Freestream turbulence measurements in icing conditionsicing.ae.illinois.edu/papers/98/AIAA-1998-96.pdfan NACA 0012 found a maximum heat transfer increase in the IRT of 10% over heat

Copyright© 1997, American Institute of Aeronautics and Astronautics, Inc.

Freestream Turbulence Measurements in Icing Conditions

C.M. Henze+, M.B. Bragg*, H.S. Kirn*University of Illinois at Urbana-Champaign

Abstract

Current understanding of the ice accretion process isbased largely on icing wind tunnel tests. Turbulentfluctuations in the freestream, which are different in thewind tunnel than in flight, have been identified as havingpotentially important effects on the results of testsperformed in icing tunnels. The turbulence intensity levelin icing tunnels with the spray cloud turned off had beenpreviously measured and found to be quite high due to thelack of turbulence reducing screens, and to the presence ofthe spray system. However, the turbulence intensity levelin the presence of the spray cloud had not been measured.In this study, a method for making such measurements wasdeveloped and used to study the effects of the spray cloudon the turbulence level in the NASA Lewis Icing ResearchTunnel (ERT). Turbulent velocity fluctuations weremeasured using hot-wire sensors. Droplets striking thewire resulted in distinct spikes in the hot-wire voltagewhich were removed using a digital acceleration thresholdfilter. The remaining data were used to calculate theturbulence intensity. Using this method, the turbulenceintensity level in the IRT was found to be highlydependent on nozzle air pressure, while other factors suchas nozzle water pressure, droplet size, and cloud liquidwater content had little effect.

Introduction

Wind tunnel testing has played and will continue toplay an important role in our attempt to understand thephysical processes behind ice accretion and its effects onaircraft performance. While wind tunnel testing is aninvaluable tool, there will always be important differencesbetween the wind tunnel environment and that which anaircraft sees in flight. For this reason, caution shouldalways be taken in applying test results, and attemptsshould be made to identify these differences and assess

their influence on test results whenever possible. Oneproblem of particular importance, which has long beenrecognized, is the turbulent fluctuations in wind tunnelflows which are often considerably larger than those in theatmosphere. In 1920 the British Aeronautical ResearchCommittee acknowledged the importance of tunnelturbulence when they proposed that, in an attempt tostandardize wind tunnel testing, a series of identical testsbe performed in wind tunnel facilities throughout theworld.1 The influence of freestream turbulence is ofparticular importance in the study of aircraft icing sinceturbulence levels in flight are generally quite low while theturbulence intensity in tunnels used for icing research isinherently high. This is due to the lack of anti-turbulencescreens, and to the turbulence generated by the sprayapparatus. Gonsalez2 has measured turbulence levels inthe NASA Lewis Icing Research Tunnel (IRT) rangingfrom approximately 0.5% at 50 mph to 1.5% at 325 mph.With the nozzle spray air (no water) operating, he saweven higher turbulence levels which varied from around0.8% at moderate speeds up to approximately 2% at aspeed of 350 mph. The turbulence measurements taken byGonsalez are shown as a function of velocity in Fig. 1. Insimilar measurements, Poinsatte3 found turbulence levelsin the IRT ranging from 0.5-0.7% over a range of 70 to210 mph without the spray nozzle air operating. He alsomeasured higher levels with the nozzle air (no water) on,but these values weren't reported due to concerns abouttemperature fluctuations which will be discussed later. Incontrast, taking hot-wire measurements in flight, Poinsattemeasured turbulence intensity levels of less than 0.1% inclear air.

Importance of Freestream Turbulence in Ice AccretionPhysics

The role of freestream turbulence in the ice accretionprocess is not well understood. While there are severalpossible ways in which increased velocity fluctuations

+ Graduate Research Assistant, Dept. of Aeronautical and Astronautical Engineering.Currently Associate Engineer, Hamilton Standard, Windsor Locks, Connecticut, Member AIAA.

* Professor, Dept. of Aeronautical and Astronautical Engineering, Associate Fellow AIAA.* Graduate Research Associate, Dept. of Aeronautical and Astronautical Engineering, Member AIAA.

Copyright © 1998 by C.M. Henze, M.B. Bragg and H.S. Kirn. Published by the American Institute of Aeronautics andAstronautics, Inc. with permission.


could affect the accretion of ice, it seems likely thatenhancement of heat transfer in the region of ice growthand over the ice shape itself would play the most importantrole.

It is well accepted that increases in turbulenceintensity can affect the structure of the boundary layer, inparticular by moving the location of boundary layertransition toward the leading edge.4'5 Changes in theboundary layer structure are likely to affect ice accretionsince heat transfer is increased in the presence oftransitional and turbulent boundary layers. Kuethe andChow4 state that the turbulence level in the atmosphere isessentially zero as far as its effect on boundary layertransition is concerned. Green6 calculated that aturbulence level of just 0.2% results in a 1% increase inskin friction coefficient, and a decrease of 0.005 inboundary layer shape factor. Bragg et al.7 found that whenthey increased the turbulence intensity level to 0.95% tosimulate the high levels seen in icing tunnels, no trulylaminar flow was observed on the model. The flow wasalready transitional in the leading edge region. In studyingthe effects of freestream turbulence on heat transfer from aflat plate with a pressure gradient, Junkhan and Serovy8

attributed the increase in heat transfer which they observedwith increased turbulence levels of 1.8 to 5% to the factthat the boundary layer was no longer laminar but hadbecome transitional.

In studies connecting the change in boundary layerstructure to increases in heat transfer, it has been noted matwhen no pressure gradient was present, increasedfreestream turbulence had a negligible effect on heattransfer as long as the boundary layer remained laminar.8'9However, for a fully turbulent boundary layer, Blairobserved an increase in heat transfer on a zero pressuregradient flat plate of up to 20% when turbulence intensitywas increased from 0.25 to 7%. This suggests thatincreased freestream turbulence intensity can increase heattransfer through turbulent, and possibly transitional,boundary layers. With turbulence intensity increased toapproximately 2.4% Van Fossen and Simoneau10 found a30% increase in heat transfer from a cylinder. In an earlier.study, Van Fossen et al.11 found similar increases in heattransfer on models of ice shapes based on actual iceaccretions measured in the IRT.

Of particular interest to the present study are theworks of Gelder and Lewis12 and Poinsattte3 in whichcomparisons of heat transfer on an airfoil in flight and inthe IRT have been made. Gelder and Lewis found anincrease in heat transfer of as much as 30% in the IRT.Poinsatte's more recent investigation of heat transfer froman NACA 0012 found a maximum heat transfer increasein the IRT of 10% over heat transfer in flight. He observeda 2-3% increase over the flight condition near the leadingedge of the model at 0% angle of attack and Re = 1.2x10 .With the nozzle air on, a 3-5% increase was seen. Furtherback on the airfoil, the increase in heat transfer was even

larger, and larger increases in heat transfer were also seenwith increasing Reynold's number. These studies allindicate mat increased turbulence has a strong effect onheat transfer, and it is logical that such increased heattransfer would affect ice accretion. It is thus important tocharacterize the turbulence level in flight icing conditionsand icing wind tunnels so that we can begin to assess theeffects of turbulence intensity on icing tunnel results.

Hot-wire Anemometry in Two-Phase FlowsWhile hot-wire anemometry is the most popular

method of measuring turbulent fluctuations, the icing windtunnel environment presents a particular problem in theuse of this method. The presence of the water droplets hasa significant effect on the hot-wire signal. In order tosuccessfully measure turbulence in these conditions, amethod must be developed for separating the dropletstrikes from the turbulent fluctuations in the free stream.Numerous researchers have addressed this problem, andmore generally the problem of phase discrimination in anytwo-phase flow.

The use of hot films and hot wires for measuringturbulence in rain storms was addressed by Merceret.13'14

By dropping rain drop size water droplets, 0.3 - 0.5 cm, onhot wires and hot films in a plexiglass tube with no flowpresent, he was able to examine the signal returned when adroplet hit the anemometer sensor. He noted a high spikein the signal due to the increased heat transfer to the waterdroplet, and referenced work by other researchers where asimilar spike had been observed in work with aerosoldroplets. Similar data were taken in rain storms with windsless than 20 meters per second from a tower on top of thelaboratory.13'14 In these conditions, Merceret observed aclearly distinguishable droplet strike signal from a wedgeshaped hot film, but found that hot wires were not useablein these conditions because the signal from large and smalldroplets could not be clearly differentiated from theturbulence signal. Goldschmidt, and Householder15

however, found hot-wire sensors suitable for use asparticle concentration and size samplers as well asturbulence sensors.

