PROCESSING EFFECTS ON THE HIGH CYCLE FATIGUE LIFE OF WELD REPAIRED CAST Ti-6A1-4V PARTS by GORDON BRUCE HUNTER SUBMITTED TO THE DEPARTMENT OF MATERIALS SCIENCE AND ENGINEERING IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREES OF MASTER OF SCIENCE and BACHELOR OF SCIENCE at the MASSACHUSETTS INSTITUTE OF TECHNOLOGY June © 1981 Gordon Bruce Hunter, 1981 The author hereby grants to M.I.T., the U.S. Army and any agency of the U.S. Government permission to reproduce and to distribute copies of this thesis document in whole or in part. 7 Signature of Author Department of Materials Science and Engineering / M;;v 8, 1981 Certified by I Thomas W. Eagar iThesis Supervisor Accepted by I Regis M.N. Pelloux J hhairman, DepartmentkGraduate AR1CHIVES a MASSACHNLUSETTSST JUL 1 7 1981 L1BRARIES ira, Committee John B. Vander Sande Department Undergraduate Committee -2PROCESSING EFFECTS ON THE HIGH CYCLE FATIGUE LIFE OF WELD REPAIRED CAST Ti-6A1-4V PARTS by Gordon Bruce Hunter Submitted to the Department of Materials Science and Engineering on May 8, 1981 in partial fulfillment of the requirements for the Degrees of Master of Science and Bachelor of Science at the Massachusetts Institute of Technology. ABSTRACT Specimens were processed according to a twelve-run Plackett-Burman statistical screening design to study the effects of weld repair on high cycle fatigue properties of cast Ti-6A1-4V. Variables studied were betatreatment before welding, weld bead size, post-weld hot isostatic press treatment, heat treatment time and temperature, heat treatment cooling rate, surface polish, and peening. Specimens were tested by R.R. Moore rotating beam fatigue machines at 72.5 ksi and 10,000 rpm, and then examined by metallography,fractography, x-ray radiography, RMS surface measurements, and chemical analysis. The results were also compared to the high cycle fatigue properties of unwelded specimens. The analysis of the results showed that weld repair is not detrimental to high cycle fatigue properties of cast Ti-6A1-4V. Surface condition was shown to be the most critical variable with a finer surface finish and a higher surface compressive stress improving fatigue properties. Mechanical polishing was superior to glass bead peening, which was superior to chemical milling. Microstructure was also shown to be a critical variable with a larger colony size improving fatigue life by slowing fatigue crack growth. The weld zone resisted fatigue crack initiation due to a basketweave structure and thin grain boundary alpha. A secondary hot isostatic press (HIP) treatment was shown to be unnecessary if the weld zone is sound. A HIP treatment may be made either before or after welding without detrimental effects. Multiple treatments were shown not to be detrimental. Multiple pass repair welds did not affect fatigue life when compared to single pass welds. Thesis Supervisor: Title: Dr. Thomas W. Eagar Associate Professor of Materials Engineering - 3- TABLE OF CONTENTS PAGE TITLE PAGE 1 ABSTRACT 2 TABLE OF CONTENTS 3 LIST OF FIGURES 5 LIST OF TABLES 7 ACKOWLEDGEMENTS 8 CHAPTERS I. II. III. IV. V. VI. VII. INTRODUCTION 10 PREVIOUS WORK 19 EXPERIMENTAL PROCEDURE 24 A. Statistical Design 24 B. Prior Microstructure - Variable 1 28 C. Bead Size - Variable 2 33 D. Hot Isostatic Press - Variable 3 35 E. Heat Treatment - Variables 4 and 5 35 F. Surface Finish - Variables 6 and 7 38 G. Testing Methods 40 RESULTS 42 DISCUSSION 65 SUMMARY AND CONCLUSIONS 70 POSSIBILITIES FOR FUTURE WORK 72 -4APPENDIX I. EFFECT OF TRIAL 5 RESULTS 73 APPENDIX II. FAILURE STRESS AT FRACTURE 76 BIBLIOGRAPHY 79 -5LIST OF FIGURES FIGURE PAGE 1 Solar prototype APU impeller in the as-cast condition. 16 2 Early DDA prototypes for a cast impeller. Impeller on right is as-cast and impeller on left has been chemically milled to final contours. 17 One of first two of mold. 18 3 UH60A main hub casting, as broken out 4 Diagram of tube furnace. 32 5 Diagram of a machined blank. 34 6 Diagram showing positions of weld passes. 37 7 Diagram of the dimensions for the specimens after machining. 39 8 Specimen 4 base metal. 48 9 Specimen 19 base metal. 48 10 Specimen 36 base metal. 49 11 Specimen 38 base metal. 49 12 Specimen 1 weld metal. 50 13 Specimen 32 weld metal. is shown at bottom. Fracture through the weld met,al 50 14 Specimen 37 weld metal. 51 15 Specimen 52 weld metal. 51 16 Specimen 1 microstructure showing transition from weld metal to base metal. Weld metal is shown in the upper part of the figure. The bottom of the figure shows the fracture plane across base metal. 52 Specimen 10 fatigue origin showing fracture facets. Lower part of figure is the outer surface of the specimen. 57 17 -6- 18 Specimen 9 fatigue origin showing fracture facets. Lower part of figure is the outer surface of the specimen. 57 19 Specimen 13 fatigue origin in weld metal. 58 20 Specimen 21 fatigue origin showing smeared appearance. 58 21 Specimen 7 outer surface showing mechanically polished condition. 59 Specimen 3 outer surface showing chemically milled condition. 59 Specimen 10 outer surface showing mechanically polished and peened condition. 60 Specimen 4 outer surface showing chemically milled and peened condition. 60 22 23 24 25 Specimen 9 outer surface showing peened condition. (a) shows the secondary electron image with the fracture origin and part of the fracture surface in the upper right section of figure. (b) is an x-ray map showing concentrations of silicon. Note that the dark areas in (a) correspond to silicon conditions in (b). 61 26 S/N plot 27 S/N plot for welded and unwelded specimens. 64 28 Plot showing the effect of various values for trial 5 on the statistical results. 75 Diagram showing the method of determining the specimen radius at the fracture. 77 29 for the screening experiment specimens. 63 -7LIST OF TABLES PAGE TABLE NUMBER 1 Twelve-Run Plackett-Burman Design 25 2 "t" Distribution 27 3a Rod Composition 29 3b Rod Mechanical 4a Process Variables 30 4b Rod Processing 30 5 Welding 6a R.R. Moore Results 43 6b Additional R.R. Moore Results 43 7a Fatigue Life Statistics 44 7b Fatigue Life Significant Effects 44 8 Specimen Surface Finishes 46 9 Slug Analysis 47 10 Specimen Colony Size 54 lla Colony Size Statistics 55 llb Colony Size Significant Effects 55 12 Comparison of Fatigue Life and Colony Size 56 13a Effect of Trial 5 on HCF Factor 74 13b Effect of Trial 5 on HCF Significant 14 Calculated Failure Stresses Properties 29 Schedule Parameters 36 Effects Effects 74 78 -8ACKNOWLEDGEMENTS Most of the work contained in this report was supported by AMMRC in Watertown, MA through the cooperative studies program of the Department of Materials Science and Engineering at MIT. Additional funding for work in MIT's weld lab was provided by Mr. Bruce MacDonald of the Office of Naval Research. The author is grateful for the support of all three agencies. Special thanks go to all the guys in the MIT welding lab, especially Mr. Bruce Russell,and everyone in the Prototype Development Division at AMMRC. Everyone has been very kind and more than helpful in the past year. The author would also like to acknowledge the expert and friendly help of Mr. Andy Zani and Ms. Carolyn Shimansky at AMMRC with the metallography and Mr. Andy Connolly of AMMRC and Ms. Libby Shaw of MIT with the electron microscopy. Ms. Rosalie Allen has been a godsend for her speedy and accurate typing job, as well. Mr. Al Hammer of Solar and Mr. Stan Silverstein of Sikorsky have been very helpful with technical advice. Professor Thomas King of MIT has also been of great service by finding this co-op assignment for me. It would be negligence to fail to mention the friendship and support of all the friends, associates, "family" and the UGOI who helped maintain my sanity and the spirit of M.O.S., Inc. The author is indebted to the advice, help and friendship of both Mr. Frank Hodi and Dr. Robert French of AMMRC. Nothing I could ever write expresses my gratitude to Professor -9- Tom Eagar. He continues to be a great teacher, employer, advisor, and, above all, friend. Ms. Linda Schaffir has been a wonderfully stablizing influence in the massive confusion of a college education. Besides aiding the author with the figures in this paper, she has been the best companion a guy could ever know. And, of course, there is Mom and Dad. They stood by me when I couldn't keep my "d"'s and "b"'s straight and when I couldn't ski down the big hill, having to walk instead. No matter how cliche it might be (and this author hates cliche), this is dedicated to them. - 10 I. INTRODUCTION In an effort to reduce the cost and lead time for the manufacture of rotating titanium aircraft components, various commercial contractors with support from the U.S. Army have been studying possibilities for re- placement of Ti-6 wt% A1-4 wt% V (Ti-6-4) forgings with titanium castings. [1-4] These parts can be cast to near net shape, reducing the amount of scrap metal and reducing or avoiding time-consuming and expensive machining