Development of a Global Heat Transfer MeasurementSystem atAEDC Hypervelocity

Wind Tunnel 9*

Inna Kuritst and Mark J. LewistDepartment ofAerospace Engineering

University ofMarylandCollege Park, MD 20742

Marvine. P. Hamner§LeaTech LLC

Frederick, MD 21701

and

Joseph D. NorrisAerospace Testing Alliance

Arnold Engineering Development Center - White OakSilver Spring, Maryland 20903-1005

Heat-transfer rates are an extremely important consideration in the design of hypersonicvehicles such as atmospheric reentry vehicles. This paper describes the development of adata reduction methodology to evaluate heat-transfer rates using global surface temperaturemeasurements on wind tunnel models at the Arnold Engineering Development Center(AEDC) White Oak's Hypervelocity Wind Tunnel 9. As a part of this development effort, ascale model of the NASA Crew Exploration Vehicle was painted with temperature-sensitivepaint (TSP), and multiple sequences of high-resolution images were acquired during a five-run test program. The calculation of heat-transfer rates from TSP data acquired in Tunnel 9is challenging because of high Reynolds number and dynamic pressure environments andthe desire to use standard stainless steel wind tunnel models that were originally designedfor force and moment testing. The authors developed an approach to reduce TSP data intoconvective heat flux while taking into consideration the challenges listed above. Apreliminary comparison of the heat flux value calculated using the TSP surface temperaturedata with the value calculated using the standard thermocouple data is reported.

NomenclatureCp = specific heatcp = specific heat of nitrogen at constant pressure, 0.248 BTU/lbm-°Rfps = frames per secondHo = calculated total enthalpy, BTU/lbmHz = Hertz

*The research reported herein was performed by the Arnold Engineering Development Center (AEDC), Air Force MaterielCommand. Work and analysis for this research were performed by personnel of Aerospace Testing Alliance (ATA), theoperations, maintenance, information management, and support contractor for AEDC. Further reproduction is authorized tosatisfy needs of the U.S. Government.Graduate Research Assistant, University of Maryland, College Park, MD

t Professor, Department of Aerospace Engineering, University of Maryland, College Park, MD§ Principal, LeaTech LLC, Frederick, MD

Project Engineer, Aerospace Testing Alliance, AEDC White Oak, Silver Spring, MD; joe.norris@arnold.af.mil

11-4244-1 600-01071$25.00 ©2007 IEEE.

K = thermal conductivityL = paint layer thicknessMoo, MINF = freestream Mach numberPoo, PINF = freestream pressure, psiaqdot, q" = heat-transfer rateRe = Reynolds numberSt = Stanton numbersurf = surfaceT = temperatureTst = steel temperatureTtsp = temperature-sensitive coating temperatureTw = wall temperatureTi, j = temperature at step i, and node jt = timeu = freestream velocityx = distance through surface of the modela = thermal diffusivityAx = node size through the model wall in the numerical modelp = freestream density

I. IntroductionFor many applications, the evaluation of heat flux at the Arnold Engineering Development Center (AEDC) White

Oak's Hypervelocity Wind Tunnel 9 currently involves acquisition of surface temperature measurements usingthermocouples or a direct-reading heat-transfer gage (i.e., these are discrete measurements). These measurementstypically are made on a large, stainless steel model designed also to simultaneously acquire force-and-moment andsurface pressure data. In the case of the thermocouples, once surface temperatures have been measured, a simpleone-dimensional (ID) numerical heat-transfer conduction model is used to determine the convective heat flux at thediscrete locations of the thermocouples. Unfortunately, most complex flow phenomena cannot be capturedadequately by using point measurements. Examples of such phenomena include boundary-layer transition, flowseparation, and shock/boundary-layer interactions. These types of phenomena typically exhibit strong gradientsalong the surface of the model, thus making it difficult to resolve them using discrete measurements. In addition,some scale models may be difficult or impossible to instrument with thermocouples in areas such as sharp leadingedges and control surfaces. Moreover, the cost and time-consuming nature of installing discrete instrumentation canbe a major driver in test planning. Hence, some of the desired advantages of a global heat-transfer mappingtechnique are a lower application cost and, in effect, increased density of instrumentation.

Previous collaborative efforts between AEDC White Oak, LeaTech LLC, and the University of Marylandresulted in the development of an intensity-based, temperature-sensitive paint (TSP) system capable of withstandingthe harsh environment of the facility and acquiring high-resolution temperature maps of a complex, three-dimensional surface. This system and experimental results from its use have been reported in Ref. 1. In that previousstudy, these temperature maps were acquired at frame rates that were too slow to extract heat-transfer data.However, the images are extremely valuable in that they provide an unprecedented level of qualitative informationabout the model surface flow patterns.

During 2006, AEDC White Oak conducted extensive aerothermodynamic testing of a model of the NASA CrewExploratory Vehicle (CEV) that was instrumented with coaxial gages to measure surface heat transfer and to providean indication of boundary-layer-transition location. Following completion of that test program, a temperature-sensitive coating was applied to the model. The TSP illumination system improvements suggested in Ref. 1 werecompleted, and a camera system capable of obtaining digital images at sufficient sample rates and resolution wasidentified. The improved TSP system (lamps, cameras, and paint coatings) enabled acquisition of multiple high-resolution images during run times on the order of 0.5 to 1.5 s experienced during the test program. This paperdescribes both the data reduction methodology developed to compute the heat flux using this system and some of thepreliminary results.