Hetsroni, Cutler, and Sokolov16 experimented withthe use of hot wires in particle contaminated flow anddescribed the method by which their data were analyzed.A threshold voltage was set such that any voltage above itwas considered to represent a droplet strike, and "clipperand pickoff' circuits were employed. The "pickoff' circuitpassed only signals above the voltage threshold whichwere then fed to instruments which counted the number ofdroplets impinging during a set period of time. The"clipper" circuit passed all voltages below the threshold toDC and true rms voltmeters for measuring time-averagedvelocity and turbulence intensity. Because of theremaining "stumps" of the droplet strikes, corrections hadto be made to these measurements.


Hetsroni, and Sokolov17 made some importantobservations when they applied this method to two-phaseturbulent jet flow. Droplets 13 microns in diameter wereinjected into the flow prior to exiting a 25 mm diameternozzle. The exit velocity was 50 m/sec (~112 mph). Thisstudy is of particular interest because this drop size andvelocity are within the capabilities of the IRT.Measurements of time-averaged and fluctuating velocitiesalong with droplet flux were taken at several locationsdownstream of the nozzle. They noted that the dropletswere almost immediately swept off the wire and the signalreturned quickly to the initial neat transfer level. They alsoobserved that some small droplets caused signals whichwere only slightly greater than the turbulent signals andthus fell under the threshold voltage. However, theyassumed that in general the signals due to droplet strikeswere much higher than those due to turbulent fluctuations.

Much of the more recent work concerning the use ofthermal anemometry in two-phase flows has involved"bubbly flows" of gas bubbles in a liquid flow. While thisis a considerably different application, many of the phasediscrimination techniques used are applicable in air flowscontaining water droplets. The relatively large size andslow speed of the bubbles in these flows allows for carefulanalysis of the low heat transfer region indicative of abubble passing the sensor. Such analyses, in which thedifferent regions of the bubble and even the approach ofthe bubble are detected, is outlined in detail in Bruun18, Liuand Bankoff19 and Farrar et al.20 The most important thingto be noted from this work is that these analyses applythreshold criteria not only to the voltage signal or theresultant velocity, but also to the slope of the signal or theacceleration. In an observation similar to that made byHetsroni and Sokolov,17 Farrar et al.20 noted that, "somepartial bubble hits or small bubbles," were not detected bya simple voltage threshold. However, they noted that theslope of the voltage at the passing of the bubble wasgenerally an order of magnitude higher than the slopesfound in the freestream data. When a threshold filter wasapplied to the slope, rather than the voltage, the detectionof most of these difficult to detect impingements waspossible.

A slope threshold method was applied to air flowcontaining liquid droplets by Ritsch and Davidson.21 Inthis work, hot-film measurements were taken in 0.6 m/sduct flow containing 2 urn atomized oleic acid particles.The average slope between data points was calculated, andwhen the slope between any two points exceeded 6 timesthe average, those points were considered to be in theliquid phase. When the time between particleimpingements was large compared to the duration of theevent itself, the signals within the spike were replaced bythe last value prior to exceeding the slope threshold.However, it was noted that it was better to remove theparticle spikes and reduce the number of data points incases where the spikes compose a larger portion of the data

set. It was also noted that a high data rate, 20 kHz wasused by Ritsch and Davidson,21 was necessary to detect therapid phase changes.

These studies suggest that thermal anemometry canbe used as a practical means of measuring turbulentfluctuations in atmospheric and icing tunnel cloudconditions provided an effective method is used todiscriminate the water droplet strikes from the freestreamturbulent fluctuations. It is apparent that applying only anamplitude threshold to the digitized signal would likely notbe sufficient in detecting small droplets or partial dropletstrikes. Several researchers have found that applying athreshold filter to the slope of the signal is more effectivein detecting these smaller amplitude signals.

Experimental Objectives and ApproachThe previous review of literature has shown that

increased turbulent fluctuations affect the boundary layerdevelopment causing early transition. These changes inthe structure of the boundary layer not only affect generalperformance measurements such as lift and drag, but havebeen shown to cause increased heat transfer. Thisincreased heat transfer is likely to affect ice accretion. Forthese reasons it is important to characterize the turbulencelevel in icing tunnels. However, the measurement ofturbulence intensity using hot-wire anemometry in icingtunnels is complicated by two factors. The first of these isthe droplet strike fluctuations in the anemometer signalwhich will have to be removed from the data if accurateturbulence measurements are to be made. Second is theconcern that the heated nozzle air may cause temperaturefluctuations which will be misinterpreted by the hot wireas velocity fluctuations.

Thus, there are three specific objectives of theresearch reported here:• Develop an effective method for measuring turbulence

intensity hi icing tunnel cloud conditions.• Use the method to measure the turbulence intensity

level in the spray cloud of the NASA Lewis IcingResearch Tunnel at various cloud conditions.

• Determine the effects of temperature fluctuations dueto heated nozzle air on these turbulence intensitymeasurements.

These objectives were investigated experimentally in threewind tunnel tests. It was initially hoped that the hot-wiresensor could be shielded from the majority of the dropletshi the flow by placing it between the boundary layer andthe droplet trajectories above an airfoil at angle of attack asillustrated in Fig. 2. The first of the three wind tunnel testswas performed at the University of Illinois, and wasdesigned to assess the effects of the presence of the modelon the turbulence intensity readings.

Two tests were performed in the NASA Lewis IcingResearch Tunnel. In the first test, the probe was shieldedby an airfoil as previously described. The primary goal ofthis test was to acquire data in support of developing the


turbulence measurement technique for use in cloudconditions. In the second IRT test, the possibletemperature fluctuations were addressed by acquiring databoth with and without the nozzle air heated. No modelwas used in this test in order to evaluate the need forshielding the sensor.

Experimental Apparatus and Methods

A brief description of the experimental apparatus andmethods employed in the tests will be given here. For amore thorough description of the experimental facilitiesand apparatus as well as the experimental method, seeHenze.

Data Acquisition and Hot-wire AnemometryThe hot-wire anemometry system used was a TSI

Incorporated IFA100. The hot-wire probes chosen wereTSI model 1210 general purpose probes. The 1.27 mmlong wires on these probes were platinum coated tungstenwith diameters of 3.8 or 5.1 microns. The hot-wire sensorswere calibrated in a small wind tunnel at the University ofIllinois. A Pentium PC with a National Instruments AT-MIO-16x analog to digital data-acquisition board was usedto acquire data from the anemometer. Through the use ofsignal conditioners, both DC and high-pass filtered signalswere recorded.

Data Reduction MethodsPrior to being converted to velocities, the hot-wire

voltages were corrected for the difference between theambient temperature at the time of data acquisition, andthe ambient temperature at which the wire was calibrated.The corrected voltage, E^, as given in the anemometerinstruction manual23 was:

_ |(TW-TC,I)— " ——————————— "

tc(1)

where E is the anemometer output voltage, Tw is the hot-wire operating temperature, and Tî and T^* are theambient temperatures at calibration and data acquisition,respectively. Velocities were then calculated from thesetemperature corrected voltages by means of a calibrationpolynomial which was derived for each wire from thecalibration data.

In a similar manner to the temperature correction, thetemperature-corrected velocities were corrected fordifferences in the air density at calibration and acquisition.The temperature and density corrected velocity, Utdc, isgiven by eq. (2).24

U,dc = UtcPcalTamb 1

(2)V Pan* Teal)

Utc is the temperature corrected velocity found from thetemperature corrected voltage. Pî and Pamb are thecalibration and acquisition ambient pressures.

After all of the temperature and density correctedvalues were found, mean values of the original andtemperature corrected voltages, and the temperature anddensity corrected velocities were calculated. The standarddeviation in the temperature corrected voltage, and thefinal velocity were calculated using the followingequation:

(\ fTCs= - x'2dt""* ^T J0(3)

In eq. (3), x' is the fluctuation of any quantity from itsmean, and T is the total time. The turbulence intensity,defined as the standard deviation of the velocity, a'™,,,normalized by the mean velocity, in this case Utest, andexpressed as a percentage was then calculated using thefollowing equation:

TI = -U,,

•x lOO (4)

It is important to note that for data which had dropletsremoved, the integration in (3) was only performedbetween points which were originally adjacent to eachother in the time trace, and not between points separatedby a droplet strike. The following equation summarizesthe turbulence intensity calculation for data which haddroplet strikes removed.