operations. While most mechanical properties of the cast material have been comparable to the forged material, the fatigue test results of the castings have generally been inferior and have been characterized by high scatter. [5-8] Weld repair to ensure the integrity of cast com- ponents is necessary to reduce the rejection rate, thus making castings more attractive economically when compared to forgings. There has been concern that welding would degrade the fatigue properties and make castings unacceptable for rotating components. In general, Ti-6-4 castings have a tendency for two types of internal defects; inclusions near the mold interface, and voids due to gas porosity and solidification shrinkage. [5, 6, 9, 10] Both types of defects have been shown to reduce fatigue strength. [7] The inclusion problem, in many cases, has been solved by casting the components slightly oversize and then etching the surfaces away by chemical milling. The milling is continued until final dimensions are attained, eliminating surface inclusions and any casting investment remaining. Chemical mill- ing is also used to remove contaminated surfaces formed during heat - 11 treating. Hot isostatic press (HIP) treatments have been successfully used to eliminate internal voids. [6, 8, 9, 11] Further heat treatments are generally needed to improve the microstructure, increase the fatigue strength, and reduce the scatter in properties. Closure of internal voids during the HIP cycle may produce dimples on the surface of the component. Further, there can be inclusions in the structure which are too deep to be removed by chemical milling and must be machined out. In either of these cases, weld repair is employed to backfill and build-up material. The most popular method of weld repair employs gas-tungsten arc welding (GTAW) with argon gas and Ti-6-4 ELI (extra low interstitial content) filler wire. [6, 11, 12] Following welding, the component is either stress relieved, or otherwise heat treated to prevent distortion in service. follows welding. A second HIP treatment often [6, 9 10, 13] Repair welding is generally accomplished by making repeated passes, laying down filler metal so that the part will meet specified dimensions.[12] Small surface defects may be closed by a single shallow pass without filler metal. [12] High purity must be maintained during welding as small amounts of oxygen may affect the mechanical and fatigue properties of the material. [14] Following weld repair, a HIP treat- ment is sometimes required to ensure closure of any porosity or voids that may have formed in the weld zone. No published studies show the effect of bead size or number of weld passes on internal stresses, microstructure, or mechanical and fatigue properties of cast Ti-6-4. Few, if any, of the applications which are anticipated would re- - 12 - quire components with the as-cast surface intact because of the variability of the as-cast finish and because of processing procedures, such as weld repair, mechanical machining, and chemical milling. Additionally, chemical milling is required to remove the surface casing if the heat treatments are conducted in a contaminated atmosphere. Many components are glass bead peened for a final surface finish, inducing surface compressive stresses which improve the fatigue properties. [15, 16] Titanium exists in two allotropic forms: a lower temperature closepacked hexagonal structure (), and a body-centered cubic structure (). Because of the relative stabilizing effects of the alloying elements, Ti-6-4 has a mixed a + structure with the -transus near 1823°F (995°C). [17] Cast material cooled from the structures. The basic morphology is a Widmanstatten structure with broad a-platelets separated by [18] -phase region has several common -films and a complex interfacial phase. The platelets are arranged in either a structure. [19] A colony is a packet of similarly aligned and crystallo- graphically oriented platelets. [19, 20] the prior basketweave or colony Long a-platelets also form in -grain (PBG) boundries, called grain boundary (GBa). [21] Additional heat treatments can form martensitic structures (a' and a"), or recrystallized primary a-grains interspersed in the Widmanstatten structure. 22] The morphology, colony size, and a-platelet width depend on the cooling rate, the concentration of B-solutionizing conditions. [18-20, 22] -stabilizing elements, and the Rapid quenching will aid the formation of a martensitic structure in preference to a Widmanstatten - 13 - structure. [22] or with increased Colony size decreases with an increase in cooling rate -stabilizer content. [20] A basketweave structure re- sults when the colony size is so small that the transformed region appears to be a distribution of a-platelets in random orientations. lets and wide GBa are formed during slow cooling. [20] Wide plate- Heat treating titanium in the presence of oxygen causes an a-casing to form due to the a-stabilizing character of oxygen. This casing usually does not excede about .010" (.25 mm) in depth. [9] Many published studies have shown that PBG size and colony size have an effect on fatigue crack initiation and growth in a + alloys. 8, 19, 20, 23-29] titanium The similar crystallographic orientation of a-platelets causes slip and shear across a colony in certain favorable orientations, thus facilitating fatigue crack initiation. [19, 23] While colonies promote crack initiation, they also promote branching and deviation of the main crack from its preferred course by the same slip mechanism thus reducing the crack growth rate, as shown by Yoder, Cooley, and Crooker. 23-27] As fatigue cracks must reinitiate at each colony boundary to propogate through the material, Eylon and Bania concluded that a smaller colony size would slow crack growth.[20] Based on these results, Yoder and Eylon suggested that the fatigue crack growth rate decreased with increasing colony size if the specimen thickness was much larger than the colony size and that the rate increased with increasing colony size if the colony size approached the thickness of the specimen. [28] This means that for a given specimen thickness, there is a colony size that corresponds to a minimum growth rate. - 14 The effective colony size is increased if two or more colonies share a weak a-basal plane and can thus shear together. [29] Because the orientation of colonies is related to the parent PBG orientation by Burgers relations, the largest effective colony size would theoretically be limited by the PBG size. [29-31] Eylon concluded from this that a larger PBG size would tend to increase the effective colony size and therefore reduce the fatigue crack growth rate in specimens with small colony size. [28, 29] Under certain conditions, GBa can promote fatigue crack initiation. Eylon found that subsurface GBa is a favorable initiation site in Ti-6-4, especially if the a-platelet is planar, is inclined at or near 450 to the tensile axis and is parallel to the long axis of a-platelets in an adjacent colony.[8,2l]Larger PBG size increases the number of large inclined planar GBa platelets. The coarse microstructure and variation of structures, cause the wide scatter bands in fatigue results of cast Ti-6-4. [6, 20] If the component is to perform reliably, the low end of the scatter band must meet the minimum fatigue specifications. scatter band is very wide. This is more difficult if the The large scatter also tends to obscure variables affecting fatigue properties. Therefore, some statistical approach to the experimental results of a cast Ti-6-4 fatigue study is necessary for determining the importance of processing variables. The structures selected for casting applications under study by the Army Materials and Mechanics Research Center (AMMRC) and the Applied Technology Laboratory (ATL) of the Army Research and Technology Labora- - 15 tories are the Solar Titan T62T-40 gas turbine auxiliary power unit (APU) impeller (see Fig. 1), [1-2] the UH-60A Black Hawk helicopter rotor hub produced by Sikorsky Aircraft, [3] and a prototype centrifugal impeller (see Fig. 2) manufactured by Detroit Diesel Allision (DDA), [4] a division of General Motors Corporation, for the Army's Technology Demonstrator Engine (ATDE). Solar Turbines International, currently an operating group of International Harvester, produces the APU with a cast impeller which would be used on the Army Black Hawk helicopter, the Air Force F-16 aircraft, and the Navy LAMPS helicopter. Currently, a 22-pound pancake forging is used to fashion the 2.6-pound APU impeller. [13] The 160- pound Black Hawk rotor hub for Sikorsky Aircraft, a division of United Technologies Corporation, may be generated from a 250-pound casting (see Fig. 3) instead of a 1,000-pound forging. [32] These studies have in- cluded castings produced by Precision Castparts Corporation (PPC), and Titech International, Inc., with additional castings by Howmet Turbine Components Corporation, and REM Metals Corporation, although all four casting firms have not been involved in studies on all three components. The objective of this study is to determine the major variables affecting the high cycle fatigue properties of weld repaired cast Ti-6-4 and to determine if such weld repaired material is inferior to unwelded cast Ti-6-4 in fatigue. The test program is statistically designed and the results are analyzed in order to provide the most information from the least number of samples and to correlate and confirm the results of previous studies. .J " ' I .J "Xr f,) UC: 4-i 1' _4 rot art .1 i >-i A +) i- .p fr, 0·r, II :il I .t: vs` -.u ·r d c) o -4 c,) 0 O r4 4 o a) H a) H *- C) rx4 oa ) C? a)Ce P) OH Oa ctiE_ @4 . If-:kL 1 I . I, k A I It i rf |L- 'L r 'i i F iS ' t VI cr, i 4 - A r ' if; C I _ _j ZIf cC : LI- I, r .' cz5 LLS O' c Ot O -<: CD 0U') Il Z: a- LL U C -, LL i' I I. .' 