Flow separation is becoming an increasing problem in highly loaded turbine blades. A gas turbine suction surface simulator has been developed for the BaylorUniversity low speed wind tunnel. A contoured top wall in the test section enables the pressure distribution to be simulated on flat plate that corresponds to actual separated flow conditions. Techniques to stop separation will be studied as well as the impact of separation on heat transfer (using a steady state gold foil liquid crystal technique). Comparison will be made with the CFD code Fluent.

Continued Development of a Turbine Blade Suction Surface Wind Tunnel

Chris Hurst & Blake Bradford

DEPARTMENT OF ENGINEERING

BAYLORUNIVERSITY

4/25/2005

OBJECTIVE:

Current turbine blade research has required the development of turbine blades that are more effective in low Reynolds Number flows. Higher altitudes continue to be achieved and lower core flows continue to decrease as bypass ratios increase producing low Reynolds number situations. In response, Pratt and Whitney has developed the Pak B turbine blade for low pressure circumstances. As Reynolds numbers become less than 200,000, which occur at high altitudes and low velocities, separation occurs on the suction side of the low pressure turbine blade. Once this separation occurs the flow across the blade becomes turbulent, adding energy to the flow. This additional energy allows the flow to reattach just before the end of the blade chord [1]. The separation region that occurs decreases the efficiency of the turbine blade making it less effective. In order to decrease the separation region and make the blade more efficient, the points along the blade in which separation occurs must be determined. The purpose of this project is to locate the points of separation along the Pak B low pressure turbine blade. By locating the separation region, future research can be performed in order to decrease this region.

In this experiment, the profile of the Pak B blade is mounted on the roof of an existing BaylorUniversity wind tunnel. Measuring local static pressures along the blade profile with the use of a flat pressure plate that is elevated above the floor of the wind tunnel, a graph of the coefficients of pressure versus the percent axial chord length may be produced. This graph can then be compared with 2-D Allison VBI theoretical code to analyze how closely the physical profile follows the properties of the theoretical pressure distribution of the Pak B turbine blade [1].

BACKGROUND:

Research in replicating suction surface boundary layer behavior of the Pak B turbine blade has been an interest for both industry and educational sectors alike. In 1992 the Turbulence and Turbine Cooling Research Laboratory of University of Texas in Austin performed an experiment which studied the suction surface boundary layer due to the effects of streamwise pressure gradients. However, the Development of a Turbine Blade Suction Surface Wind Tunnel has a brief history at BaylorUniversity. This project was started in the summer of 2004 by David Clubbs, a Baylor mechanical engineering student. Clubbs had developed, through his research, a preliminary design for the experiment but was not able to complete construction or run experiments with his design due to time constraints. The goal this semester was to continue the development of this project through to the construction and experimentation phase.

In recent years the jet turbine industry has strived to increase bypass ratios, higher stage loading, and achieve higher efficiencies at altitude [2]. In an effort to increase the overall efficiency of the low pressure turbine (LPT) experimentation has been preformed on the Pak B blade design which has shown that at Reynold's numbers > 200k and turbulence intensity ≥ 1% the blade performs well. However, at lower Re number flow, the Pak B blade suffers from flow separation due to adverse pressure gradients on the suction surface side of the blade [3]. Due to this result, research in low Reynolds (Re) number flows have become necessary to increase the efficiency of the LPT. To account for this separation it is important to study the location along the axial chord where the flow transitions from laminar to turbulent [2]. By developing a Turbine Blade Suction Surface Wind Tunnel Baylor University will be capable of reproducing and studying the affects of low Re number flow over a flat plate that is similar to the flow over the Pak B blade. With the wind tunnel design completed, research on separation and methods of flow reattachment can be done.

To simulate the Pak B blade pressure gradient on a flat plate it is necessary to replicate the flow velocity and acceleration patterns over the plate. By contouring a ceiling profile, similar flow characteristics can be achieved. The goal for this project is to reproduce similar coefficients of pressure (C_p) as those found in Figure 1 [1]. The solid line represents the VBI-2D code which is a theoretical model for the Pak B blade which will provide the preliminary pressure coefficients for calculating an approximate profile.

Figure 1 C_p vs. % Axial Chord [1]

Computing the approximate profile begins by determining an initial freestream velocity, V∞, based on a desired Reynolds number. This value can be calculated from the Re number equation, Equation 1.

(1)

Where l_axial is the axial chord length, ρ is the air density, and μ is the air viscosity. With V∞ the total inlet pressure, P_o, can then be calculated from Equation 2.

