AN INVESTIGATION OF SEVERAL VARIABLES IN THE STRENGTH OF SHELL MOLDSrf bv

Adolphus Lloyd Lotts

Thesis submitted to the Graduate Faculty of the

N · 2

-·N.P.as.e..1Lo.tI. Title Page 1

II. Table of Contents 2III. Introduction 4IV. Review of Literature 7V. Plan of Research 10

A. Construction and Adaptation of Equipment 10B. Influence of Several variables on Flexural Strength 10C. Influence of Grain Size Distribtion 10D. Correlation of Density and Flexural Strength 11

VI. Description of Apparatus and Materials 12A. Mixing Apparatus 12B. Flexural Test Bar Meld 12C. Furnace, Magen, and Table 12D. Flexural Testing Device 13E. Silica Sand 13F. Shell Melding Resin 14

VII. The Experimental Design 15A. Design to Detemine the Effect of Several variables 15B. Design to Determine the Effect of Grain Size Distribution 16

VIII. The Experimental Procedure 19A. Weighing and Mixing 19B. Preparation of the Flexural Test Bars 19C. Testing the Flexural Test Bars 20

IX. Results 21A. Calculation of Flexural Strength 21B. Calculation of Bulk Density 22C. Effects of Several variables on Plexural Strength 22D. Effect of Grain Size Distribution on Flexural Strength 23E. Correlation of Blk Density and Plexural Strength · 24

X. Discussion of Results 25A. The Effect of Several variables on Flexural Strength 25B. Effect of Grain Size Distribution on Flexural Strength 27C. Correlation of Bulk Density and Flexural Strength 28D. Experimental Precision 28

XI. Conclusions 30XII. Acknowledgements 31

XIII. Suggestions for Future Werk 32XIV. Bibliography 33

XvaAppendix A 35Appendix B 36Appendix C 37Appendix D 43Appendix E 45

- 3 „

List of Tables

1ah1.¢...H.¤... £.ass..kL¤...1. Statistical Factorial Design of One Replication 152. Sand Compositions for the Factorial Design 163. Cmpositions of the Sand Dlstributions l84. Flexural Strengths from Replicate No. 1 of the Factorial

Design 465. Flexural Strengths from Replicate No. 2 of the Factorial

Design 476. Flexural Strengths from Replicate No. 3 of the Factorial

esign 487. Analysis of Variance 498. Flexural Strengths and Bulk Density from the Investigation

of Grain Size Distribtion 509. Bulk Density and Average Flexural Strength in the Factorial

Design 5110. Five—way Classification of T, t, C, D, and R 4011. Four•way Classification of t, D, C, and R 4112. Three•ney Classification of D, C, and R 4213. Two—uay Classification of D and R 42

List of Figures

£lsu:r..Hs.„1. Flexural Test Bar Hold Assebly 522. Flexural Testing Device 533. Cumlative Plot of Sand Distributlons 544. Sand Composition vs. Strength Curves 555. Sand Composition vs. Strength Curves 566. Flexural Strength vs. Grain Size Distribution 577. Bulk Dnsity vs. Graln Size Distribution 588. Flexural Strength vs. Bulk Density 59

1

One of the more recent developments of technique in the feundry

industry has been the development of the shell molding process, which

was a process kept secret by the German government until the end of

World war II, The patentla for this process, being declared in the

public dmain at the end of World War II, was of such immediate practical

value that within four months after patent application had been filed,

the process was being used in production,

The process is used for making hollow cores and meld shells with

thin and approximately uniform wall thickness, to be used in metal cast-

ing as a replacement for the more cumbersome sand molds, The ingredients

for making these shell molds in present practice are dry sand or other

refractory material, dry pheno1•formaldehyde resin, and dry hexamethylene

tetramine,Shell molding, as it is used today, is essentially a process for

making small precision castings on a production basis, A typical se-

quence of operations in making a shell meld would be:

1, Preheat pattern to an appropriate temperature; for example,

500 degrees fahrenheit,

2, Place the hot pattern so that a sand-resin mixture may be

dumped on it,

3, Allew the sand-resin mixture to "dwell” on the pattern for a

length of time dependent upon pattern temperature and thickness

of shell desired,

4, Remove the excess sand•resin mixture,

5, Place the pattern and adhering shell meld in an even to ”cure",

• 5 — {

6. Remove the mold and pattern from the curing oven and separate

the mold from the pattern

7. The pattern may be immediately re·used for several cycles before

reheating is necessary.

Heat is the force which causes the shell mold to form fro the sand-

resin mixture. The heat supplied by the hot pattern does two things to

the two—step resin; it causes the phenol·formaldehyde resin to flow and

it breaks the hexamethylene tetramine down into formaldehyde and ammonia.

The formaldehyde causes the resin to cross-link, or set; the ammonia is

a catalyst for this reaction.

For several years after the shell moldlng process was adopted in

this country the information that could be found in open literature con-

cerning the process was only general. Recently, results of experimenta-

tion with the process have become available, and much literature can be

found concerning shell molds.

A recent reportlz published was concerned with:

1. Bonding efficiency of various phenollc resins, and effects of

variations in amount and size of resin.

2. Effects of variations in grain size, grain size distribtion,

purity and angularity of sand.

3. Effectiveness of other molding refractorles than silica sand ln

making shell molds.P

4. Surface treatment of the sand to improve resin-sand adhesion.

5. Adaptability of vibration packing for preparing shell molds.

6. Performance of various sand-resin mixtures used in making shell

molds.

r _ r

„ 6 -

7. Surface characteristics of aluminum and copper base alloys cast

against shells of various mixtures of sand and resins.8. Heat transfer in shell molds and breakdown time of molds (to

accommodate contraction of castings).

9. Means to prevent segregation of binders in practice.10. Mechanism of bonding of sands with resin binders.

In this report several conclusions were made which were of importanceto the foundry industry. In particular, one of interest was, "the effectof grain size distribution upon packing density and the strength of ashell is small when making shell molds by the dumping method". To thisdate, there have been no experiments to check the interaction effects ofthe variables affecting the strength of shell molds.

This thesis presents experimental work done to show the single effectsand combined effects of curing temperature, dwelling time, amount of resinused, average sand grain size, and distribution of grain size on the flex-ural strength of shell molds; and the correlation between bulk density andflexural strength. The importance of continuing the investigation of theseveral variables presented in this thesis is that possibly the cost ofmaking shell molds may be reduced by decreasing resin consumption wdthouthaving subsequent decrease in their strength. The correlation of bulkdensity and flexural strength provides a means of measuring the strengthof shell molds without destructive testing. If such correlations for

various shell molding mixtures and treatments were available, the flex-ural strength of molds made from any mixture could be inferred from ameasurement of bulk density.

I

-7In

reviewing the literature on shell melding, much information wasfound concerning the effects ef various variables on the properties ofshell molds. A few of the later publications were quite comprehensiveand gave a good survey of the status ef shell molding to date. It wasevident, however, that there are several factors or combinations offactors relating to the strength of shell molds which are little under·stood and in some cases essentially uninvestigated.

