ci? e Sectuns, or printed in ks ptd lications. Dec sssion is pri ted only if the paper is published in an ASME Journal. Authorization to photocopy

Transcription

1 THE AMERICAN SOCIETY Of MECHANICAL ENGINEERS 345 E. 47th St., New York, N.Y GT'505 S Society shall not be respons^le!or s ftments or opinions advanced in papers or lion at meetings of the Society or of i3s Divisions or ci? e Sectuns, or printed in ks ptd lications. Dec sssion is pri ted only if the paper is published in an ASME Journal. Authorization to photocopy for internal or personal use is granted to libraries and other users registered with the Copyright Clearance Center (CCG) provided $3/article or $4/page is paid to DCC, 222 Rosewood Dr., Danvers, MA Requests for special permission or bulk reproduction should be addressed to the ASME Technical Publishing Department. Copyright C 1998 by ASME All Rights Reserved Printed in U.S.A. TWIN WEB DISK - A STEP BEYOND CONVENTION Ronald R. Cairo Pratt & Whitney Kathleen A. Sargent Air Force Research Laboratory ABSTRACT This paper will discuss a study of an innovative design for an advanced turbine rotor, that could have a great impact on future engines. The design challenge is to provide a minimum weight turbine rotor system that can withstand beyond state-of-the-art levels of AN2 (turbine annulus area multiplied by speed squared). An AN 2 limit has been reached for High Pressure Turbine (HPT) disks configured in conventional (single web) geometry with state-of-the-art nickel alloys. The problem has reached the point where increased AN2 has been declared a `Break-Through" technology. The twin-web disk has the potential to provide this break through. This paper will present the history of this turbine rotor design, analytical results, material/component processing and concept validation results. All work was performed under an Air Force sponsored program entitled Composite Ring Reinforced Turbine (CRRT). INTRODUCTION The CRRT program was structured to expand upon a concept that was conceived during the last phase of the prior Air Force sponsored Advanced Turbine Rotor Design (ATRD) program. The concept, an internal ceramic matrix composite (CMC) ring reinforced nickel alloy disk capable of satisfying advance engine demonstrator requirements, showed significant weight and stress reductions relative to a conventional nickel alloy HPT disk during initial trade studies. The CRRT program started with this concept and progressively added more details to the design study and began to include manufacturing considerations as they impacted the disk geometry. The program also remained flexible to the increasing demands of the demonstrator engine. However, the turbine rotor inlet temperature and associated HPT body temperature increased to the point where contemporary CMCs could no longer offer a structural advantage due to the thermal incompatibility between the nickel and CMC. This concept was subsequently replaced by a non-ring reinforced twin-web disk (TWD) that eventually became the focus of the program. The TWD continued to offer significant advantages over conventional HPT disks in terms of structural load path efficiency and AN2. DESIGN STUDIES Design Criteria/Philosophy Design constraints included a 1000 cycle demonstrator life, a 1.94 inch (4.93 cm) minimum bore radius, and a 5.2 inch (13.2 cm) maximum bore width. The boundary conditions reflected simultaneous application of maximum T4 (turbine rotor inlet) temperature and AN2. These consisted of a bore temperature of 1200 F (649 C), a live rim temperature near 1350 F (732 C), and an AN2 of 600x 10*s in2rpm2 (3871 x 10's cm2 rpm2 ). The temperatures at other disk locations were interpolated using a simple conduction analysis. All structural analyses were performed using ANSYS 5.0. Allowable stresses were based on surface flaw crack growth. The disks were required to meet 1000 Total Accumulated Cycles (TACs) of life based on an advanced engine duty cycle. Lives for metallic components were computed for a range of stresses using Pratt & Whitney's crack growth analysis program. Inspectable areas of disks were required to meet 1000 TAC cycles of crack growth life to support a 500 TAC inspection interval. Internal areas (twin-web cavity) which could not be inspected after disk-assembly were required to have 2000 TAC cycles crack growth life. Presented at the International Gas Turbine & Aeroengine Congress & Exhibition Stockholm, Sweden June 2 June 5, 1998