(2)

Where P_atm is the atmospheric pressure, and with both P_o and V∞ the initial static pressures P_s for given distances down the chord can be calculated from Equation 3.

(3)

With the static pressure distribution computed the local velocities, V_local, for each distance down the chord can be calculated from Equation 4.

(4)

Finally combining the information from these equations and the principle of continuity the height, h_local, between the tunnel roof and the profile could be computed. Equation 5 expresses how each height was calculated, where h∞ is the freestream height.

(5)

This solution provides an approximate solution for the overall profile shape that was initially used to begin the construction of the apparatus.

EXPECTED RESULTS:

The primary measurements for this experiment are of the differential pressure between the total pressure, P_o, and static pressure, P_s, and determining the freestream velocity. After computing theoretical values for the differential pressure from the Equations 1-5 from the previous section, using both standard air properties and Re number of 50,000 (Re₅₀) and 100,000 (Re₁₀₀), the expected magnitudes are displayed in Figure 2. Figure 2 presents the static pressures vs. position along the chord.

Figure 2 Calculated P_o – P_s vs. % Axial Chord (suction surface)

These values, when converted to in H₂O, are a magnitude of around 0.005 in H₂O for Re₅₀ and 0.0158 inH₂O for Re₁₀₀. These values are low due to the low air velocity and dynamic pressure influence is low therefore the static pressure is nearly equal to the P_atm. The velocity measurements for Re₅₀ are expected to be 0.789 m/s (2.59 ft/s) and for Re₁₀₀ is 1.579 m/s (5.18 ft/s). The corresponding pitot-static pressure readings are expected to be on the order of 0.0015 inH₂O. These values were also calculated from Equation 1 defined in the previous section.

REQUIRED EQUIPMENT AND SUPPORT:

The required equipment for data collection in this experiment is the MKS Instruments 223BD-000.2ABB Pressure Transducer in conjunction with an MKS Instruments type PDR-D-1 Power Supply/Digital Readout. Since this experiment deals with low Reynolds number flows the velocity through the test section will be low, this low velocity produces low static pressures across the profile chord. The MKS Instruments equipment, with serial number of 000692780 for the pressure transducer and 000532287 for the Power Supply/Digital Readout, is an excellent instrument for this experiment given that its full-scale range is 0.2 TORR, which is equivalent to a range of 0.107 in H₂O. This low range will ensure that the data collected is sufficiently accurate and is not outside of the range of the instrument.

Figure 3 SolidWorks drawing of profile support apparatus

The experimental apparatus used in this experiment required the design and construction of several pieces of equipment. The first of which was the profile support assembly. This assembly rests on two adjusting rails that are mounted on the top of the wind tunnel test section (Figure 3). With the use of these rails a series of five profile adjusters penetrate the top of the test section and attach to the profile material within the wind tunnel. Holes were cut in to the top plexi-glass of the test section of the wind tunnel to allow the profile adjusters to enter in to the wind tunnel. The profile adjusters, Figure 4, were previously fabricated. They are placed on the rails using the two holes in either side of the upper nylon block. In order to adjust the profile, the knobs are rotated which force the lower support to extend or retract.

Figure 4 Profile Adjuster

For proper support of the profile material a piano hinge was purchased and mounted to the front, top portion of the test section to allow the profile to be adjusted without producing permanent deformation in the profile material. BaylorUniversity owned a sheet of 1/8 inch thick abs plastic which was used for the profile material. Through research it was determined that this material is flexible enough to produce the desired profile contour. For proper support of the profile material to the profile adjusters, double sided tape was used between the lower portion of the profile adjuster and the profile material. The profile material was cut to a very close tolerance so that it would fit in the test section snugly. This snug fit allowed the rear portion of the material to not need a support apparatus to be developed. Instead the material is, for the most part, held up from the tension across the test section with some help at the very end of duct tape to ensure it stays up against the roof of the tunnel.

In addition to the profile adjusting assembly, a flat plate was designed and constructed for the bottom of the test section. In previous research it was determined that the plate be 38 inches long, 1 foot wide, 0.5 inches thick, and a 4:1 ratio ellipse at the leading edge. It was also suggested that it be elevated three inches above the floor of the test section in order to avoid any boundary layer conditions. The flat plate has a series of staggered pressure taps. It is necessary to stagger the taps in order to avoid any flow disturbance for the taps that will be further along the chord length. There are 18 taps in the plate. This number was decided on since it is not too time-consuming while taking data but it is still precise enough to see the pressure distribution. The placement of the taps was critical since it was predicted that the latter 60% axial cord is where separation is to occur.

