S AIAA 2003-3764 A Thermal Analysis Approach for the Mars Odyssey Spacecraft's Solar Array John A. Dec and Ruth M. Amundsen NASA Langley Research Center Hampton, VA 36th AIAA Thermophysics Conference 23-26 June 2003 Orlando, Florida For permission to copy or to republish, contact the copyright owner named on the first page. For AIAA-held copyright, write to AIAA Permissions Department, 1801 Alexander Bell Drive, Suite 500, Reston,VA, 20191-4344. A THERMAL ANALYSIS APPROACH FOR THE MARS ODYSSEY SPACECRAFT’S SOLAR ARRAY John A. Dec and Ruth M. Amundsen National Aeronautics and Space Administration Langley Research Center Hampton, VA 23681-2199 ABSTRACT axis stabilized spacecraft to perform this type of multipass aerobraking1. Mars Global Surveyor, the There are numerous challenges associated with placing second to use multipass aerobraking, gradually reduced a spacecraft in orbit around Mars. Often, trades must its orbit from an elliptic 45-hour period to about a be made such as the mass of the payload and the 380km 2-hour circular orbit. amount of fuel that can be carried. One technique employed to more efficiently place a spacecraft in orbit To control periapsis altitude, Magellan planned on while maximizing payload mass (minimizing fuel use) using thermocouple data to signal the need to perform a is aerobraking. The Mars Odyssey Spacecraft made use periapsis raise maneuver, which would raise the of aerobraking to gradually reduce its orbit period from periapsis altitude, lower the maximum atmospheric a highly elliptical insertion orbit to its final science density experienced, and thus lower the aerodynamic orbit. Aerobraking introduces its own unique heating on the spacecraft and solar arrays. In the challenges, in particular, predicting the thermal literature, it was noted that Magellan experienced at response of the spacecraft and its components during least 5 thermal sensor failures prior to the start of each aerobraking drag pass. This paper describes the aerobraking2. It is unclear from the available literature methods used to perform aerobraking thermal analysis if any of these thermal sensors were located on the solar using finite element thermal models of the Mars arrays and were to be used during operations. It is also Odyssey Spacecraft’s solar array. To accurately model unclear whether or not there were any thermal models the complex behavior during aerobraking, the thermal developed and used during the Magellan aerobraking analysis must be tightly coupled to the spatially process to make up for the inoperable thermal sensors, varying, time dependent aerodynamic heating analysis. but there are references to heat rate and surface Also, to properly represent the temperatures prior to the temperature calculations being performed using a direct start of the drag pass, the model must include the orbital simulation Monte Carlo particle method3,4. In any solar and planetary heat fluxes. It is critical that the event it is clear that in the early 1990’s the limitations thermal behavior be predicted accurately to maintain of computers would have prohibited the use of a the solar array below its structural flight allowable sufficiently detailed finite element or finite difference temperature limit. The goal of this paper is to describe model that could have been run in a timely enough a thermal modeling method that was developed for this fashion to be used during operations. During the Mars purpose. Global Surveyor operations, the heat rate encountered during previous drag passes was reconstructed using a INTRODUCTION 1-dimensional thermal model at various locations around the solar arrays. This model used the spacecraft and solar array thermal sensor temperature data as input The Mars Odyssey spacecraft was launched on a Delta to determine what heat rate the spacecraft and solar II launch vehicle from Cape Canaveral Air Force Station on April 7th 2001. On October 23rd 2001, after a arrays experienced. A thermal model to predict the solar array temperatures for future drag passes and to 197-day cruise, the spacecraft performed a propulsive reconstruct the solar panel temperatures for past orbits maneuver to insert itself into an 18.5-hour elliptic orbit was not available. Originally, during operations, the around Mars. To place itself into its 2-hour, 400km plan for Mars Odyssey was to use a 1-dimensional circular, sun-synchronous mapping orbit, Odyssey used thermal model similar to that of the Mars Global the multipass aerobraking technique, the same Surveyor model. Unlike Magellan and Mars Global technique utilized by the Magellan spacecraft around Surveyor, Mars Odyssey was the first multipass Venus and the Mars Global Surveyor spacecraft