A. Tunnel 9 Facility DescriptionTunnel 9 is a unique blowdown facility that utilizes pure nitrogen as the working fluid and currently operates at

Mach numbers of 7, 8, 10, and 14. An operational envelope showing Reynolds number equivalent altitudes vs.

2

Mach number for Tunnel 9 operating conditions is presented 200

in Fig. 1. The unit Reynolds number range for the facility isfrom 0.05 x 106/ft (useful for high-altitude/viscous interaction*simulation) to 48 x 106/ft (duplication of flight dynamic H5pressure). IPresent unne

The test section is over 12 ft long and has a diameter of 5 Ope ation Cosntnns100 E g _g _1Simulationft, enabling testing of large-scale model configurations that - I L Duplicationcan include simultaneous force and moment, pressure, and l lopheat-transfer instrumentation. The test cell features a model 50 H RetrBneodpessupport system that is capable of dynamically pitching large Flight veopiesPP Y PY Y P g g _ ~~~~~~~~~~~~~~~~~~~~~~~~~~Flight Vehiclestest articles through an angle-of-attack sweep from -5 to +45 ENDO Interceptorsdeg at rates of up to 60 deg/s during a typical run. The Mach a

10 and 14 nozzles are 40 ft in length with a 60-in.-diam exit. 4 8 12 16 20 24 28Mach Number

The Mach 8 nozzle is 40 ft in length with a 35-in.-diam exit Fig. 1. Hypervelocity Wind Tunnel 9and operates as a free jet when it is mated to the 60-in.-diam operational envelope.test cell. A photo of the Tunnel 9 Mach 10 nozzle and test cellis provided in Fig. 2, and a schematic of the entire facility is shown in Fig. 3. Note that flow is from left to right inthese figures.

Fig. 2. AEDC White Oak Hypervelocity Wind Tunnel No. 9 Test Cell and Mach 10 Nozzle.

During a typical run, the vertical heater vessel (left side of Fig. 3) is used to pressurize and heat a fixed volumeof nitrogen to a predetermined pressure and temperature defined by the desired freestream conditions. The test celland vacuum sphere are evacuated to approximately 1 torr (mmHg) and are separated from the heater by a pair ofmetal diaphragms located upstream of the throat. When the desired temperature and pressure are reached in theheater, the diaphragms are ruptured. The gas then flows from the top of the heater vessel, expanding through thecontoured nozzle into the test section at the desired freestream test conditions. As the hot gas exhausts from the topof the heater, cold nitrogen gas from the pressurized driver vessels enters the heater base. This cold gas drives thehot gas out the top of the heater in a piston-like fashion, thereby maintaining constant conditions in the nozzlesupply plenum and in the test section during the run. A run is completed once the supply of hot, pressurized gas isexhausted. A more complete description of the Tunnel 9 facility and its capabilities can be found in Ref. 2.

B. Temperature-Sensitive Paint BackgroundTSP systems have been successfully applied to the study of flows from low subsonic to hypersonic speeds over a

wide range of Reynolds numbers and under a variety of test conditions, e.g., cryogenic and high-temperature (high-enthalpy) conditions. The primary benefit of these types of global mapping systems is their ability to acquire high-resolution, quantitative global surface temperature maps for complex, three-dimensional models. With anappropriate data reduction algorithm in place, these temperature maps can be converted into heat-transfer rates.Convective heat transfer associated with aerodynamic heating of a vehicle traveling at hypersonic speeds is anextremely important factor in the design of hypersonic vehicles. Moreover, a TSP system is potentially less costlyand time consuming than is the use of discrete measurements currently employed at facilities such as Tunnel 9.

3

r ~~~~~~~~~~~~~~~VacuumSphere

Heater pW NzlNozzle

Fig. 3. Tunnel 9 facility schematic.

In general, a temperature-sensitive coating consists of a binder and photoluminescent (fluorescent orphosphorescent) probes. The photoluminescent probes, or luminophores, are temperature-sensitive moleculesdispersed within the binder. The binder is a host matrix, usually a polymer, that forms the coating material. Thetemperature-sensitive molecules undergo a photochemical reaction when excited at the appropriate wavelength andsubjected to a change in temperature. The resulting photoluminescence, or emission from the coating, is red-shiftedrelative to the excitation wavelength, and its intensity depends on the temperature. The emitted light intensity isdetected by a photodetector, e.g., a photomultiplier tube, photodiode, or charge-coupled device (CCD) camera, andis converted into a temperature history by processing the detected intensity in the image-processing software using aknown calibration curve. For an in-depth description of the photochemical process, see Ref. 3.

There are a few unique challenges associated with developing a TSP system for use in Tunnel 9. Among themare a high dynamic and thermal loading environment (i.e., large freestream Reynolds numbers), long run timesrelative to other hypersonic facilities of similar freestream conditions, and transient heating profiles that result fromdynamically pitching the model during the run. These conditions impose limitations on the types of models tested atTunnel 9 and dictate data acquisition and reduction requirements.

