DESIGN AND TOOL ANCHORING FOR A 120-KILONEWTON EXPANDER CYCLE ROCKET ENGINE LIQUID OXYGEN TURBOPUMP

As part of a German nationally funded research programme ¤TARES,¥ a turbopump initiative has been started in recent years within Airbus DS GmbH. The aim of this study is to design a liquid oxygen (LOx) turbopump assembly (LOx-TPA) for a 120-kilonewton thrust class expander cycle rocket engine. To realize this objective, Airbus DS GmbH builds on in-house heritage, notably the turbopumps of the P111 and the H20 staged combustion engines. This experience serves as input for the design of the 120-kilonewton LOx turbopump. The current paper details the §uidic design of the turbopump, including the design philosophy and the anchoring on the heritage hardware. Discussed are the pump and turbine predesign starting from the con¦guration trade-o¨, the preliminary design, the §ow path and blade design, and the design of inlet/outlet and the volute. Finally, the performance (nominal and o¨-design) is characterized by means of three-dimensional (3D) computational §uid dynamics (CFD) simulations.


INTRODUCTION
In recent years, a turbopump initiative has been started at Airbus DS GmbH as a part of a German nationally funded research programme ¤TARES.¥ The study is aimed at the design of an LOx-TPA for a 120-kilonewton thrust class expander cycle rocket engine. To achieve this objective, a turbopump design process has been established, building on available software packages and capabilities with the acquisition of additional options and packages wherever necessary. For the anchoring of the design and simulation tools, Airbus DS GmbH draws on in-house heritage, notably the turbopumps of the P111 and the H20 staged combustion engines. Available design and test data have been used, with a focus on the redesign of the heritage hardware based on existing construction drawings and the original hardware itself. The CFD simulation results of the P111 axial turbine were compared to available historic data in terms of performance and studies have been initiated to recharacterize the H2O LOx inducer, both numerically and experimentally.
This experience serves as input for the design of the 120-kilonewton LOx turbopump. A trade-o¨has been performed [1] to de¦ne di¨erent con¦gurations of the LOx-TPA, evaluating several alternatives to establish the most promising baseline design variant. Criteria include the pump and turbine e©ciency, the cavitation margin, and the number of turbine stages. The selected baseline turbopump design consists of a high head inducer, radial impeller, and vaned di¨user on the pump side driven by a single stage subsonic axial turbine on the same shaft.
The paper details the §uidic design of the turbopump, including the design philosophy and the anchoring on the heritage hardware. Discussed are, with the trade-o¨in [1] as starting Figure 1 Established turbopump design work §ow point, the preliminary design based on one-dimensional (1D) theory and semiempirical correlations (inlet/ outlet diameters, blade heights, number of blades, e©ciency estimate, and cavitation characteristics), the §ow path and blade design (twodimensional design of meridian contour, blade mean lines, and blade pro-¦ling), and the design of inlet/outlet and volute. Finally, the performance (nominal and o¨-design) has been characterized by means of 3D CFD simulations and the suction performance of the inducer is characterized experimentally.

DESIGN PROCESS
As a part of the ¤TARES¥ research program, one of the principal aims of the current initiative within Airbus DS GmbH has been to establish the turbopump design process for the development ( Fig. 1), with the aim of de-veloping a 120-kilonewton LOx-turbopump. The design §ow builds largely upon available software packages and capabilities. The design process has now been established by means of anchoring on available heritage hardware data and a full design loop for the 120-kilonewton LOx-turbopump including manufacturing and testing for the inducer hardware.

ANCHORING ON HERITAGE DATA
To anchor the design process that has been established as a part of the TARES research program, Airbus DS GmbH has drawn on its turbopump heritage. Most notably, the H2O 200-kilonewton and the P111 50-kilonewton stage combustion engines have been used as validation cases, for which detailed design and test data are available [1].

