CFD Analysis of Flow in CEV Radiator PanelsKumud Ajmani

ASRC Aerospace / NASA Glenn Research CenterCleveland, OH 44135

kumud.ajmani-1@nasa.gov

1.0 Abstract

A computational effort was undertaken to analyze the details of fluid flow in the variouscomponents of a radiator panel of the CEV Service Module. The effort focused ondetermining the envelope of environmental conditions and flow conditions at which theflow in the radiator would approach “stall” or near-freezing conditions. The FLUENTcode was used to perform conjugate heat-transfer computations for a single radiator panelwith a face-sheet and upstream and downstream headers. The results from the single-panel computations were used extensively to provide feedback to designers performingthermal analysis with the Thermal Desktop software package. The CFD analysis providedseveral insights into the correlation between areas of “stall” in the radiator and coolantmass flow rate.

2.0 Background

NASA and its contractors have chosen to develop a single-loop, non-toxic, stagnatingactive pumped loop thermal control design for the Active Thermal Control System(ATCS) of NASA’s Orion/CEV program. A stagnating radiator design was chosenbecause it provides self-regulation of thermal dissipation parameters. As heat flow to theradiator is reduced, less cooling capacity is required; stagnating radiators incorporatefluids that gradually change properties from Newtonian fluid to non-Newtonian fluid. Ifincreased heating loads are encountered, the radiator working fluid changes again fromnon-Newtonian fluid, or solid, to fluid to increase the heat transfer capability of theradiator.

The Apollo program successfully used a stagnating radiator design in which the fluidflow in all but one tube of the external radiators “stagnated” or came to a stop due toincreased fluid resistance to flow at low temperatures1. This feature allowed a veryelegant “shut down” of the radiator without having to rely on complicated bypassfunctions in low power, cold conditions. During such mission phases as long-termdocking on the Skylab Space Station2, this feature allowed for effective low poweroperation of the Apollo Command and Service Module during the extended on-orbitstay—just as will be required of the Orion when docked to ISS or perhaps in low lunarorbit for extended periods of time. However, the fluids used in Apollo units are notcompatible with today’s human-rated flight requirements, as they could be consideredtoxic and/or flammable.

NASA and its contractors for Orion are designing ATCS radiators that allow for singleloop systems that can operate in highly variable environments3. A single loop system waschosen as it eliminates the additional logistics of a second fluid, the temperature loss

47th AIAA Aerospace Sciences Meeting Including The New Horizons Forum and Aerospace Exposition5 - 8 January 2009, Orlando, Florida

AIAA 2009-836

Copyright © 2009 by K. Ajmani. Published by the American Institute of Aeronautics and Astronautics, Inc., with permission.

through the interchanger, and the additional pumps and instrumentation needed for a two-loop system4. The new design will use a 50/50 mixture of propylene-glycol and water(50/50 PGW), which is a non-toxic, non-flammable fluid acceptable for human-ratedspacecraft interior use5,6.

The computational work described in this paper is a two-part effort towards 1. Providing IV&V data and analysis of the ATCS stagnating radiator performance

with the “new” fluid (50/50 PGW).2. Providing a high-fidelity conjugate-heat transfer model to perform parametric

studies of radiator design variations.

3.0 Computational Model

A computational model was developed for a single-panel of a multiple-panel radiator forthe Orion ATCS. The model has been designed for use with the FLUENT CFD code forconjugate heat-transfer (fluid-solid interaction) analyses. The model accounted forconvective heat-transfer (fluid-flow passages), conductive heat-transfer (fluid to solid andacross the face-sheet) and radiative heat-transfer (face-sheet to external environment).Figure 1 shows the geometry and mesh for a single panel with ten-tubes, the face-sheetand one of the upstream headers. Flow enters the system at the bottom left and exits at theupper right via a downstream header (not shown).

Figure 1: Geometry and Mesh for a single-panel (ATCS Radiator)

A detailed view of the junction between the header and the tube is shown in Figure 2. The

header and the L-shaped connector use an unstructured mesh while the flow passageswithin the tube, the solid portions of the tube and the face-sheet connecting the tubes usea structured mesh.