'I I 0 ;z O Lu A i q 4' t . It _ _ I _ ' . - . . . iw tjt I 4 I, t 4 , * AI i I k S I -19 II. PREVIOUS WORK In many weld repair studies, a single weld pass with filler wire is employed to fill a notch in the specimen. Miller of TiTech [33] and Nagon of PCC [34] used a "V"-notch normal to the bar axis. Each conclu- ded that welding was not detrimental to fatigue strength. Morton and Lane [35] tested bars machined from welded plate and found that welding was detrimental to fatigue strength. Barice and Allen [6] found no significant differences in fatigue of unwelded and welded samples with spot, plug and seam welds. Ewing and Green of DDA [10] en- countered no adverse fatigue effects by using both a backfilled a shallow fillerless weld. notch and Hammer of Solar [13] concluded that weld re- pair was detrimental to fatigue strength based on results from blade samples welded at the vane base and tested by reversed flexing. Also, work by Harrigan [36] using a normal V-notch demonstrated no change in creep mechanisms in weld repaired bars, and work by Kelly of Howmet 37] showed no decrease in tensile strength of satisfactorily weld repaired bars. The study by Kelly included backfilled normal and parallel to the axis of the bars. notches both He did find a decrease in tensile strength when filler metal did not completely fill the notch. In practice, typical weld repair procedures often require multiple passes to build-up metal to specification. Early passes are heat treated as later passes are laid down, perhaps changing internal stresses and the microstructure. The effect of multi-pass repairs was unknown for cast Ti-6-4 and none of the studies mentioned above reported using multi-pass - 20 - welds. Because of the slow cooling rate after casting, Ti-6-4 components have a very large PBG and colony size. The effect of PBG and colony size on HCF properties was discussed in the previous section. It was not reported whether the weld zone properties would or would not be affected by altering the microstructure prior to weld repair by heat treatment. It is generally accepted that HIP processing improves the HCF properties of cast Ti-6-4 by eliminating voids. [8] Almost all present and future applications for dynamically stressed cast Ti-6-4 require a HIP cycle following weld repair to ensure the integrity of the weld metal. No published data reported the effect of this second HIP. As mentioned above, Kelly [37] reported that tensile strength is not affected by weld repair if there is complete backfilling of the notch. Other investigators found weld repair to be acceptable in terms of HCF strength when not using either a secondary HIP [6] or a primary HIP cycle. [33, 34] These studies, plus a study by Taylor, Burn, and Clarkson [38] show that porosity is not a problem in the weld zone given proper preparation, which implies the second HIP is unnecessary. on work by Hammer, Based 13] there is evidence that the heat cycle during HIP might have a deleterious effect on HCF properties, implying that the second HIP may even be undesirable. Repair welding induces residual stresses in Ti-6-4 which can distort the repaired structure and can have a negative effect on various mechanical properties. [12, 13, 39] A stress relief treatment between 900-1200°F C480-6500°C) is necessary to improve the properties. [12] The - 21 - stress relief is often replaced by a higher temperature heat treatment to alter the microstructure and further improve the properties. Anneal- ing for titanium alloys is carried out between 1300-1550°F (700-845°C). Solution treating in the (760°C) and the + -transus. [40] region is performed between 1400°F Aging follows a solution treatment or $-treatment to relieve cooling stresses and to further alter the microstructure. The heat treating temperature, time, and cooling rate all affect the microstructure as mentioned previously. At high temperatures, atmospheric purity is very important to prevent contamination of titanium. For this reason, heat treating is usually accomplished in a vacuum or an inert atmosphere (usually argon). If heat treatment is accomplished in air or if a water quench is employed, a subsequent vacuum anneal is necessary to remove excess hydrogen while machining or picking is necessary to remove the oxygen contaminated c-casing. In a water or oil quench, some components will distort before stress relieving or become contaminated. A furnace or air (argon) cool is often too slow to be an acceptable alternative. Therefore, a rapid argon cool is employed for these applications to maintain purity and to rapidly cool the part without distortion. The process involves blowing or otherwise circulating argon through the furnace and around the heat treating charge. Bass [5] reported that solution treating with a water quench produced better HCF properties than annealing with a rapid cool. Sim- ilarly, Hammer [13] found that solution treating with a rapid cool was - 22 - superior to annealing with a furnace cool. While rapid cooling was more desirable than slow cooling, Bartlo [41] reported HCF properties deteriorated as annealing temperature increased. Morton and Lane [35] made no mention of a post-weld stress relief, which may explain their conclusion that weld repair is detrimental. Harrigan, [36] Miller, [33] Nagan, [34] and Ewing and Green [10] all annealed specimens following weld repair and found no decrease in HCF properties. [9] Silverstein and Conrad -treated their samples and reported that rapid cooling is superior to slow cooling. Few studies have been published comparing the affects of various surface finishes on HCF of cast Ti-6-4. Bass [5] reported notch sensi- tivity was less for annealed material than for solution treated material. Bartlo [41] reported that heat treatment had much less effect on HCF properties in notched bars. Lucas [16] showed that for forged Ti-6-4, the effect of peening on fatigue properties was only evident between 100,000 andlo million cycles and that it had no effect on the strength limit. Most of the studies in the literature have used reversed bending (R =-1.0 = maximum stress/minimum stress) methods for HCF studies, although some used stress cycles of R =-0.1. Bass [5] compared the two stress cycles at both room temperature and 500°F (2600 C) and reported that fatigue limits for R =-1.0 were lower than for R = 0.1, which is to be expected. Based on this, the rotating beam method would seem to be a harsher test for HCF than tension-tension. This study will examine many of the conflicts or omissions men- - 23 - tioned above. While most investigators have reported that weld repair of cast Ti-6-4 is unharmful to HCF properties, there are published findings to the contrary. No reports were located examining the effects of multiple pass welding, microstructure prior to welding, and surface finish of weld repaired material. Despite the general acceptance of a post-weld HIP treatment, no reports were located showing the necessity of this secondary HIP, and some evidence exists that implies it may even be detrimental. well understood. The effects of post-weld heat treatments are also not - 24 III. A. EXPERIMENTAL PROCEDURE Statistical Design Previous studies of fatigue behavior of cast Ti-6-4 have been ac- complished under a wide variety of material conditions and testing procedures. Consequently, there are some disagreements between the various investigators as to the relative importance of particular variables, including the effect of weld repair. This variety of opinion coupled with the limited number of test bars available for this study led to the choice of a statistical screening experiment. The design aided in de- termining the major variables influencing high cycle fatigue (HCF) behavior of weld repaired cast Ti-6-4 with the fewest number of samples. The screening design used in this study is a 12-trial PlackettBurman design. [42] The design contains 11 factors, resembling the degrees of freedom in a regression analysis (see Table 1). A factor is assigned to each variable to be studied, leaving the unassigned factors for determination of scatter in the data. coded "+" and "-". "high" and "low", Each factor has two levels, For continuous variables, these levels correspond to and for discrete variables, the levels correspond to "on" and "off" or "type one" and "type two". For this design, each variable can be studied using only two levels. With the design, the condition of each trial sample is determined from the coding of factors assigned to the variables. The samples are then tested for the property under study so a quantitative result can be assigned to each trial. The results are used to analyze the variables - 25 I O + I + + + I + + I + + + I + I + + I + + + I I + I + + I + + + + I + + I + + + I + I + + I + + I + + cml I cO1[I I I + I I r- I I I-) .olI ro L + LnlI + + I I I + I + + I + I 1-I + + + I I I + I + + I I z C Co, 0 : i, w F -j LI -J mcII c'j -1 + + + I I I + I + + I + I + + + I I I + I + I + + I + + + I I I + I I CO cn C a, * - C - - r- - COJ - 26 and determine the scatter of the data in this way: a. For each factor, the sum of all the "+" trial results is subtracted from the sum of all the "-" trial results. b. The difference is divided by 6 (the number of "+" or "-" values for each factor) to determine the "effect" of the factor. The effect is simply the difference of the mean "+" results and the mean "-" results. c. The root mean square (RMS) value for the unassigned factor effects is calculated. d. The appropriate "t" distribution value for the desired confidence limits and degrees of freedom is determined (see Table 2). The number of degrees of freedom is equal to the number of unassigned factors. e. The "t" value is multiplied by the RS value to obtain the minimum significant factor effect. f. Any variable with the absolute value of its factor effect greater than the minimum significant factor effect is considered significant within the chosen confidence limits. g. The absolute values of the significant variables are ranked. The larger the absolute value, the greater the relative influence that variable has on the property studied. While the design is able to estimate the significance of each variable separate of the other variables, it is unable to estimate the interactions that may exist between the variables. The estimates may also be correlated to groups of two-variable interactions. When a factor - 27 - Table 2 "t" DISTRIBUTION Degrees of Freedom 80 Confidence Limits (%) 90 95 99 1 3.08 6.31 2 1.89 2.92 4.30 9.93 3 1.64 2.35 3.18 5.84 4 1.53 2.13 2.78 4.60 5 1.48 2.02 2.57 4.03 6 1.44 1.94 2.45 3.71 7 1.42 1.90 2.37 3.50 8 1.40 1.86 2.31 3.36 9 1.38 1.83 2.26 3.25 10 1.37 1.82 2.23 3.17 11 1.36 1.80 2.20 3.11 12 1.36 1.78 2.18 3.06 12.71 63.66 - 28 - effect is found to be significant in comparison to the minimum signifi- cant factor effect, it means the factor stands out over a pool of background noise of both interactions and experimental error. For the purposes of this study, 7 variables were chosen to be analyzed leaving 4 unassigned factors. The variables included micro- structure prior to welding, welding procedure, post-weld HIP, post-weld heat treatment temperature and cooling rate, finishing process and peening. B. Prior Microstructure - Variable 1 For the purpose of this study, separately investment cast test rods were purchased from Howmet of Whitehall, MI. Each rod was approx- imately 4.5" (11 mm) long and .625" (15.9 mm) in diameter. The rods had been given a HIP treatment of 1650°F (900°C) for two hours at 15 ksi (100 MPa) and chemically milled by Howmet before shipment. The rods had no observable surface contamination and met the specifications for cast Ti-6-4 (see Table 3). The rods were labelled for identification and were examined by x-ray radiography to ensure that they were sound. Rods were then as- signed processing schedules according to the statistical design (see Table 4). To vary the microstructure, a -solutionizing treatment followed by a rapid argon cool and a stress relieving anneal was employed. This treatment was intended to refine the colony size and platelet width. Half the specimens were -treated and the other half were left in the -29- (v S- 4T r_ aJ 0, (a 0 r-. o ¢x C) 0 s o' C C U I co a V) C o w C C z U) S .4 7:1 v X a, S.- C C C CD >- LO C .0 cn CO C Ln X D X I0 4E z .~ I---n CV r* C) o 0 rO B_ n LU F-, o C) Cl O0 X C 00 X C\ nO LO C O . CD *4) a) - L E X X E ,LO- O0 F- C z I- c: u 4-)Cu C a) (L) C) Cu00 (J C. O O (o r-( U.) = O 'J 4D C\J a C O V cOa, (_) ~C ( r O C_ CD 4- Ure ,--- S* ._ vJ CV (V a U ,'"- O Cr v I LO O :: a) cF.' gO --(MM~D H'" E Cu (/, 0 · ,c? .0 (n 0d c S.- (0 =dI C-) v,~~~~~~( C) -0 4-) Cu H - 30 Table 4a PROCESS VARIABLES Variable "+" Condition "-" Condition Initial Structure s-treat As-received Bead Size Large Small Pressure Treatment HIP None Heat Treatment Solution Treat and Age Anneal Cooling Rate Rapid Slow Polish Mechanical Chemical Final Treatment Peen None Table 4b ROD PROCESSING SCHEDULE Trial Rod Numbers Processing* 1 1, 21 s-treat/Large Bead/STA/RC/MP 2 2, 32 B-treat/Small Bead/HIP/STA/RC/CM 3 3, 13 Large Bead/HIP/STA/SC/CM 4 4, 14 $-treat/Large Bead/HIP/@/SC/CM/Peen 5 5, 15, 25, 45 B-treat/Large Bead/@/SC/MP 6 16, 36 $-treat/Small Bead/@/RC/CM/Peen 7 37, 47 Small Bead/STA/SC/MP/Peen 8 7, Small Bead/HIP/@/RC/MP 9 9, 19 52 10 10, 20 11 11, 12 38, 48 *STA Large Bead/STA/RC/CM/Peen -treat/Small Bead/HIP/STA/SC/MP/Peen Large Bead/HIP/@/RC/MP/Peen 31 Small Bead/@/SC/CM = Solution = Anneal Treat RC SC = = Rapid Cool Slow Cool MP = Mechanical Polish CM = Chemical ill anmd Age - 31 - as-relieved condition. The S-treatment used was 1900°F (10500 C) for two hours with a rapid argon cool followed by an aging cycle at 1300°F (7000 C) for two hours and a rapid cool. The heat treating was accomplished using a tube furnace (see The samples were cleaned with acetone and wrapped three in a Fig. 4). bundle with titanium foil. Two bundles were heat treated in each load. The bundles were loaded into an open-ended quartz tube along with Ti-6-4 chips. The foil and chips were used to protect the rods from contamination in the system. The quartz tube was slipped into a vycor tube with a ball joint on one end and a tapered joint on the other. The ball joint was joined to a sealed glass rack with valves which lead to a vent, to pressure gauges, and to a vacuum pump. A cap with thermo- couple leads and a valve to an argon supply was attached to the tapered joint. The argon supply was welding-grade quality. hot junction was placed between the two bundles. The thermocouple A fiberfrax heat shield with heat reflecting aluminum foil ends was placed in the vycor tube near the cap. The apparatus was then placed in a standard three-zone tube furnace controlled by a separate thermocouple. After the load was placed in the vycor tube and the tube sealed from the atmosphere, the tube was pumped down and backfilled with argon twice. Following the second backfill, the tube was sealed and the load brought to temperature. During heating, argon was vented from the tube to maintain a nearly constant pressure of one atmosphere. Furnace cooling was accomplished by switching off the furnace. When rapid argon cooling was required, the glass rack vent was opened and E zr o CC .. c 0 o o 0 o ,4.XI 1 E, : i \\ Co 0 0O m 0,. 0 , 0 - c ;E .. cl'J m _,...__ I ,m C CD 0 a 0 i , r 1. I- o , 0 l, c 0 1 *00 0 CU o -r C., C 400 .*0'4- 0 co o I .a Co ,,, , 0 QD 0. 0. , Q'-//x 3 0ft CO - a0 ,,._'" ._ ii - 33 and argon was forced through the tube from the argon supply, while the The cap thermo- furnace top was lifted to increase radiation loses. couple was used to record the cooling rates (see the Results section for actual rates). Prior to welding, the rods were machined to oversized blanks (see Fig. 5). The resulting shavings were vapor degreased with 1,1,1- trichloroethane and used for heat treatment chips. The slugs removed from the rods were used to check for contamination during processing. One slug was included with the rods during each treatment. Each slug was then analyzed by Luvak, Inc. of Boylston, MA for oxygen content by vacuum fusion and hydrogen content by hot extraction. C. Bead Size - Variable 2 To determine the effect of repeated weld passes on HCF life of Ti-6-4, half of the bars were welded using a single large weld bead pass and the other half using six smaller bead passes. Because past work had shown that incomplete notch filling was detrimental, no notches or filler metal were used in this study. The weld beads were made circumferen- tially around the reduced section normal to the bar axis. This reduced the risk of contamination and incomplete filling, while allowing greater control on weld zone and bead size. These results were compared to re- sults obtained with unwelded specimens to find detrimental effects. The simulated weld repair was done in MIT's welding laboratory. The blanks were held by one end in a rotating chuck connected to a variable speed motor. The torch was clamped to a stand and positioned --' OO' i 0 ~. i* ' C, o rX o C) , I. is 4 [ 0 .2i o E .0 coL r~~~~~~~~~~~~ ~~-·---.X------------~~ 