Solidification Modeling of Complex-Shaped

Single Crystal Turbine Airfoils

K.O. Yu*, J.A. Oti.*, M. Robinson* and R.G. Carlson**

* PCC Airfoils, Inc., Bzmhmod, Ohio ** GEAiroraftEngines, Cincinnati,Ohio

Abstract

Single crystal ~ineairfoilscastingintegritydependsonstructuresas well as inherent defects. Pee Airfoils, in conjunction with the Air Force MANIEUI program and GE Aircraft Engines (GFAJZ), correlatedthe solidifi- cation conditions to the microstructures and defects of single crystal castings. !Ihe approach cmnbined finite-element therm1 analysis and experimental results. Simple-shapedcylinderswerefirstcastandmodeled to establish a defects prediction map. Solidification conditions of complex-shaped airfoils were then modeled, to predict the evolution of casting microstructures as well as the using the pre-established defects map.

occurrence of casting defects by Cmparisonofmdelpredicztionsto

experimental inspection results show that finite-elementmdelingisan effective tcol for understanding the solidification sequence and predictiq the miorostruoture, defects and chemistry variation of investment cast, cmplex-shaped, single crystal turbine airfoils.

Superalloys 1992 Edited by S.D. Antolovich, R.W. Stusrod, R.A. MacKay,

D.L. Anton, T. Khan, R.D. Kissinger, D.L. Klarstmm The Minerals, Metals & Materials Society, 1992

135

Introduction

Although single crystalcastingsprcanisehigherperformingp~~injet * they p==n tfoundrymanufacturing problen-~~.Foracomplexsingle

zz' blade or vane design, optimizingalloastingvariablesis almost impossible with state-of-the-art,oastingmethods. Finite-elementthermal modeling offers the capability of simulating casting solidification and understanding the solidifioationseguenoe. Effortswereundertakn inthe Air Foro.e/GEAE MANTECH program and at Pee Airfoils, to develop a software (Sixqle Crystal Modeler, SCM) arxdestablishamethodologyto sirmlatethe solidification of single crystal castings. TheScMtakesgeometrioaldata, oonstructs a 3-D model of the part, adds sating and mold material, emeshes the mold-metal system, perform solidification simulation, displays the results graphically, andpermitsvariationsbythecastingengineersothat thebestproduotionprooess c!anbeseleoted.

The approach oombined finite-element themal analysis andexperimental results. Simple-shapedcylinderswere firstcastandmdeledtoestablish a defects prediction map. The solidification conditions of two corqlex shaped airfoils, high pressure turbinevane (HPTV) andlowpressur eturbine blade (LPIB), werethenmodeledtopredictthecastingmirrostructures and defects, by using the pre-established defeotsmap. Ihemodelpredioted results were validated by ocmparing them to experimental inspection results. To further validate the applicabilities of the modeling method- ology and defects mapestablished intheMANTECHprogmm, anotherairfoil (high pressure turbine blade, HPTB) was modeled byusingaommeroial finite-elementoode ProCAST for a GFAE internally supported program.

Modeling results and the method for developing defects map frm simple shaped cylinders have been reported elsewhere 8) Thepuqoseofthis paper is to present results from the modeling of a0mpi=-shaped airfoils.

Amroach

The approach combines finite-element thermal analysis andexperimntal results. Cyl' ers prediction map1 . t"sz

were first cast andmdeledto establishadefeots Several molds of fxnplex-shaped airfoils were then

cast and the solidification conditions of those airfoils were modeled, to predict the evolution of casting miorostructures and the occurrence of casting defects by usiq the pre-established defects map. Finally, these mdelpredictionswereccqaredtocasting inspection results.

ExtxrimentalPmce43~e

Clusters (Figure 1) of three different omplex-shaped airfoils (HPTV, LPIB and HPlB) wereoastatspeoified furnaoetemperatureandwithdrawalspeeds. Thermocouples were placed inthemold, metalilndooreto reootithetemp- eratureasa fun&ion of time at specified locations. After shake-out of the mold, castings were inspected by grain etching, F'PI (fluorescent penetrant inspection), X-ray and Laue' X-ray for grain defects, porosities and primmy dendritedirections. Numerousspecimens werecutfrmas-cast castings to measure the primaryandsecondarydendriteamspxing(PDAS and SW). For one particular airfoil configuration (HPIB), samples were taken frcm root and tip regions for Y, Al, Si and Zr analysis.