Figure 2: Detailed Geometry and Mesh near tube entrance

Fluid properties like density (ρ), thermal conductivity (λ) and specific thermal capacity(Cp) may all be calculated by engineering data available in the Refrigeration Handbookfor Brines7. The property, Px (for a given temperature, T, and concentration of thesolution, ε) is represented by

Px = A1 + εA2 + A3 (273.15/T) + εA4 (273.15/T) + A5 (273.15/T)2

The parametric coefficients A1–A5 are given in Table 1. The same applies to thecalculation of the dynamic viscosity (µ) and Prandtl Number (Pr), with a slightlydifferent equation

ln(Px) = A1 + εA2 + A3 (273.15/T) + εA4 (273.15/T) + A5 (273.15/T)2

The freezing temperature of the solution (TF) is determined by the equation

TF = 273.15 (A0 + εA1 + ε2A2)

All of the fluid properties for the 50/50 mixture were computed dynamically as a function of temperature (T) and ε=0.5 at each node during the CFD analysis.

Parameter r Cp l m Pr T_F

Order [kg/m3] [kJ/kg K] [W/m K] [Pa s] [-] [K]

0 1.0

1 508.411 4.476 1.189 -1.028 6.661 -0.0374

2 -182.408 0.608 -1.491 -10.033 -6.994 -0.4005

3 965.765 -0.715 -0.697 -19.935 -18.551

4 280.291 -1.938 1.136 14.658 12.046

5 -472.225 0.479 0.067 14.621 14.477

Table 1. Parameters of the mathematical model for fluid properties of Propylene Glycolmixtures

4.0 Computational Results

A simplified CFD model was first validated by performing a grid-independence studywith a three-tube model derived from the full-panel, ten-tube configuration. The focus ofthe study was on resolving the boundary layer mesh in the flow tubes. The CFD modelwas considered grid-converged after there was a less than 1% change in the computedpressure-drop across the inlet/outlet boundaries (for a specified mass-flow rate) betweentwo mesh-refinements. The boundary conditions used were representative of the specifiedcoolant-flow rate, coolant inlet temperature and radiator face-sheet temperature for thefull radiator panel. All solid walls (except the face-sheet and the solid-fluid interfaces)were treated as adiabatic walls and a radiation boundary condition was applied at theface-sheet. Results from the grid-independence study were used to create the final meshfor the ten-tube panel (1.8 million nodes).

4.1 Stall Prediction with CFD

The CFD model was exercised with the FLUENT code in order to determine its ability topredict “stalling” of fluid-flow in one or multiple tubes for a given heat-load (asdetermined by the environmental temperature, the coolant fluid flow-rate and the coolantfluid inlet temperature). An initial, steady state solution was obtained for a nominal flow-rate (160lb/hr, 95F) and a fixed environmental temperature at the face-sheet (-460F). No“stall” was expected or observed in any of the tubes for these conditions.

The computations were then re-started by lowering the mass flow rate at the inlet of thesystem by 5% of the nominal flow-rate. It was theorized that if the flow rate reduction(and consequent change in pressure drop across the system) did not produce “stalling” inany of the tubes, the computations with FLUENT would yield a steady-state, convergedsolution. The 5% reduction in mass-flow rate was performed repeatedly until the pressuredrop across the system was low enough to force one or multiple tubes to show “stalling”behavior. The final result was that the flow rate had to be decreased to about 50% of the

nominal flow-rate before “stall” was observed in the tubes. Figure 3 shows the variationof tube centerline velocity for tubes 1-10 for various mass-flow rates.

Variation of Tube Centerline Velocity with Mass Flow Rate

0.08

0.13

0.18

0.23

0.28

0.33

0 1 2 3 4 5 6 7 8 9 10 11

Tube Number

Velo

city

(m

/s)

144lb/hr128lb/hr112lb/hr 96lb/hr 80lb/hr 76lb/hr

Figure 3: Variation of Tube Center-Line Velocity with mass-flow rate

For high mass flow rates of 96lb/hr or more, the velocity drops from tube 1 to tube 4 andthen increases uniformly to tube 10, thus creating a parabolic profile for the velocityvariation across tubes. Even though all the tubes are seeing the same flow-path length, theinertia effects of the upstream and downstream headers are responsible for this effect.Tube 4 is the minimum velocity tube, and would be expected to stall first.