0 0) .. ; - 35 - The entire apparatus was above the area on the blank to be welded. placed in a glove box with an argon atmosphere. The blanks were cleaned prior to welding using acetone and methylethyl-ketone. To simulate weld repair, a single bead was run around the reduced section of the blank using a ramped and pulsed arc program (for The gas used to fill the glove box and used for parameters see Table 5). Atmosphere purity was the shielding gas was ultra-pure (99.9%) argon. confirmed by checking for discoloration in a spot weld made on a test The blanks were allowed to cool in the glove box atmosphere be- slug. fore subsequent welding or removal from the box. For the multi-pass simulation, subsequent weld passes were made at lower current, but otherwise using the same welding conditions (see Table 5 and Fig. 6). D. Hot Isostatic Press - Variable 3 To study the effect of the secondary samples were given a HIP treatment HIP cycle, half the using standard practice for Ti-6-4 The blanks designated for a HIP treatment were following weld repair. sent to Industrial Materials Technology Corporation (IMT) of Andover, MA. The blanks were given a standard 16500 F (900°C) treatment at 15 ksi (100 MPa) for two hours. E. Heat Treatment - Variables 4 and 5 Because Sikorsky had reported no difficulty in -treating and water quenching the cast rotor hub to produce the desired microstructure without distortion, [32] the heat treatment portion of this study concen- - 36 - Table 5 WELDING PARAMETERS For all welds: Electrode diameter = 3/32" (1.6 mm) Electrode type = Tungsten with 2% Th Electrode tip angle = 90 deg. Electrode tilt angle = 0 deg. Tip to cup distance = .15" Tip to work distance X .1" (2.5 mm) Cup diameter = 15/16" Torch gas flow = 20 cfh Ar (9.4 l/min) Welding speed = 10 rmp = 10 ipm (4.2 mm/sec) Arc initiation = high-frequency Peak voltage = 15 volts Peak current (first pass) = 200 amps Initial current = 5% Time at initial current = 5.5 sec Ramp-up time = 1 sec Time at peak current = 6.2 sec Ramp-down time = 2 Final current = 2% Time at final current = 4 sec Background current = 15% Pulses = 2 per sec Time at pulse peaks = 15% = 150 amps Peak current (fourth, fifth and sixth pass) = 100 amps (3.8 mm) (1.6 cm) Ramp function: sec Pulse function: For multiple pass welds: Peak current (second and third pass) Cooling time between passes = 1 hr (minimum) L) Cn en c CY) cU co C, (L at. ,) a. ._to 0. o c'0 0. CD o) ac1 0) IL. - 38 - Solar [13] and DDA [43] both trated on the more sensitive impellers. reported difficulty with distortion on cooling from the B-treatment temperature and during water quenching from a solution treatment. There- fore, this study was limited to annealing and solution treating the specimens, with the maximum cooling rate being that of a rapid argon cool. Variable 4 was chosen as the time and temperature variable with variable 5 the cooling rate variable. To study time and temperature, To half the bars were annealed and half were solution treated and aged. study cooling rate, half were furnace cooled and half were rapid argon cooled. All blanks were given their designated heat treatment using the tube furnace and procedure already described. The solution treated blanks were heat treated for 1.5 hours at 17500 F (9550 C), cooled, and then aged at 950° F (510°C) for four hours. heated to 1350F F. The annealed blanks were (730°C) for two hours and cooled. Surface Finish - Variables 6 and 7 Variable 6 was chosen to study the effect of a chemically milled surface finish. All bars were polished to a similar finish before Thus half the samples remained in a mechanically chemical milling. polished condition. Variable 7 was chosen to study peening effects. Half of all the bars were glass bead peened after the final polish and half were left in the polished condition. All blanks were machined to final dimensions at AMMRC (see Fig. 7). 0 0 s--- o l4- L. D 0 N ac cC) N .C cQ 0 a 3I I i I ~ ,'-- / i1 _I- N U) / N co 4, '4. i, cO (0 i~~a C ca, . C, f i 0, T E 0. a, '0 a, '-I N,. C 'C, i; ii ii m aO. 0 co c cm c E '- C ¢D O a, ' CD E a, ._ I- E ._ '0 CM I. - 40 - The specimens were then hand polished in order to achieve a fine surface finish and to eliminate transverse scratches. Specimens which were to be chemically milled to final dimensions were machined slightly oversize before milling. Chemical milling was done in a heated acid bath. [44] The acid consisted of 10% nitric and 3% HF in deionized water and was heated to 115°F (45°C). Each specimen was milled separately using a small rotat- ing tumbler to provide uniformity. Designated bars were sent to Metals Improvement Company (MIC) of Lynn, MA for peening. The peening was accomplished using a slurry with AB size glass beads, attaining an Almen intensity of .010-.012 N2 , as per specification AMS 2430H. G. Testing Methods The fatigue mode of impeller blades is considered to be closest to a flexing motion. 13, 43] The rotor hub has a variety of different fatigue modes including flexing motions. [45] The limitations of the fatigue testing equipment provided the choice of testing in tensiontension or on a rotating beam fatigue machine. It was desired to increase the fatigue cycle frequency as high as possible both to better simulate the actual operating conditions, and to decrease the testing time. This desire plus the flexing motion in impeller blade fatigue were the reasons for choosing R.R. Moore rotating beam machines. Prior to testing, the RMS surface finish of the specimens was - 41 - measured with a profilometer and internal defects were located by x-ray radiography. The specimens were each tested on one of two R.R. Moore machines at 72.5 ksi (500 MPa) and 10,000 rpm (167 Hz). Broken specimens were sectioned for metallography. One half of each specimen was sectioned logitudinally to observe the internal microstructure, while the other was sectioned transversely to observe the fracture surface. The microstructure was revealed by mechanically polishing and etching in a nitric-HF solution. The fracture and specimen surfaces were examined in a scanning electron microscope (SEM). Colony size measurements were made from micrographs of the base metal in each trial specimen pair. A line intercept method was employed to calculate the mean intercept length (L3 ), which is a measure of the average colony size. [46, 47] A colony boundry was assumed to be located wherever the prevailing platelet orientation changed in the micrograph. For comparison to the welded specimens, ten as-received rods were machined and mechanically polished. Two rods were fatigue tested at each of five stress levels: 72.5 ksi (500 MPa), 65 ksi (450 MPa), 55 ksi (380 MPa), 50 ksi (345 MPa), and 47.5 ksi (325 MPa). - 42 - IV. RESULTS Two specimens were processed according to trial 5 of the statisti-treat, large bead, no HIP, anneal and slow cool, cal design, namely mechanically polished and no peen. They were each tested at two low stress levels, 62.5 ksi (430 MPa) and 65 ksi (450 MPA), until run-out and then run until failure at 70 ksi (480 MPa). Both failed at approximately 2.5 million cycles, one failing in the threads. This procedure made these specimens invalid for inclusion in the design, so two additional specimens were prepared to replace them. The processing of the additional specimens only approximated the processing of the original specimens since much of the welding and heat treatment equipment had to be reassembled. The results of the R.R. Moore testing (see Table 6) could not be analyzed directly using the screening design because the replacement trial 5 specimens ran-out. A variety of different values for trial 5 (see Appen- dix I) did not greatly affect the results of the statistics, therefore a value of 10 million cycles was chosen. By calculating the statistics (see Table 7), surface finish is well over 90% significant as a variable in HCF life with heat treatment time and temperature nearly 80% significant. Mechanically polishing is shown superior to a chemical milling for HCF life and annealing is shown superior to solution treatment. X-ray radiography showed no indications of voids either in the as-received or post-processed material. The as-received material showed indications of what may have been inclusions or segregation, however the post-processed material showed no imperfections. RMS surface measure- - 43 - a) (A F- -j l . - O L. H LL. (A ox LUI r-' O r cC\- 0 - o .0 M ~ CO --- O L n co r-P" n 1n r_ O M - cm c C r- r-. ~ LO r-- r-- O c -::Im - C- CO -co C C O Ln cn o CD rLoT r- - O O r-o O OO c M r~~~~~~~~~~~~~~~~~~~~~~~c-, a C LUJ 0y C) oo .r- aIrH0 -J C: Z LO COCLO C-- C00 Co o CO C C (a o LO O LLI O i- .v- S. ct 03 o 1 *r- L ( : ( L r- LO O - L C M O D :f (A (C n C M 0 (- Co J M i N :* 0:z aj + co r) ) r- .r- o - CD S.. N LOL O o oo co O 0 Cj C N- 0 N C) N- O k CO O CO CD i Ln Ln A (A CO Ln , (A O - r- lO O CO -J V) tU LUL "-' O n:: 0 ox :'I CJ - CO .LL or- 00 -LOO O00Cld" CO:lCOC'JC) CO 00 rAJ 0ooN o Ln l "LO O0 f~- 0~. D-N f~- C) - O C OO :-0 CDc CD 00 CO cor-t . -o,. \O D r_-COCOLOL-- LOCO 00 .