around aerobraking mission to make use of detailed 3- Mars. Aerobraking makes use of atmospheric drag on dimensional finite element thermal model during each orbit pass to gradually reduce a spacecraft’s operations to predict the temperatures for future orbits velocity at periapsis, which then reduces the altitude at and reconstruct the solar panel temperatures for past apoapsis. The Magellan spacecraft was the first three- orbits. This model, used in addition to the 1- while the mid panelhas a unique geometry. Figure 1 dimensional model, provideddetailed,3-dimensional showsa 3-dimensional geometricrepresentation ofthe temperature profilesof the solar array, transient plotsof solar array. This geometric representation was maximummaterial temperatures, and transient plotsof developed inMSC/PATRAN and wasused in the thermal sensor temperatures. It alsoprovided the aerobraking thermal analysis. Eachpanel consists of ability to identify thehottest spots on the solar array, five layers; the first three make up the structural and provided a means to develop thermal limits based components andprovide the structural integrity during onheat rate or atmospheric density. launch and throughout aerobraking. Specifically, the solar array structure is a sandwich construction with a THERMAL MODEL DEVELOPMENT 0.190mm (0.0075”)facesheet of M55J/RS-3 graphite composite, a 19.05mm (0.75”)5052 aluminum honeycomb core, and another 0.190mm (.0075”) The thermal analysis of the solar array had three main M55J/RS-3 composite facesheet. components: the view factorand orbital heating analysis, the aerodynamic heating analysis, and the computationof solar array temperatures. The calculations were performedintwo flight regimes. One regime was the vacuum phase where the spacecraftwas in orbit around Mars,butoutof the atmosphere. The thermalenvironmentinthis phase was dominated by the solar and planetaryheating withnegligible aerodynamic heating. The secondregime was the aerobraking phase, or drag pass,where the spacecraft made its excursion into the atmosphere. The thermal environment in thisphase was dominated bythe aerodynamic heating (roughly60 times greater than the orbital heating). Temperatures were calculatedforboth flightregimes. Accurate calculationof the Figure 1. 3-D geometric representation of the Mars solar/planetary fluxduring flight,as well as Odyssey solar array. aerodynamicheating during drag passes,requires that the thermal analysis be tightly coupled to the flight Thenext layeris a 0.051mm (0.002”)Kapton sheet that mechanics, aerodynamics, and atmosphericanalysis. is co-cured to the M55J/RS-3graphite beneath it. The next layer is the solar cell layer and is made up of Solar ArrayDescription severalsublayers: 0.190mm (0.0075”)Gallium Arsenide solarcells, 0.152mm (0.006”) cover glass, and The function ofMarsOdyssey’s solar array is two-fold. 0.229mm (0.009”) of adhesives6. Figure2 shows a First, its primary function is toprovide a stable power cross-sectional view of the solar array at a sourcefor the spacecraft andscientific instruments representative location. during all phases of the mission. Second,while in the stowed configuration, the solar arrayprovides a large surface areaon which the aerodynamic forces can act during eachdragpassof aerobraking. Following Mars orbit insertionthe Odyssey spacecraftbegins aerobrakingby slicing through the Martian atmosphere at a relativevelocity of about4.57 km/s. Despite low atmosphericdensities of about 80kg/km3, these high velocities, along with thelarge exposed surface area, produce significant aerodynamic heatingon the solar array, and spacecraft. Careful design and construction of the solar array allows it to withstand this aerodynamic heating.6 Figure 2.Cross-section of Mars Odysseysolar array The Mars Odyssey solar array is a three panel, layered Toreinforce certain areasof the array,sectionsof the constructionof low density materials. Thepanels are standard 19.05mm (0.75”) thick,1.0 lb/ft3 aluminum commonly referred to as the +X,-X, and mid panels. honeycombcore were replaced with a higherdensity The +X and–Xpanels are mirror images of each other, core of the same thickness. Also, to reinforce the 1 American Institute of Aeronautics andAstronautics array’shard points,a doubler sandwichstructure was On the engineeringdrawings6, the thermal sensors are used. Thedoubler sandwichstructure consistedof designated T1,T2, T3, T4, and T5. T1 andT4 areon 0.381mm (0.015”) layersof M55J/RS-3graphite the +/-X panelon the exposed facesheet side. T2, T3, composite on either side of a18.161mm (0.7125”) and T5 are on the solar cell side; Figure1 from above aluminum honeycomb core. Thedoubler core densities shows the locationsof