2 Initial Design studies with CMC Rina The initial configurations of disk designs used geometric parameters that were output from a Taguchi experiment. This study suggested that some of the critical variables were disk bore width, composite ring width, disk cavity width, disk bore height and thickness at ring inner diameter, disk bore height at appendages for flange and bearing attachments, and appendage structural flexibility. The unique loading situation of transferring radial load into the composite ring results in an unusual state of stress in the rotor system. The studies all revolved around transferring load from the nickel disk to the CMC reinforcing ring(s) by radial compression, loading the ring inner diameter (ID). Load transfer by this mechanism avoids the weak, interlaminar tension and shear load paths of the CMC. The rings were designed with fiber architectures that specifically avoid axial splitting and local ID crushing. The rings were developed and demonstrated under the ATRD program. The emphasis at each stage of the design iteration was to minimize the appropriate components of stress in order to satisfy durability concerns consistent with generating the lightest weight rotor system possible. As the design evolved, added realism was built into the finite element model. This was to ensure that the analytical results most appropriately reflected the rotor response. The stresses for the initial and final CRRT configurations are presented in Table 1.0 for comparison. disk half geometries. At the completion of this study it was decided to drop the CRRT and continue with the TWD design. The design effort now focused on meeting two additional objectives: (1) analyzing the current TWD design to determine stress levels to set bond strength requirements and (2) determining interface requirements for the spin demo disk/spin arbor system, the details of which will not be discussed in this paper. However, for the first objective, the Finite Element Analysis (FEA) revealed several critical stress locations. One location is the bond joint itself which is axially stressed by virtue of the twin webs and the corresponding secondary moment, brought about by the eccentricity in load path induced by the blade load line of action and the centroid of each web. Axial tensile stresses of 65 ksi (448 MPa) are predicted. Another critical stress location is the web, which is currently designed to a maximum allowable stress of 135 ksi (931 MPa) to minimize weight. This stress can be reduced, as required, if producibility becomes a concern. The third critical location is the bore since it is required to carry all disk loading beyond the "selfsustaining" radius (the radius beyond which a rotating mass becomes parasitic rather than load carrying). This area is currently designed to conventional design allowable stresses for weight optimization. Again, stresses can be reduced, if required, by the addition of bore volume at the expense of weight savings. Final Design Studies without Rina As operating temperatures for the engine increased due to cycle changes, the viability of the CRRT HPT came under question. The thermal property mismatch between the CMC and nickel alloy was pushed beyond the threshold for structural efficiency. The additional temperature caused the higher expansion rate nickel bore to grow too much relative to the CMC ring resulting in excessive axial bending. Design modifications such as an initial gap between the bore and ring to modulate ring loading were tried but proved inefficient. As a result, a design comparison was completed between the CRRT and another design concept that removed the CMC ring, the TWD. The design requirements discussed in the design criteria section were used for this study. Both disk configurations were iterated upon until strength and durability requirements were met. Comparing hoop stress plots, as seen in Figs. 1.0 & 2.0, the bore stress distribution for the TWD was more uniform than the CRRT disk since the CMC ring provided localized restraint due to a combination of reduced inertial and thermal response relative to the nickel disk. This restraint induced a small degree of axial bending, in the nickel bore, about the ring. Also, there was a higher peak radial stress at the bore/web juncture for the CRRT due to the limited space available to develop a smooth geometric transition via a generous "blend" radius. Both configurations had desirable axial stress distributions, however, the TWD had preferred distributions. The bore interface stresses for the TWD were entirely in compression at 30 ksi (207 MPa), compared to a gradient of 45 ksi (310 MPa) compression to 8 ksi (55 MPa) tension for the CRRT. Another result of the design change was that the TWD was 7 lbs (3.2 kg) lighter. Figure 3.0 shows an overlay of the TWD and the CRRT PROCESSING Several different processing methods were considered for producing this HPT disk. Initially, for the CRRT design, the fabrication process was selected to minimize damage to the encapsulated CMC ring and involved a consideration of inertia welding and forge joining to join the rim, and electron beam (EB) welding to join the bore. Characterization and development involved identification of essential joining