The strength of shell molds may be inferred from the results of oneof several different tests. The two which have been used most frequentlyare the tensile test, which uses the ”dog—bone” type specimenl1*l2 andthe flexural test, which requires a flexural test barllvlz. It appears,however, that a flexural test more closely simulates stress conditionsin a shell meld while casting metal. There is a standard flexural test2used for ceramic materials, which can also be used for shell moldsbecause of the similarity between the physical structure of ceramicmaterials and those of shell melde.

Mch work has been done on the strength of shell molds as a functionof numerous variables. It appears that the most comprehensive coverageof these variables to date was done by investigations at M.I.T.l2. Theyconclude that:

1. In shell molding proper choice of sand and resin is importantfrom the point of view of performance in making shell moldsand in production of castings.

2. Resin is wasted in ceating the surfaces of sand grains.

N

- 3 - !

3. Vibration packing produces stronger shells.

4. Separation of resin from the surface of sand is the mechanism

of fracture of shell molds.

5. Zircon yields an unusually strong shell, apparently due to favor-

able affinity for resin.

6. Breakdown time of a shell meld is a function of the rate of heat

transfer, initial strength, and shell thickness.

According to Sprinklelö, the strength of shell molds is a function of the

work of adhesion for phenol-formaldehyde resin and the metallic oxide in

question.

Some disagreement was found between investigators concerning the in-

fluence of grain size distribution on density and strength. Anspachl

mentions that greater grain size distribution yields significantly greater

bulk density and strength in green sand molds; the investigators at M.I.T.l2

conclude that the effect of grain size distribution is small when making

shell molds by the dumping method. Even though one is considering the

effect of grain size distribution on green sand molds and the other, on

shell molds, the effect should be generally the same in each caee.

Dallavalleß shows that the bulk density increases with wider grain size

particles in an idealized situation.

Braybrook and waters4*5 also consider the effect of grain size.

They write,”It may be mentioned that none of the work described here

‘or carried out recently by others supports any suggestion that the grain

size of a sand alone is a primary factor in determining the amount of

resin needed to give a particular strength”. They state further that

more important factors are sand contamination and resin segregation.

9 -

Precoating sand with resin has been mentioned by several authors as a

means of minimizing segregation.

In considering appropriate curing temperatures, Braybrook and

Waters4*5 indicate that curing temperature is important. Childs and

Hymané have investigated the variation in temperature of shell molds

during the dwelling and curing cycles. They show that appreciable burnout

of resin on the shell surface is probably at the curing temperatures that

are used.No information on the synergistic effects of variables affecting

the strength of shell molds was disclosed in the literature.

- 10 — V

VT

^·To investigate the effects of the variables on the flexural strength

of shell molds in this work, it was necessary to contruct equipment or

adapt existing equipment to make and test the flexural test bar, This

included adaptation of a furnace, construction of a table and metal wagon

to facilitate the handling of the mold, construction of a flexural testing

device, and construction ef a mold,B•lt

was planned to investigate the effects of dwelling time, curing

temperature, sand composition, and amount of resin by using a statistical

factorial design for the experiment, In this design two levels of dwell-

ing time, two levels of curing temperature, two levels of resin amount,

and four levels of sand composition were planned, The reson that the

factorial design was used in this investigation was to show the inter-

action effects of the variables, if any existed, It was found after the

experimental work was started that the temperature of the curing furance

interior decreased from the front to the rear of the muffle, This was

evidenced by the difference in color of different areas on the test bar,Therefore, to decrease the experimental error another variable, approximate

position of the test bar area in the curing furnace, was added to the

experimental design,

¢•It was planned to investigate the effect of grain size distribution V

on the strength of shell molds by using a separate experiment in which Tthe sand distributions had been prepared artificially, The other varia· T

ble in this part of the investigation was the effect of the variation

oftemperaturein the curing furnace,

JJ • 11··Üe

.•a ·& ..1.2. S .1... . .· #.112.1

To coxreleta density md the flexural strength of shell molds, a

density meuurmsnt aus planned for each test bar,

J

· 12 • ‘n

Mixing of the various mixtures wes acconpllshed by using e standard

ball mlll with the belle rmoved to prevent the sand fron being orushed.

The ball mill container was sight lnches in diameter and ten inches in

length. Better mixing would probably have reaulted had e mulling machine

been used for mixing, but the volume: of the mixtures used were too small

to be used in the mulling machines which were available. One attmpt to

get better mixing in the ball mlll eas he use of beffles. However, it

was found that a greater amount of segregation resulted from the use of

baffles than without them.

Figure l lllustrates the fleural test bar meld, which is similar1 te one used to make flexural test specimens in a previos investigation.

It was designed to make a apecien 13 inches long, 1 inch wide, and .25

inch thlck. Machlned frm .25 inch steel plate, it consisted of three

_ pieces, which were held in position by a slamp at either end. The out-

side dimensions of the mold were 2' x 1/2* x 14* when it is assbled.

Io prevent the test bars free stloklng to the meld a thin coatlng of Dow

silioone grease was used as a partlng agent.

C. ;«

For all heatlng done in this investigation a Hoskins muffle type

furace, having a temperature range up to 2300°P, was used. It has a

muffle measuringl5”

x 7.5* x 5.5*, and its temperature was controlled

with a Leeds and Northrup recording pyrometer.

• 13—The

principal difficulty encountered with the furnaca was the

fluctuation in temperature of 2 l5°F. In view of the total heat con-

tent of the meld, this variation in temperature probably contributed

little to the experimental error. Another difficulty was the variation

in temperature inside the furnace, but this error was removed by ac-

counting for it in the design of the experiment. A small wagen was

built to facilitate handling the meld. A metal table was also con-

structed and placed on e level even with the furnace door so that the

wagen containing the meld cold be rolled in and out of the furnace

easily.‘

A device which was designed and constructed for making flexural

tests on the test bars was quite similar to the standard flexural test•

ing device for ceramic uaterialaz. This device is shown in Figure 2.

It consisted of a steel plate mounted on four logs. On top of the steel

plate were two steel blocks whose distance spart could be adjusted by

removing the two bolts in each and moving the blocks to the desired holes

in the plate. The blocks were then secured by inserting the bolts through

the holes in the blocks and plate. A piece of tubing, which was two in-

chas long and one•half inch in outside diameter, was used es the bearing

surface. The bcket for loading was connected to the tubing by a chain

which passed through holes in the steel plate. In actual testing, the

test bar was placed on the steel blocks with the tubing er bearing sur-

face on top of the bar and in the center between the two blocks. It was

then loaded with lead shot until the bar failed.

¤·£1l.Lca..§.a¤.dThe sand used in this investigation was round silica sand obtalned

from Hedron Sand Company in Hedron, Illinois. lt contained at least 99.9%

silica and less than 0.02% iron. The sand from Iedron was separated

intofractions,using a Ro—Tap Sieve Shaker, according to screen sizes which T

are shown in Tables 2 and 3. All sand mixtures were thancompoundedfrom

these fractlons. A11 sand sizes presented in this investigation

are expressed in microns because auch measurements of grain size are

eeemingly easier to interpret and use than the familiar AFS Grain Fine-

ness number. For purposes of cmparison, the approximate AFS graln

fineneas numbers are shown in Tables 2 and 3. In order to find the average

grain size of a sand, the following procedure was used:

1. Average the size of openings of the last screen through which

the sand passd and the screen on which the sand was retained.

