P 51967 A

2.3 Seal Flow System

The first stage vane blade gap features a planar seal as shown in Figure 2.10. Since the experimental turbine facility is a cold flow facility and was not initially designed to model seal flow behavior, there was no inherent need for the cooling flow passages that would normally be integral to a high temperature turbine. In this original configuration, the inner cavity that would exist on a turbine engine was instead replaced with an open-ended rotor drum facing the upstream center-body and the hollow inlet nosecone. Thus addition of internal walls and flow passages was required to allow for the seal flow studies.

2.3.1 Seal Flow Path and Cavity Geometry

A dimensioned drawing of the seal geometry is presented in Figure 2.11. The nominal width of the seal at the measurement location is 18 mil (0.45 mm), and the axial gap to disc radius ratio is 2.4x10-3. The vane-rotor gap varies during rotor rotation due to run-out of the rotor shaft. This run-out is measured statically as 3 mil (0.0762 mm) on the 6 in. (0.152 m) diameter drum/shaft mounting point. Further variation in seal gap size exists during operation due to dynamic effects and minor variations in drum-shaft mounting. To account for these variations, the actual seal gap has been measured both with optical techniques and with feeler gages.

The optical measurements are dependent on the optical set-up described within Section 2.4.1. Using the PIV camera, a view of the seal and measurement plane is digitized, and each image pixel has a given length for the specified optical magnification. The number of image pixels that fit within the seal gap is then the measure of the seal width. A sampling of this value was taken for both static conditions and under operational conditions. The range of seal widths as measured optically is 18-26 mil

(0.45-0.65 mm). For comparison, the range of seal widths measured with feeler gages ranged from 17 - 19 mil (0.43-0.48 mm). It is postulated that the optical measurements did not measure the actual minimum seal width, but rather a wider portion that exists within the imaged area, due to rounded edges of the seal.

The model in Figure 2.12 shows the cavity form, which is below the seal gap. The instrumentation ring is shown with three spokes supporting a center disc. The open space in this ring is sealed with a rigid, heat shrinkable plastic film that is applied over the entire surface of the ring. At the center of the ring an annular pattern of electrical connectors are raised from the surface by 0.11 in. (2.8 mm) and are located at a 5.56 in. (141 mm) diameter at the outer most point. Thus, the plastic sealing film forms a conical shape leading up to the planar annulus at the outer ring of the wheel. On the opposite side of the seal, an acrylic sheet is mounted to the seal ring and forms the fixed facing wall of the seal cavity. The maximum cavity width is 0.35 in. (8.9 mm) at the outer radius.

At the center of the fixed seal wall is an annular opening of 2.5 in. (63.5 mm) outer diameter and 1.375 in. (34.9 mm) inner diameter through which the pressurized seal flow is delivered. The seal flow is delivered to this opening through a section of telescoping PVC pipe mounted within the inlet nose cone. Any misalignment between the center of the nose cone and the center of the pipe coupling is absorbed in the o-ring that seals to two pipe sections. At the upstream edge of the pipe section, duct putty is used to seal the pipe with the rough interior wall of the nose cone.

This seal flow feed pipe is attached to a centrifugal blower powered by a 5-hp (3.75-kW) electric motor. This blower is only required for the highest seal flow rate used in these experiments, as in the other two cases the natural pressure differential and rotor pumping action was sufficient to supply the required flow rates. A throttle plate on the blower inlet, a throttle valve on the delivery pipe inlet, and pipe friction were all used to control the seal flow rate. For the three seal flow rates investigated, two different size seal flow supply lines were used. For the larger flow rate a 3 in. (76.2 mm) supply pipe was used, and for the smaller flow rates a 1 in. (25.4 mm) supply pipe was used. A photograph of the seal flow delivery system is given in Figure 2.13, and a schematic drawing of the details of the system is provided in Figure 2.14.

2.3.2 Seal Flow Rate Measurement

Seal air flow rates of 90-120 cfm (0.042-0.056 m3/s) are inferred by velocity measurements made with a United Sensor model PAA-"a"-KL 1/16 inch (1.6mm) Pitot-static probe. A small probe diameter was selected for the purpose of minimizing end wall boundary layer interference effects. At a distance of over 10 probe diameters from the flow passage wall, there is no measurable effect of the endwall on accuracy. Interference of an accurate measurement would normally arise from a venturi affect developed within the probe - wall space. Additionally, the flow Reynolds number could be of concern. For low flow rates delivered to a small impact hole, viscous effects may be an issue. United Sensor Corporation publishes 12 ft/sec (3.65 m/s) as the critical velocity for air at standard atmospheric conditions impacting a hole of 0.010 in. (0.25 mm) in diameter. The smallest velocities are measured around 40-50 ft/sec (12 - 15 m/s) in the 3 in. (76.2 mm) supply pipe, which is far above the speeds that result in error.