the thermal sensors on the solar varieddepending on the locationwithin thearray and array. the expected loading in the regionof concern. Examples of such reinforced hard points on thearray Orbital Heating Modeling are the hingemountingpoints. Overall, thearray designmakes use offour different densityaluminum Theorbital heating and viewfactor analysis,or honeycomb cores. Figure3graphically displays the radiation model, was performedusing Thermal locations of the different cores used and Table 1 lists Desktop, a commercially available software package.7 the density and thicknessof aluminum coreusedwith The radiation modelwas developed usingengineering respect to the colors in Figure 3. The table entries drawingsof the spacecraft and solar array.6 View where the corethicknessdeviates from the nominal factor and heat rate calculationswere performed for 0.75” thickness also show the locationsof the M55J severaldifferent solar array configurations and doublers. Thedoubler thickness is such that it brings spacecraftorientations. First, calculations were made the sectionback to thenominal panel thickness. with the spacecraft and solararray in its vacuum phase configuration; the spacecraft orientedwith its high gain antennapointed towards Earth and thesolararray normal to the sun. In transitioning to theaerobrake configuration, both heat ratesand viewfactors to space for the solar array were calculated as the solar array articulated to its stowed position.As the spacecraft slewed to the aerobraking configuration, thesolar array’s viewfactors to spacedidnot change, so view factors did not haveto be recalculatedforthat maneuver. Figures4 and5 show the spacecraft and solar array in Figure3. Variation in aluminum honeycomb core the aerobrake configurationandvacuum phase density (lb/ft3) respectively. Thispartof the analysis requireddetailed knowledge of the orbit andspacecraft orientation,and Table 1. 5056 Aluminumhoneycombcore thus was highly dependenton theflight mechanics analysis. Theorbital elements in thevacuum phase Marker Density(lb/ft3) Thickness(in) were obtainedfrom a mission trajectoryrun-out Red 1.0 0.75 performed by JPL8. For thedrag passphase, theorbital Yellow 2.0 0.75 elements wereobtainedfrom POST simulations and Green 3.1 0.715 outputin unit vector form. Blue 12.0 0.715 Toprotect thearrayfrom aerodynamic heating, a 0.072mm (0.00283”) thick layer of multilayer insulation (MLI)wasplacedon the facesheet exposed to the aerodynamic heating and around theedgesof the M55J/RS-3 graphitefacesheet. Thebulkof the MLI wason the facesheet surface,and a small portion wrapped around the edges and terminated on the solar cell side. Thewidth of the MLIon thesolarcell side ranged from 50 –148mm (1.96-5.83”). Velocity Vector The solar array has five thermal sensors. There isone sensoron themid panelon the solar cell sideof the Figure 4. Mars Odyssey Thermal Desktop model in array. The +X and –Xpanels each haveone sensoron the aerobrakeconfiguration. the solar cell side and one on the “hot”facesheet side. 2 American Institute of Aeronautics andAstronautics Againwith the spacecraftremovedfor clarity, Figure 7 shows a representative resultof the solar/planetary heating nearperiapsis. Figure 5. Mars Odyssey Thermal Desktop model in thevacuum phase configuration, view from Earth. Thevacuum phaseorbital elements were input into Figure7. Typical orbitalheat flux nearperiapsison Thermal Desktop via the “Keplerian Orbit” input the facesheet exposed to aerodynamic heating in the option. Thisorbital inputform requires theperiapsis aerobraking configuration.(W/m2) altitude,orbit eccentricity, right ascensionof the sun, After completing theorbital heating and view factor right ascension of the ascending node, and the argument calculations, this datahad tobe input into PATRAN of perigee. The vectoroutput from POST is input into Thermal for use in the temperature calculations. This Thermal Desktop byusingthe “Vector List” orbit process will be explained in detail later in the boundary optionwhere theuser supplies the time, the unitx, y, conditions anddata exchangesection. and z componentsof thevector from the probe to the sun, and the unit x, y, and z componentsof the vector Aerodynamic Heating Analysis from the probe to the planet. Theuser also supplies the ratioof thedistance to thespacecraftfrom the planet center and theplanetradius. With the spacecraft The aerodynamic heating analysis consistedof calculating theatmosphericdensity, the spacecraft removed forclarity,Figure 6 showsthe viewfactors calculatedwith the solar array in the stowed,or velocity relative to the atmosphere, and theheating aerobraking configuration. The view factorsshownare coefficient for