2. Complete step 1 for all screens in the series which are used.

3. height the sand on each screen and find the fractional portion

of the total sample retained on each.

4. Multiply the declmal fractions found in step 3 by the average

size retsined on each screen and add the products to obtain the

overall average grain size.

F. st~. near ss ::= «

The resin used in this investigation was a commercial shell molding

resin,'¤urex” 736 Resin Compound, Serial 6825, fro the Monsanto Cheni-

cal Company. This ls a two·step resin, the chmistry of which was dis·

cussed in the Introduction. v

• •

!1L T}!

Ae Ü1?.»„.€;t .=! Hi ¢~ E1; ? (,1* 1.1 4 * „¢ #1

4.Thestatistical factorial design shown in Table 1 was used for deter-

mination of the effects of dwelling time, curing temperature, amount of re-

sin used, and sand copositien on the flexural strength of shell molds.

=..e= . =1 • .2 Lw- fl • *1,+ •. . 1; (A.!

Curing Temp., °F T1 (500) T2 (600)

Dwellinge r1ee,se¤. 1; (20) 1 (40) 1 (20) 1; (40)1 2 1 2Furnace 1poaiticn D1 D2 D3 D1 D2 D3 D1 D2 D3 D1 D2 D3Sand Gem- Resinposition Amount Specimen Number

nl (6%) 1 2 6 6C

1112 (4%) 2 4 7 9

al (6%) 9 10 13 14

112 (4%) ll 12 15 16

(623 17 18 21 22

122 (4%) 19 20 23 24

121 (6%) 25 26 29 30

122 (4%) 27 28 31 32

Table 2 is a tabuletion of sand coposition. The sign, (-), indi-

cates the screen through which the sand passed, while the sign, (+),

indicates the screen on which the sand was retained. D1, D2, and D3

•16•

repreeent respectively that part of the specimen which was located in thealddle of the nold and hence in the middle ef the muffle furnace duringcurlng, that part of the spectmen which was located near the front oropening ef the furnace, and that part of the epectnen located neer therear ef the furnace.

U.S„ Steve Average Size Send CoupositlonNumber Hlcrone C1 C2 C3 C4

• 20 + 30

~ 30 + 40 505 50 50

· 40 + 50 359 50— 50 + 70 24• 70 + 100 180 100•l00 + 140 127 50 50•l40 e 200 90 50

Average Greln

Approxtnete AFSGreteThree

repltccttens were made for each possible combination of thevariables shown in Table 1. One denetty and three flexural strengthueasureeents were taken on each ef the 32 speoinens ln each replicatton.B. ¤»< »v ++ m„ „„„ t« ·„ ·" *".« - ·

„„ + „· · 0 an „„ •„

To determine the effect of grain size dietribttcn on the strengthef shell melde, five different dletrlhuttons were used, all having

1 - 17 „

approximately the same average grain size. These different distribution:

ware obtained by plotting the available grain sizes on probability paper

and drawing straight distribution lines through the chosen average grain

size, which is located on the 50% probability line? ‘This plot is shown

in Figure 3. Probability paper was used because coanerical sand die-

tributions plotted on this paper results in a nearly linear curve, a curve

quite similar to that obtained by plotting on 1og•log graph paper. Also,

it is most difficult to find the average graln size on log~log graph

paper. The conpesition of each distribution was easily read from the

distribution line. Esch of theesuwumäätions le shown in Table 3.

For·al1 of these specimens the ouring taperature was 500°F.. the

duelling time was 20 aeoonds, and &% resin was used. For this part of

the investigation, two replications were used.

u

··18··

O Om CO r~·--1--1

0* -1ä ·<r gg gggr N

ä 3 ° "uhPI

$-1 6* ··1°’ 1- m <r wa 1:5 cS c>ä

<‘ M -1 ä \*~:0 N•-1

IC (*7 *0 •-4 O 4O 6* «—1

-1-« P1-1-4

O ä N 6* 6* O W4 O anN N ·-4 ·—-• QS °°‘ E 9O 6* O O ab in N

E ä Q g t~ 10 M «·-e--1us--1Q

E ä "' °‘ F3 93 E $3 .92 ISig 0* O— --1 (*4 I'* $-1O 0 O 0 O <f‘ N 2E g Q Q- Q Q Q 8 $1 C Q

+·* ·-1 L4-1~1« ep -4-*O N Gvsä

*-4o* Q r~·· O ~¤ us m22 1D 624 N 0* N $-1 mgm AE l~ 10 M N --1 -1 LD wgg ua ms-1 >-1; C äm ··-4+*co .¤ -•-*1:

O O O O O O O C gz ,3g M v 10 h- O ·¢ O w UE w gu• <b$~• + + -4- + •·-4 --1 N ¤- am S4M -«~• 1- + ·•· + mc: -»·•1.¤ UU) Ö)•r4 XG}

cu w Q1: IIU; O O Q O O äI IE Q 9* 9* '? *979 T 7* • äö-T Qéi ä

a,]_ga,

„. .> a, ,a.l e a·;·e.Por

each raplication ef the experiment, individual samples weighing

300 grams each, of appropriate sand and resin mixtures were made, The

cmpositions of the mixtures used are given in Tables 2 and 3,

These mlxtures were each mixed one•half her in the bell mill which

yielded nearly homogeneous sanples, Visual inapection indicated that

longer mixing perieds did not improve the honogenuity, After mixing, the

samples were stored in closed containers until they were made into test

ars, It should be pointed out that the time reqlred to make each test

bar lade it impossible to use the mixturea inmediately upon miring, It

is probeble that this error in the experient had little effect upon the

results since all the test bars of an individual repllcate were made in

a short period of tin,

To make the flexural test bars the cold flexural test bar meld

was placed on the wagen, which had been preheated to the curing tepera·

ture, and rolled into the furnace which was set for the appropriate cur-

ing temperature, After renaining in the furnace for 15 minutes, which was

sufficient time for the nold to reach the curing temperature, the wagen

containing the meld was reeoved, Tha sand and resin mixture was poured

fte! ¤ tlask onto the meld immediately, and the appropriate dwelling

time eas allowed, Buring the uellin period, the mixture in the meld

was leveled so that it eas approxilately 0,25 inch above the top of the

meld, This lade the overall depth of the mlxture approximtely 0,50

h

p • 20 •

inch. After dwelling, the meld containing the 'set° eixture was rolledinte the furnace and cured at the appropriate curing teperature for oneminute. The meld containing the flexurel test bar specieen was than re-moved freu the furnace, with the test bar being removed free the meld af-ter it had cooled for two nintee. The uold was further cooled in coldwater end cleened if necessary. Several test bars were made withoutcleanin the meld er coating it with slllcone grease. Starting with thecold cold and the preheated wegen for each specieen, the procedure wasrepeated for other epecieens.