- L"-- -- u>, E= E~ aLzz (A (N ( -. - - c: :- C" ' ' -':C (An . t' - CO M")cf' (A1 VI + - 44 LO Lr) cO C) - j C LO C?) co I I C CY) m (" Cr) Mt C C CO L r- Cm L I'"' o o on c?) C\ 0 C?) N- C?) m m co N-, Ln~ C r- o- C C ( c) M C Cr O I-) m C: CO C 0) cn C? 0a ' co C?)e CY) o O- c: C-) LL LL U -, N_ ko - ,o- - I'D - C) Cm F- H U v4 -4 LO -- L0 LL a) r0) C CMi cO C co Z CD LL LL -4 C?) LLU C9 P-4 0 4- c?) Co C CN o IN- o a) C I cCL L. - C? -:I C CO HL I H 1- Cn c) I o CD I p-4 O U_ A C O CO V v- HC Ca) c N 0) *t C C? C? C) ;.0 C\ C) C) LO IC) % I C?) N- Ln 11 A| Al I C r - CO .U, r- 0 ) -r' S- (z CO m -p~ C') co In C?) C OL U C-) CO Q 0D CD e0 CC) 0:I coJ c5- O 0 --) (n O m, C" ul ( u a) LU 4- 1: , O C oE O C C a + I E E V) V) :: :3 +I q- orm C-) LU E3 CO cn a 45 - ments (see Table 8) showed the peened surface to be much rougher than the chemically milled or mechanically polished surfaces. Mechanical polishing was only slightly finer than chemical milling. Chemical analy- sis of the slugs (see Table 9) showed no significant oxygen or hydrogen contamination in any of the processes. The time to cool from the treatment temperature to 1000°F (540°C) was measured for each heat treatment. was slightly over 3 minutes from 1900F The time of rapid argon cooling (1050°C), slightly under 2 minutes for 1750°F (955°C), approximately 1.5 minutes from 13500 F (730°C) and approximately 1 minute from 1300°F (700°C). The time of furnace cooling was under 2 hours from 1750°F (955°C) and under 1 hour from 1350°F (730°C). Approximately 45 minutes was required to cool from the HIP temperature of 1650°F (900°C). Microstructural analysis revealed that all failed specimens, except the specimens for trials 2 and 3, fractured outside of the weld zone. Trials 2 and 3 were both chemically milled and had the lowest HCF lives. The microstructure of the weld metal (see Figs. 8-11) was similar among all the specimens. The PBG size of the weld metal was generally very large and the morphology was a basketweave structure with little or no GBa. The base metal in contrast had thicker GBa, larger colonies, and a smaller PBG size (see Fig. 12-15). Weld penetration was complete through the thickness of the reduced section. There was no visible change in microstructure across the weld metal due to the multiple passes or in the base metal due to the heat affected zone (HAZ), although a transition from weld metal to base metal was observed (see Fig. 16). - 46 - Table 8 SPECIMEN SURFACE FINISHES Specimen Number RMS Range 7 5-6 6-8 4-5 15 5-6 21 5-6 4-6 1 5 52 2 3 13 32 38 48 4 9 14 16 19 36 10 11 20 31 37 47 15-19 18-22 15-23 11 -14 13-18 11-14 40-65 45-65 50-65 35-45 35-45 30-50 45-60 60-80 50-80 40-65 35-55 55-70 - 47 - Table 9 SLUG ANALYSIS Analysis Treatment* (%) 0 H As-received 1 .21 .0048 As-received 2 .18 .0051 As-received 3 .19 .0050 S-treat (load 1) .21 .0050 $-treat (load 2) .21 .0074 HIP .20 .0089 STA/RC .19 .0052 STA/SC .18 .0060 @/RC .21 .0060 @/SC .19 .0042 Chemically Milled .19 .0054 *STA = Solution Treat And Age @ = Anneal RC SC = = Rapid Cool Slow Cool i Figure 8: Specimen 4 base metal. (optical, HF-nitric etch) . mm l i'ur e ): Specimen 19 base metal (optical, HF-nitric etc } imrrm 1..' :·,·. ?I ··;; · :·T'--:-· i ·- ic-:··-·:-.1,s ,, 5-·? X '".: -) ···'i.·i-": ;··;··'""··i.'-1;''" · nt ·· :lf;· 2:d=" c: · Figure 10: Specimen 36 base metal FiLire e 11: Sp)ccimten 38 bas( rritali (optical, HF-nitric --- ('_) t i<(i, etch) 11-mll if'-n i tric ( t(. ) Figure 12: Specimen 1 weld metal Figure lj: Specimen 32 weld metal (optical, HF-nitric etch) -- I.rmm Fracture through thie weld metal is shownat bottom (opt.ical, HF-nitric (tlch) -- $ 1mm Figure 14: Specimen 37 weld metal (optical, IIF-nitric oetch) ~.- ,; v' - O-7 11 :i \;;? \\) Figure 1: Specimen 52 weld metal (optical, IIF-nitric -- -. I ll .~tch) Figure 16: Specimen 1 microstructure showing transition fr(mn weld metal to base metal Weld metal is shown ill the upper part of the figure The bottom o!' th, figure shows the fracture plane across the ba~se metal (optical, }iTF-nitric etch) b---~ -11'!1 - 53 - Colony size measurements of the base metal for each trial (see Table 10) were analyzed with the statistical design (see Table 11). Cooling rate and HIP were both over 80% significant variables with slow cooling and HIP producing a larger colony size. The largest colony sizes were observed in HIP processed solution-treated and slow-cooled specimens 3 and 10). (trials By ranking the trials by average HCF life and comparing this rank to mean intercept length (see Table 12), a trend is revealed. For the mechanically polished and the peened specimens, the larger colonies with larger intercept lengths tend to be associated with longer HCF lives. A trend for chemically milled specimens is not observed. The fracture surfaces were generally sufficiently intact to make fracture origin determinations. All fracture origins were located on or near the surface and usually exhibited facet-like cleavage (see Figs. There were no indications of fractures originating at voids or 17-19). inclusions, however most of the origin sites had been somewhat smeared during the fatigue testing (see Fig. 20). Chemical milling removed the directionality on the outer surfaces characteristic of the mechanically polished surfaces (see Figs. 21, 22). Peening obscured the previous polishing procedure (see Figs. 23, 24). X-ray emission analysis in the SEM revealed that some of the glass beads used in peening had become imbedded in the outer peened surfaces (see Fig. 25). By measuring the distance of the fracture surface from the minimum diameter in the reduced section, the stress at the point of failure - 54 - Table 10 SPECIMEN Trial COLONY SIZE MeanIntercept Length (in. xl000) 1 3.6 (0.09 mm) 2 3.9 (0.10 mm) 3 6.5 (0.17 mm) 4 5.6 (0.14 mm) 5 4.7 (0.12 mm) 6 4.4 (0.11 mm) 7 3.8 (0.10 mm) 8 3.4 (0.09 mm) 9 3.4 (0.09 10 6.4 (0.16 mm) 11 5.7 (0.14 mm) 12 5.1 (0.13 mm) mm) 55 - O U.) LO CM CD 7- CMi CM CO m (-3.) I NCM I CN Ln oCIM CO co CM o CQi r- Lo CM 1.0 C: 3-Y. C a) cn CM !l_ 0. O) CM a) * 7 Ln CM I- CM 0 . C.) I- U, .r. 0 F- U, o L O LLJ LL cu U- * CM C m CM FLL ) I I 0~ co I-Q OCr CM c CD V I tU) 0u 4-) CM >-j C_ o o) CO V A av L L _ I ¢U.) ( Ln =: 0 CO O LDU: In I LO C0 CO O 4- CMu e - CY- .- .7. il o N Q) , 1 m mu CM .4-) F Al I u, U.L 0 o- C) I S.- N- o CMj u 4-) CMj CS (z W I (0 Al ( os CM CM N 4- U, (n C 4-) -) O) O a au CV C 5: - CD U ') ( r.- co + E :3 I : a) (E .-1 .4-) a) 4(4-. Lii 0 0n 0 *- (A re, co (A - 56 - Table 12 COMPARISON OF FATIGUE LIFE AND COLONY SIZE Finish* Trial HCF Life (log cycles) Mean Intercept Length (in. x1000) MP 5 Run-out 4.7 MP 8 6.607 3.4 MP 1 6.058 3.6 PE 10 5.917 6.4 PE 4 5.899 5.6 PE 11 5.863 5.7 PE 9 5.818 3.4 PE 6 5.752 4.4 PE 7 5.682 3.8 CM 12 4.809 5.1 CM 2 4.733 3.9 CM 3 4.625 6.5 *MP = Mechanically Polished CM = PE = Chemically Milled Peened F i ,u re 17: Specimen 10 fatigue specimen. Fi< ire I ' origin showing fracture face- Lower part of figure is the outer surface of tlet- S pec imen )(:cfr[Ftl(f sart 0 <z LYlI 'IlT (SEM) 9 ftt 1mm i gue { o 1 'hl:I i uer S or iin sh(~wing L nT t heLfue fcra fr;1 1,ilal'c -- 1'tu r I i (- X t'l Figure 19: Specimen 13 fatigue origin in weld metal, Lwer part of figure is the outer surface of the specimen. (SEM) i., 51m t'aLiu:ti 2: 6pecirnen 21 fatigue s r tace Lo'wer part surf ace Of rigi.n sho)wing sinInred m:'e: (f' titulrt' is tit c. iltttP tthe spec inetin ( SEM ) ... }....)> ,_1 ll Fi-Lure 21: Specimen 7 outer surface showing mechanically polished condition. (SEM) --- [,' : ,, . 0. 1111 1'~~~~~~~~~~~~~~~~ 1.9.1e I':ft< ile~ s11 (.Clt ,~1)91l .i--(l t.lt'ai'l ! ......... I'a., UXs- '~ I )c(:I 'len .5 mi L l (I uter , ( tildi Jt i (-)in sun ur SLI ce ) i)w, 2 I1h Figure 23: Specimen 10 outer surface showing mechanically polished and peened condition, (SEM) --,2mln Plg urc }A: 5peet.lmen 4 luter mi.led sulrtace .showi.ng cthlerTli cal Iv and p)eened c ondit ion- ( S:N ) (a) (b) tiuro 2W'-: t i ,11t,(Fl t 9 OUL r S1 fr.Ca (( - I1 Wir 1 lld t tI )(- ci . (il l I1 \ i 11 ( j~l ) -i>tfDW.;£tiej e ~*ridivN )lf c tt l lN',v t'tci Urf ()ri. in iIld )art )1 ttLh I r t<.-tt II ' ¼t Ir ( . i tI',: t I t i: 't _ ri'~& st:iti.ot il Ith u1ppert i -i n ls t1, -)ll:t ,. c. showi _l r(APT.1[) -Li}: tl t,, ,rr1 , in ()1,<n '"()L iIaL the dark area ' -. 11 c ';n1 ind(1icat ionIs int ( bv) (S>EW1) §f - - 62 - was calculated (see Appendix II). A stress-log cycles (S/N) graph was then used to plot the results (see Fig. 26). By this method, it can be shown that the peened specimens had HCF lives between those of the chemically milled and most of the mechanically polished specimens. The results of the welded specimenswere also compared on an S/N plot with results of the unwelded specimens (see Fig. 27). II '0 C <c -c .0 a-o 1o *40 ino-uni 4®a l Cr, I 4 I 4 I C) C> 0 I 4 0 4 ar0 E 0 O0 I U I 0 *e 9 a, E L0 0 CL) - xX 0- I co 4.I- c.. co - 0) C .a t0 Cu 3 tO · V 0 a J a a a 0 a z I I: I tO N C,, 6N,- - Lb co (!Sl) SS3E1S c co 3UllOVuE- U) 0 c 01 I I 4 v f-..I wouniJ / 0 / a ult / / 0 a) ; -o zY 0 . (D O Cu a) 0 E uJ U- 0 0 a) 0 S 0. 