TIfi,,=1 n:-l

'filt(5)

In this equation, TJU, is the total time remaining afterdroplets have been removed, m is the number of segmentsof data which don't contain a droplet, n; is the number ofdata points hi the i* such segment, uj is the deviation ofthe j* voltage from the mean, and At is the time betweendata points.

Model Shielding Effects TestThe test to determine the effect of the presence of the

model on turbulence measurements was performed in theUniversity of Illinois laboratory. This test was performedin a 3x4 foot low-speed, low-turbulence tunnel. Themodel used was a 21-inch chord NACA 0012. In order togenerate increased freestream turbulence levels ofapproximately 0.5% and 1.0% similar to those present inthe IRT, turbulence generating grids were installed in thetest section 30 inches upstream of the model leading edge.These are the same screens used by Lee25 and Bragg et al.7The actual hot-wire mounting apparatus used in the IRT,which will be describe in more detail later, could not be


mounted to the fiberglass model. Therefore, a traversesystem was used to position the probe at locationsavailable with the IRT mounting apparatus. Thiscorresponds to approximately 0.5 to 6 inches above themodel surface at the 85% chord location. For a morecomplete description of the test section setup and traversesystem, see Kerho.26

In order to keep the size of the test matrix reasonable,nearly all of the data were taken at a velocity of 100 mph,since this was the velocity at which most of the icingtunnel data were acquired. The majority of the dataacquired for this test were taken at 10 kHz for 3 seconds.Six, three second sets were acquired at each condition.The low-pass filter on the anemometer was set at 5 kHzwith a 3 Hz high pass filter used to remove the DCcomponent and the influence of low-frequency velocityoscillations. Tests were also performed with a boundarylayer trip near the leading edge of the model. The tripconsisted of 0.012 inch beads sparsely scattered on a 0.25inch wide piece of double-sided adhesive tape. Thetrailing edge of the trip was placed at the 5% chordlocation. Table 1 gives the conditions tested at each probelocation and angle of attack.

Conditions 3, 4, and 5 in Table 1 were tested only at8° angle of attack since this was where most of the IRTdata were taken. For the same reason, additionalmeasurements at varying velocities and filter settings werealso taken at 8° angle of attack with the probe at 1.5 inchesfrom the model.

Icing Research Tunnel TestsAs mentioned earlier, testing in the IRT was

performed on two separate occasions. The first testinvolved shielding the hot-wire sensor by placing it abovean airfoil at angle of attack. Figure 3 is a schematicdrawing of the probe support used to position the sensor.The distance of the sensor from the model surface wasadjustable from 0.5 to 6 inches in 0.5 inch increments byremoving and replacing different size spacers above andbelow the probe. This entire fixture was bolted to thelower surface of a 21" chord, NACA 0012 model asshown.

With the tunnel velocity set at 100 and 250 mph, themodel at 0° angle of attack, and the probe 6 inches fromthe surface to place the probe in the freestream flow,measurements were taken with nozzle air pressures (nowater) varying from 0 to 80 psig. The probe was thenmoved to 1.5 niches from the surface. At a velocity of 100mph, data were acquired at angles of attack of 4, 6, and 8°to assess the effects of model angle of attack. A set of datawith varying nozzle air pressure but no water present wasalso acquired with the probe 1.5 inches from the model at8° angle of attack and 100 mph. Air pressures of 0, 10, 30,and 50 psig were used.

For the measurements taken hi the presence of thespray cloud, a data acquisition rate of 100 kHz was chosen

in order to make sure that high-frequency droplet strikescould be resolved. Unless otherwise noted, the model wasat 8° angle of attack, and the probe was 1.5 inches from themodel surface. A small set of data was taken attemperatures below freezing to assess the usefulness ofthis technique in flight icing conditions.

In order to investigate the effects of MVD and LWCon the turbulence intensity level, data were acquired at theconditions summarized in Table 2. As noted, data werealso acquired with only the nozzle air operating, and nowater present at each condition. In a similar manner, thedependence of turbulence intensity on nozzle air and waterpressure was investigated as summarized in Table 3. Theeffectiveness of the model shielding was checked at the100 mph, MVD = 30 urn, LWC = 1.5 case by decreasingthe angle of attack to 6,4, and 0°.

A speed of 100 mph, droplet size of 20 microns, andLWC of 0.7 g/m3 were chosen for testing the operation ofthe hot wire while ice was forming. Total temperatures of25, 20, 15, and 10° F were tested. Several data sets weretaken at each condition, until the ice accretion preventedthe anemometer from functioning properly.

In the second test, no model was used, but two wireswere operated simultaneously in an attempt to measuretemperature fluctuations. This attempt was unsuccessful,therefore those results are not presented here. For detailsof this method, see Henze.22 As this was a one day test, alldata were taken at 100 mph. To facilitate the investigationof the possible temperature fluctuations data were acquiredwith no water present at nozzle air pressures varying from0 to 80 psig. With the goals of supplementing the data setfrom the previous test, and assessing the need for shieldingthe sensor with a model in mind, a set of data was acquiredin the water cloud without heating the nozzle air.Relatively low liquid water content levels were chosen,since that is where the droplet filtering methods hadworked best previously. The cloud conditions chosen aresummarized in Table 4.

Uncertainty AnalysisAn analysis of the uncertainty in the measured

quantities, the details of which are given in Henze,22 wasperformed based on the general uncertainty analysispresented by Kline and Mclintock27 and Coleman andSteele.28 The uncertainty due to bias errors in thevelocities measured in the IRT was found to beapproximately 4 ft/s at conditions typical of thoseencountered during testing. The fact that the meanvelocities measured by the hot-wire sensors were lowerthan those measured by the pitot probes in the IRT raisedsome concern about the accuracy of the resultingturbulence measurements. However, the turbulenceintensity measurements were found to agree well withthose estimated by a King's law analysis which relied onlyon the hot-wire voltages, and not the velocity calibration ofthe sensor. It was also determined through a statistical


analysis that a considerable amount of data could beremoved by the droplet filtering techniques, and theturbulence intensity calculated from the remaining datawould still be valid.

Results and Discussion

Effect of Model ShieldingThe effect of the presence of the model on the

turbulence intensity readings was investigated in theUniversity of Illinois tunnel by positioning the proberelative to the model using the traverse system. The modelangle of attack was varied from 0 to 12°, and the distanceof the probe from the model surface ranged from 0.5 to 6inches (the same range of locations which was availableusing the IRT mounting system).

The results obtained from moving the hot-wire sensorprogressively closer to the model at 8° angle of attack areshown in Fig. 4 where turbulence intensity is plotted as afunction of distance from the model. Data at all threeturbulence intensity levels, and with and without aboundary layer trip are plotted. The boundary layer triphad no observable effect on the turbulence intensitymeasurements. The fact that the probe was inside theboundary layer at 0.5 inches was indicated by the largeincrease in turbulence level shown in Fig. 4 for all cases.The turbulence levels at each location and condition aregiven in Table 5 along with the change in turbulenceintensity from 6 to 1.5 inches. Outside of the boundarylayer a very slight increase in turbulence was observed asthe probe approached the surface. For the lowerfreestream turbulence levels (clean tunnel and 0.5% grid)the increase in turbulence apparent at one inch was alsolikely due to the influence of the model boundary layer.With no grid present, the turbulence level increased from0.15% at 6 inches to 0.25% at 1.5 inches (average ofvalues with and without the boundary layer trip), anincrease of 0.1%. With the 0.5% grid installed, theturbulence intensity 6 inches from the model averaged0.56%, and only increased to 0.59% at 1.5 inches.Similarly, the difference in turbulence from 6 to 1.5 incheswhen the 1% grid was installed was just 0.06% from0.99% to 1.05%.

It was apparent that the increase in turbulenceintensity was larger when the freestream turbulence waslow. This radiation of turbulence from the boundary layeradded like the sum of the squares and therefore has asmaller contribution to the total when the freestreamturbulence was high.

