Impeller Flow Field Laser Veloc meter Measurements

Development of Computational Fluid Dynamics (CFD) computer codes for complex turbomachinery affords a complete three-dimensional (3-D) flow field description. While significant improvements in CFD have been made due to improvements in computers, numerical algorithms, and physical modeling, a limited experimental database for pump CFD code validation exists.


INTRODUCTION
evelopment of Computational Fluid Dynamics (CFD) computer codes for complex turbomachin- ery affords three-dimensional (3-D) flow field calcula- tions.However, existing data for pump CFD code validation is limited.Accurate databases are required to validate the CFD codes to enable evaluation of existing turbulence models and grid number requirements.
Traditional flow field survey information from pressure sensors and directional probes are intrusive and typically yield an uncertainty in excess of 0.5% of full range and thus do not yield quality benchmark data.Pressure sensors and directional probes only yieldcir- cumferentially averaged information.Laser velocimetry is nonintrusive and yields accurate flow velocity and angle data.More importantly, laser velocimetry provides information at specific circumferential locations blade to blade.
The objective of the test program undertaken was to obtain benchmark quality data, flow velocity and angle at key locations in a generic pump operating at the impeller design flow rate.These data along with data from a similarly tested impeller [1, 2] will be used by the Consortium for Computational Fluid Dynamics Application in Propulsion Technology Pump Stage Team [3] to validate pump CFD codes.
The tested configuration consisted of a four-bladed unshrouded inducer, a shrouded impeller with six full blades and six partial blades, in conjunction with a diffusing crossover and discharge plenum.The flow field survey consisted of 10 radial stations along one axial plane at the impeller inlet, 11 axial stations in a radial plane immediately downstream of the impeller dis- charge, and 9 axial stations close to the crossover inlet.The fluid medium for the laser velocimeter surveys was ambient water.H is head gc is the gravitational constant (2) Utip is the impeller discharge tip speed Nondimensional Velocity: V Vnondim Utip where: V is velocity Utip is the impeller discharge tip speed (3) TEST

ARTICLE
The test article was a Rocketdyne-designed shrouded impeller that met the operational requirements for the Space Transportation Main Engine (STME) fuel pump and designated the "Consortium baseline impeller."The impeller was tested with an axial inlet and a four-bladed unshrouded inducer.The axial spacing be- tween the inducer and impeller was intended to minimize the inducer wake effects at the impeller inlet while still providing enough critical speed margin for rotordynamic stability.Testing the impeller in conjunction with an axial inlet and inducer provided geometry typical of current rocket engine pumps.
While the test article geometry reflected current rocket engine designs, the impeller shroud wear ring clearance was purposely atypical of an engine.The nondimen- sional radial clearance between impeller shroud and the polypropylene seal was nominally -0.000166.The ra- dial interference fit minimized impeller discharge recir- culation flow that may be more useful for validation of current CFD codes.This arrangement affords less com- plex CFD modeling since both the front and rear shroud recirculation zones can be eliminated.
The test configuration, presented in Figure 2, included a crossover discharge to minimize asymmetric flow at the impeller discharge.

TEST FACILITY
The test program was conducted in the Engineering Development Laboratory (EDL) Pump Test Facility (PTF) located at Rocketdyne's main facility in Canoga Park, California.A schematic of the flow loop is pre- sented in Figure 3.The test article was driven by a 1,200-rpm, reversible, synchronous electric motor rated at 2,984 kW.The motor was coupled to a 2,984-kW gearbox capable of producing speeds of 6,322, 8,013,  and 10,029 rpm.The pump CFD code validation con- figuration was tested at 6,322 rpm.
Water was supplied to the closed flow loop from a 28.769-m stainless steel tank.The tank was rated at 1.0342 MPa with a vacuum capability of 0.0962 MPa.A heat exchanger located adjacent to the tank maintained a uniform fluid inlet temperature.The flow rate to the tester was regulated by a throttle valve downstream of the pump.After passing through the throttle valve and into the tank, the flow passed through a series of baffles in the tank and was recirculated through the facility.

TEST INSTRUMENTATION
Two four-tap static piezometer rings were used to mea- sure the inlet static pressure.They were located approxi- mately 13.5 and 8.7 inlet pipe diameters upstream of the inducer.Water temperature was measured in the inlet line with a platinum wire filament temperature detector to allow calculation of the fluid density and vapor pressure.Three impeller inlet static pressure measurements were aligned in one axial plane, but distributed around Parameter the circumference (90 deg apart) to identify any circum- ferential pressure variations of impeller inlet flow.This identification is important because the impeller inlet laser survey occurs at one particular circumferential location.Circumferential variations are not reflected in the laser survey.The impeller inlet static pressure tap plane was located a nondimensional axial length of 0.0766 down- stream of the laser survey plane.
Similarly, identification of circumferential variations at the impeller discharge was accommodated by three static pressure taps located around the tester circumfer- ence in a plane midway between the two impeller discharge laser survey planes.One static pressure tap was included at each of the two laser survey radial planes.A differential static pressure measurement be- tween the lowest radial impeller.dischargesurvey plane and the tester inlet provided redundancy.All pressure measurements were obtained with Taber full bridge strain gage transducers with an accuracy of _ _ .0.5% of full range.Flow rate was measured with one 20.32-cmturbine-type flowmeter in the inlet line with measure- ment redundancy provided by another 20.32-cm turbine- type flowmeter in the discharge line.Additional instru- mentation was employed for facility monitoring and computer redlines.

