Steven M. Whitaker

A Survey of Airborne Particle Impact Characteristics Using High Speed Particle Shadow Velocimetry

A review of the effect of particle, surface, and impact characteristics on the coefficient of restitution for quartz particles impacting an aluminum plate has been performed using highspeed Particle Shadow Velocimetry. Over one million rebounds, obtained from the analysis of approximately 2.6 million images, were analyzed. Particle impacts occurred at speeds between 0 and 100 m/s, with impact angles ranging from 0 to 90 degrees. The roughness of the target surface was varied from Ra = 0.8 – 8.0 micrometers. Observed particle sizes ranged between 150 and 1000 micrometers, with shapes ranging from roughly spherical to oblong. A code was developed that incorporated image processing, particle identification, particle tracking, rebound identification, size and shape analysis, out-of-plane velocity estimation, and calculation of restitution coefficients in a manner sufficient as to allow the efficient analysis of substantial numbers of particle images. Results match trends observed in the literature that have examined variations in impact velocity, impact angle, particle size, and surface roughness independently. The total coefficient of restitution is shown to vary strongly with both impact velocity and angle, while restitution coefficients in the normal and tangential directions are primarily functions of impact angle alone. Particle size appears to have no effect on the restitution coefficients, while the shape of the particles adds randomness to the rebound characteristics. Increased surface roughness was observed to slightly lower restitution coefficients, likely due to plastic deformation of the larger asperities.

Evaluation of Elastic-Plastic Rebound Properties of Coal Fly Ash Particles for Use in a Universal Turbine Deposition Model

Three particle impact models have been evaluated to determine their ability to predict particle material properties and restitution coefficients using experimental data for the coefficient of restitution of particles impacting a 410 stainless steel plate. The particles consisted of PMMA and three coal fly ashes: JBPS, Bituminous, and Lignite. Particle speeds ranged from approximately 20 to 120 meters per second, and the nominal impact angle was approximately 85 degrees. Flow temperatures for the ash particulate experiments were set at 295 K and 395 K. The impact models were applied to the experimental data via curve fitting to evaluate the yield stress of the particulate, which was known for the PMMA. For the ash particulate, a linear law of mixtures was used to approximate the modulus of elasticity and Poisson’s ratio for use in the yield stress determination. A Hertzian mechanics model was shown to over-predict the yield stress of the PMMA particulate, indicating that, for known material properties, they would under-predict the coefficient of restitution. A Plastic-JKR model and a finite element based model by Wu et al. showed good agreement between the calculated yield stress and known range of yield stress values for the PMMA particulate, indicating that the model would accurately predict restitution coefficients for particulate with known material properties (or could be used to accurately determine the material properties from experimental coefficient of restitution data). However, some questions remain as to the ability of these models to be used for non-spherical, conglomerate type particulate. A thorough overview of the impact process is provided, and the application of the results of the study to the development of a physics-based universal impact and deposition model is presented.Copyright

A Comparison of Techniques for Particle Rebound Measurement in Gas Turbine Applications

The rebound characteristics of 100–500μm quartz particles from an aluminum surface were imaged using the particle shadow velocimetry (PSV) technique. Particle trajectory data were acquired over a range of impact velocity (30–90 m/s) and impact angle (20°–90°) typical for gas turbine applications. The data were then analyzed to obtain coefficients of restitution (CoR) using four different techniques: (1) individual particle rebound velocity divided by the same particle’s inbound velocity (2) individual particle rebound velocity divided by inbound velocity taken from the mean of the inbound distribution of velocities from all particles (3) rebound velocity distribution divided by inbound velocity distribution related using distribution statistics and (4) the same process as (3) with additional precision provided by the correlation coefficient between the two distributions. It was found that the mean and standard deviation of the CoR prediction showed strong dependence on the standard deviation of the inbound velocity distribution. The two methods that employed statistical algorithms to account for the distribution shape [methods (3) and (4)] actually overpredicted mean CoR by up to 6% and CoR standard deviation by up to 100% relative to method (1). The error between the methods is shown to be a strong (and linear) function of correlation coefficient, which is typically 0.2–0.6 for experimental CoR data. Non-Gaussianity of the distributions only accounts for up to 1% of the error in mean CoR, and this largely from the non-zero skewness of the inbound velocity distribution. Particle rebound data acquired using field average techniques that do not provide an estimate of correlation coefficient are most accurately evaluated using method (2). Method (3) can be used with confidence if the standard deviation of the inbound velocity distribution is less than 10% of the mean velocity, or if a linear correction based on an assumed correlation coefficient is applied.Copyright