The Pitot-static probe was mounted at 80 in. (2.03 m) from the pipe leading edge for the 3 in. (76.2 mm) supply pipe, and a development length of 80 in. (2.03 m) was used for turbulent boundary layer calculations. For the smaller pipe, the Pitot static probe was mounted at 50 in. (1.27 m) from the pipe leading edge. The probe location was selected so that the flow was fully developed far upstream of the core velocity sampling point. In all three seal flow cases, the pipe Reynolds number was sufficient to use turbulent pipe flow correlations for predicting the boundary layer thickness and displacement thickness. Displacement thickness at the probe location was determined to represent less than 0.4% error in flow rate calculations.

The two pressure measurements for the Pitot-static probe are taken with a PDCR23D Scanivalve pressure transducer, which is made for a maximum 5 psi (34.5 kPa) differential measurement and outputs a ±12 V signal. The static accuracy of the device is published as ±0.06% or ±0.003psi (20 Pa). The differential pressure measurements were taken relative to atmospheric conditions, and were over 200 times greater than the smallest measurable pressure, so the certainty of the individual measurements is not in question. The differential between the two measurements of total pressure and static pressure within the seal flow delivery pipe must be greater than the accuracy of the device. The predicted pressure differential for a flow at 45 ft/sec (13.72 m/s) is 0.0173 psi (199 Pa), resulting in an uncertainty in the velocity measurement of 8.7% or ±3.9 ft/sec (1.19 m/s). An empirical study of this measurement revealed that the measured pressures were consistent with micro-manometer measurements, and thus the true accuracy was higher than predicted.

Pipe flow temperature is measured with an unshielded type-T thermocouple bead. The measurement is electronically cold junction compensated using an Omega Electronic Ice Point model MCJ. The compensated signal was acquired through a National Instruments data acquisition board NB-MIO-16XL on a Macintosh Quadra 950 computer.

2.4 Particle Image Velocimetry

The Dantec Inc. PIV system utilized for these experiments consists of a 30 mJ NewWave Research Minilase III Nd:YAG laser, a high-resolution Kodak Megaplus ES 1.0 digital camera, and a dedicated PC controlled PIV 2100 Processor. The laser has twin oscillators operated in single Q-switch mode and are capable of delivering a 5-7 ns duration pulse with a wavelength of 532 nm (visible green light) at a repetition rate of 10 Hz. The digital camera has a 1008 x 1018 CCD array operated in cross-correlation mode, with the images corresponding to the first and second pulse of the laser recorded separately, with the minimum allowable time between these frames of 1 ^sec. Both images are then transferred to the PIV 2100 Processor which provides near real-time vector processing of the images using Fast Fourier Transform (FFT) correlation techniques. This unit also synchronizes the camera and laser, and is capable of resolving the particle displacement to within one-tenth of a pixel through the use of sub-pixel interpolation schemes. PC controlled software is used to perform off-line validation and post-processing of the vector maps, with directional velocity information unambiguously determined since the initial and final particle positions are recorded as separate images.

2.4.1 Optical Path

To image the seal flow region, a 39.4 mil (1 mm) thick laser light sheet is introduced through the bellmouth inlet section using a combination of cylindrical lenses. Two angled surface mirrors deflect the laser beam to a height such that the beam is at a 15% span location within a radial-axial plane oriented 45 degrees from horizontal. Alignment of the beam with the turbine axis was achieved by mounting a pair of targets at parallel upstream and downstream locations. The bellmouth mounting flange and the diffuser exit flange were mount points for these targets, which carried angular and spanwise demarcations. Thus a gage was established for aligning the laser sheet at a known angular and radial location, on a plane that passed exactly through the axis of the turbine shaft.

The cylindrical lenses chosen are of focal length: f=+80mm, f=-250mm, and f=+250mm, listed in order from upstream to downstream. All three are positioned such that the upstream surface is the planar lens surface. The three lenses are spaced at approximately 8 in. (200mm), but precise adjustment must be made to establish the desired sheet thickness and height. The positive lenses are used to control the sheet thickness, and the negative lens is used to spread the beam into a sheet that is of a designed height. The height of the light sheet must be sufficient to cover the entire measurement region without over-expanding the beam. Over-expanding the beam will result in intensity reduction so that particle reflections can not be distinguished from background noise.

This light sheet is brought from the turbine inlet through the first vane row to the seal region. Due to the typical large turning in this vane row, the removal of a single vane was necessary, thus introducing a local reduction in solidity. The measurement plane, typical of that shown in Figure 2.15, spans the distance between vane endwall trailing edge and rotor leading edge. The flow is illuminated in the axial-radial plane, with the measurement thus resolving those components of the flow. The measurement plane dimensions are 0.8386 in. (21.3 mm) in the axial direction by 0.8464 in. (21.5 mm) in the radial direction.