Water and Liquid Oxygen Turbopump Inducer
The ¦rst case for anchoring the design tools was the H2O LOx-inducer from the 200-kilonewton H 2 /O 2 staged combustion engine developed in Ottobrunn in the late 1960s and early 1970s for the Europa III rocket. Design and test reports are available from this period, documenting the thorough characterizations that were made on this inducer, including parameter studies concerning di¨erent geometrical properties. The tests were performed in both water and LOx at rotational speeds between 20 000 and 40 000 rotations per minute (rpm), the latter being the nominal value for the LOx-pump. Table 1 shows the nominal inducer characteristics. The relevant quantities are de¦ned as follows: with the speci¦c speeds being de¦ned using the units rpm, m 3 /s, and m for, respectively, the rotational speed, the volumetric §ow rate, and the head.   The original hardware is still available today and has recently been scanned as input for 3D-CFD simulations (Fig. 2). The existing hardware and design data have served as a basis for the current redesign of the inducer (Fig. 3). In addition to an anchoring of the design tool, it serves as a basis for further CFD investigations, in particular, for cavitation modeling studies. These were undertaken by the DLR G ottingen in the context of the Propulsion2020 cooperation. In addition, the original inducer hardware has been recharacterized in a test campaign at the University of Kaiserslautern.

P111 Axial Turbine
The P111 LOx/kerosene 50-kilonewton staged combustion engine, developed in the 1960s in Ottobrunn, was chosen as the second anchoring case for the design tools. In particular, the single stage reaction turbine, driven by an oxygen rich preburner, was a valuable validation case. The original hardware, including design and test documentation, is available and has served as input for the redesign (using the original geometry) and the 3D CFD simulations.
The original P111 power-pack hardware, an excerpt of the construction drawing of the axial turbine, and the blade pro¦les are shown in Fig. 4. The ¦gure also shows (at the bottom right) the numerical mesh used for the 3D CFD simulations of the redesign of the P111 axial turbines. The computational grid has a total size of ∼ 1.4 million cells. It is divided into two domains: the stator domain with 500 000 cells and the rotor domain with 900 000 cells, both resolving a single blade. The applied wall resolution is y + = 100 using wall functions to estimate the viscous losses. The appropriate periodic boundary conditions were used to represent the complete turbine behavior. The CFD results are shown in Fig. 5, depicting the static pressure ¦eld normalized by the total inlet pressure (Fig. 5a) and displaying the relative Mach ¦eld (Fig. 5b). Shown are for both ¦elds a blade-to-blade view generated at a relative blade span of 0.5 (top ¦gures) and a mass weighted meridional view (bottom ¦gures).
The performance results from the 3D CFD simulation are compared to the heritage hardware data in Table 2. The original e©ciency has been determined from the test data. To enable a proper comparison with the current simulation results, the mechanical losses and the disk friction losses have been corrected  for in the heritage data based on the original design estimates for these loss contributions. A good agreement between the original data and the simulation results has been obtained with respect to the pressure ratio and the e©ciency, hence providing an empirical validation of the aforementioned loss models as well as of the CFD simulations.

120-KILONEWTON TURBOPUMP DESIGN
The redesign experience, consecutive tool anchoring activities, and the planned test campaign with the heritage hardware all serve to establish the work §ow for the Airbus DS GmbH turbopump design activities. This section details the current §ow path design of the 120-kilonewton turbopump that shall establish a baseline for further optimization studies.

Design Requirements and Con¦guration Trade-OB
ased on the operating requirements (for the nominal conditions, see Table 3) for the 120-kilonewton turbopump, a dedicated trade-o¨has been performed (for details, see [1]) to de¦ne di¨erent con¦gurations of the 120-kilonewton LOx-TPA, evaluating several alternatives to identify the most promising design variants. The selected baseline turbopump con¦guration consists of a radial di¨user pump with a high head inducer driven by a single stage subsonic axial turbine on the same shaft with a rotational speed of 25 000 rpm.