As mentioned above, feasibility studies previously conducted at Tunnel 9 demonstrated the survivability of thepaint in the extreme conditions of the tunnel as well as the ability to acquire global maps of complex flowphenomena. From these studies the need for improvements in camera/detection and illumination systems wasidentified. A study was conducted to assess various possible illumination sources for their intensity, stability, andoperational qualities. A detailed description of this effort is provided in Ref. 1.

C. Review of Global Heat-Transfer Techniques for Hypersonic Wind TunnelsA range of global temperature and heat-transfer acquisition techniques has been utilized at various hypersonic

facilities. In all cases, reducing global temperature data into heat flux is a nontrivial task; therefore, simplifyingassumptions related to specific test conditions usually have to be made to develop a practical data reductionmethodology. In other words, the choice of simplifying assumptions that define the heat-flux data reductionalgorithm depends on the facility and the types of models tested. For instance, a two-color thermographic phosphortechnique has been successfully developed and applied to ceramic wind tunnel models by Buck4 at NASA LangleyResearch Center for several years. The test articles are injected into the flow, eliminating the need for inputting atransient heat flux in their mathematical heat-transfer modeling. Two factors greatly simplify the heat-transfercalculations: 1) step input heating due to model injection, and 2) semi-infinite wall assumption, which is valid forceramic models of appropriate thickness.

An example of luminescent paint techniques successfully applied in a hypersonic ground-test facility includesthe research at the JAXA Hypersonic Shock Tunnel facility by Nakakita et al. The approach taken in this facility inorder to simplify the heat-transfer data reduction is to utilize a very thin layer of the polymer so that the influence ofthe typically insulative TSP layer on surface temperature response can be neglected. Using typical polymer material

4

properties, the authors estimated that the paint layer can be ignored in the data reduction if the thickness is less than1 ptm thick and a 2-percent error in heat-transfer rate calculation is acceptable. Then, an assumption of uniform,semi-infinite media can be made in the heat-transfer rate calculation, making the data reduction straightforward.Ohmi et al.6 conducted a followup experimental study to evaluate this assumption. They tested ceramic modelspainted with a very thin TSP layer (0.2 to 3 ptm) and used the same simple, 1D, semi-infinite heat conduction modelto calculate the heat-transfer rate. They ignored the TSP layer in the data reduction and calculated the errorassociated with this simplification. They concluded that the paint layer can be ignored in the data reduction if it isless than 0.5 ptm, and not 1 ptm, as previously estimated by Nakakita et al. because of the differences between theactual TSP material properties and handbook tabulated polymer material properties. Additionally, the error incalculated heat flux changed nonlinearly with the change in paint layer thickness.

Hubner et al.7 used TSP to measure full-field surface heat-transfer rates in short-duration hypersonic flow (runtimes under 10 ms) at the LENS1 shock tunnel at CUBRC. A thick, insulating polyurethane layer (100 to 150 ptm)was applied between the thin (approximately 5 to 10 ptm), active TSP layer and the metal model surface. The heattransfer was calculated assuming adiabatic wall condition, constant step input heat-transfer rate, and temperature-independent thermal conductivity, K, and thermal diffusivity, a. These assumptions are applicable only because ofthe short run time of the facility.

The initial approach for evaluating heat transfer in Tunnel 9 from global surface temperature measurementsfollows a somewhat different path from that of previous research. This is partly because of the operational behaviorof the facility and the need to use structurally robust stainless steel models that are well suited for force and momenttesting in the high Reynolds number environment. The goal is to be able to use the same test articles for TSP tests asare used for force and moment testing in order to reduce complexity and cost resulting from multiple models for asingle test program.

In general, the approach at Tunnel 9 was to apply the same transient, ID, finite-difference conduction algorithmas was used for reducing coaxial thermocouple data in the standard Tunnel 9 method. However, a second layer wasadded to the heat-transfer model representing the temperature-sensitive coating on the model surface. Additionalfactors that differ from methods outlined above include tunnel startup time (on the order of 200 ms), which is non-negligible. Heating profiles during startup resemble a ramp since the model is located in the flow (not injected)while the facility is started. This means that a step change in heat-transfer rate cannot necessarily be assumed as inthe cases of short-duration hypersonic facilities or model injection into the flow. Furthermore, it is desired toeventually acquire TSP data while dynamically pitching the model during a single run. As a result, the heating inputto the model for this application is by nature unsteady since the heating profile is a function of angle of attack. Otherfactors that must be accounted for in the data reduction include the effects of nonlinear thermal conductivity of thepaint formulation over the range of temperatures encountered at Tunnel 9 and paint layer thicknesses ofapproximately 2 mils (approximately 51 ptm). This thickness, which is larger than that used by most facilities, isdesired in order to increase paint emission so that good signal-to-noise ratios can be obtained. It is currentlyconsidered too thick to be ignored in the heat-transfer modeling.

II. Experimental SetupThe test article was a 7-in.-diam model of the NASA CEV capsule constructed out of 15-5 stainless steel. The

geometry is similar to that of the Apollo capsules flown in the 1960s. During the Tunnel 9 tests, the pitch angle wasfixed at 28 deg for all TSP runs. The heat shield and aft cone of the model were coated with temperature-sensitivepaint.