However, when the mass flow-rate drops to 80lb/hr, the velocity curve changes from aparabolic profile to a uniformly increasing profile, with tubes 1 and 2 being the minimumvelocity tubes. This indicates that at this lower mass-flow rate, the inertia effect has beenminimized. A further reduction in mass-flow rate to 76lb/hr (which is the flow rate atwhich the tubes start to stall, starting with tube 1) leads to a non-linear profile for thevelocity variation across the tubes. The flow velocity in the downstream tubes (6-10)increases to compensate for the reduced flow in the stalling tubes (1-4). In summary,when the (lower) flow rates for stalling are approached, the inertia effects from theheaders disappear, and the tubes behave with a linearly increasing velocity.

The inertia effect due to the headers is shown in more detail in figure 4, which plots thecross-sectional contours of flow-velocity for tubes 1 and 10 at the entrance (top), halfwaylocation (center) and exit (bottom) planes. The flow-velocities have been non-dimensionalized by the maximum centerline flow velocity in tube 10. The flow velocity

is considerably lower in tube 1 as compared to tube 10. Both tubes show a thickening ofthe (laminar) boundary layer and the transition to fully developed flow, as the exit of thetube(s) is reached.

Figure 4: Contours of flow-velocity at three cross-sections for 160lb/hr flow rate

4.2 Flow Characteristics in Tubes at Stall Conditions (76lb/hr)

Figure 5 shows the variation of static (gage) pressure along the centerline of the varioustubes at the incidence of stall.

Figure 5: Static Pressure in various tubes at stalling flow-rate (76lb/hr)

The static-pressure curves show that tubes 1,2,3 are stalling and tubes 4,5 areapproaching stall (slope of curve is beginning to change as compared to tubes 6-10). Astubes 1-3 go into various stages of stall, the fluid tries to keep pushing against the“stalled” flow and this raises the local Δp across these tubes by an order of magnitude ascompared to the “un-stalled” tubes (4-10). Tubes 4 and 5 show local changes in the Δpcurve as they approach the stall condition. The nominally flowing tubes (6-10) show afairly linear pressure drop for a nominal Δp of 7800Pa (1.1psi) across the length of eachtube. Note that the data for the “stalled” tubes is truncated beyond Δp=10000Pa.

Figure 6 shows the variation of static temperature along the centerline of the varioustubes. Note that for the 50/50 propylene-glycol/water mixture modeled here, the freezingpoint, TF = -27F (240.7K). Hence, tubes 1-3 are mostly “frozen” or stalled because theyare below the freezing point of the mixture. The downstream end of tube 4 is approachingthe freezing point or incipient stall and tube 5-10 are flowing nominally. The temperaturedecreases with increasing axial distance because the heat capacity of the fluid in the tubeincreases with axial distance. This heat is then rejected to the surrounding metal (viaconduction) and eventually to the space environment (-415F, 25K) by the face-sheet (viaradiation).

Figure 6: Static Temperature in various tubes at stalling flow-rate (76lb/hr)

Figure 7 shows the effect of flow stalling on the overall performance of the radiator panel. The temperature contours on the face-sheet show that the stalled flow in the tubesclosest to the inlet of the panel affects the heat rejection capability of the panel. The panelwith un-stalled or nominal flow through all the tubes shows fairly uniform heat-rejectioncharacteristics along the various tubes.

Figure 7: Temperature contours on radiator face-sheet for un-stalled and partially-stalledflow

4.3 Radiator Performance Analysis with CFD

The pressure drop across the panel drops (almost) linearly as the mass flow rate decreasesuntil the point of stall, where the pressure starts to show a slight increase with decrease inmass-flow rate (see Figure 8). This is a typical characteristic of stalling behavior. The exittemperature of the coolant (Tex) also shows an almost linear drop with decreasing massflow-rate (since less heat is removed by the fluid as the flow-rate decreases) until thestall-point, where the exit temperature drops dramatically.