0 LL S - I a) Qa 'e - - 0 c 0 0 o LU 0 4r 0 0 0 z -. -~~~~~~ 0 --- -· 0CD CD (!S) SS38S r~~~~~~ LD U) 0 L. L- Q. .) - 65 - V. DISCUSSION The most important result of this study is that weld repair is not detrimental to the HCF properties of cast Ti-6-4. Because most trials (all but 2 and 3) failed in the base metal instead of in the weld metal, it is valid to extend the findings of the study to cast Ti-6-4 in general as opposed to just the weld repaired material. While this study did not examine the effects of weld repair on the HAZ, the colony size and S/N data indicates that this region is also not affected adversely. The failure of the -treatment to have a significant effect on the base metal microstructure was disappointing. ure are unknown. The reasons for this fail- It is possible that the heat treat furnace was unable to heat the rods above the -transus, but this seems unlikely since both thermocouples were reading approximately the desired temperature. is also an unlikely possibility that the cooling rate of the There -treatment was slow enough to create the same colony structure as in the as-received material. A more likely possibility is that the change in microstructure was not very great and subsequent heat treatments obscured this change. In any case, the lack of microstructural alteration would explain the lack of any significant effect of -treatment prior to weld repair on fatigue properties. Weld bead size had no significant effect on HCF properties. Again, there was no observable effect of bead size on microstructure, which would explain this result. Subsequent heat treatments probably obscured any differences that may have existed. The benefit of this finding is - 66 - that in actual practice, the welder may use whatever procedure, within reason, that is needed to satisfactorly repair the component. The second HIP treatment also had no significant effect on fatigue properties according to the HCF life screening statistics. However, the HIP treatment did significantly increase the average colony size of the base metal according to the colony size screening statistics. A large average colony size was correlated to a longer HCF life for peened and mechanically polished specimens. If the secondary HIP treatment does influence HCF properties as the colony size statistics imply, the effect was swamped by interactions with other variables, by data scatter and by the effects of much more significant variables. No voids were observed using both x-ray radiography and microscopy in either the as-received or post-processed material. The HIP treatment was therefore unnecessary, which may help explain the HCF results. These results demonstrate that only one HIP treatment is normally necessary in Ti-6-4 casings and the treatment can be made either before or after weld repair without any adverse effects (_assumingthe weld zone to be sound). However, any additional HIP treatment, if necessary, would not be detrimental to HCF properties. The correlation between larger colony size and longer fatigue life confirms the results of Yoder and Eylon [28] for the fatigue growth rate in a specimen with a colony size much less than the specimen thickness. The base metal fracture origins also exhibited the expected planar colony slip facets. 29] The correlation between slower cooling rate and larger colony size also conforms to previous findings. [20] - 67 The majority of specimens showed no indications of fatigue cracking in the weld metal despite a smaller colony size (a basketweave structure) and a larger PGB size than the base metal. The basketweave morphology and the absence of GBa in the weld metal, both conditions unfavorable for fatigue initiation according to work by Eylon [8 ], are the most likely reasons for this result. Even if fatigue crack growth rate is expected to be faster in the weld metal, crack initiation is more favorable in the base metal. The heat treating results of the HCF life statistical screening are difficult to interpret. The numerous interactions and the predominance of the surface finish variable cloud the statistics. It is unjustified to use these results to say much more than that heat treating variables have an effect on HCF properties and that for the procedures followed, annealing was more beneficial than solution treating. Since previous investigators concluded that solution treating with a rapid argon cool was more beneficial than annealing [5, 6, 13], this last finding could be a result of the interaction of the time and temperature variable with the cooling rate variable. For example, if anneal- ing with either cooling rate was much more beneficial than solution treating with a slow cool, and slightly less beneficial than solution treating with a rapid cool, then the HCF life statistics would show solution treating less beneficial than annealing. This tendency would increase if the rapid argon cool was not rapid enough to get the full benefit of a solution treat and age procedure. Unfortunately, this study did not generate enough information on heat treating to either confirm or deny this - 68 - hypothesis. A similar hypothesis can be proven for the peening variable, however. The HCF life screening statistics show no significant effect for peening, yet an S/N plot (see Fig. 26) reveals that peening is signifiantly more beneficial for chemically milled specimens and less beneficial for mechanically polished specimens. These results are explained by the effects of surface finish and surface residual stress. The fine, directional surface finish and probably slight compressive stress of the mechanically polished specimens achieved a longer HCF life than the rougher, chemically milled specimens which probably had no surface stress. The peening produced a large compressive surface stress while also greatly roughening the surface and implanting glass inclusions in the surface layers. The induced stress in the peened specimens was great enough to offset the extra surface imperfections in comparison to the chemically milled specimens, but not in comparison to the mechanically polished specimens. The fracture behavior of the chemically milled specimens without peening was quite different from the other specimens. Four specimens frac- tured in the weld metal and many exhibited some major surface cracking not associated directly with the final failure. This indicates that chemical milling has a deleterious effect on fatigue crack initiation resistance. Two possible causes for this behavior are grain boundary attack and hydrogen embrittlement. The chemical milling solution used for this study had no surfactant and was much less in total volume than called for in the specifications. Although the actual milling removal was within specified - 69 - rates, it is possible that there was some hydrogen pick-up on the surface layers that could not be detected by hot extraction. would be grain boundry attack by the milling solution. A more likely cause Microscopy of the surface (see Fig. 21) showed the underlaying platelet structure had been revealed. While this sort of localized attack was mild, it may have been enough to help initiate fatigue cracks. - 70 - VI. SUMMARY AND CONCLUSIONS Rods were processed according to a statistical screening design to study the effects of weld repair on HCF properties of cast Ti-6-4. Variables studied were -treatment before welding, welding bead size, post- weld HIP treatment, heat treatment time and temperature, heat treatment cooling rate, surface polish and peening. Specimens were tested by R.R. Moore rotating beam fatigue machines and examined by metallography, fractography,x-ray radiography, RMS surface measurements, and chemical analysis. The conclusions drawn from the results of this study and other previous work are: 1. Weld repair is not detrimental to HCF properties of cast Ti-6-4. 2. Surface condition is the most critical variable in HCF life of flexing cast Ti-6-4 components. A finer surface finish and a higher surface compressive stress improve HCF properties. Mechanical polishing is superior to glass bead peening, which is superior to chemical milling. 3. Microstructure is also a critical variable in HCF life of cast Ti-6-4 components. A larger colony size in the base metal improves HCF properties by slowing fatigue crack growth provided the colony size is small compared to the specimen thickness. The weld zone resists fatigue crack initiation due to a basketweave structure and thin GBa. 4. No secondary HIP treatment is necessary in cast Ti-6-4 if the - 71 - weld zone is sound. A HIP treatment may be made either before or after welding without detriment to HCF properties of cast Ti-6-4. Multiple HIP cycles, if necessary, are also not detrimental. 