To image this region, the digital camera views the flow from an oblique angle through a window located over the first rotor row. The optical access window is centered over vertical and the measurement location is 45 degrees off of the vertical. Because of the imaging angle and the window thickness, the camera cannot focus on the entire object plane, with the resulting astigmatic distortion shown in Figure 2.16. For the purpose of correcting the astigmatism, a calibration target was designed and built, as shown in Figures 2.16 and 2.17. The target was a square grid print, mounted on an aluminum block which was clamped to a single rotor blade and fitted to the hub radius such that the target is set in a radial-axial plane. The target was positioned within the plane of the laser sheet by rotating the drum until the rotor-clamped target swept through the laser sheet and the target print itself was aligned with the laser light.

Adjustment of camera focus was accomplished through an adjustment ring on the lens extension body which would lengthen or shorten the separation between the lens and the image focal plane. Focusing the camera on the grid resulted in either the axial grid lines or the radial grid lines coming into focus at two different focal distance settings, but never would they both be in focus simultaneously. As previously stated, this behavior is due to an astigmatism as shown in Figure 2.16, and is a result of the optical access window and the angle with which it is viewed.

To correct the astigmatism, three different window designs were attempted, since a proper design of the window outer surface could completely correct the distortion. The original window design had a planer outer surface, with the inner surface matching the outer radius of the flow annulus. The second iteration on window design was based on the basic lens maker formulas and used a window surface with the same radius as the inner surface. This second design resulted in an astigmatism of opposite sign and half strength of the astigmatism generated by the first window, i.e. the focus points changed order for the axial and radial focus distances. Previous experience had shown that concentric window surfaces minimized distortion, so a third window was attempted which used concentric curves. This final window design achieved a reduction of the astigmatism sufficient that a simple optic element could be used to provide the final correction.

The corrective optic element required to eliminate the remaining astigmatic distortion is a planar acrylic sheet slanted at the opposite angle from that made between the window plane and the image plane. The case window which gives access to the flow path is 1.25 in. (31.75 mm) thick with concentric curvature on the inner and outer window surfaces. The corrective window is 1.4 in. (35.56 mm) thick and skewed at 33 degrees from the camera axis. Because of some difficulty in finding a supplier for thick acrylic and the prohibitive cost of custom glass, the corrective optic was not a single piece but rather, three separate pieces, each approximately 0.46 inch thick, welded together. This three part optic was a compromise as it introduced error in certain measurement situations. The number of material transitions that exist in the light path affects the magnitude of the distortion, and thus the resulting quality of the image. The interior interfaces of the three part optic acted as light tunnels, introducing shadow images in the measurement. Figure 2.19 is a sample of the reflection issues which were introduced. The rotor leading edge image which is located in the lower right hand corner of all the images is duplicated by a shadow image at the mid-height of the image. This reflection was however only present over a small range of rotor positions.

2.4.2 Flow Seeding

Reliable PIV measurements require that the flow be seeded with tracer particles small enough to accurately track the flow and large enough to scatter sufficient light to be detected with the imaging system. These particles must be introduced in a manner that does not significantly disturb the flow field and at the same time provide a sufficient and uniform seeding density in the region of interest. Due to the high flow rate through the turbine, a Rosco 1600 Fog Machine is used to generate the seed particles. This is a thermal aerosol generator that produces a high volume of seed particles by discharging a heated and pressurized glycol based mixture into the atmosphere where it immediately vaporizes and then condenses into a fine mist of monodisperse particles. A uniform seeding density in the test section is achieved by introducing these seed particles upstream of the bellmouth inlet and allowing them to disperse into the ambient air prior to being drawn through the facility.

One stream of high seed density flow from the fogger and another of dry air is pumped into a mixing plenum using a small blower, with the ratio between the two streams setting the final seed density. One side of the mixing plenum is open and covered with a honeycomb flow straightener, allowing seeded flow to discharge at low velocity. The exit of the mixing plenum is located approximately 9ft (3 m) from the turbine inlet, and the position adjusted until the appropriate stream tube feeding the turbine is intercepted.

In the PIV experiments, seeding of the seal flow stream was not attempted. Although the seeding of all streams is common for PIV so that full field measurements can be made, the degree of mixing that is present in the turbulent unsteady flow field of an intra-stage turbine space was sufficient that seeding of only the main stream was needed to provide the particle density required for full field PIV measurements. This was advantageous, as flow visualization studies performed with a seeded seal flow showed that run time was quite limited due to seed fluid deposition. The glycol seed fluid quickly collects on any surface that has become wetted by the fluid particles. Such accumulation is especially problematic in small spaces where the flow path takes sharp turns, and thus significant accumulation occurred in the seal flow delivery path during these studies.