¢·Using an abrasive eheel nonted on a drill press, the test barswere greun to an approxieete thlckness of 0.25 inch. In grinding arough grind wee used te remove the bulk of the excese, end a finishgrlnd was used to ebtein the final thickneas.

Each specieen wes weighed to the nearest 0.1 gran, an the thicknessof each speciuen was obtained to the nearest 0.001 inch by averaging fiveneesurenents cn each. After theee neasureaents, the test bar was placedin the flexural testing device with the roller on the center of the barand loaded until it failed. The weight of the leed shot required tocauee failure was detennined to the nearest 0.1 gran. These weight: werecorrected by adding the tere weight of roller, chain, and bucket asseblyof the flerural testing device. After making the flexural test of thecenter of the bar, the two pieces that reeind were tested in the saeenenne: ae the former. It should be noted here that the variables, Di,D2, and D3 were thus eccounted for by a test et the center and testsnear the two ends.

1l

• Q1 «.

.Ih...Bu1L1.¤.

A-The neesurenent made fer flexurel strength cn the flexurel testing

device nec night in peuns. As previously described, e thickness measure•

ment nes made en each specimen. Ueing these aeesurments, the flexurel

strength was celculsted free the bending stress formulezzS..„·¥‘where $m·x is the maximn bending stress, M is the bending mcment, C is

ene•he1f the height (thickness) ef the specinen, end I is the nament ef

inertia. I is given by the equetions

I ¤.Eä212

ehere b is the base of the specinen end h is the height. M is given by

the equatiensn¤ 1/2 1W

where I ie the distance between the ten fixed ends cf the specimen end H

is the weight of the load required tc break the spesimen when loeded in

the center between the twe fixed ends. Since 1 and b, 3.5 inches and 1.0

inch respectively, were the ssee in all specinens, e mere sinplified fere~

ula may be used to celculete the flexurel strength end is given by the

equeticn: $..„="§ä‘Appendix A shews a sample calculaticn cf flexural strength using this

formale.

rl * 22 *

B•The meeaurement made for bulk density was weight in grams, whichwas converted to pounds, Using the weight, w, thickness mesurment,h, the length, L, and base neasuremant, b, the bulk density was calcu·lated from the formulas

nur names erbä-

w end h were variable, while b had a sonstant value of l inch and L aconstant value of 13 inches, A sample oalclation of blk density isgiven ln Appendix B,C• é » ··' ~v;. = 2t•. -. an äl „.< we J .:111

Results of tests of flexurel strength made to determine the effectof dwelling time, curing temperature, anunt of resin, and sand compesi•tion on flexural strength are given in Tables 4, 5 and 6, These datewere analyzed using the statistics! method ehon in Appendix C; the re-sulting analysis of variance is shown in Table 7 of Appendix E, Frnthis analysis of variance it is evident at the 99,5 per sent level thatreplication, curlng temperature, amount of resin used, sand copositlon,and the interaction of temperature and sand composition is significentin considering the strength of shell melde, At the 97,5 per cent level,in addition, two other effects are eignlflcant; the interaction ofamount of resin used and oomposition of sand and the three—way lnterectionof dwelllng tie, send composition, and amount of resin used, All othervariables or combination of variables are negligible since sme are signl·fioant only et the 75 per sent level and others are significent at evenlower levels,

I ,,,, 23 „,, I

Data, es follows, to show the effect of curing temperature, sandcompositin, and the oomined effect of curing temperature and sand ccm-position on the flexural strength cf shell nolds was obtained by averagingacross replicates, dwelllng time, and distance in the furnsce:

Sand Cempesition C1 C2 C3 C4Average Size, Microns (298) (316) (180) (243)

T1 (sooss) 720 psi 721 psi 669 psi 688 psiT2 (600°F) 592 psi 676 psi 600 psi 691 psi

er »r’ •e. »*• e-3The eurves shown in Figure 4 give a grephicel illustration of theseresu1ts•

Data to show the single effects, sand cmpcsition (C1, C2, etc.)ad anont of resin (R1, R2) and their conbined effect were obtained inthe same nenne: as for the case above and are es followsz

j Sand Cmpositien C1 C2 C3 C4Average Size, Micrens (298) (316) (180) (243)

R1 (6% resin) 775 psi 856 psi 763 psi 801 psiR2 (4% resin) 537 psi 542 psi 507 psi 579 psi

These data are shown graphieelly in Figure 5.0. sl ß«„The

grain size distributione used have been previously described,

ITable 8 of Appendix E shows the values obtained for flexurel strength

•Q_4••

and blk density fer this part ef the investigation,

Graphically illustrated in Figure 6 ls the effect of graln size die-

tributien o the flexural strength of shell melde, while Figure 7 shews

the bulk density vs, grain size distribution curve, In making these

speeimens, s curing treatment of 500°F fer one minute, dwelling time cf

20 secends, and 5 per cent resin were used,

E, ·, „ «„„«„ « es „ ee; ~ age .„. s » e„« „

Bulk density measurements were made on all specinens; there were

three flexural strength neasurements made on each, Tables 8 and 9 of

Appendix E show the average flexural strength and bulk density fer each

specimen, The data shown in Table 8 were used for a regressien analysis

ef the dependency of flexural strength on bulk density, which is shewn

by the curve in Figure 8, The regressien analysis, which is described

· in Appendix D, yielded an equation which can he used te predict flexursl

strength ef shell molds when the density is known and the treatnents in

making the melde are identieel to these used in the part ef the investi-

(gation ccncerning grain size distribution, This equetion is es follewss

Y ¤ 670,2 e 36410,9 (X · ,05213)

where Y is the flexural strength ef the shell meld in psi and X is the

bulk density in pounds per cubic inch,

Yn

- Q5 -

, n ;..;~.nl «' ze: Vu;

A, „l„ h"·l ·; ;„ ex e.•

Prem the results ef this experimental werk, it is evident that

the euring temperature does have an appreciable effect upon the f1ex—

ural strength ef shell mlde• In emparing the mean values of flex~

ural strength for the tue eurlng temperature: used, it was obvloue that

greater strength reeulted from shell meld test bars which were cured

at 500°F than those which were curad at 600°F, This result indlcates

that a curinq temperature near sooßr is suffleient in getting the re-

actione in the resin to proeeed• It ie deubtful that the maximum

strength occurs at a curing temperature appreclably below 500°F, It

is clear that a curlng treatmnt of 600°F results in burn—up of the

resin, Besid•a decreaeing the flexural strength of shell melde,

resin burn—up cause: poerer surface finish•s en castinge, since most

ef the burn-up ocrura on that part ef the shell which will be in con—

test with the castin9•

In practice, the hot strength of shell molds may be of primary

importance, However, it shauld hn peinted out that when a metal ls

cast ln e meld, which ie net pr•heat•d, selidification occurs rapidly

near the meld wall and thencefotth proc••ds from that surface with the

metal first selldifled centalning the liquid metal, This means that

the shell meld must reaist the m•tallestatic pressure fer only a short

period of tlme• In sme prectices, it ls custmaxy to build up a

thlek shell by using several cyeles of dmping an curing, This, of

course, sacriflcas casting surface fer increased shell strength; fromf

l

— 26 -

·this experimental work it appears that the increase in strength would

be small because of resin brn·up•

There is considerable interaction between curing temperature and

sand compositien as shown by Figure 4, The major source of this inter-Ä

action 16 sand compositions, C4 und C2, and temperature. (No exp1ana·

tion can be given for this phenome¤on•)