5. Multiple pass repair welds, when compared to single pass welds, have no effect on HCF properties of cast Ti-6-4. - 72 VII. POSSIBILITIES FOR FUTURE WORK Surface condition is the most critical variable in HCF properties of flexing cast Ti-6-4 components. Different surface finishes and resi- dual surface compressive stresses could be compared to find an optimum finishing procedure. Microstructure is also a critical variable. Different heat treat- ments could be compared as to their effects on HCF properties. Cor- relations between the microstructures produced could show the relative importance of crack initiation as compared to crack growth rate on fatigue properties. This study concentrated on reversed bending fatigue. Other studies could concentrate on different testing conditions such as varying fatigue cycle frequency and testing temperature, or on different testing methods, such as tension-tension. Studies on other titanium alloys may allow broader generalizations about the effects of processing on HCF properties to be developed. Of particular interest would be generalizations about effects on cast alloy fatigue properties. - 73 - Appendix I EFFECT OF TRIAL 5 RESULTS The replacement specimens for trial 5 ran-out during fatigue testing. A value for the number of cycles to failure for trial 5 (n) was necessary for the statistical analysis, however. The statistics were therefore calculated for each of several assumed values of "Log n" (see Table 13). These values were plotted against the resulting factor effects to show that over a wide range of assumed values, the surface finish variable is over 90% significant and the time and temperature variable is approximately 80% significant (see Fig. 28). For the purposes of this study, a value of log n = 7.0 was assumed as this is nearly run-out and the actual value has little affect on the results. - 74 - C\j *c .- r- Ln - 0c 0 -- O l I CO <- 0i 0 tI CO c i i C ,C cO O C O OCCY CO COLo -- L- C Lf V) C.) LL LL LL C..) LL U LL- A I.L) W 00 a, zJ. O F- LO Cm Lt rC r- - r-11 U 1.4 I_ q. 0 nO Lc CCCI LOI C) cO .--O U, 4-) 0 .0 CO LIL. !--) - F- L- 0 a, 44C 1. Ln C -i N- Cn Cj -- C0 ILO r- Z U cC- CY) LOI' I co rEo u · (.rar cn C), .. I- C-, CJ O CO .o C rJ, V) LL LU -. r l cjl - - LU LL L- LU LL. a C .- I- (A n U) * O CYC:O r- - S- 1.0 CD CD LC CD CO C) C) O C) n r-. N cco o ro 10 O q L O · -CO <· CO 00· O)0 C,- a, aj -r 0 0 CO r- aI- E - = LO 1.0 - co 0 Il.- (AO (.-. c:c U, Llc s. r h T- Co la LO co d 0) co O o0 .: 4" o 4. 0) 0) 0. I- -a o CO Co (0 N. aa. U) 1-0 0 U) _.: c (anleA sqn) 133-13 0 0, 4" 'I Cu 0 013OOV 0) - 76 - Appendix II FAILURE STRESS AT FRACTURE Almost all specimens fractured in the base metal. Although the proper presentation of the fatigue results would show all points on an S/N plot at the maximum stress (72.5 ksi), it is convenient to plot the results at failure stress for comparison to unwelded results and previous work. To determine the failure stress, the diameter is calculated at the fracture plane (see Fig. 29) and compared to a list of constants. [48] These constants are multipled by the weight applied to the ends of the specimen in the R.R. Moore machine to get the applied stress. A constant moment along the reduced section is assumed for the analysis. [49] By measuring the distance of the fracture from the minimum cross-section and using this method, the failure stress for each specimen was calculated (see Table 14). L. 0 L. 0o 0i 0) co x 0 0 . ._1. 0 -7A I X E,) . ax ._ c l co m, E x c 0E 0 c- -o C E X rr 0) L. cc LL 0 II 1. - 78 - ., LO Lr It- OC o M Si-W C\JC\.CO o M (/) v-i-- O v- CK COrO COu 1\ LO C\i C\'rj r-- On O 'CIO 0U)O )O LO LC\ LOCO 1 O' r- Co,0 CJ- 0C tD OCM 00 0 r- r-r0 0 - C\ ln1 0 C\J C\J Or C 1- Ci LOr- 1- - rl C - o M 1- 1- r- C\ LOC C\JCQ C\J 1- 1- t -ko O k-O0 0Ck OIO1, Co . r tDl r- 0 r-- C\JC\J I- co :jC0 n 1- MLS) d- 1- C\J C\J - 1- O ±' Cr LO D 1O C'J 1- - LU v) 0)n -.a LU -- , F O c 0.. O C)LA ,--C0 ,--- M,----. ., .' . . o C Cooo co t ko r- CO rrr_ r r_ r-r- r r-_r-_ r r - Q0 C -"O r- r In . -.. cO 'c.LO CO. O0Cc)O O 00 . . . . . . . t r-r- r- LAtO 0rr_rr-~ coCO r- r- I'~,..0 LOI.r o..I r-. CO M ' O O M00' -zCM CO) CMO0 CD O o o0 =_ r- lt<J -" c a) C0 LO c - I"-.: r O0 LO M M CD M0,-MO0C.00~ 0C C) O0 M0 . CC <COC0 M M0", - 'r M C.D 'C - r." C 0O OC C CD CCC 00' 0 J _ ,~f''.O-CMC -- LO O O00 · ·L:._ O0(N. = 0-- - c i) CD 0.--OO L-Cr-"O C Com 0o i ,':", Lo- C i-- I'~ O ,O:~ ,f'~ ,0:CO :iO c: ,-0COO 0 O:L "O- 'O, -O" J:' o-: c-a-: O- LO c::t'00 of-0 C---, ooO O-C v- C) OO C -t - 79 - BIBLIOGRAPHY 1. "Cast Titanium Compressor Impellers", contract DAAK51-78-C-0019, Solar Turbines International, San Diego, CA (1978). 2. "Cast Titanium Compressor Components", contract DAAG46-76-C-0042, Solar Turbines International, San Diego, CA (1976). 3. "Cast Hot Isostatically Pressed Titanium Main Rotor Hub", contract DAAG46-79-C-0043, Sikorsky Aircraft, Stratford, CN (1979). 4. "Cast Titanium Compressor Impellers", contract DAAK51-78-C-0020, Detroit Diesel Allison, Indianapolis, IN (1978). 5. C.D. Bass, Air Force Materials Laboratory Report: AFML-TR-69-116 (1969). 6. M.M. Allen, W.J. Barice and W. Grant, Air Force Materials Laboratory Report: AFML-TR-76-192 (1976). 7. D. Eylon and B. Strope, 8. D. Eylon, 9. S. Silverstein and F. Conrad, Sikorsky Aircraft Laboratory Report: MER-K1-0004 (1977). 10. B.A. Ewing and K.A. Green, DDA Interim Phase I Report on contract DAAK51-78-C-0020 (1980). 11. G.M. Glenn, SAMPE Quart.,8(1976), pg. 1. 12. D. Johnson, Precision Castparts Corp. Process Control Procedure for the Welding of Titanium Castings, Portland, OR (1973). 13. A.N. Hammer, unfinished report on contract DAAG46-76-C-0042, Solar Turbines International. 14. G.R. Yoder, L.A. Cooley and T.W. Crooker, Met. Trans. A, 8A (1977), J. Mater. J. Mater. Sci., Sci., 14 (1979), 14 (1979), pg. 345. pg. 1914. pg. 1737. 15. J.E. Campbell, Metals and Ceramics Information Center (Battelle) Report: MCIC-71-02, Columbus, OH (1971), pg. 31. 16. J.J. Lucas, Titanium Science and Technoology, Vol. 3, R.I. Jaffee and H.M. Burte, eds., Plenum Press, New York, N.Y. (1973), pg. 2081. - 80 - 17. R. Castro and L. Seraphin, Mm. Sci. Rev. Metallurg., 63 (1966), pg. 3. 18. C.G. Rhodes and J.C. Williams, Met. Trans. A, 6A (1975), pg. 1670. 19. J.C. Chesnutt, C.G. Rhodes, and J.C. Williams, ASTM STP 600, C.D. Beachem and W.R. Warke, eds., ASTM, Philadelphia, PA (1976), pg. 99. 20. D. Eylon and P.J. Bania, Met. Trans. A, 9A (1978), pg. 1273. 21. D. Eylon and W.R. Kerr, ASTM STP 645, B.M. Strauss and W.H. Cullen, Jr., eds., ASTM, Philadelphia, PA (1978), pg. 235. 22. J.C. Williams and M.J. Blackburn, ASM Trans. Quart., 60 (1967), pg. 373. 23. G.R. Yoder, L.A. Cooley,and T.W. Crooker, Met. Trans. A, 8A (1977), pg. 1737. 24. G.R. Yoder, L.A. Cooley, and T.W. Crooker, J. Eng. Mater. Tech, 99 (1977), 25. pg. 313. G.R. Yoder, L.A. Cooley, and T.W. Crooker, J. Eng. Mater. Tech., 101 (1979), pg. 86. 26. G.R. Yoder, L.A. Cooley, and T.W. Crooker, J. Eng. Mater. Tech., 101 (1979), pg. 805. 27. G.R. Yoder, L.A. Cooley,and T.W. Crooker, Naval Research Laboratory Memorandum Report: 4232 (1980). 28. G.R. Yoder and D. Eylon, Met. Trans. A, 10A (1979), pg. 1808. 29. D. Eylon, Met. Trans. A, 10A (1979), pg. 311 30. W.G. Burgers, Physica, 1 (1934), pg. 561. 31. J.C. Williams, Titanium Science and Technology, Vol. 3, R.I. Jaffee and H.M. Burte, eds., Plenum Press, New York, N.Y. (1973), pg. 1433. 32. S. Silverstein, J. Lucas and D. Markey, private communications (1980). 33. G.E. Miller, TiTech Research Report: 138, Pomona, CA (1972). 34. R. Nagan, PCC Compendium of Cast Properties of Ti-6A1-4V (1975). 35. A.G.S. Morton and I.R. Lane, Jr., Titanium Science and Technology - 81 - Vol. 1, R.I. Jaffee and H.M. Burte, eds., Plenum Press, New York, N.Y. (1973), pg. 119. 36. W.C. Harrigan, Jr., Met. Trans., 5 (1974), pg. 565. 37. M. Kelly, Howmet Technical Report: "Properties of Welds Deposited on Titanium Investment Castings" Whitehall, MI. 38. E.A. Taylor, A.H. Burn, and H.R. Clarkson, The Science Technology and Application of Titanium, R.I. Jaffee and N.E. Promisel, eds., Pergamon Press, New York, N.Y. (1970), pg. 1037. 39. J.M. Sprauel, R. Rodriguez, J.L. Lebrun and G. Maeder, J. LessCommon Met., 69 (1980), pg. 107. 40. ASM Metals Handbook, 8th ed., Vol. 2, ASM, Metals Park, OH (1964), pg. 301. 41. L.J. Bartlo, ASTM STP 459, ASTM, Philadelphia, PA (1969), pg. 144. 42. R.L. Plackett and J.P. Burman, Biometrika,33 (1946), pg. 305. 43. B.A. Ewing, private communication (1980). 44. R. Dickey, PCC Process Control Procedure for Titanium Chemical Milling and Pickling Solution (1977). 45. S. Silverstein, private communication (1980). 46. E.E. Underwood, Quantitative Stereology, Addison-Wesley Publishing Co., Reading, 47. MA (1970), pg. 34, 81. ASM Metals Handbook, 8th ed., Vol. 8, ASM, Metals Park, OH (1964), pg. 42. 48. Information supplied with R.R. Moore machine, Baldwin-Lima-Hamilton Corp., Philadelphia 42, PA. 49. S.H. Crandahl, N.C. Dahl, and T.J. Laydner, An Introduction to the Mechanics of Solids, McGraw-Hill Book Co., New York, N.Y. (1972), pg. 423.