Finite-Element Modeling

Details of the prooedur e for finite-element tieling of single orystal castings have beendesoribed inReferenoe1, and interestedreaders should

136

reference that paper. In general, metal and core gemetriescanbe generated from any cmmercial software such as PATRAN, IDEAS and ANSYS. The mesh for the ceramic mold around the metal was semi-autmatically generatedthroughtheuseof EXTRUDEprograminIDEASforHprvandLmB models. A Pee Airfoils proprietary software was used to g enerate mold for the HPTE3 model. l%e finite-elementsolutionswereobtainedbyTOPAZ/ SDRC forHPl?VandLPIBandbyProCAST for HPIB. isothems,

Moaelingmzsultsincluded isochmns (time toreachthespecifiedtemperatums), teqera-

ture gradients G, solidification rates R, combinations of G and R, cooling rates, local solidification times (LST) and cooling curves.

The finite-element m&ielofHPWis shown inFigure 2. Figure 3 shmsthe cluster configurations ofHPWandLPl73. furnace,

Itcanbeseenthatthepart,the and the copper chill are all incorporated intothese finite-

element models. The initial temperature conditions of the metal, core arCi mold were assigned based on the th ermcouple mental castings.

readings of those experi- ?heboundaryconditions of the fumacewall temperature

werealso appliedbasedonth ermocouple data.

FIGURE 1. Experimental Cluster of HPTV. FIGURE 3. CLuster Configuration of HPTV and LPTB Models.

i ; GlttO !Gi /&‘

Aidoil pi

Ramps g; j &i ,a.

HPTV LPTB

FIGURE 2. Finite-Element Model of HPTV. FIGURE 4. Model Calculated Isotherm of LPTB at Various

Uithdraual Times.

The calculated isothems of WIBatvariouswithdrawaltims areshown in Figure 4. Figure5~~thecalculatedcoolingcurvesof~. lkmo- couple data at the location of Node 2689 is also shown.

Results and Discussion

Emed on the pre-established defects map, microstructures and defects of the cmplex-shaped airfoils are predicted and compared with casting

137

inspection results. The change of Y, Al, Si and Zr contents of I-iPTB is also correlated with model calculated solidification conditions.

DendritekmSmcing

For dendritic solidification, D7G is the most convenient parameter f representixqthedegreeofmi crosegregation of castings. me fmC1rT of DW is directly related to LST or average cooling rate during solidifi- cation. As heat is extracted faster frcmcastings (i.e., shorter ISTard higher average cooling rate), thesizeofthe resultedDAS andthedegree of the associatedmicrosegregationarealso smaller. TheDASpredictorwas established by correlating lIIeasWed~valuesfrrm-lexperjnleIltalcastings to model calculated LST. Figures 6 and 7 shm this correlation for FIXS and srvs. Dasedontheseresults, Figure8 presents a cmparisonof LPIB model predicted DAS values versus measured DAS values.

GrainDefects

Nurmmus graindefectSwerefoundinexperimentalcastings,including~~ low angleboundaries(HABandLAB),bigrains,equiaxedgrainsandfreckles. The formation of these grain defects are related to casting solidification conditions. Misoriented dendrite (MOD) is not a grain defect but its for- mation is clearly related to the contour of the solidification front, and is also discussed in this section. On the other hand, sliver is a type of grain defect which is primarily related to metal-ceramic reaction and is discussed in that section.

0 2000 1000 6.500 WITHDRAWAL TIME, SECONDS LOCAL SOLIDIFICATION TIME, SECONDS

FIGURE 5. Model Calculated ard Experimentally FIGURE 7. Measured SDAS as a Function of Model

Recorded Cooling Curves of LPTB. Calculated LST.