Fig. 5 shows the variation in turbulence level as themodel angle of attack was increased with the probe at 1.5inches from the model. Results with and without the 0.5%turbulence generating grid in the University of Illinoistunnel, and with and without nozzle air on in the IRT areshown for comparison. While there was a clear trend ofincreased turbulence with increasing angle of attack, it was

difficult to quantify the effect of angle attack due to theamount of scatter in these data. However, at angles of 10°or less, the increase due to model angle of attack was quitesmall. These results all indicated that with the model at10° angle of attack or less, and with the probe at least 1.5inches from the model surface, the model had only a slightinfluence on the turbulence measurements. With themodel at 8° angle of attack, and the probe 1.5 inches fromthe surface, as it was for most of the data taken in the IRT,this increase should be 0.05% or less at the freestreamturbulence levels typical of the IRT.

Effectiveness of Sensor ShieldingIt was obvious in initial tests that a considerable

number of droplets were still striking the sensor even withthe wire at 1.5 inches from the model surface, and themodel at 8° angle of attack. With the probe at 1.5 inchesfrom the surface, it could not be moved closer to the modelwithout risking significantly increased turbulencemeasurements due to the boundary layer. Therefore, themodel angle of attack was increased to 10°. This appearedto have no significant effect on the number of dropletsstriking the wire, so attempts were made to increase themomentum of the droplets, and thus there distance fromthe model by increasing the droplet size and tunnelvelocity. Again, no significant change was observed. Itwas likely that smaller droplets in the cloud were passingclose to the model, and/or droplets were being swept closerto the model by turbulent fluctuations.

Despite the fact that numerous droplets were stillstriking die sensor, the model was apparently serving toshield the probe to some extent. Figure 6 shows the resultsof varying the model angle of attack in cloud conditions.The standard deviation of the velocity, plotted here as afunction of model angle of attack, was calculated using allof the data from each set without removing droplet strikes.The significant decrease in standard deviation withincreasing angle of attack indicated that the model wasindeed shielding the sensor.

The shielding effect of the model was also apparentin ice accretions on the probe support during the testswhere the temperature was below freezing. Figure 7 is aphotograph of the probe and its mounting support (see Fig.3 for a diagram of this mounting apparatus) from abovewith ice accreted on the leading edge of the airfoil shapedsupport. It was apparent from the decreased ice accretionnear the model that a significant amount of water wasbeing deflected away from the model surface.

Despite the evidence that the model was indeedshielding the sensor from some of the droplets, this wasnot immediately apparent in the filtering results. Neitherthe percentage of data removed by the filter, nor thenumber of droplets detected, was observed to decreasesignificantly with increasing model angle of attack. (Thedroplet filtering and counting techniques will be explainedin detail in the next section.) One possible explanation for


this observation is that the number of small dropletsstriking the wire was much higher in comparison to thenumber of large droplets being deflected by the model.Thus, no noticeable decrease in the number of dropletsstriking the wire was apparent as the model angle of attackincreased.

In support of this theory, droplet size distribution datafrom Papadakis29 et al. for an MVD of 22.6 microns wasexamined to determine the number of very small dropletspresent in a cloud relative to those at the MVD and larger.These data were given as the percentage of the total liquidwater comprised of droplets of a given diameter, and areplotted in Fig. 8a. Figure 8b shows the distribution of thenumber of droplets in the cloud at a particular sizecalculated from these same data. It is important to notethat these numbers are per cubic centimeter of total water,not air as is the case with LWC measurements. Thisanalysis revealed that the droplet cloud containedsignificantly more very small droplets than droplets at theMVD and larger.

Assuming that larger droplets caused larger spikes inthe data allowed an analysis of the size distribution of thedroplets striking the wire. Larger voltage spikes wereobserved at lower angles of attack indicating that largerdroplets were striking the wire. However, there were farmore droplets spikes in the data with relatively smallvoltage peaks. This indicated that the number of verysmall droplets striking the wire .was large compared to thenumber of larger droplets.

The duration of the droplet spikes was observed to berelatively constant regardless of the amplitude of the spike.Thus, a nearly constant number of data points wereremoved due to each spike. Therefore, because thenumber of droplets striking the wire was not varyingsignificantly with model angle of attack, neither was theamount of data removed by the droplet filter. However, asthe angle of attack was decreased, larger droplets werereaching the sensor resulting in larger spikes. These largerspikes were causing increased signal rms while notsignificantly increasing the number of droplet strikescounted or the percentage of data removed. The decreasedice accretion observed near the model was due to the factthat, as indicated by the droplet distribution in Fig. 8a, thedeflected large droplets contained a large fraction of theliquid water in the cloud.

These observations made the need for the modelshielding questionable. Therefore, as stated previously,data in the second IRT test were taken with no modelpresent. It appears that at lower liquid water contentlevels, the model shielding may have been more effectivethan at the relatively high (1.5 g/m3) LWC conditionexamined by the angle of attack sweep performed in thedroplet cloud. This was indicated by the fact that a largepercentage of data had to be removed even at very lowwater contents when the model was not present. With themodel shielding, significantly less data were removed at

these low LWCs. Unfortunately, the angle of attack sweepat a fairly high liquid water content was the only oneperformed in cloud conditions. The effectiveness of themodel shielding could have been more clearly determinedif similar data had been taken in various cloud conditions.

Droplet Filtering TechniqueAn acceleration threshold method was developed for

identifying the droplet strikes in the hot-wire data. Theterm "threshold method" refers to removing data whichexceed some preset level in a measured or derivedquantity. The following explanation of this filteringmethod will rely on the time traces of velocity andacceleration data shown in Figs. 9 and 10. It is importantto note that the calculated velocities and accelerationspresented in these plots were not actual velocities oraccelerations when a droplet struck the wire, but are theacceleration or velocity calculated based on the hot-wirecalibration in air - not in water. The high heat transfer dueto the water leads to unrealistically large "sensed"velocities and accelerations which can be removed using athreshold filter. The freestream velocity for these timetraces was approximately 100 mph in a droplet cloud ofMVD = 30 microns and LWC = 1.5 g/cm3. The modelangle of attack was set at 8°, and the sensor was 1.5 inchesfrom the model surface. Tune traces for the sameconditions with no water present were also plotted forcomparison. Figure 9 is a 0.01 second time trace, whileFig. 10 is a plot of a 0.002 second segment of the samedata. The importance of using an acceleration thresholdfilter as opposed to a velocity threshold, as noted by Farraret al.20 in their work in bubbly flows, was apparent in thesetime traces. While the larger spikes due to droplets wereclearly apparent in the velocity plots, some of the smallerspikes, apparently due to small drops or partial dropletstrikes, were difficult to differentiate from freestreamturbulence. Plotting the accelerations made even thesmaller spikes considerably more apparent, and thus mucheasier to filter using a threshold method.

The acceleration threshold filtering method used isillustrated in Fig. 11. An upper and lower threshold,indicated by the two horizontal lines, was set just abovethe maximum absolute acceleration of a correspondingdata set with the nozzle air operating at the appropriatepressure, but no water present. At 100 mph, a thresholdlevel of 100,000 ft/s2 was found to be appropriate fornearly all cases. When any acceleration value exceededthe threshold, it was considered to be part of a spikeindicating droplet impingement. That point along with allpoints 0.0001 seconds (10 points in the case of dataacquired at 100 kHz) before and after it were then markedto be excluded from turbulence intensity calculations.These additional points were removed to avoid theremaining "stumps" of the droplets spikes noted byHetsroni, Cutler, and Sokolov.16 Dotted lines were used toindicated the excluded points in Fig. 11. With the data


filtering complete, the turbulence intensity calculation wasthen performed on the remaining data. It can also be seenin this plot that some small spikes apparently indicatingsmall droplet strikes fell below the threshold and were notfiltered. However, in general the acceleration peakscaused by the droplets were much larger than those due toturbulent fluctuations in the air. As an indication that thiswas the case, for the data from which the time traces inFigs. 9 and 10 were extracted, the magnitude of themaximum acceleration in the no-water case was around8xl04 ft/s2, while the maximum sensed acceleration in thewater-on case was much higher at almost IxlO7 ft/s2. Thedata taken at 250 mph was not successfully filteredbecause the rate of droplet strikes was too high relative tothe data acquisition rate.

