A., and Beckermann, C., 2015, “Characterization of the

Thermophysical Properties of Riser Sleeve Materials and Analysis of Riser Sleeve

Performance,” (5), pp. 1–28.

[74] Lekakh, S., Richards, V., and Druschitz, E., “New Method of Dynamical Measurements of

Mold Thermal Properties and Applications for Casting Processes.”

[75] Ravi, B., Srinivasan, M. N., Ravi, B., and Srinivasant, M. N., 2016, “Casting

Solidification Analysis by Modulus Vector Method,” 0461.

[76] Xu, Y., and Chung, D. D. L., 2000, “Effect of Sand Addition on the Specific Heat and

Thermal Conductivity of Cement,” 30, pp. 59–61.

[77] Coniglio, N., Sivarupan, T., and Mansori, M. El, 2018, “Investigation of Process

Parameter Effect on Anisotropic Properties of 3D Printed Sand Molds,” pp. 2175–2185.

[78] AFS 5100-00-S, Schaumburg.

[79] 2015, “Standard Test Method for Determining Specific Heat Capacity by Differential

Scanning,” pp. 1–6.