I PREDICTED MEASURED - io30

E 1_1A\̂

96 pm 6 SDAS L 105 )~rn SDAS = 88 “in *

616um sPDASL924um “DAS = 896 “iI f siiu

2 g h-00 .- I_ CL 200 0 S!YPLE SHAPED CYLIfIUERS

,- CC’VLEX SHAPED AIPF”I!S ,

c 800 1600 :-,. ;c::

LOCAL SOLIDIFICATION TIME, SECONDS

FIGURE 6. Measured PDAS as a Faction of Model Calculated

LST.

138

SDAS = 64 MI

: 67 Zpm

Hiqh/Lm Amle Boundaries andBiqrains. Graindefects that are functions of the contour and orientation of the solidification front include HAB, LAB, big-rains and multigrains. Dendrites grow opposite the direction of the heat flow, perpendicularto themushy zone contour. Whendendrites of different orientation meet, HAB or LAB may form depending on the magnitude of theanglebetweendendrites. In Figure 9, the trailing edge of the HPTV cools much faster than the leading edge side. Consequently, the mushy zone forms an "L" shapedcontour,ardagrainboundaryisexpe&ed to form at the location where dendrites with different grmth directions meet. In the -tins, the dendrite grcrwth direction is consistent with the model prediction. The two dendritic orientations meet to form a grain boundary. Because this is a single crystal alloy, therearenograinbomdary strengthenirqconstituents, andacrack forms alongthegrainboumlaq.

Sametimes a second solidification front is alsopresentinthecasting. Usually this second solidification front is locate3 near the trailing edge of thecastingbecauseitisthinthereand,hence,ittendstocoolfaster than the rest of the casting. As a result, two grains with different cry~llogmphic orientations form. This type of defect is called bigrains. If more than two grains are presentsimultaneouslyinthe casting, it is calledmultigrains.

C*8TIHG BTACI(INO 1 3010 BBEculdS ,180 seconda ISOTHERM PLOT

----------__

FIGURE 9. Conparison of Model Predicted and Experi- FIGURE 10. Calculated Isotherms at Various Uith-

mentally Inspected HAB for HPTV.

t CASTING STACKING AXIS

0 .i- iI 6,

: k CA p-y

drawal limes for Lot 15 of HPTB.

G/R AT ‘!‘, PLI

/ EDUIAXED

GRAINS I

FIGURE 11. Primary Dendrite Growth Direction FIGURE 12. Conparison of Predicted and Inspected

of a HPTB Casting of Lot 15. Equiaxed Grains for HPTV.

Misoriented Dendrite Direction. Since thedendritegrmthdirectionis directly opposite to the heat flow direction and perpendicular to the mushy zone contour, the mushy zone contour is also used to predict the MOD.

139

Laue' back reflection x-ray technique is used to determine the deviation of the primary dendrite growth direction frm the stacking axis ofthe airfoil. When deviations are beyond a specified value, castings are rejectable. In Figure 9 themshy zonecontour indicatesthattheangle between the dendritegrowthdirectionardthe stacking axis oftheHPIVis high and thus the castings haveahi~tendencytoberej~byLaue' inspection. Laue' results of these castings show a high casting rejection rate, which is consistentwithmdelpredictions. Figure10 showsthatthe calculated isotherm contours of HPTB are always peqendicular tothe casting staoking axis, indicating the primarydendritegrmthdirection should be parallel to the casting stacking axis. This prediction is consistent with the results frm microstructure evaluation (Figure 11). Laue' resultsalsoshc~thatallcastingspassthe inspection.

Eauiaxed Grains. INring solidification, the solidification front morph- ology is controlledbythe ccmbi.nationofthetemperaturegmdientG inthe liguid phase just ahead oftheadvanc'

(3r4Y interface, andbythe interface

velocity or solidification rate R Inthisstudy,eguiaxedgrains are predicted by theG/Ratthelic&lustemperature. ThehighertheG/R value, the lower the eguiaxedgrains formationtendency. Figure12 cm- pares a G/R plot with a grain photograph of an HlTV casting. The low G/R patches on the leadingedgeofthemdel (which indicatesequiaxedgrains will form) are consistentwiththe eguiaxedgxains found onthe casting.