The physical setup of the model, TSP system, and test cell is sketched in Fig. 4. Three illumination sources werelocated on top of the test cell, and there were two on the side to ensure the entire model surface was illuminated asevenly as possible, with sufficient radiant intensity to provide 0.75-to-0.80-percent full-well potential, or maximumlight capacity, of the CCD cameras. Both of the CCD cameras were initially mounted on top of the test cell toprovide images of the heat shield (CAMI) and aft body (CAM2). Later in the program one camera was relocated tothe side of the test cell in an attempt to map the flow over the side of the aft cone. The test conditions and camerasettings for each run of the NASA CEV test are listed in Table 1.

Similar to TSP systems used in other research facilities, the system developed for use in Tunnel 9 consistsessentially of four main components: an illumination system, a detection system, the temperature-sensitive coating,and the data-processing algorithms. A brief description of each system component is presented below.

5

CEV Test Article

Test Cell

Light 4 Light 5

Top View

CAM 1

CAM 2

Light 2Light 1

Heat Shield

Flow

Aft Cone

Side View

Fig. 4. Schematic of lights and cameras for NASA CEV test in Tunnel 9.

Table 1. Run matrix.

Unit Nominal Binning of the CCD Frame Rate, fps ExposureRun M/,0 Re, Good Flow {aTime

1061ft (s) Heat Shield Aft Cone Heat Shield Aft Cone (ms)(CAM 1) (CAM 2) (CAM 1) (CAM 2)

1 10 5.00 0.4-1.3 4x4 4x4 61 61 1.92 10 10.00 0.25 - 0.8 2 x 2 1 xI 42 25 1.93 10 5.00 0.4 - 1.3 2 x 2 2 x 2 42 42 1.94 10 10.00 0.25 - 0.8 2 x 2 2 x 2 42 42 1.95 10 5.00 0.4 - 1.3 2 x 2 2 x 2 42 42 1.9

A. System Components and ConfigurationPhoton Technologies 200W mercury-xenon arc lamps were chosen as the optimal illumination source based on

their superior stability, intensity, and operational qualities assessed in the study described in Ref. 1. To providesufficiently intense and near-uniform lighting, five lamps were used to illuminate the model. The lights were filtered

6

Il

with a broadband bandpass filter centered at 365 nm. To protect the paint from potential photodegradation by theultraviolet illumination, the light from each lamp was blocked during periods in which the model did not need to beilluminated. The output of the lights was monitored by photodiodes to ensure stable output for the duration of eachrun. The lamps were operated continuously, i.e., not flashed.

Two PI/Acton PhotonMax 512B cameras were used to enable acquisition of the continuous, high-quality imagesrequired during each run. Per the manufacturer's description, the 512Bs are low-noise CCD cameras with on-chipmultiplication gain via electron multiplication CCD (EMCCD), which multiplies photoelectrons by an impactionization process prior to readout. The 512B features a 512 by 512 pixel CCD array. As a result of pretestanalytical comparisons, it was thought that these cameras would be capable of providing the short exposure timesand high frame rates necessary to obtain surface temperature data of suitable quality for calculating heat transfer inTunnel 9.

The camera exposure time was set to 1.9 ms for all of the runs. A variety of camera settings were tested,including 2 by 2 and 4 by 4 pixel binning to increase the frame rate. For example, with the full 512 by 512 CCDarray, the effective maximum frame rate for the 512B (including exposure time and readout rate) is 25 fps. Pixelbinnings of 2 by 2 and 4 by 4 resulted in rates of 42 and 61 fps, respectively. The required frame rate for heat-transfer calculations will vary depending on test conditions and is currently under study. For example, for a statictest (no pitching) with relatively low heating rates, the required frame rate may be as low as 30 fps.

The temperature-sensitive coating used in this test was developed and applied to the test article by LeaTechLLC. Application of the TSP to the test articles used in Tunnel 9 consists of airbrushing a white basecoat directly tothe model surface and a temperature-sensing layer on top of the base coat. Both layers were kept as thin as possiblewhile still creating a uniform coating. The coating thickness was measured using a magnetic induction probe. Onehundred measurements were made on the heat shield and aft body to assess the uniformity of the coating for thepaint job. The average paint layer thickness was found to be 2.1 mils (approximately 53 pim) with a standarddeviation of 0.15 mil. The white basecoat is used to create specular reflection of the excitation light through thepaint layer and thus increase paint emission intensity. It is important to note that the basecoat is not used to create an

718insulating layer, as is done in other TSP systems used with metallic wind tunnel models in other facilities.The TSP formulation used for the NASA CEV test utilizes a Europium complex as the temperature-sensitive

luminophore. This paint formulation has a broad absorption spectrum (relative to Europium alone) with excitationcentered at 365 nm. This formulation's emission is centered at 614 nm.1 Utilization of Europium gives thisformulation very good temperature sensitivity, on the order of tenths of a degree Fahrenheit. This lumiphore iscombined with a high-temperature polyurethane developed for use in Tunnel 9 that easily withstands temperaturesup to 360°F. Moreover, there is no uncertainty associated with the paint's acting as a pressure sensor via oxygenquenching since the facility uses nitrogen as the working fluid.