SETUP

As discussed above, the profile support apparatus is placed on the roof of the test section of the wind tunnel with the flat pressure plate on the floor of the test section as seen in Figure 5.

Figure 5 Apparatus Schematic

There is plastic tubing connected to each pressure port in the pressure plate that runs out of the test section so that measurements can be taken externally with the MKS Instruments equipment. Since BaylorUniversity does not own a pressure transducer that is large enough to connect all of the pressure taps during the experiment, each tap is measured individually by connecting, measuring and disconnecting each port for each trial. A pitot static tube is mounted in front of the profile in order to determine the free stream velocity through the test section so that the desired Reynolds number can be confirmed.

CALIBRATION INFORMATION:

The MKS equipment that was used in taking data for this experiment has recently been calibrated. The MKS Instruments pressure transducer was calibrated in compliance with ISO 120012-1 using MKS transfer standard S/N: 000172289 versus a reference standard S/N: 295020 which was calibrated with a CEC Air Dead Weight Tester, traceable to the Institute of Standards and Technology. This calibration was performed February 15, 2001. There is no calibration information that BaylorUniversity possesses for the Type PDR-D-1 Power Supply/Digital Readout. Beyond the above instrumentation there will be no other calibration needed for data acquisition.

EXPERIMENTAL PROCEDURE:

Before beginning the experiment the approximate profile shape was calculated in Microsoft EXCEL. The preliminary profile calculations were performed using the published C_p values, the associated Reynolds number, and standard atmospheric air properties [1]. The physical dimensions can be seen in Table 1 below.

Table 1 Profile Dimensions

The row labeled "distance from leading edge" has highlighted cells which refer to locations the profile supports were placed. From this table it was possible to determine the approximate shape the profile needed to be in order to produce the theoretical C_p data provided in Figure 1 above. With these physical dimensions for the profile the roof assembly was constructed and adjusted according to the preliminary profile calculations. When the construction was completed on the flat test plate, all the necessary tubing to the static pressure taps and to the pitot - static tube were attached. After completing these preliminary tasks the experiment then began.

For this experiment there were two target Re numbers, 50k and 100k. The corresponding freestream air velocities are 2.4819 (ft/s) for Re₅₀ and 4.9642 (ft/s) for Re₁₀₀. Due to the low pressure differentials at these Re numbers the experiment was run using the Re of 100k. However because the measurements at Re₁₀₀ might fluctuate as well higher Re number flows might be used to quantify the performance of the design. Therefore step 1 was to power the wind tunnel and allow it to reach the target upstream air velocity in the test section. To ensure Re number similarity the upstream air velocity was verified by a pitot – static tube. The pitot-static readings were inserted into the test matrix which was designed to input the differential pressure reading and output the freestream velocity. To reduce random error the pitot-static measurements were read 5 times and then averaged. Once the appropriate Re number is verified via the air velocity measurement step 2 began.

Step 2 was reading the differential pressure (P_o – P_s) measurements between the pitot-static total pressure port and the static pressure ports of the flat plate. In order to develop the pressure profile pressure transducer had to be plugged and unplugged to each associated static pressure tube. These measurements were also recorded 5 separate times for each pressure taps to reduce the affects of random error. The C_p plot for this data was constructed by dividing the differential plate pressure readings by the average freestream dynamic pressure. Through comparing these C_p values with the published C_p data both the accuracy of the profile shape and any possible profile adjustments were determined [1]. All experimental data are compiled in the test matrices provided in Appendix C. Finally step 1 and 2 will be repeated for the Re₅₀.

RESULTS

The preliminary experiments performed with the wind tunnel were done for Re numbers higher than 100,000. This approach was taken because at the lower air velocities the readings experienced wide fluctuations and by increasing the air speed steadier readings were achieved. There were two initial experiments run with the preliminary profile shape. The Re number for trail #1 was 144,950 (2.242 m/s, 7.357 ft/s) and for trial #2 148,566 (2.298 m/s, 7.540 ft/s). The data from these trials can be seen in Figure 6 below, which shows three trial data and the published data.