Freckles. Interdendritic fluid flow in the mu@qqone ~~~~f&r cation is responsible for freckle-type defects shcweii that the formation of this interdendritic fluid flow is controlled by the casting averagecoolingrateduringsolidi ' tion. viously established freckle formation criterion 87

Rasedonpre- model predictions

indicate that'no freckles should be present in &asecastings. This conclusion is consistent with casting inspection results.

shrinkDefects

There are two specific types of shrink;macroshrinkandmicroporcsity. Macroshrink is controlled by bulk heat transfer conditions, whereas microporosity is related to the feeding conditions in the mushy zone.

Macroshrink. Inthe casting, ifthereis anisolatedreg'onwhichislast to solidify, then mmmshrink forms in that region lo . t ) Eguiaxed castings usually have a higher tendencyto formmcroshrinkthandirec- tionally solidified and single crystal castings do. The formation of macroshrink is predicted by the contour of the solidus isochron. An isochron is a contour plot which illustrates regions of constant time for castings tocooltoa specifiedterqerature. In the solidus isochron plot, an isolated spot with a higher time to reach solidus teqeraturemeans metal in that spot is thelastto solidifyandmzroshrinkis expe&edto foKm inthatspot. There arenoisolations in Figure13 which couldcreate macroshrink, and this prediction is verified by casting inspection results.

Microrm-os.itv. Microporosity forms near the end of solidification ' e inte&andritic regions when capillary feeding becomes insufficient l . rsr Thus the propensity of the microporosity formation is related to the casting feeding ability during the last stage of s@Q$jcat~,l~~~ controlled by the G/R at the solidustemperature . G/R at solidus value, the higherthetendencythatmicroporosity form. Microporcsity can showuponorbeneaththesurfaceofthe on previously established microporosity formation criterion results indicate very lowtendency forsurfacemicroporosity foLation in

140

the HPTV castings, and verifies this with FPI inspection results. Figure 14 compares the internalmicmporosity seenatx-rayontheHPIVwitht.he model predicted results. The model prediction indicates that all the castings should have internal microporosity. This prediction is not coqletely consistent withcasting inspectionresults. However, themodel predicted relative trends (i.e., a bigger light color area indicates a higher tendency) to fonn microporosity in castings of these four molds are veryconsistentwith inspection results.

Effects of Metal-Ceramic Reaction

Chemical reactionbetween ceramics (mold and core) and active elemants such as Y inthemetalresultsinadecreaseofmetalYcontentandinrreaseof metal inclusion content Thelatereffectthen increases the tendency for the form&ion of slivers.

Active Elements Control. For each mold of HPIB, samples were taken from root and tip regions of castings for Y, Al, Si and Zr analysis. Results are shown inTable1. Fmnthetable itcanbe seenthat:

1. For all molds, the Y content in the tip region is always higher than that of the root region, whereas tip Si content is always lower than the rc0tSiconter-k. This indicates thattheprimaryreason foryloss is due to the reaction between Y and SiO correlation between Al and Zr contents and Y &lt?we

seems no clear con

2. I&s 15 and 16 have a 40% lower initial Y content than Lot 5, but the casting Ycontent~orthesethreemoldsispradicallythe same. If we define the ratio of the casting final Y content to its initial Y content as Y retenticm efficiency, Table 1 indicates that Lots 15 and 16 have a much higher Y retention efficiency than I& 5.

3. Casting Y content of Lots 11-13 is significantly higher than that of Lot 5 (all four molds have the same casting condition). Although Lots 11-13 have a 10% higher initial Y content than Lot 5, the magnitude of the difference of the resultant casting Y content is still a surprise.

TIME TO REACH TL

TIME TO REACH TS

No. of Castings With Defects / No. of Castings Inspected

-

Long Time

319 ME& J- Mi3

s/9

MOLD 1 MOLD 4

Long Time

- Short Time

G/R AT T, PLOT

FIGURE 13. Isochron PLot of LPTB for FIGURE 14. Comparison of Model Predicted and X-Ray

Prediction of Macroshrink. Inspected Internal Shrink for HPTV.