The paint was applied over the majority of the coaxial thermocouples that were included in the model to measureheat transfer at discrete locations during the non-TSP runs of the test program. However, a few thermocouples wereleft unpainted on the heat shield for comparison with symmetrically located painted thermocouples. The locations ofpainted and unpainted thermocouples are indicated in Fig. 5. Black circles were added to the left side of the pictureto indicate the locations of the painted thermocouples. The black dots on the surface of the paint are the registrationmarks used to align the images in the image-processing software.

III. Methodology for the Evaluation of Heat TransferIn general, the algorithm used to calculate the heat flux from TSP data at Tunnel 9 is based on the same analysis

as that applied to reducing coaxial thermocouple temperature data into heat transfer and is driven by the transientnature of the facility. In essence, a time history of the surface temperature is applied as a boundary condition in atransient, ID, heat-transfer conduction model. This model employs a second-order, Euler-explicit, finite-differenceapproximation method to solve the transient ID heat equation to obtain a ID temperature distribution at nodes atvarying depths through a steel model wall of finite thickness at each time step of the algorithm. The localconvective heat-transfer rate is calculated based on Fourier's Law using a second-order derivative approximation ofthe temperature profile at the model's surface. At the beginning of the run (initial condition), the model is assumedto be at a uniform initial temperature. Zero heat transfer at the back wall inside the model is the remaining boundarycondition required to solve the equation numerically. This assumption has been validated for the thick-walledmodels (0.375 in.) that are typically tested at Tunnel 9. The calculated heat-flux uncertainty from coaxialthermocouple data using this approach is quoted ±6 percent for fully laminar or fully turbulent regions. A detaileddescription of the coaxial thermocouple data reduction methods used at Tunnel 9 can be found in Ref. 9.

7

To develop an analogous data reduction methodology for evaluating heat transfer using TSP data, a second layercomprising the temperature-sensitive coating was added to the ID heat-transfer model. In reality, the temperature-

Fig. 5. Heat shield of the CEV model painted with temperature-sensitive coating.

sensitive coating consists of two layers: the base coat and the active layer. However, the two layers can be treated asone in the data reduction algorithm since they are made of the same host matrix material.

The numerical model is represented schematically in Fig. 6. In this case the TSP data become the input boundarycondition (T1) at the surface of the model. Then, the heat-flux balance at the interface between the two materials (theTSP and the model wall material) is enforced using Fourier's law of conduction [Eq. (1)]. One additionalassumption is made to simplify the algorithm: the temperature gradient through the paint layer is assumed to belinear. This assumption allows for a very simple discretization of the heat-flux balance equation at the interface ofthe two materials [Eq. (2)].

qg' = -K(8T/Ix)

T - T T - TK =

L 21 L 2 Ax

(1)

(2)

Moreover, this assumption eliminates the need for knowledge of the overall thermal diffusivity of thetemperature-sensitive coating. However, thermal conductivity (K) of the coating material still needs to be known.Currently, lab measurements of K are being conducted. In the interim, a method for estimating K usingthermocouple data available from the NASA CEV test is described in the Results and Discussion section of thispaper.

Subsequently, the temperature history at nodes through the metal model wall and the local heat transfer at thesurface are found using the same numerical method as that described above for coaxial thermocouples. The one-dimensional, transient heat equation [Eq. (3)] is solved numerically using second-order, Euler-explicit, finite-difference approximation [Eq. (4)]. Once the temperature distribution through the model wall is known, Fourier'sLaw of Conduction [Eq. (5)] is applied at the surface of the model to calculate convective heat transfer usingsecond-order derivative approximation of the temperature profile at the model' s surface [Eq. (6)].

8

8T/It = u(82T/Ix') (3)

T. T. (T.. 2T..+T..1+ j i,jj i j+1 i i ii (4At Ax2

K20C2 =-

P2.CP2

q" = -K(8T/8x)surf (5)

q"= 1-3LTil I+ 4K(Ti+1, + Ti+1,2) Ti+12J (6)

lqA ,T TSP:

L =_ K1, Pi, Cpl

Ax TiSteel:41 K$F T2, P2, Cp2

411Ts|

Fig. 6. Schematic representation of the numerical model.

The linear temperature gradient through the paint layer assumption is justified via an analysis performed bymodeling the problem in ID using ANSYS, a powerful, commercially available finite element modeling tool forstructural and thermal analysis. It was used extensively throughout the modeling process as a means of datareduction algorithm development and validation. The ANSYS simulation was designed to closely represent theactual test article and test conditions: 0.375-in.-thick stainless steel model wall (200 nodes) painted with a 0.002-in.-thick paint layer (six nodes). A 0.002-s time step was used to meet the convergence criteria. The actual heatingprofile experienced during the TSP test and representative thermal properties for the temperature-sensitive coating(based on Ref. 10) were applied as inputs to the ANSYS model. The results are depicted in Fig. 7. The locationcorresponding to x = 0 in. represents the interface between the steel model wall and the temperature-sensitivecoating, and x = 0.002 in. represents the surface of the coating exposed to the flow. The curves in the plot are thetemperature profiles through the paint layer at different times during the simulation. As can be see from thissimulation, the temperature gradient is linear at lower temperatures and can be closely approximated as linear athigher temperatures. The benefit of assuming that the temperature gradient through the paint layer is linear isapparent above.