TEST PROGRAM
The test program comprised a fiber-optic laser two-focus (L2F) veloimeter survey in one axial plane at the impeller inlet and two radial surveys at the impeller discharge.Table 2 lists the nondimensional radial positions of the impeller inlet survey referenced from the shaft centerline.The positions are also identified in terms of the percentage of the laser survey plane annulus height.With this designation, 0% is the hub and 100% is the laser window face on the inner diameter of the inlet housing.The nondimensional axial length between the laser survey plane and the impeller leading edge tip was -0.1290.The negative sign indicates an axial location upstream of the axial reference zero.The laser veloci- meter is capable of partitioning any given circumferen- tial section into 16 separate data zones called "windows."For the tested configuration inlet survey, data were acquired over two adjacent 90-deg segments.The survey was performed over 180 circumferential degrees because the inducer had four blades and the impeller had six full blades; it was not known a priori if inducer or impeller effects would dominate the flow field at the survey location.With 16 windows, data were segregated into distinct flow zones of 5.625-deg (90 deg/16 windows) angular sweep.An axial projection of a blade leading- edge tip intersection with the impeller shroud was used as the centerline of an impeller inlet timing mark which was machined into the impeller hub.This mark allowed triggering the start of the laser survey to be coincident with an impeller full blade.The center of the timing mark was used to commence circumferential laser velocimeter surveys at all annulus heights.Since adjacent 90-deg survey arcs were not similar in terms of impeller blade location, data were not ensemble averaged.
The impeller discharge surveys were performed at multiple axial locations along two radial planes.The nondimensional radial locations of the axial surveys were 0.5138 and 0.5597.The plane located at a nondi- mensional radius of 0.5138 was immediately down- stream of the impeller discharge.The plane located at a nondimensional radius of 0.5597 was the furthest down- stream survey possible and at an approximate location of downstream components in actual hardware.Table 3 lists the laser survey nondimensional axial positions refer- enced from the impeller discharge tip shroud, 0.0% of impeller B 2 width.A negative value occurs when the axial location of the survey point is forward (toward the tester inlet) of the impeller discharge tip shroud.A value greater than 100% is indicative of a laser survey position at an axial distance toward the tester aft end, in excess of the impeller B2 width.
Encompassed in a 60-deg segment between two adja- cent full blades at the impeller discharge is a partial blade.Circumferential partitioning over two adjacent 30-deg segments provided better circumferential flow definition between adjacent blades.With 16 windows, data were segregated into distinct flow zones of 1.875 deg (30 deg/16 windows) of angular sweep.The impeller discharge timing mark centerline was located from an axial projection midway between the impeller suction and pressure surfaces at the impeller tip, 0% impeller Be width.
The center of the timing mark was used to commence circumferential laser velocimeter surveys at all axial survey locations.Adjacent 30-deg segments were not similar and ensemble averaging of the impeller discharge data was not performed.
Although the impeller inlet and discharge laser sur- veys spanned multiple test days and long time durations, a statistical analysis indicated that test-to-test parameter variations were negligible.
The tester shaft was marked and served to trigger the start of the laser velocimeter survey in the circumferen- tial direction.This shaft mark was not necessarily coin- cident with the circumferential location of the timing mark machined at the inlet and discharge of the impeller.Therefore, the angular distance or equivalent time delay between the desired survey start and the laser start trigger were determined.A photonic sensor was used to generate a once-per-revolution (OPR) pulse from the drive shaft of the tester.A timing mark was machined onto the impeller surface as described in the section entitled Test Program.The L2F was positioned so that the measure- ment volumes were able to detect the timing mark and generate a scattered light signal.The light signal was converted to an electronic signal via the L2F signal processing electronics.The phase angle between the occurrence of the OPR pulse and the timing mark pulse was determined.
The phase angle between the OPR pulse and the impeller inlet timing mark signal was also performed as the assembly rotated at full speed.Good agreement existed between the static and dynamic phase angle methods at both the impeller inlet and discharge loca- tions.TEST