Assuming a constant initial Ycontent, theamuntof Ylossduringcastirq depends on the~~calreactionratebetweenmoltenmetaland surrounding ceramics. Threemostimportant factors controlling the kinetics of this metal-ceramic reaction are casting geometry (surface to volume ratio), reaction temperature and contact time. For the data shown in Table 1, casting gemetry and reaction tfmperature (all six molds are cast at the same preheat and pour temperature) are practically constant, andonly

141

metal-mramic contact time is the importantcontmlling factor. Alonger contact time results in a bigger Y loss and a lmer Y retention efficiency.

FIGURE 15. lsochron Plot for Lot 5 and 15 of FIGURE 16. Transverse and Longitudinal Vieu of a Sliver

HPTB.

Figure 15 shms the isochron plot for castings of Lots 5 and 15. In this figure, the isochron isthetime forthemetaltocoolfrmpourtmpera- ture to themetalsolidustemperature. l%us,theisochronshmninFigure 15 represents metal-cerami c contact time before metal is completely solidified. Fran the figure it can be seenthattheisochmnofthe castings of Lot15 arelowerthanthatof L&5, indicating castings of Lot 15 should have a higher Y retention efficiency than that of Lot 5. A similar conclusion was reached for L&16. Thesemmlelpredictions are consistent with inspection results shown in Table 1.

TABLE 1. y content of HPTB castings with various casting conditions and initial Y content.

A - Index for Initial Y Content C - Metal-Ceramic Contact Time (set) B - Index for Y Retention Efficiency * - Average Value

FYom a geometrypointofview, thecastingrootregionhas almersurface to volume ratio than the tip region, thuswewould expect the root region to have a higher Y retention efficiency than the tip region. On the other -, Figure 15 shows that the root region always has a significantly higher isochron than the tip region, indicatingtheY contentinthe root region should be lower than the tip region. Table 1 shows, for all molds, the actual Y content and the retention efficiency in the rcot region is lmer than that of the tip region. These resultsindicatethatthe influence of isochron (metal-ceramic contact time) on Y retention effi- ciency is more important than that of geometry (surface to volume ratio).

142

Slivers. Slivers (Figure 16) are grains forming streaks in the microstructure. They are usually aligned c 0

IT- to the primary direction,

but misoriented in the transverse direction ' . The formation of slivers is believed to be related to the metal inclusion contentduringthe cast-. These inclusions serve as the nuclei for the formation of slivers. A higher inclusion content results in a higher sliver formation tendency* Assminga constantinclusioncontentofthe input material, the actual inclusion content during the castirq depends on the chemical reaction rate between metal andceramics (mold andcore). Thus, casting sliver formation tendehcy, similar to Y retention efficiency, is predicted by the casting solidus isochron plot. A higher solidus isochron (i.e., higher metal -ceramic contact time) indicates ahigher inclusioncontent a, h-t a higher sliver formation tendency. The isochron of castings of Lot 15 is lower thanthatofLot5 (Figure15),which indicates that castings of Lot 15 should have fewer slivers than castings of Lots 5 and 11 to 13 (these four molds have the same casting conditions). A similar con- clusion is obtained for Lot 16. Grain inspectionresultsofthesemolds show that castings of Lots 5, ax-ii 11 to 13 have 5% slivers, whereas castirqs of Lots 15 and 16 have none.

Conclusion

Finite-element tieling proves to be aneffectivetcol forunderstanding the solidification seguence and predict- the microstructures, defects and chemistry variation of investment cast, complex-shaped, singlecrystal turbine airfoils. Comparison of model predictions to experimental inspectionresultsshows:

1. Both PINS and SDAS are afunctionoftheLSTortheaverage,cooling rate between the alloys liquidus and solidus temperatures. Dasedon the previously established relationship between DAS and LST, the distribution of casting I can be predicted.

2. Primary dendrite growth direction is directly opposite to the heat flow direction and is perpendicular to the mushy zone contour. Thus, the casting isotherm plot (especially the mushy zone contour) can be used to predict various types ofgraindefectssuc.hasIAEJ, HAl3, bigrains andmultigrains.