points spatially across the array. Using this information, the incident aerodynamicheating was for the facesheet exposed to the aerodynamic heating; the solar cell sidehas anunobstructed view to space so calculated usingequation 1, the viewfactor is equal to 1.0. Notice that the spacecraftprevents most of the middle panel from 1 Q (cid:32) (cid:152)(cid:85)(cid:152)V3 (cid:152)C (1) seeing deep space (low viewfactor); fortunately, the i 2 H spacecraft also shields the middlepanel fromthe aerodynamic heating thus,preventing a potential where(cid:85)(cid:3)is theatmosphericdensity, V is therelative thermal problem due to inadequateradiative cooling. velocity, and C is the heat transfer coefficient. This H calculationwas made at 5-second intervals throughout the aerobrakingpass togivea transient representation of aerodynamic heating. A databaseof theheat transfer coefficientswas developedusing a Direct Simulation Monte Carlo(DSMC)particle method -- specifically the programs DAC andDAC free wereused. From the POST simulations, the flightmechanics team was able to provide theatmosphericdensity and the relative velocity as a function of time(during operations, to reconstructpast orbits, the density was calculated using acceleration readings fromthe flight accelerometer)9. Using thedensity and the relative atmospheric velocity, Figure6. View factors to spacefor the facesheet the heating coefficientswere calculated. Due tothe exposed to aerodynamicheating in the aerobraking longrun timesneeded toperform the DSMC configuration. computations,an aerothermodynamic database was 3 American Institute of Aeronautics andAstronautics developedprior to aerobraking. Then for each time coverglass, wire, solder, andadhesives thatwere step, thedensity was calculatedby interpolating on modeled as anaveraged mass spread evenly across the either the accelerometer data, or the calculateddensity layer. Thiswas an engineering approximation since the versus time outputfrom POST. Forboth cases, the wire, solder, etc.,were not evenly distributed across the heating coefficients across the array were calculatedby panels. The titanium hinges and dampers that interpolating betweendensity andrelative velocity. physically connect the three solarpanels were also The interpolation error on the heating coefficients was included in the model and provided a thermal link calculated to be about2%,whichwas within the between the threepanels. The model was highly accuracyof the DSMC calculations. Theheat transfer detailed and included the variations in the aluminum coefficients ranged fromapeak value of 0.90 toalow honeycomb coredensity as well as thevarying value of 0, andincluded the surface accommodation thicknessof M55J composite facesheet doublers. For coefficient. allmaterials, propertieswereincluded as functionsof temperature;for the aluminum honeycomb core and Reflectedheatis a phenomenon which occurswhen M55J facesheets, orthotropicmaterial properties were particles striking thesurfaceof the solar array leave the included. Thefive spacecraftthermal sensors were surfacewith more energy. This occurs when the modeled as bar elements whichhad mass and were surface temperatureof the solar array has increased thermally connected to the spacecraftwith acontact enough to impart some of its thermal energy to incident resistance. Overall, the PATRAN thermal model was a particles which are at a lower energy level. The physically accurate 3D representation of the solar array. thermal modelaccountsforreflected heat, which was Compared to the as-built mass of the solar array of 32.3 approximated empirically for this mission using kg, the mass of the PATRANthermalmodel was 33.4 equation2 andis a functionof the incident heat flux kg, a difference ofonly3.4%. and surface temperature. The PATRAN thermalmodelrequiredinput boundary conditions fromthe two analysis components (cid:170) (cid:167) Q (cid:183)(cid:186) (cid:167)T (cid:183) Q (cid:32)1(cid:85)V3(cid:171)0.015(cid:14)0.1(cid:168)1(cid:16) i (cid:184)(cid:187)(cid:152)(cid:168) wall(cid:184) (2) mentioned earlier. The model included radiation to ref 2 (cid:171)(cid:172) (cid:168)(cid:169) 12(cid:85)V3(cid:184)(cid:185)(cid:187)(cid:188) (cid:169)300(cid:185) space, incorporating the view factors that were calculated from Thermal Desktop. Theorbital and planetary heat fluxes calculatedfrom Thermal Desktop where T is the surface temperatureof thesolar array. wall were applied to the surfacesof the model as well. The Equations 1 and2reveal the couplingof the thermal aerodynamic heating and reflected heating boundary analysis to the flight mechanics, aerodynamics and conditionswere applied tothe exposed M55J facesheet atmospheric analysis. These