4.4 Comparison of CFD Results with Spreadsheet Analysis

The computational results from the FLUENT model in this work were compared withspreadsheet analysis performed by the NASA contractor8 (PARAGON, Inc.) responsiblefor the overall radiator design (see Figure 9). The comparison shows that the“spreadsheet” analysis tends to under-predict the severity of the mass-flow variationwithin each tube. The CFD analysis predicts a flow-rate variation of 2.5lb/hr(17.25-14.75) as compared to 1.8lb/hr for the spreadsheet analysis. In addition, the CFDpredicts a slightly lower exit temperature (60F, 288.7K) as compared to the spreadsheetanalysis (63.5F, 290.7K). However, the spreadsheet analysis does qualitatively agree withthe CFD analysis in predicting a parabolic distribution for the mass-flow rate across thetubes.

Pressure Drop vs Mass Flow Rate

0

1000

2000

3000

4000

5000

6000

7000

8000

70 80 90 100 110 120 130 140 150

Mass Flow Rate

Pre

ssu

re D

rop

274

276

278

280

282

284

286

288

290

292

294

296

dPTex

Figure 8: Panel pressure-drop(Pa) and exit-temperature(K) vs mass-flow rate for inlet temperature of 95F (308.3K)

Total flow rate=154.5lb/hr, Inlet=95F, Outlet=60F(63.5F)

14.5

15

15.5

16

16.5

17

17.5

0 2 4 6 8 10 12

Tube Number

Mass

Flo

w R

ate

(lb

/h

r)

FLUENT 3DPARAGON

Figure 9: Comparison of the predicted flow-rate in each tube for 154.5lb/hr case

5.0 Conclusions

A computational model has been successfully developed, validated and exercised forpredicting the onset of “stall” in stagnating radiators based on propylene-glycol/watermixtures for the Orion Active Thermal Control System. The model incorporates aconjugate heat-transfer methodology to capture fluid-solid heat transfer andenvironmental radiative effects. The CFD methodology designed, developed andvalidated in this work can be extended to analyze future designs of radiators for LSAMand other space applications.

6.0 Acknowledgments

This work was supported by the Service Module (Thermal) Project Office at NASAGlenn Research Center.

7.0 References

1. “Apollo Experience Report – Command and Service Module Environmental ControlSystem”, Frank H. Samonski, Jr. and Elton M. Tuchker, NASA TN D-6718, 1972. (http://ntrs.nasa.gov/archive/nasa/casi.ntrs.nasa.gov/19720012252_1972012252.pdf)2. “The Solar Dynamic Radiator with a Historical Perspective”, K.L. McLallin,M.L.Fleming, F.W. Hoehn and R.L.Howerton, NASA TM 100972, August 1988. (http://ntrs.nasa.gov/archive/nasa/casi.ntrs.nasa.gov/19890000746_1989000746.pdf)3. “Analytical Investigation of Flow Stagnation in a Pumped Fluid Loop Radiator”, G.Reavis, SAE Paper 2007-01-3260, July 2007.4. “Investigation into the Understanding of Flow Stagnation in Fluid Tube Radiators”, N.Hahn and G. Reavis, AIAA Paper 2007-529, 45th AIAA Aerospace Sciences Meeting,January 2007.5. “Tube Stagnation Experiments and Modeling Using a Safe, Non-Corrosive DielectricFluid for Radiator Thermal Control Systems Near Stagnation Regimes”, C. Iacomini, J.Dang and G. Anderson, SAE Paper 2008-01-2005, June 2008.6. “On Stagnation Flow in a Tube Radiator”, B. Motil, D.F. Chao, J.M. Sankovic and N.Zhang, AIAA Paper 2008-0825, 46th AIAA Aerospace Sciences Meeting, January 2008.7. “Properties of Working Fluids – Brines”, M. Conde Engineering, 2002, Zurich.8. Personal Communication, N. Hahn, Paragon, Inc.