3. The mushy zone contour is also used to predict the magnitude of the angle between the primary dendrite growthdirectionandthecasting stacking axis. 'Ihis type of defect is called a misoriented dendrite

WD) l The larger the angle ofthemD, thehigherthetemdency for

the casting to be rejected by the Laue' x-ray inspection.

4. Freckles are predicted by the average cooling rate between the alloy's liguidus and solidustemperatures. Thecastingcoolingrate shouldbe higher than a specified value to avoid the formation of freckles.

5. The casting's G/R at liguidus temperature value should be higher than a specified value to avoid the formation of equiax&i grains.

6. Macroshrink is predicted by the contour of the solidus isochron. If there is an isolated spotwherethetimato reach the solidus tempera- ture is higherthanthe surrounding area (i.e. mtalinthe spot is the last to solidify), macroshrink forms in that area.

7. Microporosity (both surface and internal) is controlled by the value of G/I? at the solidus temperature. The lower the G/R value, the higher the tendency for microporosity to form.

143

8. Isochron is also used to predict the formation tendency of sliver grains and the retention efficiency of active elements such as yttrium. Both phenomena arerelatedtothechemical rmctionratebetweenmlten metal and ceramics (mid andcore) which is then relatedtothemlten metal-ceramics contact time. The higher the~metal-ceramiccontact time, the lower the Y retention efficiency and the higher the sliver formation tendency.

Acknowledcrments Part of this work (HPTV and LPTB) was supported by contract No. F33615-85- C-5014 (GEAE was the primary contractor and Pee Airfoils was a subcon- tractor toGEAE) witht.heAirFoxceWrightResearch and Develowt Center, Manufacturing Technology Directorate (WRDC/MT), Wright-FWtersonAFB. The HPTB portion of the work was sponsored by GEAE (P.O. No. 200-14-lOW95668).

1.

2.

3.

4.

5.

6.

7.

8.

9.

10.

11.

12.

13.

References K.O. Yu et al.: "Solidification Modeling of Single Crystal Investment casting", AFS Transactions, 1990, 417-428.

R. Mehrabian; Proceedinqs of the Int. Conf. on RaDid Solidification Processinq, (Reston, VA, 13-16 Nov 1977), 9-27.

M. Cohen, B.H. Kear and R. Mehrabian; Proc. 2nd Int. Conf. on Rmid Solidification Processirg, (Reston, VA, Mar 1980), l-23.

M. McLean; Directionally Solidified Materials for Hich Tenmerature Service, (The Metals Society, Lmdon, 1983).

M.C. Flemiqs and G.E. Nemo; "Trans. Met. Sot.", m, 239 (1967), 1449-1461.

M.C. Flemings, R. Mehrabian and G.E. Nemo; "Trans. Met. Sot.", m, 242 (1968), 41-49.

S. Lm, D.R. mirier, and M.C. Flemings; "Met. Trans. B.", s, (1978), 771.

J.A. Dmingue, K.O. Yu and H.D. Flanders: Characteristics of Macroseqr~ation in ESR IN-718, (Symposiumof Flm%mentalsofAlloy Solidification Applied to Industrial Processes, NASA, Cleveland, OH, Sep 1984), 139-149.

M. Gel1 and D.N. Ihhl; Processinu andPro~rtiesofAdvanc&Hiqh T'3mmratm Allovs, s. Allen et al. editors, (Metals Park, OH: ASM, 1986), 41.

M.J. Beffel, K.O. Yu, M. Robinson and K.R. Schneider; "Computer Simu- lation of Investment Casting Processes", Et!!, 41, (2) (Feb 1989), 27-30.

L. Quichou, F. Lavard and G. Lzsoult; Smerallo~s 1980, Prcc. of the 4th Int. Syn'p. on Superalloys, J.K. Tien et al. editors, (Metals Park, OH: EN, 1980), 205.

E. Niyama, T. Uchida, M. Morikawa and S. Saito; CIATF, Paper No. 10 (1982).

J~cq&@!eckers; "Study of Microporosity Formation in Nickel-Base It, Met. Trans. A, 19A (Sep 1988), 2341-2348.

144