Knox T. Millsaps

Effects of Reynolds Number and Surface Roughness Magnitude and Location on Compressor Cascade Performance

An experimental investigation has been conducted to characterize the influence of Reynolds number and surface roughness magnitude and location on compressor cascade performance. Flow field surveys have been conducted in a low-speed, linear compressor cascade. Pressure, velocity, and loss have been measured via a five-hole probe, pitot probe, and pressure taps on the blades. Four different roughness magnitudes, Ra values of 0.38 lm (polished), 1.70 lm (baseline), 2.03 lm (rough 1), and 2.89 lm (rough 2), have been tested. Furthermore, various roughness locations have been examined. In addition to the as manufactured (baseline) and entirely rough blade cases, blades with roughness covering the leading edge, pressure side, and 5%, 20%, 35%, 50%, and 100% of suction side from the leading edge have been studied. All of the tests have been carried out for Reynolds numbers ranging from 300,000 to 640,000. For Reynolds numbers under 500,000, the tested roughnesses do not significantly degrade compressor blade loading or loss. However, loss and blade loading become sensitive to roughness at Reynolds numbers above 550,000. Cascade performance is more sensitive to roughness on the suction side than pressure side. Furthermore, roughness on the aft 2=3 of suction side surface has a greater influence on loss. For a given roughness location, there exists a Reynolds number at which loss begins to significantly increase. Finally, increasing the roughness area on the suction surface from the leading edge reduces the Reynolds number at which the loss begins to increase. [DOI: 10.1115/1.4003821]

Dynamic Forces From Single Gland Labyrinth Seals: Part II—Upstream Coupling

The standard lumped parameter model for flow in an eccentrically offset labyrinth seal, which assumes constant upstream and downstream boundary conditions, has been extended to include the effects ofa nonuniform upstream cavity flow due to coupling. This new model predicts that the upstream perturbations in pressure and azimuthal velocity caused by this coupling can have a very strong impact on the pressure distribution in the seal gland itself. Augmentation by a factor of four, over the uniform inlet model, is predicted under some circumstances. Although no precise comparison to the experimental data with this new model was possible, due to the lack of control over the face seal venting the upstream cavity to the center hub plenum, the calculated effect of this coupling was shown to be approximately what was required to restore quantitative agreement between the data and theory

Effect of Surface Roughness Location and Reynolds Number on Compressor Cascade Performance

An experimental work has been conducted in a linear compressor cascade to find out the effect of surface roughness and Reynolds number. Surveys were conducted with different roughness size and Reynolds number. The k s /c value of each roughness is 0.0006, 0.0090, 0.00150, 0.00213, and 0.00425. The range of Reynolds number is 300,000~600,000 and conducted with roughened blade, which roughness Ra is 2.89 microns. Flow pressure, velocity, and angle have been found out via 5 hole probe. Pressure loss and deviation increased with increasing roughness. In the low Reynolds number under 500,000, tested roughness does not affect to the performance of compressor cascade. However, roughness is very sensitive to pressure loss in high Reynolds number over 550,000.

Impact of Non-Uniform Leading Edge Coatings on the Aerodynamic Performance of Compressor Airfoils

Abstract : A computational fluid dynamic (CFD) investigation is presented that provides predictions of the aerodynamic impact of uniform and non-uniform coatings applied to the leading edge of a compressor airfoil in a cascade. Using a NACA 65(12)10 airfoil, coating profiles of varying leading edge non-uniformity were added. This non-uniformity is typical of that expected due to fluid being drawn away from the leading edge during the coating process. The CFD code, RVCQ3D, is a steady, quasi-three-dimensional Reynolds Averaged Navier-Stokes (RANS) solver. A k-omega turbulence model was used for the Reynolds Stress closure. The code predicted that these changes in leading edge shape can lead to alternating pressure gradients in the first few percent of chord that create small separation bubbles and possibly early transition to turbulence. The change in total pressure loss and trailing edge deviation are presented as a function of the coating non-uniformity parameter. Results are presented for six leading edge profiles over a range of incidences and inlet Mach numbers from 0.6 to 0.8. Reynolds number was 600,000 and free-stream turbulence was 6%. A two-dimensional map is provided that shows the allowable degree of coating non-uniformity as a function of incidence and inlet Mach number.

ORIGIN OF SPLIT RESONANCE AND BACKWARD WHIRL IN A SIMPLE ROTOR

Split resonance and backward whirl in a simple rotor, were investigated both analytically and experimentally. A two degree-of-freedom rotor model was developed to simulate the steady state, lateral vibration characteristics of a simply supported, single disk rotor. This model included the effects of direct and cross coupled, linear damping and stiffness. The computer model was used to quantify the influence of bearing characteristics on rotordynamic response. In the absence of gyroscopic effects, split resonance is due to separate and distinct natural frequencies in the two orthogonal lateral directions created by non-equal direct stiffnesses. Backward whirl can occur between these two frequencies if the direct damping is sufficiently low. The model was able to predict the observed response of a simply supported rotor, including split resonance and backward whirl. The cause of the asymmetric direct stiffnesses in the experimental apparatus, which created split resonance and backward whirl was investigated. In particular, the influence of geometric imperfections in the plain bearing sleeve, gravitational forces, degree of imbalance and bearing support stiffness asymmetries were isolated using the experimental apparatus. It was determined that the bearing asymmetry was caused by the gravitational influence. However, larger imbalances increased the asymmetry and large damping was able to suppress backward whirl.© 1996 ASME

Reducing Lateral Vibrations of a Rotor Passing Through Critical Speeds by Acceleration Scheduling