This methodology for evaluating heat flux offers a simple way of calculating convective heat transfer from TSPdata in transient situations where heat conduction into the test article can be considered one-dimensional, thetemperature-sensitive coating is too thick to be ignored, and material properties of the coating, such as thermalconductivity and thermal diffusivity, are functions of temperature. While a ID heat conduction model may seemlimiting, it can be applied to many practical geometries tested at Tunnel 9 and other hypersonic facilities. In fact, theheat-transfer models used in Refs. 4 through 8 all assume ID heat conduction through the model wall. Nevertheless,

9

"I "I

a technique for solving more complex 2D and 3D conduction problems is currently being developed for areas ofhigh radii of curvature and thin walls.

1.30

t=G s

t=1 sO t=1 4 s

so~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ E110|1) 1

U 32z Bid OBiB Bi 12 IA 16 18i 2

Fig. 7. ANSYS simulation of temperature gradient through the temperature-sensitive coating.

IV. Results and Discussion

A. Processing of TSP imagesReference (wind-off) and run (wind-on) images for surface temperature data are acquired using PI/Acton's

Winview32 software. Appropriate prerun and postrun images are acquired and archived. Reference and run rawimage data are mapped to a 3D grid, ratioed, and converted to engineering units by using the Greenboot software."Surface temperature data are output globally from Greenboot and interrogated using Matlab for reduction to heattransfer.

Figure 8 illustrates the time history of ratioed intensity, which is inversely proportional to paint surfacetemperature, acquired using the TSP system at a location coincident with a thermocouple installed in the CEV testarticle. The freestream Reynolds number is also plotted in green to give the reader an indication of the tunnelconditions that the model experiences during the course of the run. Each blue symbol represents an image takenduring the run. The curve fit of the ratioed intensity history, shown as the black line in Fig. 8, was used as an inputto the heat-transfer data reduction algorithm to obtain local heat flux. The scattered data points in the beginning ofthe run (before the "good flow" period) are not representative of the actual paint emission during the startup. Theextreme drop in recorded intensity is caused by a condensation cloud passing through the test section andobstructing optical path. In order to correct for this corrupted data at the beginning of the run, temperature profilesfor TSP were inferred using paint material properties and thermocouple data underneath the paint at this location.

Figure 9 illustrates when the TSP images were acquired compared with data acquisition fromthermocouples. The red line represents the surface temperature rise recorded by the thermocouple. Note that thethermocouple output is captured at 500 samples/s. Each gray bar corresponds to a TSP image acquired during therun, where the width of the bar represents the camera exposure time and the spacing between the bars represents thecamera frame rate. It can be seen that the temperature change during each exposure is insignificant (i.e., it isreasonable to assume that the temperature captured by each frame represents an instantaneous reading at the time ofthe frame exposure.) It is also evident that the 42-fps frame rate is sufficiently high to resolve the heating rateencountered during this test.

B. Global Temperature EvolutionThis paper is not intended to summarize the results of each run of the program. Instead, the goal is to

present the data reduction methodology. Therefore, only results from Runs 3 and 5 are presented in the following

10

sections. Sequences of high-quality images of the heat shield and the aft cone that show the evolution of the heatingprofile in time were acquired during each run. Images of the heat shield from Run 5 are shown here (Fig. 10) in

Fig. 8. TSP ratioed intensity data at a single location on the heat shield(blue symbols) and freestream Reynolds number (green line) for Run 3.

|| Frame Rate

Thermocouple Surface TempCamera Exposure

Exposure Time

L~~~~~~~~~~~~~~~~~~~~~~~~~~75-

70

-0.2 0 0.2 0.4 0.6 0.8Time, s

1 1.2 1.4 1.6

Fig. 9. Time history of TSP image acquisition for Run 3.

order to demonstrate the capability of the composite Tunnel 9 TSP system during the good flow period. For eachimage, the time at which it was acquired is shown in the figure. The reader should note that even though theseimages are acquired over a very short exposure time (1.9 ms) and relatively high frame rate (given good gray-scaleresolution of the 14-bit A/D converter), the image quality and signal-to-noise ratio are very good. This is attributed

11

100

95-

90

6. 85

80 _

to the high quantum efficiency of the paint and the high intensity level of the illumination system. As expected, theoverall temperature of the model surface shows an increase during the run because of aerodynamic heating.

0.20 s

1.10 S 1.4U S

70°F 160°F

Fig. 10. CEV heat shield; TSP images during Run 5, Mach 10, Re = 5.0 x 106 /ft.

The dark blue color on the left-hand side of the images is an area out of the field of view of the camera; thus noTSP data are available. The two unpainted thermocouples in the top right quadrant of the heat shield are also blue,

12

indicating that no TSP data are available in those locations. The deepest red color appearing in the lower right cornerof the heat shield in some of the images does not represent aerodynamic heating; it is caused by overlapping of theillumination from two light sources on the model surface, which saturated the CCD array. The streaks appearing insome of the images result from increased localized heating induced by particles impacting the model surface andeffectively developing discrete roughness elements in the paint layer, which in turn disturbs the boundary layer andcreates localized higher convective heating.