There was some initial concern that setting such athreshold would essentially set the turbulence intensity byremoving not only droplet strikes but any additionalfluctuations caused by the presence of the water dropletswhich were not due to the water striking the wire.However, while the maximum acceleration in the water-offdata plotted in Figs. 9 and 10, was 80,600 ft/s2, applying athreshold level as low as 50,000 ft/s2 to these data onlyremoved 0.617% of the data reducing the turbulenceintensity to 0.662% from 0.663%. This indicated that themajority of the accelerations in the water-off data wereactually considerably lower than the maximumacceleration and thus considerably lower than the thresholdlevel. Examining histograms of accelerations alsorevealed this type of distribution as shown in Fig. 12. Thishistogram was developed by dividing the total range ofaccelerations into "bins" and counting the number ofmeasurements that fell into each. For this water-off casewith a nozzle air pressure of 15 psig, the largestaccelerations were found to be about 150,000 ft/s2,therefore a threshold of 200,000 ft/s2 was used to filter thecorresponding water-on data sets. As shown in Fig. 12, thevast majority of accelerations were small compared to themaximum acceleration. If we make the assumption thatany accelerations caused by the spray cloud which weren'tdue to droplets striking the wire were on the order of theaccelerations in the water-off data, then the large majorityof those accelerations were also well below the threshold.

The results of applying this filtering technique areshown in Fig. 13. This figure shows the results ofapplying the filter to the same data from which the timetraces (Figs. 9-11) were extracted. The resultingturbulence intensity as the filter threshold was decreasedwere plotted here, along with the percent of the dataremoved by each filter setting. The turbulence level of thecorresponding water-off data set was also plotted forreference. The turbulence intensity level was observed toapproach that of the corresponding water-off case as thefilter threshold decreased. The threshold level used to findthe final droplet-filtered turbulence intensity in this casewas 100,000 ft/s2 which was found to be an appropriate

filter threshold for the majority of the data taken at 100mph. The decrease in turbulence intensity at levels belowthis was quite small indicating that most of the dropletstrikes had been removed, but that the threshold was not solow that it was removing a considerable number offluctuations which were not due to droplet strikes. Themaximum acceleration in the water-off data was alsoindicated in Fig. 13. In this case, it appeared that somesmall droplets still circulating in the tunnel caused a fewsmall spikes in the water-off data. Applying a 100,000ft/s2 threshold filter to this water-off data only removed 2droplet strikes from the 1 second data set, which accountedfor 0.047% of the data and resulted in a drop of 0.001% inturbulence intensity. Therefore, a threshold of 100,000 ft/s2

was still considered appropriate in this case and others likeit.

The amount of data removed from the water-on dataset in Fig. 13, approximately 77% at a threshold level of100,000 ft/s2, was quite large. However, with the modelshield present, applying the filter to data sets where theLWC was lower resulted in far less data being removed.Figure 14 is for the same conditions as Fig. 13 except thatthe LWC was 0.9 g/m3 instead of 1.5 g/m3. In this caseonly about 12% of the data were removed by a threshold of100,000 ft/s2.

ERT Turbulence Variations Due to Cloud ConditionsThe effect of the variation of nozzle air pressure on

the tunnel turbulence intensity level was examined in bothIRT tests by acquiring data at varying nozzle air pressureswith no water present. The results of the tests are shown inFig. 15. The data marked with open symbols andconnected by solid lines are those taken at 10 kHz with 5kHz low-pass and 0.1 Hz high-pass filters. All of thesemeasurements were taken in the free stream with noshielding. In the first test the model was present, however,it was at 0° angle of attack, and the probe was six inchesfrom the surface. The solid circles are the results ofvarious water-off data sets taken at 100 kHz with themodel at 8° angle of attack and the probe at 1.5 inchesfrom the model throughout the data acquisition period inthe first test. Additional high-frequency oscillationspassed by a higher low-pass filter used at higheracquisition rates, along with the increase due to the modelshield, may explain why these turbulence measurementswere generally higher than those taken at 10 kHz. Thesolid triangles represent the data acquired at 50 kHz fromeach of the two wires in the second IRT test. These datawere reduced using the standard single-wire data reductiontechniques for each wire, and the two resulting turbulencelevels were averaged. These measurements were inreasonable agreement with the 10 kHz data.

The cause of the overall increase in the measuredturbulence intensities observed in the second test asopposed to those from the first is not clear at this time. Anew tunnel spray bar system was installed in the IRT


between these tests. However, it uses the same nozzles,and was designed to reduce the turbulence level. It islikely that there were considerable spatial variations in theturbulence level throughout the IRT test section due to thepresence of the spray bars and nozzles. This was probablythe cause of the observed difference since the sensor wasin different locations for the two tests.

Overall, these water-off turbulence measurementsagreed reasonably with those of other researchers.Poinsatte3 found turbulence levels of 0.6%, 0.52%, and0.7% at velocities of 70, 140, and 210 mph respectivelywith no nozzle air pressure. Gonsalez2 also measuredturbulence intensity levels which increased with nozzle airpressure and tunnel velocity. At 100 mph, he found aclean tunnel turbulence level of approximately 0.6%increasing to about 1.1% with the nozzle air pressure at 80psig. These measurements agreed very well with the datafrom the second IRT test (with the nozzle air heated)where the corresponding turbulence levels were 0.58%with the nozzle air off and 1.06% at 80 psig nozzle airpressure. All of these data are plotted in Fig. 16 alongwith the two corresponding water-off data points from thecurrent test.

It was apparent that increasing nozzle air pressurecaused a significant increase in the tunnel freestreamturbulence level. This was not surprising, as the air fromthe nozzles entered the free stream with a considerablecross-flow component. Small, low-pressure jets (5 psig.)directed upwind were used to increase turbulence in a windtunnel to levels up to 0.25%.30 Thus it was a logicalconclusion that the cross flow component of the IRTnozzle jets at high pressures would cause increasedturbulence. Frequency spectrums of the hot-wire voltagewith the nozzle air off and at 80 psig were compared asshown in Fig. 17. This comparison indicated that atfrequencies below 1000 Hz, the increase in turbulenceintensity due to nozzle air was a broad band increase - notan increase in fluctuations in any specific range offrequencies.

The effect of nozzle water pressure on turbulencelevel was explored in both of the two IRT tests by takingmeasurements at varying nozzle water pressures for agiven air pressure. The results are shown in Fig. 18 whereturbulence intensity levels at constant nozzle air pressuresare plotted as a function of nozzle water pressure. Againan overall increase in turbulence level with increasednozzle air pressure was seen, and the measurements fromthe second test (no model) showed higher turbulence at agiven air pressure. Unfortunately, all of these data wereacquired at relatively high LWCs or without the model.As noted earlier, large amounts of data were removed inthese conditions. This was likely the reason for the scatterin the data. There was an apparent increase in turbulencewith water pressure seen in the two 30 psig nozzle air and15 psig nozzle air, no model case. An increase in thenumber of small or partial droplet strikes missed by the

filtering method was probably responsible for this trendand may have also been contributing to the scatter in all ofthese data. However, in each of the other three cases thenozzle water pressure had little effect on the turbulenceintensity. The effects of air and water pressure will beaddressed more thoroughly later in this section.

In the first IRT test turbulence intensity data weretaken throughout the droplet size and water content rangeof the IRT. These results are presented in Fig. 19. Water-off cases at the same nozzle air pressures were plotted forcomparison. The general trend which emerged was anincrease in turbulence intensity as droplet size decreasedand liquid water content increased. This agreed withearlier observations since an increase in air pressurecaused a decrease in droplet size, and as water pressurewas increased to increase LWC, nozzle air pressure mustalso increase to maintain a given droplet size. Based onthis, it was apparent that this turbulence trend wasprimarily due to higher nozzle air pressures at higherLWCs and lower MVDs. In general, the turbulenceintensities calculated after filtering water droplets from thespray-on cases were slightly higher than those from water-off data. This could have been partially due to waterdroplet interactions with the freestream flow, but wasagain likely due to small spikes in the hot-wire signal dueto small droplets or partial droplet strikes which fell belowthe filter threshold as was observed in Fig. 11.