A method is presented for reducing the lateral response of an imbalanced rotor accelerating or decelerating through its first lateral bending critical speed by using a variable acceleration rate. A lumped parameter model along with a numerical integration scheme is used to simulate the response of a simply supported, single disk rotor during fast acceleration and deceleration through critical speed. The results indicate that the maximum response and/or the total vibrational energy of a rotor passing through the critical speed can be reduced significantly by using a variable acceleration schedule. That is, reducing the acceleration rate after the nominal critical speed is passed. These predictions were verified experimentally for a single disk rotor.Copyright

Effect of Leading Edge Roughness and Reynolds Number on Compressor Profile Loss

An experimental investigation has been conducted to characterize the influence of leading edge roughness and Reynolds number on compressor cascade profile loss. Tests have been conducted in a low-speed linear compressor cascade at Reynolds numbers between 210,000 and 640,000. Blade loading and loss have been measured with pressure taps and pneumatic probes. In addition, a two-component laser-doppler velocimeter (LDV) has been used to measure the boundary layer velocity profiles and turbulence levels at various chordwise locations near the blade suction surface. The “smooth” blade has a centerline-averaged roughness (Ra) of 0.62 μm. The “rough” blade is roughened by covering the leading edge of the “smooth” blade, including 2% of the pressure side and 2% of the suction side, with a 100 μm-thick tape with a roughness Ra of 4.97 μm. At Reynolds numbers ranging from 210,000 to 380,000, the leading edge roughness decreases loss slightly. At Reynolds number of 210,000, the leading edge roughness reduces the size of the suction side laminar separation bubble and turbulence level in the turbulent boundary layer after reattachment. Thus, the leading edge roughness reduces displacement and momentum thicknesses as well as profile loss at Reynolds number of 210,000. However, the same leading edge roughness increases loss significantly for Re = 450,000 ∼ 640,000. At Reynolds number of 640,000, the leading edge roughness decreases the magnitude of the favorable pressure gradient for axial chordwise locations less than 0.41 and induces turbulent separation for axial chordwise locations greater than 0.63, drastically increasing loss. Thus, roughness limited to the leading edge still has a profound effect on the compressor flow field.Copyright

Dynamic Forces From Single Gland Labyrinth Seals: Part II — Upstream Coupling

The standard lumped parameter model for flow in an eccentrically offset labyrinth seal, which assumes constant upstream and downstream boundary conditions, has been extended to include the effects of a non-uniform upstream cavity flow due to coupling. This new model predicts that the upstream perturbations in pressure and azimuthal velocity caused by this coupling, can have a very strong impact on the pressure distribution in the seal gland itself. Augmentation by a factor of four, over the uniform inlet model, is predicted under some circumstances. Although no precise comparison to the experimental data with this new model was possible, due to the lack of control over the face seal venting the upstream cavity to the center hub plenum, the calculated effect of this coupling was shown to be approximately what was required to restore quantitative agreement between the data and theory. The new theory can explain the anomalously large pressure non-uniformity previously found by other authors in short seals as well as the first few glands of multi-cavity seals.Copyright

MEMS-Scale Turbomachinery Based Vacuum Roughing Pump

This study forms part of a program to develop a micro-electro-mechanical-systems (MEMS) scale turbomachinery based vacuum pump and investigates the roughing portion of such a system. Such a machine would have many radial stages with the exhaust stages operating near atmospheric conditions while the inlet stages operate at near vacuum conditions. In low vacuum such as those to the inlet of a roughing pump the flow can still be treated as a continuum however the no-slip boundary condition is not accurate. The Knudsen number becomes a dominant non-dimensional parameter in these machines due to their small size and low pressures. As the Knudsen number increases slip flow becomes present at the walls. The study begins with a basic overview on implementing the slip wall boundary condition in a commercial code by specifying the wall shear stress based on the mean-free-path of the gas molecules. This is validated against an available micro-Poiseuille classical solution at Knudsen numbers between 0.001–0.1 with reasonable agreement found.The method of specifying the wall-shear stress is then applied to a generic MEMS scale roughing pump stage that consists of two stators and a rotor operating at a nominal absolute pressure of 500 Pa. The zero flow case was simulated in all cases as the pump down time for these machines is small due to the small volume being evacuated. Initial transient two-dimensional simulations are used to evaluate three boundary conditions, classical no-slip, specified-shear and slip-flow. It is found that the stage pressure rise increased as the flow began to slip at the walls. In addition it was found that at lower pressures the pure slip boundary condition resulted in very similar predictions to the specified shear simulations. As the specified-shear simulations are computationally expensive it is reasonable to use slip-flow boundary conditions. This approach was used to perform three-dimensional simulations of the stage. Again the stage pressure increased when slip-flow was present compared with the classical no-slip boundaries. A characteristic of MEMS scale turbomachinery are the large relative tip gaps requiring three-dimensional simulations. A tip gap sensitivity study was performed and it was found that when no-slip boundaries were present the pressure ratio increased significantly with decreasing tip gap. When slip-flow boundaries were present this relationship was far weaker.© 2013 ASME

Analysis of Stall Precursors and Determination of Optimum Signal Processing Methods for an LM-2500 Gas Turbine

This paper presents an analysis of data taken from several stall initiation events on a GE LM-2500 gas turbine engine. Specifically, the time series of three separate pressure signals located at compressor stages 3, 6, and 15 were analyzed utilizing various signal processing methods to determine the most reliable indicator of incipient stall for this engine. The spectral analyses performed showed that rotating precursor waves traveling around the annulus at approximately half of the rotor speed were the best indicators. Non-linear chaotic time series analyses were also used to predict stall, but it was not as reliable an indicator. Several algorithms were used and it was determined that stall wave perturbations can be reliably identified about 900 revolutions prior to the stall. This work indicates that a single pressure signal located at stage 3 on an LM-2500 gas turbine is sufficient to provide advance warning of more than 2 seconds prior to the fully developed stall event.