RESULTS
The impeller inlet laser velocimeter survey, which may serve as a boundary condition to CFD codes, also serves as a check for data integrity.Laser surveys over 180 circumferential degrees (two inducer blade passages or three impeller blade passages) were performed because of the dissimilar number of inducer and impeller blades.Laser data indicated inducer effects were prevalent and no significant impeller effects were evident in the flow.
Data integrity was checked with a flow continuity match, i.e., a comparison of integrated flow based on the laser velocimeter velocities to flow measured with two tur- bine-type flowmeters.A continuity match of 98.6% was achieved at design flow coefficient.Discrepancies be- tween integrated and measured flow may occur because of grid coarseness and/or use of a trapezoid integration scheme that does not include interpolation of the flow velocity to the walls.The continuity match was consis- tent with those calculated during previous test programs.Table 4 presents a summary of the impeller inlet flow characteristics for one of the 90-deg surveys.The aver- age values at each radial position were essentially the same for both 90-deg circumferential surveys.The sum- mary compiles nondimensional absolute velocity (C/ Utip) nondimensional absolute tangential velocity (Cu/ Utip), nondimensional absolute axial velocity (Ca/Utip) absolute flow angle, nondimensional relative velocity (W/Utip) nondimensional relative tangential velocity (Wu/Utip) and relative flow angle at each radial position surveyed.The values presented reflect flow-weighted nondimensional velocity, flow-weighted nondimensional tangential velocity, and area-weighted axial velocity.The other values were derived from these parameters.Bulk flow quantities were derived over all radii and represent average values for the entire flow.The hub to tip distribution of axial velocity is linear over most of the flow annulus, between 25 and 90% of impeller inlet annulus.Figures 4 and 5 present contour plots of the impeller inlet axial and tangential velocity components.The relative tangential velocity, axial velocity, and inci- dence angle at the impeller inlet are the parameters that affect the impeller discharge flow characteristics and hydrodynamic performance.
No significant circumferential pressure variations were detected at the impeller inlet.The variations mea- sured were within transducer accuracy.
At the impeller discharge, the flow continuity match, calculated across the data locations within the impeller  B2 width, immediately downstream of the impeller, is 115.2%.A number greater than 100% is attributed to laser survey grid coarseness coupled with the trapezoid integration method used to calculate the flow rate.At the plane located further downstream, calculated flow devi- ated only by 1% from the measured flow.This excellent continuity match is attributed to obtaining data that on the average represented the overall flow characteristics of the discharge channel.
Tables 5 and .6 present summaries of the impeller discharge flow characteristics for the two radial planes.Figures 6 and 7 present contour plots of flow absolute radial velocity and absolute tangential velocity components at the impeller discharge.Each of the contour plots contains data for the two radial planes.The plane on the left side is nearest to the impeller discharge.
From the presented plots, the impeller discharge wakes are clearly evident in the plane immediately downstream of the impeller.Regions of low radial velocity, absolute flow angle, and high tangential veloc- ity extend over most of the axial region, hub to shroud, in the wake defect areas.At a radial plane further downstream from the impeller, this becomes less distinct since mixing causes the flow to become more uniform.The nondimensional radial velocity clearly shows sec- ondary flow downward into the impeller shroud and hub clearance.
A circumferential variation in impeller discharge static head coefficient of 1.1% was measured.This value is slightly outside the range of transducer accuracy.The majority of circumferential variation in impeller dis- charge static head coefficient may be attributed to trans- ducer accuracy with a slight contribution caused by an asymmetric discharge flow condition.The crossover discharge was effective in eliminating the typical dis- charge flow asymmetry.
The impeller discharge laser survey traversed the impeller discharge channel while commencing in the same location circumferentially for each axial location.No attempt was made to alter the circumferential start location to coincide with the impeller full blade; thus, a rectangular grid results.A calculation was performed that defined which data windows corresponded to the indi- vidual flow passages in the plane immediately down- stream of the impeller.In this manner a flow split among the two blade passages encompassed by the laser survey was performed.The passage between the first full blade and the partial blade was calculated to contain 49% of the total flow.The percent of calculated flow in the passage between the partial blade and the adjacent full blade was 51% of total flow.This type of uneven flow distribution, driven by secondary flow along the suction surface, was similarly exhibited in Reference 1].pressures were included at key locations to provide boundary conditions for CFD code validation.Laser surveys at the inlet and discharge of a Rocketdynedesigned impeller were performed.Repeatability of test conditions was demonstrated over the entire test duration.
The success of this empirical study adds significantly to the limited database of accurate centrifugal impeller flow field measurements.Further laser velocimeter sur- veys throughout various pump components will permit additional CFD code benchmarking data to be .acquiredand, ultimately, to a better understanding of fluid phenomena.A test program to obtain benchmark quality data for typical rocket engine pump geometry was successfully performed in Rocketdyne's EDL PTE Data were ob- tained nonintrusively with a L2F velocimeter, static success of the project and Michael R. Yandell for his design contributions and support during fabrication.

CONCLUSION
Table summarizes the impeller geometry and test conditions.Figure provides a layout,of the impeller blades.

TABLE 2
Impeller inlet laser survey radial locations

TABLE 3
Impeller discharge laser survey axial locations