The amount of resin used ie highly significaut in its influenceon the strength of shell mo1ds• This 16 to be expected since greater

amounts of resin provide’more points of bonding between sand and resin

or more interfacial area between the two. The principal reason fox de·

signing the experimental work eith amount of resin as a variable was to

determine its interaction effects. In this investigation there are only

two in which the amount of resin is involved. One is the interaction

between amount of resin and sand eomposition, the major source of which

is the intexaction of sand compesitien, C2, with amount of resin used,The other is a three-way interaction between dwelling time, sand campe-

sition, and amount of resin. No explanatien has been found for either

of these interact1o¤s• _

The means of flexural strength calculated for sand composition do

not give much information about the nature of the influence of average

grain size en shell meld strength. Contained in these means are strong

interaction effects of sand composition with other variables. It does

appear, howver, that the strength of shell molds, as inferred from the

flexural bar tests, inczeases with increase in average grain size of

the sand, as has been shewn by other inveetigators. The effect of grain

fI

l ..27-

size distribution cannot be reselved frcm this part of the investigation

since it was not anticipated in the original experimental design. All

measuremente of flexural strength, therefore, are confused with grain

size distribution which effect is most likely contained in the effectof sand compositio. In coplete analysis, since grain size distribution

and average gxaln size both effect the flexural strengthe obtained fromthe various sand coposltlons, it is evidant that no quantitative meas-

urement can be made of the effect of average grain size on the strength

of shell molds. However, the variation in grain size distributione

among the eands was small, so that some information about the effect

of average graln size was confldently obtained•B• , ~. 9 „ °l.¤l..Q.•1 ·z„ .„ :4..;. 1

Figure 6 shows geuerally that the wider the distribution of graln

size the greater will be the flexural strength of the shell meld, Itls evident that the lowest stxengths were obtalned with distxibutions

No. 3 and No, 5, It is believed that the strength for No. 3 ie slightly

lower than normal, and for No. 4 it ie sllghtly higher. The dashed

curve in Figure 6 shows what ls believed to be the normal variation of

flexural strength with gralu size distribution. It should be obeervedalso that the average grain size of distribution No. 5 is eignificantly

different frem that of the other distxibtlons• Henoe, if a correctionwere applied, distribution Ne. 5 would have a slightly higher flexural

strength•

Fo: an explanatien of veriation in strength with gxain size distri-V

bution, it was found that the bulk deueity of the epecimens increaeed

V

V

[ - 28 — [with wider distribtions, as shown in Figure 7. A greater bulk density

indicates that there is less void space and better packing in the shell

meld specimen• With this better packing there is a smaller average

distance between sand grains so that less resin is needed to fill the

void between any two typical grains. This makes available more resin

to actually adhere to the surface cf the sand particles, so that there

is more interfacial area between the sand and resin. This greater ef-

ficiency of the resin in bonding then increases strength.

C. ..„.• 11 =.2 *.21. •' =. .1

An equation was used to describe how flexural strength of shell

molds varies with bulk density. This equation is suitable only where

the treatments are identical to those used for this part of the inves•

tigation• The flexural strength was found to be a linear function of

density, this being substantiated by a statistical test for linearity

of regression. The value of this correlation is that it shows that if

a series of such correlations can be found for various treatments,/

shell molds may be tested for flexural strength non—destructively•¤•In

the investigation of the several variables affecting the strength

of shell molds for which the factorial design was used, it was found

that the effect of replications was significant. This indicates that

the experimental precision was low. However, the effect of this lack

of precision was removed frcm the other variables in the analysis of

variance so that they were essentially unaffected•

[

- 29 ·

The large difference in the means of the replicates is believed

to be caused by partial break-down of the resin while in storage. The

experimental work for replicate No. 1 and replicate No. 2 was done during

the same month, and there is only a small difference between the means

of their flexural strengths. But the experimental work for replicate

No. 3 was done approximately four months after the other two. The mean

of the flexural strengths of replicate No. 3 is quite different from

the means of the other two.

The error term of the analysis of variance was larger than expected,

lndicating that some additional variable was not anticipated in the de-

sign. Probably this large error was caused by contributing interactions

of the period of time separating the completion of each replicate and

the variables which were known. This has been partially substantiated

by analyzing the variance of just two of the replicates. Replicate No.

l and Replicate No. 3 ware used for this analysis which yielded compare-

tively large interactions of time between replicates and the other var-

iables. This means that true replication did not exist insofar as repli-

cates No. 1 and No. 3 are concerned. Since replicates No. 1 and No. 2

were done in the same month, they can be considered true replicates of

the experimental design.

The error term evidently could have been made smaller by doing

all three replicates at the same time, which probably would have given

more information in the analysis of variance. The large error term in

this investigation then means that estimation of all results from the

analysis of variance are conservative.

T

- 30 •

‘Within the limits ef this investigation, the werk of this thesis

leads to the following conclusions:

l, There is a significant difference between the effects ef

different curing temperatures on the strength ef shell melds,

A curing temperature of 500°F is more desirable than one of

6oo°1=.2, The strength of shell molds increases with increasing resin

consumptien,

3, The strength of shell molds increases with increasing average

sand grain size,

4, A wider distribution of sand grain sizes present in a single

mixture yields a shell meld with s high flexural strength

than a meld with the same average graln size but narrower dis-

tribution,

5, The following interaction effects have a significant influence

on the strength of shell molds:

a) Sand compesitlon and curing temperature,

b) Sand cemposition and amount ef resin,

c) Sand cempesition, delling time, and amount of resin,

6, The correlation between bulk density and flexural strength is

11„,ar_ plexural strength decreases with decreasing density,

nt

-.31-V

Sincere appreciation is expressad to the many people in the

Engineering School of V.P.I. whose advice and direction made this

thesis possible; in particular, to Professor John F. Eckel for his

supervision of the thesls, Professor J. K. Sprlnkle for his help in

planning the investigation, and Professors A. J. Metzger and P. S.

Dear for their helpful counsel.

Appreciation is also expressed to the Monsanto Chemical Com—

pany for furnishing the resin use in this investigation.

VV

V

- 32 —

Frm work to date, it appears that the following research shouldbe conducted:

1. Determine the influence of several other levels of thevariables used in this investigation on the strength ofshell molds.

2. Determine the influence on the strength of shell melds ofvarious sand distributions having other average grain sizes.

3. Investigate the interaction effects found in this work todetermine why they exist.

4. lnvestigate other means of non-destructive testing shellmolds.

MMM

1. Anspach, J. M. °*Foundry Molding Sand", _E_ggg_g;;y_, gg (1956), No. 8, p.100.2.ASTM Committee C-8 on Refractories. ßmgg _g_f_ ßgm gtmdgggggPhiladelphia: American Society for Testing Materials,1952, pp. 54-56. .