Pump Fluid Path Design
Based on the nominal pump requirements and the selected rotational speed of 25 000 rpm, a preliminary sizing was performed. The relevant nondimensional parameters are de¦ned in Eq. (1). The ¦rst pump diameter and e©ciency estimate were made based on the N s D s -diagram [2]. In addition, an estimate of the minimum required shaft diameter was made. The number of blades Z = 6 was based on the correlations from [3,4]. The suction characteristics of a pump stage without inducer were evaluated based on the P §eiderer criterion [3]. This yields the ¦rst estimate of NPSHR, con¦rming indeed the need for an inducer. These estimates give the ¦rst idea of the overall dimensions of the pump and the requirements in terms of suction speci¦c speed and inducer head. Based on these predesign estimates, a preliminary design of the inducer was made. The inducer inlet was sized based on the Brum¦eld criterion to attain the minimal NPSHR value that provides su©cient cavitation margin. Following the approach from [5], an optimal §ow coe©cient of � ind,in = 0.1 was selected, in line with the values for other realized LOx-inducers [2]. This is conservative in terms of the attained minimum experimental values speci¦ed in [5]. The inducer inlet hub-to-shroud ratio was determined based on a geometrical comparison with existing inducers (amongst others, the H2OLOx inducer). The outlet hub-to-shroud ratio and the outlet angle were determined based on a parametric study, requiring that net positive suction head available (NPSHA) > NPSHR at the impeller inlet with a margin that was anchored on existing hardware. As a baseline, the inducer outlet diameters at the hub and shroud and the impeller inlet diameters were ¦xed to the same value, verifying that the impeller suction diameter was close to the ideal value in terms of NPSHR. The blade number was set to Z = 3 and the solidity were selected based on the recommendations from [5]. A sweep angle was selected for increased suction performance and to decrease the bending load on the ¢corner §ap£ of the blade leading edge. The value for the tip clearance was based on the H2O experience.
The shrouded impeller was designed to attain the required total head rise at a nominal head coe©cient value of ā imp = 0.85 using semiempirical design functions and simultaneous anchoring on available reference data for existing pumps. The outlet blade height to impeller diameter ratio was set to yield a meridional deceleration close to unity. The axial extension of the impeller was sized according to G ulich [6]. The outlet angle was set in agreement with the recommendation from [4]. The blade mean lines were chosen such as to obtain a monotonous static pressure rise, while satisfying the Ackeret criterion for the relative velocity. The wrap angle was set to the maximum value that still enables manufacturing by means of milling.
The vaned di¨user was designed following recommendations from [4,79] for the geometrical sizing of diameters, gaps, and clearances. The number of di¨user vanes was taken equal to the optimum value based on considerations of periodicity to avoid interaction with the impeller blades. The volute was designed for the maximum expected o¨-design §ow with a circular cross section, folded inwards for compactness and §ow stability. The cross-sectional area was determined using the P §eiderer approach [3].
The resulting geometrical design (Fig. 6) and the design estimates for the nominal total and static pressure and the absolute velocity progressions through the pump (Fig. 7) are shown assuming Figure 6 The 120-kilonewton LOx-pump design a linear loss distribution between volute and di¨usor (losses of both parts are not discriminated). The overall pump design is summarized in Table 4.
A 3D CFD-analysis was made as a basis for further optimization loops. The computational grid encompasses the full 360 degree pump geometry and contains a total of ∼ 5.8 million cells divided over 4 domains: 1.7 million for the inducer, 1.6 million for the impeller, 2.2 million for the di¨user, and 350 000 for the volute. Concerning the inducer and impeller, a wall resolution with y + < 300 was applied. Grid re¦nement studies on a single channel model have shown that this approach is su©cient for the prediction of the head rise. The wall in the volute was treated with wall functions as a ¦rst approximation since the focus in the current investigation was on the characterization of the rotating parts. Figure 8 shows the CFD result for the static pressure ¦elds normalized by the total outlet pressure.   Figure 8 The 120-kilonewton LOx-pump nominal pressure iso-contours. The particle traces are visualized in Fig. 9 for several §ow coe©cient values. Figure 10 shows the obtained o¨-design performance computed for several operating points (OPs), including a polynomial ¦t. The §ow visualizations for the lower OPs show clear signs of §ow breakdown, which is re §ected in the e©ciency. The nominal OP shows a regular §ow behavior through the pump with a single vortex in the volute and the presence of a tip vortex at the inducer inlet. It is caused by an interaction between the main §ow, the boundary layer at the casing, and the tip leakage §ow due to a positive incidence of the §ow on the inducer blade tip as has been explained in [10]. For §ow rates superior to the foreseen application range, the inducer inlet tip vortex disappears whereas elevated velocities start to occur within the volute (as is to be expected since it operates outside its design range). The §ow coe©cient varies within approximately ±10% from the nominal value for the foreseen operating envelope, which is where the maximum e©ciency is obtained. The CFD predictions indicate a 5 percent higher hydraulic e©ciency for the whole pump design than estimated in the preliminary design. The current design and simulation results provide the baseline for developing a physical understanding of the relevant §ow phenomena as input for further optimization studies.