2.5 Vane Relative Measurement Plane Positioning

The measurement plane location was fixed by the optics set-up. Thus to resolve different measurement locations relative to the airfoil positions, the first vane row was indexed. The vane position is measured by the fixed position of the laser generated light sheet on a gauge block which is attached to the indexed vane support ring in place of a single vane. The gauge block mounting surface is contoured to fit the vane ring and to align the center of the gage block with the mounting hole. The center position of the gauge is then precisely known, relative to the vane circumferential location. The gauge surface features a series of angular reference marks spaced at half degree increments, over which a removable piece of photo paper is mounted. When the gauge was in position the laser would expose the photo paper, and thus allow the exact light sheet location relative to the vane to be resolved. Within the half degree marks finer resolution was achieved with standard rule gauges. The gage blocks and associated laser image marks are shown in Figure 2.20.

2.6 Data Acquisition and Processing

The seal flow rates employed in these experiments were selected by dimensionless flow rate (or seal flow Reynolds number) defined as:

ß ro where the actual mass flow rate has been scaled with the fluid viscosity and outer radius of the seal.

Typical values for turbine seals range from Cw=5x10 to 2 x 104 (El-Oun et. al., 1988). For these experiments, three different seal flow rates, Cw= 1.2x104, 9.2x103, and 4.6x103 were chosen. The turbine operating point during the experiments was near the design point at a flow coefficient of ^=0.515 and a loading coefficient of y=1.59. The rotational speed was 2500 rpm. This results in a mean inlet velocity to the turbine of 94.2 ft/sec (28.7 m/s).

The PIV images corresponding to the first and second laser pulses are divided into rectangular interrogation areas, with cross-correlation software used to determine an average particle displacement for each region. To obtain a high signal-to-noise ratio, the interrogation area must be small enough so that the flow velocity is homogeneous within each region and at the same time large enough to encompass a sufficiently large population of particle pairs. The FFT processing algorithm that computes the cross-correlations generates artificial cyclic background noise at the edges of each interrogation area since this approach assumes that the sampled regions are periodic in space. This can result in the loss of particle pairs due to a low signal-to-noise ratio at the boundaries, with particles near the edges not used in the velocity calculation. However, this information is recovered by over-sampling the images using overlapping interrogation regions. This process does not increase the fundamental spatial resolution, but generates additional vectors as suitable interpolations. For the present investigation 32 x 32 pixel interrogation areas with 50% overlap are used to process the image maps resulting in 3844 raw vectors per image.

Due to the combination of a physically small measurement area and moderately large velocities, the minimum time between laser pulses was required. With the pulse delay set to 1 ^s, the finest measurement resolution was limited by the displacement of particles within each interrogation region, resulting in particle displacements of approximately 30% of the length of an interrogation region. A resolution of 32 x 32 pixel measurement regions was chosen. Smaller interrogation regions resulted in too many particles traveling across regions during the time delay between pulses, resulting in an inability to resolve the mean velocity from the background noise.

To characterize the unsteady flow field generated by the rotor potential field, the instantaneous flow field is imaged at a fixed time delay from a known position on the rotor drum. This is achieved by triggering the PIV acquisition sequence with a signal generated from a photo-optic sensor on the turbine shaft. A LaserStrobe 165 Phase Delay Generator introduces the appropriate phase delay calculated from shaft trigger readings. Vernier marks on the rotor drum are also imaged directly with this system to verify the accuracy of the triggering system. The timing system in conjunction with the Dantec PIV system was used to activate a Kodak E.S 1.0 type 10 camera and a Spectra Physics PIV 200 as the flash for capturing the instantaneous rotor position at the specified delay time. The laser beam passed through two cylindrical lenses at a 90° offset thus producing a conical light pattern for illuminating a large area over the top of the first rotor row, including marks on the rotor drum and the fixed centerbody. The instantaneous images for specified delay times were used to determine the rotor blade position nearest the measurement plane, and the variation in rotor position with delay time.

The locations where the data are taken are indicated by the delay time, as shown in Figure 2.21. The accuracy of the rotor relative measurement position is quantified as ±0.25^s (±0.00375 degrees) for a rotor speed variation of ±15rpm about the design speed, and at the longest delay period.

Figure 2.1 Purdue 2-Stage Research Turbine.
Figure 2.2 Turbine blade row configuration.

Stator 1 Rotor 1 Stator 2 Rotor 2

Figure 2.3 Turbine blade sections.

Purdue Low Speed Turbine Rig Simulations

RotoM Midspan Blade Static Pressures

0 0

Post a comment