calculationsprovided an and MLI surfaces. Heatingon the edges of thepanels accurate representation of theaerodynamic heating as was included as 5% of the incidentheatingof the nodes well as reflectedheating thatwas a functionof time as closest to the edge. Edgeradiationwas included around wellas position onthe solararray. the outer-most edgeswith a view factor to space of 1.0. Radiationback to the spacecraft bus was included,with Temperature Calculations the spacecraftbus simplymodeled as a node with a constant temperature. The temperatures for the solar array were calculated with the commercially available software Boundary Conditions andData Exchange MSC/PATRAN Thermal10. Like the Thermal Desktop model, a 3-Dmodel of the solar arraywasdeveloped The processof applying the aerodynamic heat loads, solely using engineeringdrawings6. Normally, it is less orbital heat loads,and integrating theviewfactorsfrom efficient to create two thermalmodels,however, since Thermal Desktop into PATRAN was developedfor existing FORTRAN codewould allow simple inclusion Mars Global Surveyor11. These processeshad to be of the aeroheatingfluxes in PATRAN andbecause modified slightly to accommodate the MarsOdysseys existing methods couldbe easily applied11, this solar array configuration andsimplified slightly to be of inefficiency was nullified. Also, the Thermal Desktop use during Odysseys’operationsphase. radiation model was necessary sinceorbital heating capabilities donot existwithin MSC/PATRAN. The processof applying the aerothermodynamic heat loads calculated by DAC to the PATRAN thermal ThePATRANmodel represented all three solarpanels, model involved the useof aMATLAB script and and includedbothfacesheet layers, the aluminum customized FORTRAN program that interfacedwith honeycomb core, theKapton, the MLI, andone layer the PATRAN solver at runtime. Incorporating the for the solar cells. The solar cell layer included the view factorsand orbital/planetary heat loadscalculated 4 American Institute of Aeronautics andAstronautics by Thermal Desktopdid not require the useof the the solution converges, the solver moves to thenext customized FORTRAN codeor the MATLAB script, time step and the process is repeated. Figure 8 shows but didrequire several preprocessingstepsin orderto the DSMC data mapped to the PATRAN modelwith import the data intoPATRAN. The MSC/PATRAN the final resultof equation1plotted. Thermal user’s guide10 describes howtoincorporate user suppliedFORTRAN subroutines, and therefore it willnot be discussed here. However,the processand flowofdatafrom the aerothermodynamic database to the PATRAN thermalmodel will bedescribed. As mentionedpreviously, an aerothermodynamic database wasdeveloped prior to the start of Odyssey’s aerobraking. Thisdatabase consistedof a spatialmap of the heat transfer coefficients over the array as a Figure8. Example of aerodynamic heating mapped function of atmospheric densityand relativevelocity. to the PATRAN model(W/cm2). The drag pass duration, which definedthelength ofthe transient in the thermalmodel, was providedby the flight mechanics team. Alongwith the drag pass Theprocess for importing Thermal Desktopdata into duration, theyprovided a time history of the PATRAN Thermal is not straightforward and is labor atmosphericdensity and relativevelocity, the time intensive. First, a nodal positionreport must be created interval ofwhich was five seconds. Priorto the from within PATRAN. This report isnothing more executionof the PATRAN thermal analysis, the than a textfile that contains the node number, X,Y, and MATLAB script was executed. The MATLAB script Z locationsof thenodes. Thenodes selectedfor read thedensity, andrelativevelocity data, as well as inclusion in this report are associatedwith the surfaces the aerothermodynamic database, and for each time in the thermal model that have theradiationboundary intervalperformed an interpolation overdensity and condition imposedon them. Since thephysical location relative velocity to obtain thespatial heating of the solar array withrespect to the coordinate system distribution. The MATLAB script’s outputwas a file is different in the PATRAN and Thermal Desktop that containedblocksof datawith the time interval as a models, some corrections to thenodal coordinates had heading for eachblock, and a table ofX, Y,and Q, to be made. These corrections were made in Microsoft whereX andY are the spatial coordinates ofthe DSMC Excel and saved as a new text file. Thedisconnect in grid and Q is the incident heat flux calculatedby the physical locationof thearray was due tothe fact equation 1. Uponexecution ofthe PATRANthermal that the spacecraftbuswas modeled in Thermal analysis, thePATRAN solver