Figure 6 Experimental Results

Examining the preliminary data showed that the C_p plots followed the published data closely, but there were slight variations between 20–40% axial cord and 90-100% axial cord. In an attempt to remedy these differences there were slight adjustments made to the profile shape for a third trial, trial #3. To reduce the C_p values between 20-40% axial cord the profile was lifted to increase the area and decrease the velocity of the air through that portion of the test section. Then to increase the C_p value between 90-100% axial cord the profile was lowered to decrease the area and increase the air velocity. There were only three profile supports changed, the first two and the last supports. The percent difference between the each set of trial data and the published data are represented in Table 2 below.

Table 2 % Difference between Published and Trial Data

When the last trial was performed the trial data gathered began to converge with the published data. This result confirmed that the design, with minor modifications can produce data that will match the data that is published. To further match the published data the placement of the flat plate could be relocated in order to achieve the proper C_p values at the beginning of the plate. Then to achieve better separation from the aft portion of the flat plate the freestream velocity needs to be lowered to decrease the amount of energy in the air which will promote more flow separation. Due to the lack of time and scope of this project there were no more trails performed for this experiment.

UNCERTAINTY ANALYSIS:

In determining the uncertainty for this experiment, the Kline-McClintock method was used in determining the fixed and random errors of this experiment. Preliminary uncertainty analysis was performed on the fixed error for this experiment. The parameters for the calculations were determined from the experimental data gathered in various trials. Listed in the Table 3, below, are the accuracies for the two separate measurements taken in the experiment.

Table 3 Uncertainty Parameters and Accuracy

After computing the fixed and random errors for the trials the average total uncertainty for the data in each of the trials is listed in Table 4 below.

Table 4 Average Total Uncertainty

These values were determined by computing the uncertainty for the C_p value associated with each of the eighteen pressure taps and averaging the results. Each of the values in Table 3 state that there is confidence that 95% (20:1 odds) of measurements taken for this experiment will fall within ± the average percentage from the mean value. For further review of the uncertainty analysis and calculations please refer to Appendix B. The average uncertainty for each trial was significantly low because of the accuracy of the MKS pressure transducer. The primary sources for error in the experiment are the static pressure port construction, the low pressure differentials measured, and human bias in reading the equipment. However from the uncertainty analysis it is proven that the data taken from the experiment has a very low error and high reliability, therefore the data should be repeatable.

CONCLUSIONS AND RECOMMENDATIONS

With the completion of design and construction of the Turbine Blade Suction Surface Wind Tunnel and successful reproduction of theoretical data the wind tunnel can now be used to continue the study of the Pak B profile at BaylorUniversity. Due to time constraints and the scope of this project the wind tunnel still has potential improvements that can be made to it in order to produce better data and decrease experimentation time.

In order to adjust the profile to a more precise shape perhaps it would be beneficial to study new mounting options for the profile material to the profile adjusters. The current double sided tape hold the profile material to the adjusters but inhibits large horizontal movements without removing the tape from each adjuster and reapplying it once the new position is reached. An additional improvement that can be made to the tunnel is the addition of a grid on the side of the test section that would allow the experimenter to adjust the profile to a more specific position without the use of a tape measure or other device that has potential variability when adjusting the height of the profile.

Additional changes that could be made to the wind tunnel are the development of additional pressure plates that have different numbers of pressure ports. Using a pressure plate that has more ports would allow more precise data to be taken for the profile and a more detailed study of the separation could then be performed, specifically in the aft 60% of the test section. In order to decrease experimentation time and allow for more experiments to be performed, the purchase of new pressure reading equipment that can have multiple pressure taps fixed to it would be beneficial. With shorter experimentation times more studies could be performed in order to fine-tune the shape of the profile.

With the previous mentioned recommendations and the current status of the wind tunnel, experimentation can continue at BaylorUniversity of the Suction Side of the Pratt and Whitney Pak B turbine blade. The development of this apparatus will procure knowledge of this subject at Baylor through its students and researchers alike.

References:

[1] Lake, James, P., "Flow Separation Prevention on a Turbine Blade in Cascade at Low Reynolds Number," Air Force Institute of Technology, Wright-Patterson AFB, Ohio, 1999.

[2] Van Treuren, Kenneth, W., et. al, "Measurements in a Turbine Cascade Flow Under Ultra Low Reynolds Number Conditions," Transactions of the ASME, Vol. 124, January 2002, © 2002.

[3] Ellis, S. L. et. al, "Simulating the Effects of Streamwise Pressure Gradient on the Suction Surface Boundary Layer of a Gas Turbine Airfoil," University of Texas at Austin, Austin, TX. 1992.