parameters, metallography and baseline mechanical properties (tensile and LCF) of welded test specimens. Full-scale ring specimens were also evaluated. Concerns relative to maintaining disk-half to disk-half concentricity and the cost of fixturing full-scale disk halves caused inertial welding to be dropped from further consideration. EB welding was also dropped due to concern over controlling penetration of the EB weld beam and eliminating an inherent notch in the process. As the TWD emerged as the HPT concept of choice, the bond process was re-evaluated. There was a need for a bond process that minimized material deformation. This need was driven by the requirement to achieve concept goals for control of part position, minimizing "fish mouthing" at bond boundaries, and maintaining an interior bond zone that had mirror image symmetry about the bond plane for spin dynamics/balance control. The remainder of the program focused on forge joining and the development of a derivative process known as Activated Forge Joining. Activated Forge Joining (AFJ) utilizes features of both Transient Liquid Phase (TLP) and Forge Joining to produce high quality metallurgical bonds while imparting low distortion. Similar to TLP, AFJ utilizes a foil interlayer to suppress the melting point of the alloys being joined. Similar to forge joining, it utilizes an interface pressure, though greatly reduced relative to pure forge joining. The key feature of AFJ, when compared to TLP, is its ability to transition to production more

3 effectively due to shorter heat cycle times, retention of parent mechanical properties across the bond, and the ability to join hardware in final or near final machined condition due to minimal material upset. Bond development activities were replanned to include AFJ using a combination of coupons, subscale rings and full-scale rings. A Design of Experiments approach (DOX) was included in this plan utilizing the AFJ parameters. A 27 factorial was set up to determine the optimum process parameters such as load, heating rate, bond cycle length, and post-bond heat treatment. Mechanical specimens were extracted from each bonded unit to assess tensile, creep-rupture and impact response. Continuous strain data to the point of rupture were recorded for creep-rupture testing. The conditions were set at 95 ksi (655 MPa) and 1350 F (732 C) to limit specimen test time to 100 hours or less for economy and to provide data that could be compared to the database available for the alloy. The results of the creep rupture tests showed an upper bound failure life of 98 hrs with greater than 20% reduction in area of the failed specimen. These results exceeded the specification requirement for that of the parent material which is 75 hrs. Tensile test results also compared favorably with the parent material baseline properties. Machining Tolerance Specimen Another parameter that was evaluated for AFJ, in addition to those in the DOX, was the machining tolerance (on the interface surfaces) required for an acceptable bond. The baseline machining tolerance was inch (5 mm) on flatness. A subscale specimen was intentionally flawed on one surface prior to bonding. Four channels were milled to individual depths of (5 mm), (10.1 mm), (15.2 mm), and inches (20.3 mm). This coupon was then bonded to a standard smooth coupon using AFJ. The intent of the experiment was to determine how well the AFJ process handled machining mismatch to establish a tolerance limit for the detail design. If the process was robust, unusually liberal machining requirements could be allowed. Photomicrographs were taken of the bond cross section and showed porosity at each channel location. The amount of porosity was deemed unacceptable. Hence, the projected +/ inch (+/- 5 mm) flatness requirement for the disk halves would be distributed over a 19 inch (48.26 cm) diameter part rather than a 1 inch (2.54 cm) diameter coupon. The experiment concluded that it is required to retain the current (0.002 inch, 5 mm) flatness requirement. Process Scale-up Up to this point, the viability of AFJ had been demonstrated with coupons, rings, scaled disks, and the designed experiment. Realizing that some of the bonding problems may be unique to the sub-scale rings that lack the heat path of a disk web and the heat sink of a disk bore, the project expanded the study to include a combination of sub and full-scale simulated disk halves. Applying the processing parameters defined in the designed experiment to full-scale geometry and controlling part positioning in addition to mechanical properties, proved most challenging. Bond Trials were run on 6 disks that were designated FSD-02 through FSD-08 (FSD-04 was not bonded) with bond integrity varying from excellent to somewhat porous. FULL-SCALE BONDING TRIALS The results for the first full-scale simulated disk bonding