makes a call to the Desktop,but was not modeled in PATRAN. In customized FORTRAN subroutine. The subroutine Thermal Desktop, the originwas the centerof the performs an initializationby reading the MATLAB spacecraft, and in PATRAN, theoriginwas along the outputfile andstoring thedata in a three dimensional bottom edgeof the array at the midpointof the middle array. The first two indices for the array are the number panel. Once the correctednodal position report is ofX and Y points onthe DSMC grid. The thirdindex acquired, the Thermal Desktop model is runto solvefor is a time indexwith its maximum value equal to the the viewfactors to space and theorbital heat rates. numberof output time points provided in the Utilizing the internal interpolationroutine inThermal density/relative velocitydata. With this initialization Desktop called “map data to locations”, thePATRAN procedure complete the FORTRAN subroutine passes nodalpositionreport is supplied as input and Thermal back control to the PATRANThermal solver, where it Desktopproduces a textfile that includes the PATRAN begins its iterative solutionprocess. For every time nodenumber and the mappedThermal Desktop data. step in the solution and at every node location thatwas Thisfile then needs tobe imported into PATRAN as a flagged as having incident aerodynamic heating, the spatial field. This is accomplishedby using the PATRAN solver makes a call to the FORTRAN spreadsheetutility in PATRAN. To import thedata, subroutine. The subroutine thenperforms two simply start the spreadsheetutility in PATRAN, then interpolations. The first is an interpolation over time to using standardwindowsfile open procedures, open the obtain the heating at the correct time in the trajectory. ThermalDesktopdata file. The opening procedure The second interpolation is over X andY to obtain the loads thedata into the spreadsheet, thenan internal heating for thecurrentnode,sincethe x and y locations PATRAN spreadsheet functionisused to create a field for the PATRAN nodesdiffer from the DSMC grid from the spreadsheet data. Thefieldgenerated is a node locations. The FORTRAN codepasses thenodal finite element field and canbe used as the inputfor the heat rate back to the PATRAN solver and the solution “FormFactor” entryinthe gap radiation boundary process continuesuntil convergence is obtained. Once condition dialogbox. All of theradiationboundary 5 American Institute of Aeronautics andAstronautics conditions to space were created in thisway. Figure 9a determine the temperatures,where the aerodynamic shows the view factors to space in the Thermal Desktop heating wasdominant. Temperature predictions were model and Figure 9b shows theviewfactors as mapped made prior to eachdragpass and temperatures were to thePATRAN model. Asimilar procedure is calculatedusing actual spacecraft telemetry to followed to import the heatrate data from Thermal reconstruct pastorbits. Desktop to PATRAN. Thisprocedurebecomes cumbersome if theheat rates varygreatly over anorbit, TEMPERATURE PREDICTIONS or if the viewfactors change,as is the casewhen the spacecraft slews and stows the solar array prior to a The initial goal of the thermal analysis was to provide dragpass. To include these changes, the transientprior an independent validation and verification of the to thestartof thedragpass had to be brokeninto stages. analysis already beingperformed on the solar array. Each stagewas solved independently and each This includedverifying the location and magnitudeof successive stage used theprevious stage’sfinal predictedhot spots on the array, andverifying thesolar temperatures as its starting temperature. The only array’s thermal limit. Thethermal limit is expressed in exceptionwasthe first stagewhere the steady state terms of theheat flux level and is that heatflux which temperatures were calculated for the solar array in the would cause the solar array to exceed its flight vacuum phaseprior to the start of the slewing and allowable temperature of 175°C. Thehighfidelity stowing operations. The temperatures calculated by the natureof this3-dimensional analysis, and the speed at final stage before thedragpass beganwereused as the which analysis results could be produced,were initial solar array temperatures for the start of the drag compelling evidence to include this analysis in the pass. trajectorydecision making process during aerobraking operations. However, inorder for the analysis tobe usefulduringoperations, temperatures needed to be calculatedand postprocessed within 4 hours. Utilizing a 1.7GHz dualXEON processorcomputer made it possible for thishighly detailed finite element analysis torunin about onehour. Although, as drag passduration