The Effects of Slot Film Cooling on Deposition on a Nozzle Guide Vane

The effects of slot film cooling on deposition in a high pressure nozzle guide vane passage were investigated experimentally and computationally. Experiments were conducted in Ohio State’s Turbine Reaction Flow Rig, using a four-vane cascade, operating at temperatures up to 1353 K. Film cooling was achieved on one of the vanes using a span-wise slot, located at approximately 30% chord on the pressure surface. The coolant’s effect on vane surface temperature was characterized by taking infrared images at various cooling levels. Deposition was produced by injecting sub-bituminous ash particles with a median diameter of 6.48 μm upstream of the vane passage. Several deposition tests were conducted with varying coolant levels. Results exhibit a strong relationship between the coolant flow rate and the amount of ash that deposits on the cooled vane. Capture efficiency was reduced by 70% at the highest coolant flow rate (1.27% of the mass flow rate in the passage). Capture efficiency reduction was compared to that achieved using discrete hole film cooling in other studies. The slot scheme showed similar or larger reductions in capture efficiency at lower coolant mass flow rates. Deposit distribution patterns are affected by regions of cooler temperature, both downstream of the slot where film effects dominate, and slightly upstream of the slot which is cooled by conduction. A computational simulation was conducted to model both the flow and deposition. The solid vane was also discretized to allow for conjugate heat transfer calculations, which produced results that were qualitatively similar to IR measurements, but over predicted the effectiveness of the coolant. An Eulerian-Lagrangian particle tracking model was utilized to track the ash particles through the flow. A sticking model was implemented to determine whether particles stick upon impacting the vane surface, from which deposition rates and distributions are obtained. The computational model under predicted the baseline capture efficiency and the capture efficiency reduction factors for each cooling level, suggesting that the model is not sufficiently sensitive to the temperature changes between tests. Inclusion of surface temperature and local shear dependencies was suggested as an improvement to the sticking model.Copyright

The Effect of Free-Stream Turbulence on Deposition for Nozzle Guide Vanes

An evaluation of the effect of free-stream turbulence intensity on the rate of deposit accumulation for nozzle guide vanes was performed using the TuRFR accelerated deposition facility. The TuRFR allowed flows up to 1350 K at inlet Mach numbers of 0.1 to be seeded with coal fly ash particulate in order to rapidly evaluate deposit formation on CFM56 nozzle guide vanes. Hot film and PIV measurements were taken to assess the free-stream turbulence with and without the presence of a grid upstream of the NGVs. It was determined that baseline turbulence levels were approximately half that of the flow exiting typical gas turbine combustors and were reduced by approximately 30 percent with the grid installed. Deposition tests indicated that the rate of deposition increases as the free-stream turbulence is increased, and that this increase depends upon the particle size distribution. For ash with a mass median diameter of 4.63 μm, the increase in capture efficiency was approximately a factor of 1.7, while for ash with a larger median diameter of 6.48 μm, the capture efficiency increased by a factor of 2.4. The increase in capture efficiency is due to the increased diffusion of particles to the vane surface via turbulent diffusion. Based on these results, smaller particles appear to be less susceptible to this mechanism of particle delivery. Overall, the experiments indicate that the reduction of turbulence intensity upstream of nozzle guide vanes may lead to reduced deposit accumulation, and, consequently, increased service life. A CFD analysis was performed at turbulence levels equivalent to the experiments to assess the ability of built-in particle tracking models to capture the physics of turbulent diffusion. Impact efficiencies were shown to increase from 21 percent to 73 percent as the free-stream turbulence was increased from 5.8 to 8.4 percent. An analysis incorporating the mass of the particles into the impact efficiency resulted in an increase of the mass-based impact efficiency from 17 percent to 27 percent with increasing turbulence. Relating these impact efficiencies directly to capture efficiencies, the predicted increase in capture efficiency with higher turbulence is much less than that observed in the experiments. In addition, the variation in impact efficiencies between the two ash sizes was much smaller than the capture efficiency difference from experiments. This indicates that the particle tracking models are not capturing all of the relevant physics associated with turbulent diffusion of airborne particles.Copyright

Evaluation of Coefficients of Restitution Using High Speed Particle Shadow Velocimetry with Application to Particle Separators for Gas Turbine Engines

The effects of several parameters on the absolute, normal, tangential, and angular coefficients of restitution were evaluated experimentally for several particle and surface combinations. Glass beads and crushed quartz particulate were accelerated towards a target plate by means of a compressed air facility at velocities ranging from 15-80 meters per second. Several plate types were tested, including aluminum with surface roughness (characterized by the Ra value) of 0.25 and 2.8 micrometers, copper with a surface roughness of 0.25 micrometers, and an alumina ceramic with a surface roughness of 1 micrometer. The angle of the plate relative to the particle flow was varied from 10-80 degrees. Individual particle impacts were recorded using a high speed particle shadow velocimetry configuration. The resulting images were analyzed using a comprehensive code, yielding velocity, size, and shape information for each of the more than 1.8 million identified rebounds. In general, glass beads were shown to have higher coefficients of restitution than the quartz particles under identical conditions due to the sphericity of the glass beads relative to the quartz resulting in generally larger contact areas with the plate surface, leading to reduced plastic deformation. The random nature of the quartz particle shapes led to increased randomization of the rebound characteristics. Surface roughness was shown to lower the coefficient of restitution slightly and reduce the standard deviation in the CoR values for glass beads. With all other quantities being equal, a change in the plate material properties, specifically a reduction of the yield strength, resulted in a nearly constant reduction of the normal coefficient of restitution. Impacts on a ceramic plate revealed a tendency for the quartz particles to fracture and break apart at high impact velocities and angles. This effect resulted in a transfer of energy from the normal to the tangential direction, and increased the standard deviation of the coefficients of restitution. Further study into the effect of particle rotation and impact plate surface roughness is warranted.