3- Bennett- C- A- and Franklin- W- 1-- .§i.11.i.a.11.salAna1xa1.ala.§211mia1¤7 andgghg Qmiggljgqggggg. New York: John Wiley and Sons, 1954.4. Braybrook, A. and Waters, B.H.C. "Some As ects of Shell-moulding Theory",.1.91.

(1950),-N6- 2097- an- 753···762•

5. Braybrook, A. and Waters, B.H.C. "Some Aspects of Shell-Moulding Theory",Esundnclradaleuzaal-.191 (1956)- Na- 2098- pp- 721-728-6. Childs, W. J., and Hyman, J. B. "Thermal Studies in Shell Mo1ding”,' ßgggggg, gg (1956), No. 3, pp. 121-127.7. Cochran, W. G. and Cox, G. M. §g§j,_g_g_s_. New York: JohnWiley and Sons, 1950.8.. Dallavalle, J. M. Micromeritics New York: Pit: 91 Publishing Co., 1948pp. 131-145 .9. Dear, P. S. :Private cormaunication

10-. Pixan- W- J- and Maaaev- Jr-- F- J- ;m.$.iail.ail.ea1Ana1x:ls·New York: McGraw-+1111 Book Co., 1951.11.. General Electric Co., Chemical and Metallurgical Division, ChemicalMaterials Dept.-ghgßßggiggßgmag Pittsfield, Mass., 1956.12., Massachusetts Institute of Technology, Dept. of Metallurgy, Metals Pro-cessing Division, Foundry section. .B£§9.§1;9.11.S211.€1Z1§.1-.1. M9Ldi.D.9.~ ‘

Final Report, Dept. of Amy, ordanance Corps, Pitr..a¤.·-La-nn Lab.,Frankfort Arsenal, Contract No. DA-19-020-ORD-1961.13. McCul1och, W. W. O.T.S. Report FIAT No. 1168, May 1947 (Croning J-.GGYYHBH Pate CaP•A•14.

Schuhmann, Jr., R. "Principles of Counxinution, I-Size Distribution and ,Surface Calculations", _g (1940), No. 4

15.. Sprinkle, J. K "Adhesion of Phenol-Formaldehyde to Various Metal Oxides",Doctor of Science Thesis (ünpublished), M.I.T., 1955, p. 26.16.. The W. S. Tyler Co. _'[hg_Q5g_g_f_1g5_;1;Lg§_j_gy_e_;_. Cleveland: 1955.17. Wedron Sand Company: Private communication, November 10, 1955.

III

ggggndi; A

The formula used for calculation of flexural strength was asfollowsz

SmaxTakingspecimen No. l in replication No. 1 at distance D, as an example,the weight, W, of the load required to produce failure was 14.864 pounds;and the average thickness, h, of the test bar was 0.270 inch. Substitutingthese values in the above equation:

(.270)2All other calculatlons of flexural strength in this investigation weremade in the same manner and with the same accuracy.

@.41::1

The formula used for calculatlon of bulk density was as followszBulk Density == ··y·-bm

Taking specimen No, 1 in replication No, 1 as an example, the weight ofthe specimen, W, was 0,196 pound; L had the constant value of 13 inches,and b the constant value of 1 inch, Substituting these values in theabove equation:

Bulk Density (1) (A70) (13) 0,0558 in'?)A11 other bulk densities were found in a manner identical to thisexample,

„ 37 -

Apggndi; Q

The method used for analysis of the statistical factorial design

in this investigation can be best shown by illustration of the various

calculations and operations done to obtain the analysis of variance,

which is shown in Täblß 7.

The first step was to obtain the various one—way, two-way, three·

way, four-way, and five-way classifications which ware needed in finding

the sums of squares. The first classification obtained was the five·way

by adding the treatments across the replicates and maintaining all other

effects intact. The result of this operation is shown in Table 10.

The four-way classification was obtained slmilarly. For example,

taking the five—way classification in Table 10 and adding across T, the

four·way classification shown in Table ll results.

The three-way, two·way, and one·way classifications were obtained

in the same manner. The three-way classification is obtained by adding

across a variable in the four·way; for example, across t, which results

in the three·way classification of Table 12. The two-way and one·way

classifications are both shown in Table 13.

After obtaining all the various classifications, the next step was the

computation of the sums of squares. To find the sum of sqares for a

single effect and using for an examle, R, refer to the one·way c1assifi·

cation of Table 13:

s.sR = · U92832)2 z 4,790,0%There örö 144 m@&$¤Y@m@¤tS 1¤ R1 and 144 in R2; there are 288 in the grand

total. Continuing with the sample classifications and referring to Table

13, the sum of squares of the two—way interaction of D and R can be foundM

as followsz

i

“ • 38 ·

$-5- = ................6......!....L(38221)2+(525“°)2+‘”“+ 25775 2 - e S -s.e. - $..m,.l.l222222 S 6 266

ISimilarly,referring to Table 12, the sum of squares for the interaction,

URC, is obtainedz

·-s.s.RC— (.,Q2:§.2.„„)2 a 15,403Higher order interactions are obtained in a manner similar to the above

procedure. The total sum of squares is obtained by adding the squares of

all measurements and subtracting the correction factor as followsz

s.s.T¤tal ; (iosv)2 + (105*1):2 + ·—·· + (476)2 — ; 7,957,817

The error sum of squares is obtained by subtraction of all interactions

which do not contain the replication factor and the single effects from

the total sum of squares.

To obtain the number of degrees of freedo for single effects, it is

simply one less than the number of levels of that variable. There are

four levels of C; so there are three degrees of freedom. For an inter·

action effect the number of degrees of freedom is the product of the

number of degrees of freedom possessed by the variables involved in the

interaction. For example, C has three degrees of freedom and D has two;

therefore, the interaction C x D, has six degrees of freedom. The error

degrees of freedom is obtained by subtracting the degrees of freedom for

all interactions not containing the replication factor and for the single

effects from the total degraes of freedom.

The mean square is obtained by dividing the sum of squares of an

effect by the number of degrees of freedom possessed by that effect. The ‘

F—ratio is the result of dividing the particular mean square by the mean

V

- gg „

square of the error. The finding of the F—ratio concludes thecmputa·tionof the analysis of variance for this design. The F-ratio was used

for tests of significance in this investigation, all critical valuesE for the F-ratio were obtained from tables of Bennett and Franklin.