Due to the scatter in the data at specific air and waterpressures (Fig. 18), the exact influence of nozzle waterpressure was not clear. Thus, further analysis of the datawas performed in order to provide more evidence that thenozzle air pressure was the primary cause of the observedincrease in turbulence. A multi-variable linear regressionperformed on the water-on data from the first IRT testresulted in the following equation for the turbulenceintensity as a function of nozzle air and water pressure.

TI = 0.522 + 0.00613xP . + 0.000145x P, (6)

The coefficient on the air pressure term is over 40 timesthat of the water pressure term, again indicating thatturbulence intensity was largely a function of air pressure.For 95% confidence in this case, the critical t value was1.701. The observed t values for air and water pressurewere 10.24 and 1.343 respectively. The fact that the tvalue for air pressure was considerably higher than thecritical value indicates that air pressure was an importantfactor in determining turbulence intensity, while the t valuefor water pressure was lower than the critical valueindicated that water pressure was not important.

Effect of Heated Nozzle Air on TurbulenceMeasurements

As noted earlier, the air exiting the IRT spray nozzleswas heated to approximately 180° F to prevent ice from


forming in the nozzles. It was likely that this hot air wascausing high-frequency temperature fluctuations whichwere being misinterpreted as velocity fluctuations and thuscontributing to the measured turbulence intensity values.

In order to examine and quantify this effect, identicalturbulence measurements were taken at varying nozzle airpressures, with and without the nozzle air heated. Theresults of these measurements are plotted in Fig. 20. Asexpected, the heating of the nozzle air did cause anincrease in the turbulence intensity readings, apparentlydue to increased temperature fluctuations increasing thefluctuation in the heat transfer from the wire. Themeasurements taken with no nozzle air pressure prior toboth the heated and cool air data acquisition matched verywell at values of 0.58% and 0.584%. At 20 psig airpressure, the increase due to the heated nozzle air was0.037%, from 0.725% to 0.762% while at 80 psig theheated air caused the turbulence intensity to increase from0.991% to 1.063%, an increase of 0.072%. This largerincrease at higher nozzle air pressures indicated that asmore heated air was introduced into the freestream flow bythe nozzles, the temperature fluctuations seen in the testsection increased.

Based on an analysis from TSI Technical Bulletin16,24 turbulence intensity values due to different sourcescan be combined as the square root of the sum of thesquares. Thus, the error in the turbulence intensity due tothe temperature fluctuation, TIT , can be found usingequation (7).

^~- ~ (?)

Tlhot and Tlî are the turbulence intensities with andwithout the nozzle air heated, respectively. The standarddeviation of the temperature, T ,̂ was then found usingequation (8) which is also based on the analysis in the TSIbulletin.

TIT , ._ *rms |TI T> \ /O\

nns =~7ZTVTw ~TaJ (8)200

Tw and Ta are the wire operating temperature and theambient temperature. The calculated standard deviationsare given in Table 6. These fluctuations were found to bequite small (0.34 °C maximum) and, as expected, wereobserved to increase with increasing nozzle air pressure.

It is also quite likely that these temperaturefluctuations are spatially dependent on location relative toa nozzle. Therefore, it is likely that this effect is dependenton location in the tunnel test section. However, this spatialvariation was not studied in this test.

Turbulence Intensity Measurements in IcingConditions

While all of the testing discussed to this point wasperformed at temperatures near room temperature, a set of

data was acquired in icing conditions to assess thesuitability of these experimental methods for use in flighttest icing conditions. Turbulence measurements weretaken until the ice accretion prevented the sensor fromoperating properly. Surprisingly, the ice accretion did notlead to sensor wire breakage. Close-up inspection of theice formations revealed that the ice growing on the wiresupports eventually became thick enough to block the wirefrom the freestream flow, but the heat of the wire kept asmall area around the wire clear of ice. The turbulenceintensity level of the droplet filtered data becameunpredictable soon after ice began to accrete. However,several of the data sets taken 1 minute or less after thecloud was turned on were successfully filtered to give atunnel turbulence level near that of the water off cases.This test indicated that the droplet filtering technique couldbe successfully employed in icing conditions if the probecould be shielded from ice accretion until immediatelybefore measurements were taken, or if the probe could beeffectively anti-iced or periodically de-iced.

Conclusions

As part of this research, a successful method formeasuring turbulence intensity in icing cloud conditionswas developed. Based on this development process, thefollowing conclusions can be made.1. A high data acquisition fate was found to be necessary

for resolving droplet strikes in the hot-wire data. Datain the first test were acquired at 100 kHz which wasfound to be more than adequate for resolving thedroplet strikes at 100 mph. A portion of the 100 mphdata acquired in the second test was acquired at 50kHz and was also successfully filtered. However, at250 mph, the data could not be successfully filtered,because the frequency of droplets striking the sensorwas too high relative to the data acquisition rate.

2. The use of an acceleration threshold as opposed to avelocity threshold was critical in identifying small orpartial droplet strikes which resulted in smalleramplitude spikes in the hot-wire data.

3. The use of an airfoil to shield the sensor resulted inslight increases in the measured turbulence intensity.With the probe at 1.5 inches from the model surfaceand the model at 8° angle of attack, this increase wasat most 0.06% at turbulence intensity levels typical ofthose in the IRT.

4. With some modification, the method could be usedeffectively in conditions at temperatures belowfreezing where ice was accreting.

The droplet filtering technique was used to measureturbulence intensity in the IRT spray cloud. The followingconclusions were reached based on these measurements.1. Turbulence intensity was found to increase

significantly as nozzle ah" pressure increased. With

10


nozzle air pressure increasing from 0 to 80 psig, thefreestream turbulence level increased fromapproximately 0.6% to 1.0%.

2. Nozzle water pressure had little if any effect onturbulence intensity. Slight increases with increasingwater pressure were likely due to an increase in thenumber of small droplet strikes missed by the dropletfilter at higher LWCs.

3. Turbulence intensity in the spray cloud was found toincrease with increasing LWC and decreasing dropsize. These trends were again attributable to increasedair pressure.

4. Nozzle air, which was heated to prevent ice buildup inthe nozzles, did cause temperature fluctuations whichresulted in artificially high turbulence measurements.These temperature fluctuations increased withincreasing nozzle air pressure. At the highest nozzleair pressure of 80 psig, the temperature fluctuationwas found to be 0.33 °C, and caused an increase of0.07% in turbulence intensity from 0.99% to 1.06%.

5. Small increases in turbulence intensity were observedwhen water droplets were present over the turbulenceintensity at a given nozzle air pressure with the wateroff (for example, .015% increase at 100 psig waterpressure). It was not possible to distinguish whetherthis small increase in turbulence was caused by thepresence of the droplets or from errors due to thefiltering technique.

References1 Dryden, H.L., and Kuethe, A.M., "Effect of Turbulencein Wind Tunnel Measurements," NACA Report No. 342,'1930

2 Gonsalez, J.C., NASA Lewis Research Center,Cleveland, Ohio, personal communication, June, 1996

3 Poinsatte, P.E., "Heat Transfer Measurements From aNACA 0012 Airfoil in Flight and in the NASA LewisResearch Tunnel," NASA CR-4278, 1990

4 Kuethe, A.M., and Chow, C.Y., Foundations ofAerodynamics: Bases of Aerodynamic Design, FourthEdition, John Wiley and Sons, New York, 1986

5 Tani, I., "Boundary layer Transition" Annual Review ofFluid Mechanics, Sears, W.R., Editor, VanDyke, M.,Associate Editor 1969

6 Green, J.E., "On the Influence of Freestream Turbulence.on a Turbulent Boundary Layer, As it Relates to WindTunnel Testing at Subsonic Speeds," AGARD ReportNo.602, 1972

7 Bragg, M. B., Cummings, M. J., Lee, S., and Henze, C.M., "Boundary Layer and heat Transfer Measurementson an Airfoil with Simulated Ice Roughness," AIAAPaper 96-0866, 34th Aerospace Sciences Meeting andExhibit, Reno, NV 1996

8 Junkhan, G.H., and Serovy, G.K., "Effects of freestreamTurbulence and Pressure Gradient on Flat-PlateBoundary layer Velocity Profiles and Heat Transfer,"Journal of Heat Transfer, May, 1967