trial (FSD-02) revealed that the interface pressure was too low. Metallographic samples and test specimens were extracted for evaluating the bond. Both the micrographs and the ultrasonic evaluation showed the best bonding occurred at the outer and inner edges of the rim, with porosity present through the center regions. Investigation showed that the single root cause for the high level of porosity was a low bond pressure at the interface. It was discovered that due to error in the equipment set-up, the load applied was approximately one third the desired level. This was corrected for all subsequent trials. A second bond trial was run successfully (FSD-03: also a simulated full-scale disk) with the proper full bonding load, followed by the first monolithic set of disk halves (FSD-05) which would produce trial full-scale bond specimens more closely matching the spin demonstrator. FSD-05 was the first fully processed, monolithic disk-half bonding trial performed. The bonded disk was subjected to ultra-sonic and dimensional inspection and then put through a stress relief and age heat treatment. There was very little dimensional change in the bonded disk before and after stress relief. A photo micrograph showing grain growth across the interface is included as Fig FSD-06 was the first full-scale, full-geometry, pedigree alloy disk bonded. Ultrasonic scans (C-scans) indicated that the rim outer region was bonded more effectively than the center region, showing a decrease in density going from the mid-bond plane region to the ID. Specimens were extracted from various locations per the test plan shown in Table 2.0. Test results for monotonic tension and stress rupture for bonded specimens were extremely impressive in terms of property level and consistency. Figure 5.0 summarizes monotonic tension test results for insitu bond strength and insitu parent alloy strength. As can be seen, tension ultimate and yield strength for both bond and parent alloy are essentially the same. As expected, elongation and reduction in area for the joint are approximately 50% of parent response. These levels are, however, more than adequate for the HPT application. Stress rupture results, shown in Fig. 6.0, are equally impressive. The test conditions were selected for comparison to established data for the parent alloy. The minimum requirement is 10 hours at this condition. As can be seen, the insitu bonded stress rupture life is clustered between 26.6 and 32.3 hours. LCF testing was completed on specimens cut from various sections of the disk to include bonded and non-bonded regions in both the rim and the bore. The LCF test data showed that the bond lives are lower than the parent lives but acceptable for the design. The LCF test data were generated for a maximum stress of 140 ksi (965 MPa) at R=0.05 and 1000 F (538 C) for comparison to the available database of parent material. In actuality, the engine HPT design exposes the bond to an average tensile stress of 10 ksi (69 MPa) and a local peak tension stress of 37 ksi (255 MPa). The spin demonstrator has an average axial stress on the bond plane of about I ksi (6.8 MPa) in compression and a peak axial tensile stress of 29 ksi (200 MPa). 3

4 LCF data is shown in Fig. 7.0 for the rim section only, since this is the main point of interest. In general, it is readily apparent that a significant debit in LCF capability is associated with the bonded specimens. It was inferred that improvements in this capability were likely to come with lower porosity levels. Therefore, reduction of bond porosity was viewed as a key element in future development work and would be pursued by optimizing the chemistry of the bond activation layer. Crack growth testing was performed on 2 specimens from the rim section, with the bond aligned with the specimen notch. The first specimen was threshold tested twice at 20 Hz, followed by a Hz Region II crack growth test, which is the region of stable, slow, controlled crack growth. The second specimen was tested in Region II only to allow sufficient crack length to obtain data at higher crack growth rates. Test results were compared to both a baseline threshold model and baseline data. The test data compared to the model favorably, with slightly slower crack growth rates in the threshold and a slightly faster growth rate in Region IL Compared to baseline data, the tests show slightly faster growth rates at the higher stress intensities. Examination of the broken specimen halves revealed the crack grew within or at the edge of the bondline, and did not propagate into the parent material. Photo micrographs revealed desirable transgranular rather than intergranular failure morphology. The bonding trial for FSD-07, which was originally targeted as the spin demonstrator disk, appeared to go smoothly. Yet, there was trouble with the function of the pneumatic