increases, sodoes therun time. This increasewas slight, andfor the longest dragpasses the solution time never exceeded1hour15 minutes. Computer speed alone was not the only meansof (a) increasing efficiency. The spacecraft’sconfiguration in thevacuum phasewas always the same so vacuum phase temperatures remainedvirtually constant. Also, the occultation duration changed very slowly, causing the initial temperatures to change at a slow,predictable rate. Therefore, to increase analysis efficiency, the viewfactor,orbital heating, and vacuum phase temperature analyses couldbe performed onan as- neededbasis. (b) Figure9. View factors to space: (a) Thermal The thermal analysis was highly dependenton all of the Desktop, (b) mapped toPATRAN otheranalyses being performed. Any problemsarising within anothergroup’s analysis causeddelays in the The temperatures obtainedprior to thedragpass were completion of the thermal analysis. Thishighly primarily dependenton theorbital heat ratesand view detailed thermalmodel generated a largequantity of factors calculatedfrom Thermal Desktop,but most information about the thermalstate of the solar array. significantly, they were influenced by thefact that the Figure10 shows the predicted temperaturedistribution spacecraftpassed into solaroccultation andwas shaded for orbitpass 40, justafter periapsis,which is typical of by Marson average, for about 4 minutes before the the majority of thedragpasses encountered. This dragpass began. This reduced the initial temperatures temperaturedistribution was calculated usinga and allowed the solar array to absorb more energy and predicteddensity, and velocity profile, and assumed a therefore increased the marginbetween the flight nominal orientationrelative to the atmosphere. Figure temperatures and the thermal limits. Finally, during the 11 shows thepredicted temperature of the thermal dragpass, a transient analysiswasperformed to 6 American Institute of Aeronautics andAstronautics sensors and the predicted heat rate as a function of time phaseof aerobraking, the magnitudeof the temperature centered onperiapsis. differencebetween the materialmaximum and the thermalsensor maximumwas dependent onthe heat flux, and the difference grewas the magnitudeof the heat flux grew. 70 0.25 60 Maximum facesheet Maximum core 0.20 50 Maximum solar cells C)40 1/2 RhoV3 2m) Temperature (°2300 00..110531/2 Rho*V (W/c 10 0.05 0 Figure 10. Predicted temperaturedistributionon -10 0.00 -250 -200 -150 -100 -50 0 50 100 150 200 250 the solar arrayfor orbit pass 40. Time (s) Figure12. Maximum predicted material A similar transient plotwas generated for each material, temperatures, orbit pass040,peak heat flux = 0.232 where the maximum predicted temperature for any W/cm2. locationon thatmaterial is tracked. Figure12 shows the In all of the trajectories thatwere encountered, the maximummaterial temperatures as a function of time. materialmaximum was always higher thanwhat the 0 0.25 thermal sensors indicated and was always located in the lower, outboard quadrant ofthe+X panel. Another -10 minordifference between the thermal sensor and 0.20 -20 maximummaterial temperatureplots is due to the fact that the locationof the maximum materialtemperature C)-30 2m Temperature (°--5400 00..110531/2 Rho V W/c ctsheheearmneg sse etdoe mtjhu rmtooupbg, eht hodeui stmc tohanxetiidmnruuaigmtipe tsae smasn.p d eI rntahtFeu itrgeeum srehp ie1frt2as twtuorheere another node in thatparticular material layer. -60 0.05 T1 facesheet side -X T2 solar cell side-X In Figures11 and 12, theheatrate is represented by a -70 T3 solar cell side mid T4 facesheet side +X T5 solar cell side +X 1/2 rho V3 smooth gaussian likeprofile. Theheating is directly -80 0.00 proportional to the atmospheric density; the density -250 -200 -150 -100 -50 0 50 100 150 200 250 Time (s) predictions,which came from an Odysseyversion of Mars GRAM2000, donot accountfor atmospheric Figure 11. Thermalsensortemperature predictions, density variations and thushave an average density orbitpass 040,peak heat flux= 0.232W/cm2. profile as a functionof altitude. Large uncertainties in Thisplotdiffers slightly from the plot of the thermal predicting theatmosphericdensity from orbit to orbit sensor temperatures. Theobvious difference is that the made the temperature predictions less valuable than maximum materialtemperatures are significantly expected,but they were still useful in that they gave a higher than the predicted thermal sensor temperatures. 