C. Estimation of the Thermal Conductivity of the Temperature-Sensitive CoatingThe data reduction algorithm outlined requires knowledge of the temperature-sensitive coating's thickness as

well as its thermal conductivity, K. To determine the effect of the coating's thickness and thermal conductivity oncomputational results, a sensitivity analysis was performed using the same ANSYS 1D transient heat conductionmodel described above to assess the effect of perturbations in thermal conductivity and paint layer thickness on thecalculated heat-transfer rate. From this simple sensitivity analysis it was found, for example, that underestimatingthe thickness of the temperature-sensitive coating by 50 percent resulted in an approximately 35-percent error incalculation of heat transfer from the paint, while overestimating the thickness by 50 percent resulted in anapproximately 21-percent error. This nonlinear trend in the calculated heat flux error with uncertainties in paint layerthickness is consistent with experimental results presented by Ohmi et al.6 As mentioned previously, the thicknessof the paint was measured on the actual model itself. Based on the standard deviation of 0.15 mils for the 2.1-milthickness, the paint's thickness is known within 7.5 percent. This gives an estimated error in calculated heat-transferrate based only on uncertainty of the paint thickness of about 4 percent.

Also, the study showed that the error in calculated heat flux increased linearly with increasing error in thermalconductivity, K: a ±10-percent error in K resulted in approximately ±5-percent change in calculated heat flux.Similarly, a ±20-percent error in K resulted in an approximately ±10-percent change in calculated heat flux. Theabove analysis indicates that the thickness of the coating cannot be ignored and that the thermal conductivity of thecoating applied to a test article in Tunnel 9 cannot be assumed constant for heat-transfer calculations. Though thepaint layer thickness can be measured directly, the measurement of thermal conductivity of a polymer-based coatingis nontrivial.

A method was developed to estimate the value of thermal conductivity, K, of the coating as a function oftemperature using thermocouple and TSP data available from the test. To estimate the thermal conductivity of thetemperature-sensitive coating used in the test, temperature data from two pairs of standard coaxial thermocouples onthe heat shield of the CEV model located symmetrically about thevertical plane on the left- and right-hand sides of the model were used Stas follows. During the TSP runs, two thermocouples on the right-handside of the model were left unpainted, while their symmetric 4',counterparts were painted along with the rest of the model, as shown in tspFig. 5. A pair of these symmetrically located thermocouples wasmodeled as is graphically represented in Fig. 11, where Tl_tsp is the K1(T) T-tsp Tl_stTSP temperature data over the painted thermocouples; T2-ts iS the K2(T)painted thermocouple data; T, st is the unpainted thermocouple data;

T2st iS the 1-D heat conduction finite-difference model calculation;K1(T) is the thermal conductivity of TSP; K2(T) is the thermalconductivity of the model material (stainless steel): St is thenondimensionalized heat input; Ax is the node size through the modelwall, and L is the paint layer thickness. The Stanton number wasassumed to be equal at the two symmetrically located points on themodel, points that corresponded to the painted and unpainted Fig. 11. Graphical representation ofthermocouples [Eq. (7)]. A linear temperature profile through the paint two symmetrically located thermo-layer was assumed once again, thus allowing the use of the discretized couples. One is painted with TSP, andFourier's Law of conduction at the surface [Eq. (8)] in the same way as the other is unpainted.in the heat-transfer data reduction model [Eqs, (9) and (10)].

Stl = St2 (7)

qdot = -K (dT/dx)surf (8)

13

T - Tqdot =K ltp 2s

tsp 1 L (9)

T1 st T2 stqdot Ax (10)

From the Stanton number definition:

St= qdot and Eqn. 1pu(H0- C T)

qdot st qdot tsp

pu(H- C TISt) pu(HCHTlCp) (11)

Substituting Eqs. (9) and (10) into Eq. (11) and solving for K1 yields Eq. (12):

( T T NJH -CT >K KLi 1-St 2.st ii p Itspl

1 2AxITTIT H - C It (12)

The linear curve fit of the time history of ratioed intensity (or temperature) at the location of a paintedthermocouple shown in Fig. 8 in conjunction with the data from this thermocouple were used to estimate the thermalconductivity (K) of the paint, as described above. The resulting K1(T) estimate agrees reasonably well with thethermal conductivity of polyurethane-based synthetic enamel paint measured by Paul et al.,'0 as shown in Fig. 12.Note that the current estimate for thermal conductivity extends the temperature range to lower temperatures than dothose measured by Paul et al. It is observed that there is a strong gradient in K as a function of temperature at theselower temperatures. This again points to the fact that K cannot be assumed to be constant for heat-transfercalculations at Tunnel 9 since the models are initially at room temperature and only reach higher temperaturescorresponding to the more "level" part of the K curve toward the end of the run.

Fig. 12. K comparison for the two polyurethane-based paints.

14

The same linear curve fit of the time history of ratioed intensity used to estimate the thermal conductivity of thepaint (Fig. 8) was then used as the input boundary condition to the heat-transfer algorithm as described in Section IIIof this paper. The convective heat flux calculated from the paint data is nondimensionalized as a Stanton numberand is compared against the heat-flux value calculated from the sylmmetrically located unpainted thermocouple. TheStanton numbers normalized by the stagnation point values are shown in Fig. 13. The Stanton number plot showsonly the "good flow" period of the test run. The Stanton number is constant during the "good flow" period of the runsince the model is at a constant angle of attack throughout the run. The red line represents the normalized Stantonnumber calculated from thermocouple data using the standard data reduction technique, and the green line representsthe corresponding values calculated using the current ID algorithm for TSP data reduction. The values are expectedto be in good agreement since the thermocouple data were used in the thermal conductivity estimation as describedabove and hence were part of the "calibration" process. However, the ±3.5-percent agreement between the twocurves does tend to lend credibility to the overall numerical algorithm used for heat-flux calculation. To furthervalidate the methodology, the analysis will be applied at other points where limited thermocouple data are available.For instance, comparisons of laminar and turbulent heating rates along the centerline of the model are to be madewith thermocouple data and CFD. (As it happens these comparisons are being conducted concurrently with thepublication of this paper.)