Axial Turbine Flow Path Design
The trade-o¨in [1] has identi¦ed a single stage impulse turbine (reaction rate ρ < 0.15) as a viable option to produce the required shaft power for driving the LOxpump while complying with the nominal operating conditions speci¦ed in Table 3. To start the design process, the turbine e©ciency is a necessary input which determines, in particular, the rotor diameter and the combination of stator exit angle and blade height. The nominal characteristics in Table 3 with a 20 percent bypass mass §ow rate directly prescribe a minimum overall e©ciency level of η = 68% to attain the necessary power output; this is achievable based on the existing hardware values presented in [2].
The isentropic velocity ratio U/C 0 is taken as the governing parameter for the turbine e©ciency, with the spouting velocity C 0 resulting from the input parameters.
To make a detailed trade between e©ciency η TS , stator exit angle α 1 , turbine disk diameter D m , and rotor blade height b r , the relation from [11] was evaluated in detail. Considering only the stator exit angles yielding the maximum e©ciency (i. e., 10 • < α 1 < 15 • , de¦ning all angles relative to the circumferential direction), the relation agrees well with the other theoretical and empirical relations that were considered in [4,1115]. All curves attain a maximum e©ciency η TS between 79% and 84% with the optimal U/C 0 in the range of 0.400.55 (Fig. 11), providing an indication of the required U/C 0 .
The e©ciency estimates for the current design case based on the correlation from [11] are shown in Fig. 12. The interdependence of the isentropic velocity ratio, the stator exit angle, and the rotor diameter and blade height is also shown. A stator exit angle of α 1 = 10 • at U/C 0 = 0.4 seems to o¨er the best compromise between e©ciency and blade height for the ¦rst design, yielding a predicted e©ciency of η TS = 81%. This is su©cient for attaining an overall e©ciency (ratio of shaft power to total-to-static isentropic power) of η = 68% based on the P111 heritage data, noting that the maximum for the total-tototal e©ciency is very §at and the optimum e©ciency value is also obtained close to U/C 0 = 0.4. Smaller values of U/C 0 would reduce the e©ciency and, hence, potentially compromise the bypass mass §ow rate requirement, and larger Figure 11 Total-to-static e©ciency estimate correlations from literature: 1 ¡ NASA SP8110 [14]; 2 ¡ Nuzel and Huang [4]; 3 ¡ Dobrovolskii [12]; 4 ¡ Ovsyannikov and Borovskii [13]; and 5 ¡ Glassman [15] Figure 12 Optimization results for e©ciency, rotor diameter, and blade height: 1 ¡ α1 = 30 • ; 2 ¡ 20 • ; 3 ¡ 15 • ; 4 ¡ 10 • ; and 5 ¡ α1 = 5 • values of α 1 would progressively reduce the blade height, leading to unrealizably small blades. The basic blade height was taken constant for the whole length of the turbine (stator inlet, rotor inlet/outlet), with only a slight contraction at the stator outlet for guidance in order to avoid spillage due to expansion of the §ow between stator outlet and rotor inlet, based on recommendations from [14].
Based on this preliminary sizing and providing the reaction rate of ρ = 0.15 selected in the trade-o¨ [1], the velocity triangles now follow directly from the 1D design. Using §ow deviation values for the §uid leaving the stator blades based on [14], the required blade angles were estimated. The blade chord was calculated based on empirical correlations and a comparison to existing hardware. Finally, the blade design was generated starting from available standard pro¦les, using the respective recommended values for pitch and stagger angle for the selected pro¦les. The ¦nal blade pro¦le was generated by means of Bezier curves, using optimization loops with CFD predictions to attain the desired hydraulic power output of the turbine. The overall axial turbine design is summarized in Table 5.
The blade and meridional design are shown in Fig. 13, together with the velocity triangles (values are normalized by the blade speed U ) and the progression of the §ow quantities through the turbine (the stations represent the stator inlet (0), stator outlet/rotor inlet (1), and rotor outlet (2), respectively).
In view of the inlet and outlet design for the turbine, a trade was made with respect to the §ow direction. A con¦guration with inlet manifold and axial outlet was compared with a con¦guration with axial inlet and outlet volute. The latter con¦guration has been selected as a baseline because it provides an axial in §ow to the turbine stator and it has the potential of minimizing losses at the turbine outlet by reabsorption of the residual swirl component. Figure 14 shows the geometry with the applied numerical mesh for the 3D CFD simulations as well as representative results of these turbine simulations.