cylinder which provides the bonding load. An unusual spread in the rim temperatures during the trial was noted and a section of the rim appeared to be overheated. Inspection of the rim bond-line seemed to substantiate a low or no-load condition during the trial. Loss of load was consistent with the unusual temperature profile seen along the rim of the part since poor surface contact along the rim causes non-uniform heat transfer which could result in localized hot spots. This part was completely unsuitable for use as the spin demonstrator or for bond property characterization. However, 32 specimens were extracted for mechanical property evaluation of the base material. The final rotor, FSD-08, is shown in Fig It was bonded successfully and has been designated as the spin demonstrator. Witness material testing revealed properties similar to FSD-06 (reference Figs ). It must be emphasized that the characterization data developed during this program were focused to support the design of an experimental Spin Disk. Much more, and costly, characterization is required to support man-rated applications. This is part of a typical evolutionary approach to evaluate an innovative concept. Once validated, the concept would receive more detailed design and characterization to enable it to transition to engine applications. RPM. Also to be conducted is an elevated temperature, interrupted low cycle fatigue test to 1000 cycles. This test will be interrupted at 300 and 600 cycles for visual and NDE inspection. Upon completion of the LCF test, the disk will be subjected to a room temperature overspeed test to 110% maximum speed and then to burst. All of the data gathered from the above tests will be correlated with analytical predictions, to validate the design and codes. SUMMARY The Twin-Web disk represents a radical departure from conventional HPT design practice and was conceived to meet the aggressive AN2 goals of the future. The disk was developed by an integrated technology team that included specialists from design, metallurgy, structures, non destructive evaluation, and manufacturing, and resulted in a novel metallurgical joining process. The joining development process utilized a "building block" approach starting with small 2 inch (5.08 cm) square blocks (in a statistically designed experiment to optimize joining parameters), progressing to scaled rings (disk rim simulation); then scaled disks; and finally, full-scale/full-geometry disks. Mechanical properties were evaluated in all critical locations to ensure application to demonstrator engine hardware and correlation made to ultra sonic evaluations of bond integrity. A full-scale spin demonstrator was fabricated and will be subjected to structural validation spin testing. Future activities will include studies to introduce shot peening to the internal web walls and rim inner diameter to enhance life and further validation in a demonstrator engine. REFERENCES 1. Advanced Turbine Rotor Design Final Report, Contract F C-2594, Report Number AFWAL-TR Composite Ring Reinforced Turbine Program, Contract F C CONCEPT VALIDATION In order to verify the TWD load path and bond structural integrity, concept validation spin testing will be performed on a spin demonstrator in the Spring of A room temperature strain survey will be conducted on the disk which will be instrumented at the bore, webs and rim. Strain measurements will be taken at speed increments of 2000 RPM up to 21,500 4

5 Table 1.0: Stresses from Design Trade Studies for CRRT Stress Location Initial Design Final Design Upper Rim Stress in Web, 202 ksi 139 ksi Radial Stress (1392 MPa (958 MPa) Bore Forward Edge Hoop 220 ksi 158 ksi Stress (1517 MPa) (1089 MPa) Radial Bearing Stress, -165 ksi -121 ksi Disk Load Ring Interface (-1138 MPa) (-834 MPa) CMC Ring Stresses -33 ksi (-228 MPa) Radial -31 ksi (-214 MPa) Radial 72 ksi (496 MPa) Hoop 79 ksi 545 MPa) Hoop Table 2.0: FSD-06 Test Specimen Summary Specimen/lest Type Quantity Test Conditions Tension F(704 C) Stress Rupture F/106 ksi 732 C/731 MPa) Low Cycle Fatigue F (538 C)/ 20 c m/r=0.05 Crack Growth Rate F (649 C R=0.1.r Y. Max Bore Stress = 158,000 psi (1098 MPa) Figure 1.0: Hoop Stress, CRRT Max Bore Stress 149,100 psi (1028 MPa) Figure 2.0: Hoop Stress, TWD 5

6 r twin Web r I Disk ^ -- 1 i Composite Ring Reinforced Turbine Disk Figure 3.0: Overlay of CRRT and TWD Geometries Growth Across Interface on FSD-05 tiiiiiiiiiiiiiii.. L.. uuiiiiiuuuii w i! ui uiiuiuuiiii iii iiumeniii Figure 5.0: Monotonic Tension Results for FSD-06 Specimens Specimens 1-7: Rim-Bonded, 8-10: Rim Parent, 11-14: Bore 6

7 Stress Rupture Results from FSD-06 Specimens Specimen p DURATION, hours p ELONGATION % A REDUC AREA % 6.0: Stress Rupture Results for W M p BONDED EDGE p BONDED CENTER pun-bonded EDGE SPECIMEN NUMBER Figure 7.0: LOW CYCLE FATIGUE DATA FOR FSD-06 RIM 7