X40;

‘° E O $3 Ä '“' 2% °"ca gz vr ä M 0~ vr äes ··-• •··4 (vl ··-1 •-4

•·-4 N «—•I0even CQ ~D M O M ü'6 6 2 2 2 ä 2 93 2

N •·-4 N «-·• 0.} ·—·•.——°’N 6 6 66OA".eq .—« eq .-4 eq •-—• ga -4

8~D ·-4 d' b- vrV $1 2 E 6 :1 ä 6 äHqqw: .-4 .-4 eq •--• eq •-e eq ~·•

ucuan ca •g-• -4 © ·-—• N6 M2 2 2 ä 6 2 9. 2N mw ·-• N ··-4N..2“ 92 2 2.2 6) :1 2 2

¤ 9°·‘r~ co m r~· .-—• cwcu $2: N ~o m.;.262 2 2 2 äelä 2 EQ •-I2 2*.0) O C') ‘¤‘ 0% N

3 62 2 .2 2 6 6 6 2H AN «—··• N ···* N •··¤ N ·-I*~,..Y‘· ‘“

"‘ 9; Q 9; " Q 2% äi $3Sg) ACI) © 0* 2 O ~O •-4 ~O·•-·• N ·-··• N ·-4 N ··-• N -42ä.%ux;2.1:: Q Q 9. R Fg ä E Q';‘“"3 2% 92 R .-. N .-. R $3ev av8 3 2 2 2 .·: 6 2 1: 6

N ä P- !'~ K7 *7 äN ·—-• N •-·4 N ·-• N ·—-4-•-•

ev 2 2 63 E 9 .6 '2 QN ···•N9+

cg; rz? mw ca? cg md cr? :::,-4 mm

Fi :6€°cu GJ _ esoc U U U U ‘

• an

[abge 11. Pouvwag Classifigagicn of t, D, C, and R,tR,2 I 1 t2

R1 4598 4708 4543 5075 4383 460091R23596 3341 3539 2943 2894 2991

R1 5263 5194 4988 5513 5113 4746C2R2

3421 3039 3221 3436 3259 3140

R1 4946 4784 4464 4099 4765 4387CaR22732 3003 2968 3135 3364 3018

R1 4657 4821 4941 4752 4772 4857C4R2

3669 3626 3463 3236 3376 3435

labga lg, Thraa··way Clagsiflgation of D, C, and R.Rap. T, t D1 D2 D3

R1 9673 9091 9143ClR2

6539 6235 6530

R1 10776 10307 9734

R2 6857 6298 6361

R1 9045 9549 8851C3R2

5867 6367 5986

R1 9427 9593 9798C4R2

6905 7002 6898

Tabga gg. _1}y_g·-gay Qgasggfigatiog of D and R.

R1 38921 38540 37526 114987

R2 26168 25902 25775 77845Grand Total 192832

1

1

Agpengix D

The regression analysis of the effect of bulk density on the flexuralstrength of shell molds was done in the following manner. Starting with the

data in Table 8; and letting X be bulk density and Y be flexural strength,

the following table was obtained:

X- Y, X·Y·gbggglß ;bg1§2 1 1.0553 799 44.1847.0530 699 37.0470.0514 593 30.4802.0509 644 32.7796.0504 607 30.5928.0549 768 42.1632.0526 704 37.0304.0517 616 31.8472.0506 672 34.0032.990g 609 90,3000

ix, : .5213 2Yi '-= 6702iXiYi'-" 350.4283

X ’.05213 Y = 670.2

zäxiz : .02720429 j§Yi2 = 4,538,316

2Xi2Y, ¤ 3493.7526The regression equation is:

v = SE + b(x - SE)where Y is the flexural strength, dependent on X,

Y is the average flexural strength,

X is a particular bulk density,

X is the average bulk density,

and 1 b is theslope.Theslope is obtained from the euatlon:

-

.whereN is the number of samples.

es 44 ••

Sclving for the slope using this equation:

__b· .02720429 - .o2v1*rse6s == 364199

The regression equation for these data ls thenY -· 670,2 + 36410.9(X ·· .05213)

The 90 per cent csnfidence limits were obtained for this equation,

the procedure for which is given in many referenceslo on statistics.

T

• an

A22e.¤¤ix.£

The following tables show the experimental data.

0 t')037S? $8 E8 S E S S F30svua t~· cn vo m ·¤· en2 S 2 2 2 S 2 2 53\•l

.5* 2 S S E 5%* Si E 2%^ 0 <*>_t·~ 4 0 an 2.0 0am

SS 2 2 2 2 -2 ··· ···2.2 2 2 2 .„ 2 2 2 S--10.;,:i‘„"" SM so G &*- «-·1 !*~2 2 2 2 2 S S S S ESS Si T3 "‘zu 0.

rdan -6-¤ c: G m co 102 ;S 2 2 2 2 S S3 SS El?•-1 wb- N O 0 CJ6 .6 QS S 1: 2 S S2 1: 2Z 'S 2 S2 2 E6 2 ·· 2 2 20 0 geo wr 0 Ü IÜ Ü••-4 -·· w •-1

2 ··· 2. 2 FI 2 2 20 C) 87 NÖ *3* l'~• <2*rs Mwu.,u>% £- M N 4*22.. 2 S S 2 2 E 2 2MJ J

••·-4 •-l

Q Q Q SI S Q $3 Q:0 0 |*• CO 0tg r·~|S “ 1: 2 2 2 2 2 2 2t~·~ t- 0 oo 10 SQ 10L1., •·—•

xi MH ns] MM az?] uf cg, 0:.74 mw-1QS --1 N M 41——• G U U Q

r-

m

t~ 0~·¤·o~o~1nNé zsäszwsas•0

Z"'2

:0 ·<rc00u-• .1:° *62N ~O

4:1.6

cz: N an cugé ggsääxäuä6-6

am

N <'>1£>Q·-1 O O9g äääääääsOm E ¤é" 65* F 65** .·£“ „?‘ ¤§"

Eru •··• 67 09 vU LD U U

6

•··-¢ W F •-—¢ ~O N ~O2 2 2 2 2 2 2 2···:¤31:: Q $1 W W <*> t*~ ····• c~ M····• ·-• -·-1 N <"> EU in2 gg F W O3 en CO 10 00 21*

fx SJ:‘ 4-* O U~N3

·GO Q § ~O F ä 2.2 W

éää··-4 N €"> N C') O uf} --422 _ 2 2 2 2 2 2 2 22 2 22 2 2 2 2 2 2 23 SI, gs Hg ~„O E W F~.Ocu•···1'ZI“ 22 2 2 2 2 2 23 2O‘~ an WFca

igä* R 8 2 "‘ —‘? ‘°

’___•·g-·l°0I> :3 ~D xO CD E CO ä2 2 ä2 "' 2* 2 22 ä 2 S "° äva Q Slbuä :0 GO F 10\nP•·4 Lu JgQ2 O O*~ db Ch F ¤D2%) _ 2 2 2 2 2 2 2 2

F 10 Q •-1 Ofi $2 2 2 2 2 2 2 2 2an S;2 2 2 2 2 2 2 2 2 2•‘ v-4 § R Q»-4 en xt': G2 ~O 00 ~03X O N ··~• F- 0*gg 2 2 2 2 2 2 2 2 2