Figure 13
Blade and meridional design (a), velocity triangles normalized by the rotor speed U (b), and normalized §ow quantities progression (c): 1 ¡ total pressure; 2 ¡ static pressure; 3 ¡ total temperature; 4 ¡ total absolute velocity; and 5 ¡ total relative velocity The computational grid has a total size of ∼ 1.3 million cells. It is divided into two domains: the stator domain with 1.0 million cells and the rotor domain with 300 000 cells, both resolving a single blade. The applied wall resolution is y + = 100 using wall functions. The appropriate periodic boundary conditions were used to represent the complete turbine behavior. Concerning the CFD results, Fig. 14b depicts the static pressure ¦eld normalized by the total inlet pressure and Fig. 14c displays the relative Mach ¦eld. Shown are for both ¦elds a blade-to-blade view generated at a relative blade span of 0.5 (top ¦gures) and a mass weighted meridional view (bottom ¦gures). The required power level and the e©ciency were surpassed by a few percent in the last design iteration. This desired stator exit angle was 1 • smaller than the design value, whereas  the §ow turning 180 • β F2 β F1 was attained to within less than half a degree. The current design and simulation results provide the baseline for developing a physical understanding of the §ow phenomena as input for further optimization studies.
To summarize the §uidic design, Fig. 15 shows the 3D CAD view of the rotating parts of the 120-kilonewton turbopump design including all principal §uidic parts (inlets/outlets, volutes, and di¨users).
Inducer characterization tests in water have been performed at the Von Karman Institute of Fluid Dynamics (VKI) in the frame of an Airbus DS funded activity. The scaling to attain the proper test conditions is shown in Table 6. The inducer has geometrically been scaled up to ¦t the test section and manufactured from a polymeric material (acrylonitrile butadiene styrene, ABS) by means of rapid prototyping (Fig. 16). The hub contour has been extended to accommodate the larger test bench interface diameter. The rotation speed and §ow rate have been scaled based on §ow coe©cient and head coe©cient similarity, verifying that the Reynolds number is su©ciently high to attain similarity of the inducer performance. In addition, the water is heated to the appropriate temperature to attain thermal similarity with LOx based on the thermal e¨ect parameter › * [8].
The tests will be evaluated to con¦rm the inducer characteristics, both in terms of achieved head coe©cient and attained suction performance. In ad-  Figure 16 The 120-kilonewton LOx-inducer design (a) compared to rapid prototyping test hardware (b) Figure 17 The 120-kilonewton LOx-inducer during cavitation testing at the VKI dition, the dynamic cavitation behavior will be characterized. The ¦rst cavitation visualization results are shown in Fig. 17, starting from the nominal conditions on the left down to a cavitation coe©cient of σ = 0.25 on the right. The ¦rst occurrence of cavitation is highlighted by the ellipse. The test campaign results shall serve as input for the validation of the design tools, the CFD simulations, and cavitation models as a baseline for further inducer optimizations.

CONCLUDING REMARKS
Within the context of the German nationally funded research programme ¤TARES,¥ several activities have been initiated to build up and reestablish turbopump design competences within Airbus DS GmbH. The main focus has ¦rstly been on the anchoring of the §uidic design and simulation tools on in-house heritage hardware; subsequently, a design was made of an LOx turbopump for a 120-kilonewton thrust class expander cycle rocket engine, including accompanying experimental and modeling e¨orts. These activities have served to attain the following milestones: reestablishing proven design heritage based on the H2O 200-kilonewton and the P111 50-kilonewton staged combustion engine turbopump hardware; anchoring of the turbopump §uidic design tools to the heritage hardware and available literature data for existing turbopumps; generating a 120-kilonewton LOx-turbopump design including all main §ow path components; performing 3D CFD simulations of the heritage hardware and the 120kilonewton turbopump design; and obtaining ¦rst cavitation test results in water of the 120-kilonewton LOxinducer design.
Further activities have been performed, such as the hydraulic characterization campaign for the heritage H2O LOx-inducer. The combined experimental results provide validation data for 3D CFD simulations and cavitation modeling.
These milestones serve to establish the Airbus DS GmbH §uidic design process and have provided a baseline turbopump design for further optimizations, experimental investigations, and modeling studies.