9 Blair, M.F., "Influence of Freestream Turbulence onTurbulent Boundary Layer Heat Transfer and MeanProfile Development, Part 1 - Experimental Data,"Journal of Heat Transfer, Vol. 105, February, 1983

10 VanFossen, G. J. Jr., and Simoneau, R. J., "A Study ofthe Relationship Between Freestream Turbulence andStagnation Region Heat Transfer," Journal of HeatTransfer, Vol. 109, February, 1987

11 VanFossen, G. J., Simoneau, R. J., Olsen, W.A. Jr., andShaw, R. J., "Heat Transfer Distributions AroundNominal Ice Accretion Shapes Formed on a Cylinder inthe NASA Lewis Icing Research Tunnel," AIAA Paper84-0017, 221984

nd Aerospace Sciences Meeting, Reno, NV,

12 Gelder, T. F., and Lewis, J. P., "Comparison of HeatTransfer From Airfoil in Natural and Simulated Icingconditions," NACA TN-2480, September 1951

13 Merceret, F.J., "An Experimental Study to Determinethe Utility of Standard Commercial Hot-wire andCoated Wedge-Shaped Hot-film Probes forMeasurement of Turbulence in Water-Contaminated AirFlows," Technical Report #40, Chesapeake BayInstitute, The John's Hopkins University, 1968

14 Merceret, F.J., "An Experimental Study to Determinethe Utility of Standard Commercial Hot-wire andCoated Wedge-Shaped Hot-film Probes forMeasurement of Turbulence in Water-Contaminated AirFlows, Part II," Technical Report #40, Chesepeake BayInstitute, The John's Hopkins University, 1969

15 Goldschmidt, V.W. and Householder, M.K., "TheHotwire Anemometer as an Aerosol Droplet SizeSampler," Atmospheric Environment,Vol. 3, PergamonPress, 1969

16 Hetsroni, G., Cutler, J.M., and Sokolov, "Measurementsof Velocity and Droplets Concentration in Two-PhaseFlows," Journal of Applied Mechanics, Transactions ofthe ASME,3we 1969

11


17 Hetsroni, G., and Sokolov, "Distribution of Mass,Velocity, and Intensity of Turbulence in a Two-PhaseTurbulent Jet" Journal of Applied Mechanics,Transactions oftheASME, June 1971

18 Bruun, H.H., Hot-Wire Anemometry: Principles andSignal Analysis, Oxford University Press, 1995

19 Liu, T.J., and Bankoff, S.G., "Structure of Air -WaterBubbly Flow in a Vertical Pipe I. Liquid Mean Velocityand Turbulence Measurements," International Journalof Heat and Mass Transfer, Vol. 36, No. 4

20 Farrar, B., Samways, A.L., Ali, J., and Bruun, H.H., "Acomputer-Based Hot-Film Technique for Two-PhaseFlow Measurements," Meas. Sci. Technol. 6, 1995, pp.1528-1537

21 Kitsch, M.L. and Davidson, J.H., "Phase Discriminationin Gas-Particle Flows Using Thermal Anemometry,"Journal of Fluids Engineering, Transactions of theASME, Vol. 114, December, 1992

22 Henze, C.M., "Turbulence Intensity Measurements inIcing Cloud Conditions, " M.S. Thesis, University ofIllinois, 1997.

23 Anon., "IFA 100 Intelligent Flow Analyzer SystemInstruction Manual," TSI incorporated, 1983

24 Anon., "Temperature Compensation of ThermalSensors," TSI Technical Bulletin 16

25 Lee, S., "Heat Transfer On An Airfoil With LargeDistributed Leading-Edge Roughness," M.S. Thesis,University of Illinois, 1997

26 Kerho, M.F., "Effect of Distributed Roughness Near anAirfoil Leading Edge On Boundary layer Devlopmentand Transition," Ph.D. Thesis, University of Illinois,1995

27 Kline, S.J., and McClintock, F.A., "DescribingUncertainties in Single-Sample Experiments,"Mechanical Engineering, Jan., 1953

28 Coleman, H.W., and Steele, W.G., Jr., Experimentationand Uncertainty Analysis for Engineers, John Wiley andSons, Inc., New York, 1989

29 Papadakis, M., Elangonan, R., Freund, G. A., Breer, M.,Zumwalt, G. W., and Whitmer, L., "An ExperimentalMethod for Measuring Water Droplet ImpingementEfficiency on Two- and Three-Dimensional Bodies,"NASA Contractor Report 4257, 1989

30 Kendall, J.M., "Boundary Layer Receptivity toFreestream Turbulence," AIAA paper 90-1504, AIAA21st Fluid Dynamics, Plasma Dynamics and LasersConference, Seattle, WA, June, 1990

12


Table 1. Matrix of test conditions for model shielding effects test

Test Conditions:1 - no turbulence grid, no boundary-layer trip2 - 0.5% turbulence grid, no boundary-layer trip3 - boundary-layer trip, no turbulence grid4 - 0.5% turbulence grid and boundary-layer trip5-1% turbulence grid, no boundary-layer trip

Angle of Attack(degrees)

024681012

Distance from model surface (inches)6

1,21,21,21,2

1,2,3,4,5. 1,2

1,2

5111

1,2

41,21,21,21,2

1,2,3,4,51,21,2

3111

1,2

21,21,21,21,2

1,2,3,4,51,21,2

1.51,21,21,21,2

1,2,3,4,51,21,2

11,21,21,21,2

1,2,3,4,51,2

0.51,21,21,21,2

1,2,3,4,5

Table 2. Summary of IRT cloud conditions - varyingLWC and MVD in the first IRT test.

Table 3. Summary of cloud conditions - varying air andwater pressures in the first IRT test

Velocity(mph)

100

250

MVD(Urn)202020303030404040202020303030404040

LWC(g/m3)

0.71.31.90.91.52

0.91.52

0.40.650.90.50.75

10.50.75

1

AirPressure

11.0335.02

7012.0428.0642.8110.6723.9335.9213.4134.160.5213.8727.4141.8912.1723.3835.22

WaterPressure

31.12130.78318.3446.08147.89282.5843.58138.53264.240.3

126.32268.1456.29142.91273.1253.06133.85255.49

Model angle of attack 8°, probe 1.5 inches from modelsurface. Data acquired at each air pressure with andwithout water on.

Velocity(mph)

100

250

MVD(urn)

1825

35.5

13.516

18.5

1214

15.5

1825

35.5

13.516

18.5

1214

15.5

LWC(g/m3)

0.911.111.29

0.740.961.14

0.580.821.01

0.460.560.65

0.380.490.58

0.30.420.51

AirPressure

202020203030303040404040202020203030303040404040

WaterPressure

060801000

608010006080100060801000608010006080100

Model angle of attack 8°, probe location 1.5 inches frommodel surface. Data also acquired at each air pressurewith no water present.

13


Table 4. Test conditions in second IRT test

MVD(urn)

17.523.6

15.519

23.5293719

16.2

15*15.9

1821*

LWC(g/m3)

0.550.78

0.660.851.011.141.271.050.96

0.871.011.231.42

AirPressure

101010151515151515 .20253030303030

WaterPressure

020300304050607060600607090110

2.5-,

2.0-

1.5-

1.0-

0.5-

0.0

•nozzle air off-nozzle air on

50 100 150 200 250 300Velocity (miles per hour)

350

Table 5. Turbulence intensity values at various distancesfrom the model surface, and at various test conditions -Model at 8° angle of attack, and velocity approximately100 mph. (*: condition repeated with nozzle air heated)

Test Conditions:1 - no turbulence grid, no boundary-layer trip2 - 0.5% turbulence grid, no boundary-layer trip3 - boundary-layer trip, no grid4 - 0.5% turbulence grid and boundary-layer trip5-1% turbulence grid, no boundary-layer trip

Figure 1. Turbulence Intensity as a function of velocityin the NASA Lewis Icing Research Tunnel (IRT)measured by Gonsalez2. Measurements taken withoutthe nozzles operating (nozzle air off), and with nozzle airpressure at 80 psig with no water present (nozzle air on).

Droplet Trajectories

Figure 2. Droplet Trajectories over NACA 0012 airfoil.Hot-wire probe shielded by airfoil.

distance fromsurface (in.)