3-dimensionalpicture of the thermal state of the array, In the case of the M55J graphite, whichwasexposed to and could identify any thermal anomalies. the flow, the figures show that the peak temperature Although theuncertainties inthe density predictions was85°Chigher than thepeakon either of the were present, they didnot impact thepredictionof the facesheet thermal sensors, T1 and T4. This was a result thermal limit lines, which turnedout tobe a very useful of having MLIcovering the facesheet in the areasnear tool. The flight corridor formain phase aerobraking those thermal sensors. The MLIprovided sufficient was chosen basedon the maximum Q dot,whichwas shielding to the underlying material from the incident the valueof the aerodynamicheating that would cause heat flux toprevent the material in thoseregions from the solar array to exceed its flight allowable reaching higher temperatures. Throughout the main temperature limit for the structureof 175°C. Theupper 7 American Institute of Aeronautics andAstronautics flight corridorwasreducedby a factorof 2 to carry of the solar array. The Lockheed limit line was more 100% marginwithrespect tothe limit. Figure13 is a conservative and as suchwas used as the flight plotof the maximum solar array temperature as a maximum. function of Q dot,covering orbitpasses 77through99. This plot was generated byrunningthe PATRAN 0.90 thermalmodelwith varying densityand henceheating 0.80 profiles. Fourdifferent density profileswere used: the low, middle, and upper ends of the flight corridor, as 0.70 LaRC Qdot limit for 190°C LaRC Qdot limit for 175°C well as one that would produce a maximumQdotof 2m)0.60 0.8 W/cm2. A Q dot of 0.8 W/cm2 was chosen as the W/c upperbounding case becauseit guaranteed the solar mit (0.50 atermrapyepraretudriectoiofn 1w75o°uCld. eBxyc euesdin tghea fplrigedhitc atleldo wQa dbolet,or Heatflux li00..3400 one derivedfrom flight data, the JPLNavigation team 0.20 could quickly determinethe maximumpredicted temperature of the solar array. Determining the thermal 0.10 limit of the solar array in terms of theQdot was 0.00 0 5000 10000 15000 20000 25000 30000 obtainedsimply by finding the value ofQdot that Apoapsis Altitude (km) corresponded to 175°C. Figure13 is a snapshot of temperature vs, Qdot over a small interval. This same Figure 14. Thermallimitline encompassing the processcould be repeated for several orbit trajectories, entire mission. and a mission thermal limit line couldbegenerated. Figure14 shows the limit line as determinedby the CONCLUSIONS PATRAN thermalmodel plotted as a functionof apoapsis altitude. The two limit lines show how significant theallowable temperature is in determining It has beenshown that a 3-dimensionalfinite element the thermal limit of the array. Simply increasing the model developed to representthe actualflight hardware flight allowable temperature to190°C will increase the of the Mars Odyssey solar array canbeusedwithout a thermal limit by about10%.As the mission progressed significant run-time penalty. Thedetail captured and the correlationof the thermal model to flight data the quantity ofdata generatedby such a highfidelity improved significantly. Since the thermalmodel and modelisnot possible usinga 1-dimensional model limit lines were updated continuously basedon the alone. The3-dimensional model showed that the hot correlations, confidence in the limit line chart increased spotswere not located near any of theflight thermal dramatically during operations. sensors. It isbelieved that such a highfidelitymodel, used earlier inthe designphase, could identify thehot 200 spots andbeused to place thermal sensorswhere they 190 are most needed. Also, the3-dimensional model 180 170 Orbit Pass 77 demonstrated that it could beused to evaluate and 160 Orbit Pass 99 150 Lower Corridor establish the thermal limit lines for the solar array in 140 130 Upper Corridor both the design andoperational phases. Byapplying Temperature (°C)1110128900000 LILmaMRmAC eLd iLmiaimtite iL tA iLncientieon Line lpreersdosucocenesssd elwes athrhineleemdt hoaend deolvu deserianvlgel l qoeuxpaimsltieitnyng to tmfi mteheteh wothdaessr amdnraadlstmicoadlleyls 70 increased. 60 50 40 30 ACKNOWLEDGMENTS 20 10 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 The authorswould like to thank RichardKriegbaum of Q dot(W/cm2) Lockheed Martin Astronautics forprovidingdrawings, Figure 13. Temperature as a functionof Q dotand details of the assembly, and advice. Also, thanksgo out orbit pass. to theJPLnavigation team for giving us theopportunity to make a contribution to thismission.Wewould also In figure 13 there are two limit lines: one is a limit line like to thank Mike Lindell forproviding supporting based onthe NASALangley 3-dimensional finite structural analysis. element thermalmodel, and theother is the limit line calculated by LockheedMartin’s 1-dimensionalmodel 8 American Institute of Aeronautics andAstronautics