TI PL----- T ermocoup e

2 6 --- - -- - --

l - l - -

A2-

XR ~~~~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~~ L ----- L- , , ,

I i I Dig Dig 1 11 12 13TDime (s

Fig. 13. Normalized Stanton numbers calculated from thermocouple and TSP data.

V. SummaryTSP surface temperature data acquired on the NASA CEV model at AEDC Hypervelocity Wind Tunnel 9 were

used for the development of a methodology for measuring global heat transfer. To develop this methodology, theCEV model was painted with TSP, and multiple high-quality images were acquired during five runs in the testprogram. TSP images acquired during the program successfully demonstrate the ability of the complete Tunnel 9TSP system (illumination, camera, paint layer) to measure the response of the paint to temperature changes at framerates necessary to calculate heat flux. A data reduction methodology to calculate convective heat transfer on thesurface of the model has been developed. Some of the assumptions of this methodology include transient IDconduction, accounting for the TSP layer thermal capacitance, and calculating heat flux during the startup process. Atemperature-dependent value for thermal conductivity of the temperature-sensitive coating was estimated using TSPand thermocouple data. The estimated value of thermal conductivity was then used to calculate the heat flux fromthe TSP surface temperature data at one location on the model to validate the overall computational scheme. Theheat flux calculated from TSP data correlates well with the value obtained from thermocouple data.

15

Further development of this methodology will include direct determination of material properties of thetemperature-sensitive coating via lab testing and finalizing the computational algorithms to reduce TSP data over theentire surface of a test article. In addition, an estimate of the system's noise and resulting uncertainty in calculatedconvective heat flux will be made. An approach to dealing with 2D and 3D heat-transfer problems is also to bedeveloped.

AcknowledgmentsThe authors would like to acknowledge the financial support of NASA (Randy Lillard of JSC and Tom Horvath

of LaRC) and the dedication of the Tunnel 9 team before, during, and after this effort. Without their ongoingsupport this work would not have been possible. As always, the assistance and technical support of Marvin Sellersof AEDC is graciously appreciated.

References'Norris, J., Hamner, M., Lafferty, J., Smith, N., and Lewis, M., "Adapting Temperature-Sensitive Paint Technology for Use

in AEDC Hypervelocity Wind Tunnel 9," AIAA Paper 2004-2191, 24th AIAA Aerodynamic Measurement Technology andGround Testing Conference, Portland, OR, 28 June- 1 July, 2004.

2Ragsdale, W. C. and Boyd, C. F., Hypervelocity Wind Tunnel 9 Facility Handbook, Third Edition, NAVSWC TR 91-616,Silver Spring, MD, July 1993.

3Rabek, J. F., "Mechanisms of Photophysical Processes and Photochemical Reactions in Polymers," Wiley, New York, 1987.Chap. 1.

4Buck, G., "Surface Temperature/Heat Transfer Measurement Using A Quantitative Phosphor Thermography System,"AIAA-91-0064, 29th Aerospace Science Meeting, Reno, NV, January 1991.

5Nakakita, K., Osafune, T., and Asai, K., "Global Heat Transfer Measurement in Hypersonic Shock Tunnel UsingTemperature Sensitive Paint," AIAA Paper 2003-743, 41st Aerospace Sciences Meeting & Exhibit, Reno, NV, January 2003.

6Ohmi, S., Nagai, H., and Asai, K., "Effect of TSP Layer Thickness on Global Heat Transfer Measurement in HypersonicFlow," AIAA Paper 2006-1048, 44th Aerospace Sciences Meeting & Exhibit, Reno, NV, January 2006.

7Hubner, J., Carroll, B., and Schanze, K., "Heat Transfer Measurements in Hypersonic Flow Using Luminescent CoatingTechniques," AIAA 2002-0741, 40th AIAA Aerospace Sciences Meeting & Exhibit, Reno, NV, January 2002.

8Matsumura, S., "Streamwise Vortex Instability and Hypersonic Boundary-Layer Transition on the Hyper-2000," Master'sThesis, School of Aeronautics and Astronautics, Purdue University, West Lafayette, IN, Dec 2003.

9Boyd, C. F., and Howell, A., "Numerical Investigation of One-Dimensional Heat-Flux Calculations," Dahlgren Division,Naval Surface Warfare Center, Silver Spring, MD, 25 October 1994.

'"Paul, K. C., Pal, A. K., Ghosh, A. K., and Chakraborty, N. R., "Thermal Measurements of Coating Films Used for SurfaceInsulation and Protection," Surface Coatings International Part B: Coatings Transactions, Vol. 87, B2, 71-148, June 2004.

"The Boeing Company, Greenboot User's Guide Version 2.10. Contract NASI-20268, NASA, 1997.

16