65432

1.51

[email protected]@6

condition1

0.140.170.230.220.260.250.352.050.11

20.550.550.560.560.570.580.612.440.03

30.16

-0.22

-0.240.250.355.820.09

40.57

-0.57

-0.590.590.664.230.02

50.99

-1-

1.031.051.015.380.06

tubes which support spacersand probe support piece -

Table 6. Standard deviation in temperature with nozzleair heated

Nozzle Air Pressure(psig)

20406080

Standard Deviation inTemperature (°C)

0.2160926540.3122748480.3432020030.330482343

j* set screws to hold spacers*"/ and probe support piece

ujustable spacers

Figure 3. Hot-wire mounting apparatus used with 21"chord NACA 0012 model in the IRT

14


6.0-,

5.0-

•B 4.0-cw

~ 3.0-

M 2.0-

1.0-

0.0

- no turbulence grid, no boundary-layer trip~ 0.5% turbulence grid, no boundary-layer trip- no turbulence grid, boundary-layer trip in place- 0,5% turbulence grid and boundary-layer trip- 1% turbulence grid, no boundary-layer trip

0 1 2 3 4 5 6Distance From Model Surface (inches)

Figure 4. Turbulence intensity levels measured in theUniversity of Illinois 3x4 foot tunnel as a function ofdistance from the model surface. 85% chord location,velocity = 100 mph, angle of attack=8°.

0.7 n

0.6-

3~ 0.4 ̂

0.3-:

0.2-

0.1

- no turbulence grid, no boundary-layer nip- 0.5% turbulence grid, no boundary-layer nip- IRT no nozzle air- IRT nozzle air pressure 14.21 psi

0 2 4 6 8 10

Model Angle of Attack (degrees)12

Figure 5. Variation in turbulence intensity with modelangle of attack. Probe 1.5 inches from surface at 85%chord location. Velocity lOOmph.

^ %•£.̂ io « oII?£.H «c'5-i

14-,

12-

10-

T3 3

$1C «

|_ 8.0-101)

1 6.0-i

4.01

Model Angle of Attack (degrees)

Figure 6. Standard deviation of velocity signal, includingdroplet strikes, at varying angle of attack in IRT spray

. cloud. Velocity lOOmph, LWC = 1.5 g/m3, MVD = urn.

Figure 7. Ice accretion on hot-wire mounting support.Velocity = 100 mph, LWC = 0.7 g/m3, MVD = 20um,To = 25 °F.

u

12.0-j

10.0-

8.0-

6.0-

4.0-

2.0-

0.0

Droplet MVD

0 20 40 60 80 100Droplet Diameter (microns)

0 20 40 60 80Droplet Diameter (microns)

Figure 8. Droplet size distributions by % of total water,and number of droplets per cm3. MVD = 22.6 urn.(Data from Papadakis et al.29)

15


280n

C 260:

.£ 240-

o 220-|

J" 200-|

I 18°;£ 160-8 :£ 140-

120-0.3

,1 1 t . IUÛt>> .i. M ̂

——— water

VJJL-A-A*1* - -" —— •*•

S. 170-.8 ':^ 160-o ;

"̂ 15°off 8 :

* 140-

JL-MI

1 uo

120-0.3"

, ,

70 ' 6.372 ' 6.374 ' 6.376 ' 6.378 6.380 1 106 .Time (seconds)

2.0 106 -,

g 1.5 106-

| 1.0106:

! 5.0 10s -

1" 0.010°-aj

[ 1 1*,r4Jttn——— wato- - - - - wato

U 1 • . 1 i ll[1 T T n

11" 5 105 .

1

onoff

-W

J 0 10° .

.9 -5105.

§' "*• -1 10

——— water on-----water off

vrt — -JSIJîaOx^t^tf'My^ <! iit^—J-V?20 0.3725 0.3730 0.3735 0.3740

Time (seconds)

——— water on- - - - - water off

1 1—— ̂ ^V^J(A

V H"

0.3720 0.3725 0.3730Time (seconds)

0.3735 0.3740

Figure 10. Shorter segments of the same time tracesshown in Fig 9.

0.370 0.372 0.374 0.376Time (seconds)

0.378 0.380

Figure 9. Velocity and acceleration time traces. MVD:

30 urn, LWC = 1.5 g/m3, velocity = 100 mph.

a££S

£

4105

310s

210s

1105

-no5

-2105

1 -310'

00a•1

——— Unfiltered- - - - - Filtered

0.372 0.3722 0.3724 0.3726 0.3728Time (seconds)

0.373

Figure 11. Illustration of droplet filtering techniqueapplied to a segment of the data shown in the previoustwo figures. Horizontal lines indicate accelerationthreshold. Dotted line indicates data removed by filter.

16


en2500-,

2000-3 S73 *-> o 1500| §1 ."2 1000-o> :>o !̂u —< .s^ " 500.0 -

0.000-200000 -100000 0 100000 200000Acceleration (ft/s )

Figure 12. Histogram of accelerations in a water-off dataset at 100 mph and 15 psig nozzle air pressure.

- % of Data Removed From Spray-on Data- Turbulence Intensity of Spray-on Data- Turbulence Intensity of Spray-off Data

2.5 r,Maximum Accelerationof Spray-Off Data

Acceleration Filter Setting (ft/s 2)

Figure 13. Turbulence intensity and percent of dataremoved as a function of acceleration threshold.Turbulence intensity and maximum acceleration ofcorresponding water-off data are plotted for reference.Same conditions as Figs. 9 and 10.

- % of Data Removed From Spray-on Data~ Turbulence Intensity of Spray-on Data- Turbulence Intensity of Spray-off Data

Maximum Accelerationof Spray-Off Data

10'Acceleration Filter Setting (ft/s 2)

Figure 14. Turbulence intensity and percent of dataremoved as a function of acceleration threshold. T.I. andmaximum acceleration of corresponding water-off dataare plotted for reference. Same conditions as Fig. 13with LWC reduced to 0.9 g/m3.

§tsuug

1.1-i.o-;

0.90:

0.80-

0.70-

0.60-

0.50-

0.40

10 kHz data from first IRT test100 kHz data from first IRT test10 kHz data from second IRT testSO kHz data from second IRT tea

10 20 30 40 50 60 70 80Nozzle Air Pressure (psig)

Figure 15. Turbulence intensity variation with nozzle airpressure at 100 mph. No water present.

2.5-,

* 2-°"Zg 1.5c

1.0-

0.5-

0.0

- Gonsalez - no nozzle air~ Gonsalez • 80 psi nozzle air pressure- Poinsatte - no nozzle air

current test - no nozzle aircurrent test - 80 psi nozzle air pressure

50 100 150 200 250 300Velocity (miles per hour)

350

Figure 16. Turbulence intensity as a function of velocityfrom Poinsatte3 and Gonsalez2. Two water off datapoints from the second IRT test included for comparison.

17


-50.0-,

-60.0-

S -70.0

-80.0-

-90.0-

-100

——— no nozzle air pressure- - - - - nozzle air pressure 80 psig

0 200 400 600 800 1000Frequency (Hz)

Figure 17. Frequency spectrums acquired from voltagesignal in IRT with no nozzle air pressure and 80 psignozzle air pressure (no water).

- 20 psig nozzle air pressure- 30 psig nozzle air pressure- 40 psig nozzle air pressure- 15 psig nozzle air pressure, no model- 30 psig nozzle air pressure, no model

0.50 12020 40 60 80 100

Nozzle Water Pressure (psig)

Figure 18. Turbulence intensity as a function of nozzle. water pressure at various nozzle air pressures

—•— water off 20 micron MVD—•— water on 20 micron MVD-x - water off 30 micron MVD-•— water on 30 micron MVD4 • water off 40 micron MVD

— -»• • water on 40 micron MVD0.9 '

0.8

0.7gj>-P

0.6

0.50.5 1 1.5

Liquid Water Content (g/m3)

0.5010 20 30 40 50 60 70 80

Nozzle Air Pressure (psig)

Figure 20. Turbulence intensity as a function of nozzleair pressure with and without nozzle air heated

Figure 19. Turbulence intensity as a function of liquid• water content at 3 droplet sizes. Filtered spray-on andcorresponding spray-off cases plotted for comparison.

18

Documents

Freestream turbulence measurements in icing conditionsicing.ae.illinois.edu/papers/98/AIAA-1998-96.pdfan NACA 0012 found a maximum heat transfer increase in the IRT of 10% over heat