O‘: nä" ¤é" Q2" :2** E MN 25* ¤é"r··1„Q:2 G uw uw ov

, - 49 •Table 7. Analysis of Variance

Souree Sum greee o anof Sggares Freedom Square F·Ratio

Rep. 112,522 2 56,261 6.512T 257,642 1 257,642 29.823t 17,641 1 17,641 2.042C 194,149 3 64,716 7.491D 17,074 2 8,537 .988R 4,790,028 1 4,790,028 554.466T x t 1,760 1 1,760 .204T x C 159,637 3 53,212 6.160T x D 23,510 2 11,755 1.361T x R 1,423 1 1,423 .165t x C 19,612 3 6,537 .757t x D 712 2 356 .041t x R 967 l 967 .112C x D 83,870 6 13,978 1.618C x R 85,430 3 28,477 3.296D x R 6,265 2 3,133 .362T x t x C 41,877 3 13,978 1.618T x t x D 12,415 2 6,208 .719T x t x R 3,917 1 3,917 .453t x C x R 96,478 3 32.159 3.723t x C x D 28,928 6 4,821 .558C x D x R 15,403 6 2,567 .297T x D x R 16,566 2 8,283 .959T x D x C 50,052 6 8,342 .966C x T x R 29,395 3 9,798 1.134t x D x R 5,146 2 2,573 .298T x t x C x D 61,306 6 10,218 1.183T x t x C x R 8,306 3 2,769 .321T x C x D x R 28,289 6 4,715 .546T x t x D x R 17,243 2 8,622 .998t x C x D x R 58,085 6 9,681 1.121

T x t x C x D x R 70,777 6 11,796 1.365

Error 1,641,392 190 8,639

Total 7,957,817 287

1

1 5C Ä

Table 8. Flexural Strengths and Bulk Density from the Investigation of

Replication No. 1Y X

Däatribvtiae Average DeneägvNumber DJ D2 D3 Strength lb/in

Flexural strength in psi1 871 166 161 799 .0553

2 574 751 773 699 .0530

3 558 572 650 593 .0514

4 610 647 675 644 .0509

5 593 582 647 607 .0504

Replication No. 2_ Y X

Average DensityNumber DJ Q2

D3Flexuralstrength in psi1 753 750 800 768 .0549

2 795 721 596 704 .0526

3 629 664 556 616 .0517

4 637 710 669 672 .0506

5 656 570 575 · 600 •0505

1

¤e· . EA „a. 2 „‘13-1 1 .1r 2. ku Q!1

Specimn Replicate 1 Replicate 2 Replicate 311,..,1;.: i .2 Dénit Ave . • j lr • :· ; —. Ä . . •.„ . « , 1 x „ I

lb in lb in lb in 18/in2 1b/in 1%(5n21 .0558 1040.3 .0536 869.0 .0541 6 5,02 .0566 919.3 .0538 921.0 .0546 756.73 .0556 668.7 .0536 742.0 .0536 494.04 .0533 490.0 .0530 k 566.7 .0543 532.35 ,0538 630.3 .0541 781.0 .0547 660.76 .0543 645.7 .0536 729.3 .0536 714.07 .0546 559.3 .0526 567.7 .0538 460.38 .0535 404.0 .0515 467.0 .0542 579.39 .0563 886.7 .0555 910.0 .0552 818.3

10 .0563 937.7 .0551 899.7 .0546 847.711 .0541 577.3 .0525 538.0 .0547 454.312 .0541 545.7 .0541 691.3 .0547 553.713 .0546 761.7 .0547 950.7 .0549 771.014 .0558 744.7 .0541 829.7 .0552 864.715 .0559 580.0 .0524 543.3 .0536 534.016 .0544 437.] .0517 561.7 .0549 488.317 .0487 747.7 .0505 865.0 .0502 865.318 .0498 735.0 .0506 773.0 .0506 740.019 .0495 544.3 .0506 459.0 .0498 569.320 .0498 532.7 .0517 652.7 .0501 536.721 .0509 840.3 .0503 619.7 .0501 793.322 .0493 720.3 .0495 741.0 .0494 707.723 .0493 473.0 .0497 444.7 .0498 410.724 .0488 475.3 .0493 481.0 .0495 494.025 .0525 814.3 .0527 801.0 .0531 801.726 .0544 766.0 .0532 780.7 .0523 804.727 .0532 598.3 .0529 603.3 .0528 612.728 .0538 514.0 .0541 554.3 .0542 604.029 .0533 908.7 .0527 800.0 .0525 686.730 .0520 763.7 .0531 803.3 .0535 875.331 .0535 614.3 .0529 598.7 .0529 558.732 .0532 623.0 .0534 565.3 .0538 488.3

I

I

CLAMP TOP PIECE

! IMOLD WALLPIECES

L··•¥

Ig g BOTTOM PLATE

if li.,1}— ·—-—— CLAMP BOTTOM

PIECE

—ASSEMBLEDMw

II

FIGURE I. FLEXURAL TEST BAR MOLD ASSEMBLY I IIII

I1 ·· 53 •

0:<I

5E<U1

ä 1.113 1-0

gf 1-0E gf, 15 1.1.1E X 9 9LU L5 cm EJ:13 1. QL9

E 0i §·~· 0 Z0 1’-1 ti}II 1 "..1° :11

D><L1J.1L1.

NLU(IIDS2LL.

0x0O.1

CDL1.1.1co< 1171D0 10<r

II

III1 I I 1 I IÄ- -I;-I I -IIF F I III1FF

1 I I I I I I IIÖFF “€

I V I;I VI 1 VVFFVV °w“‘ **1 F F 6*I-I ..II III

IIII:

6 ———° F

IF

I I II EF I- I- I---- I II é I I F IIII I I

II IIVFFIGURE3. I¤I.0T 0I= SAN0 ¤IsTRIauTIONs V

I

780_ • 500°F.U)¤. 0 600°F.

740EII-QD2 700u.!II-U)

660I3><uJ.Iu. 620

580I80 220 260 300 340AVERAGE GRAIN SIZE IN MICRONS

FIGURE 4. SAND COMPOSITION VS. FLEXURALSTRENGTH CURVES

•%•

900

EB¤.E 740:1;1-eszLU

E 660 • 6% RESIN°° 0 4% RESIN..1<:S111.1[L

500I8O 220 260 300 340AVERAGE GRAIN SIZE IN MICRDNS

FIGURE 5. SAND COMPOSITION VS. FLEXURALSTRENGTH CURVES

I

800 •

O750

E0.E 700 •EE-? \ •E 650 \

\ IIB \,\\

-J • Ü‘~

g ÜD 600 •Xu.1.1LL

550

500I 2 3 4 5

GRAIN SIZE DISTFIIBUTION NUMBER

FIGURE 6. FLEXURAL STRENGTH VS. GRAINSIZE DISTRIBUTION

I

.056

.055 •'°.Excd.054.J

EE.053 •U) OZLIJQ

.052ac__]O’ IJ

Q •

.05I _

.050I 2 5 4 5

GRAIN SIZE DISTRIBUTION NUMBER IIIFIGURE 7. BULK DENSITY VS. GRAINSIZE DISTRIBUTION III

II

*

•§9•

I

//

800 / •/

/ ./ I

750 /" /E / ,

/ /

E / / / // 0

/ /G / /Z0/E

650 / / /m ° /

/_J /4 · /X 600 • /{ ·I.1„I.1 //U" /

/550 /

500.050 .05I .052 .053 .054 .055

BULK DENSITY IN LB./IN.3

FIGURE 8. FLEXURAL STRENGTH VS. BULKDENSITY

I

I