NUMERICAL SIMULATION OF EXACT TWO-DIMENSIONAL GOVERNING EQUATIONS FOR INTERNAL CONDENSING FLOWS

Transcription

1 Presented at the COMSOL Conference 2010 Boston COMSOL Conference 2010 NUMERICAL SIMULATION OF EXACT TWO-DIMENSIONAL GOVERNING EQUATIONS FOR INTERNAL CONDENSING FLOWS By: Soumya Mitra and Ranjeeth Naik Advisor: Dr. Amitabh Narain Michigan Technological University October 7, 2010

2 Condensing Flows in Applications Electronic/Computer Cooling Space Based Applications Image Courtesy: Image Courtesy: Heat pipes, Rankine Power Cycles, Thermal Management Systems, Design of ISS-based two-phase flow facility, etc. Research Purpose: Facilitate more effective design of thermal systems miniature or not. Significantly enhance chance of successful operation of condensers in ground/space based applications. 2

3 Overview: Problem Formulation COMSOL / MATLAB Implementation Validation of Computational Results By Other Computational Tools With Experiments Differences between Gravity Dominated and Shear Driven Flows 3

4 Background Literature Available knowledge for exact and approximate model equations for two-phase flows and interface. Classical solutions of external condensing flow problems. Experimental data and correlations for internal condensing flow problems. Experimental data and correlations for external flow problems. Analytical and semi-empirical data and theoretical results for internal condensing flows. Theoretical results on dynamic instabilities and turbulence. Level-set methods and its implementation Next.. 4

5 Basic Simulation Strategy Based on first-principles Continuum governing equations Interface conditions (Kinematics, Mass, Momentum, Energy Transfer & Thermodynamic) Vapor P m Wall or Far field Boundary Liquid P m Heat released Other conditions Wall conditions Conditions at infinity (if any) Inlet/outlet conditions Initial conditions (t = 0) Special features Latent heat released with huge increase in density Interface conditions bring in additional non-linearities they connect the vapor and liquid flows, and also determine its time varying location 5

6 p in COMSOL/MATLAB Based Simulation Tool Vapor Domain U, Vapor T sat (p in ) Outflow u 2 i Interface Condition v 2 i τ 1 i Liquid Domain Outflow y x p 1 i T wall (x) is specified/known or Heat flux is specified/known 6

7 Input/output Data Processing through Equation Data Extraction MATLAB Liquid Domain COMSOL Modules Fluid Dynamics Heat Transfer Deformed Mesh Level-Set??? Vapor Domain 7

8 Current Simulation Capabilities Annular Internal Condensing Flows Boundary value problem (BVP) for steady solutions Initial boundary value problem (IBVP) for unsteady solutions 8

9 Validation of Computational Results 1. The computational results from the COMSOL / MATLAB tool is compared with the following computational tools: 2-D simulation tool based on SIMPLER algorithm Independently developed 1-D analytical tool 2. The computational results are also compared with the experimental results: Internal condensing flows inside an inclined channel (Lu & Suryanarayana) 9

10 Validation by Comparison with Other Computational Results Film Thickness, (m) Steady Solution Obtained by Different Computational Tools Initial Guess Steady Solution Initial Guess Steady Solution Solution from COMSOL based tool 1D solution Solution from Fortran based tool Dimensional distance along the channel, x (m) 10

11 Gravity Driven and Shear Driven Flows Gravity driven and shear driven flows are quite different with regard to Film Thickness Values Obtianed from the Unsteady Simulation Film Thickness, (m) Initial guess Non-dim. time = 7 Non-dim. time = 15 Non-dim. time = 17 Non-dim. time = 19 1D steady solution Distance along the length of channel, x (m) 11

12 Comparison of the Comsol Simulation with Lu and Suryanarayana for Run Plot of Film Thickness - R113, Run Film Thickness(mm) Lu & Suryanarayana Computaional Results Channel Length(mm) 12

13 Validation by Comparison with Experiments of Lu and Suryanarayana Comparison of Lu & Suryanarayana Experimental data with Computaional Results for R113 with an Inclination of 1 degree Run Fluid U Delta_T Film Thickness Experimental,mm Film Thickness Computaional, mm Error between Exp m/s C and Comp (%) 208 R R R R R R

14 Relevant Results Annular flow regime is responsible for rejecting most of the heat from a condenser. The associated liquid condensate motion is strongly affected by the orientation of the gravity vector g if the duct s hydraulic diameter D H 2mm. 1g thinner condensate thicker condensate 14

15 Gravity Driven and Shear Driven Flows Gravity driven and shear driven flows are quite different with regard to film thickness, velocity, and temperature profiles. Film thickness, d and distance along y direction u-velocity profile for gravity driven and shear driven flows Film thickness for horizontal condenser u-velocity profile for horizontal condenser Film thickness for vertical condenser u-velocity profile for vertical condenser Temperature Temperature profiles profiles (@ x = 5) Linear profile temperature profile for horizontal condenser temperature profile for vertical condenser U-velocity profiles u-velocity (@ x = 20) profiles Parabolic profile Non-dimensional distance along x direction, non-dimensional u-velocity, and temperature (ºC) 15

16 Gravity Driven and Shear Driven Flows Gravity driven and shear driven flows are quite different with regard to cross-sectional pressure variations as well Interfacial Pressure profile for gravity driven and shear driven flows Non-dimensional interfacial pressure, p non-dimensional pressure for horizontal condenser non-dimensional pressure for vertical condenser Pressure difference for flow in a vertical condenser at a modest ΔT Pressure difference for flow in a horizontal condenser at a modest ΔT Non-dimensional distance along x direction 16

17 Gravity Driven and Shear Driven Flows 1g Effect of time-varying gravity vector ( ) on film thickness g Time Varying Tilting of Horizontal Plate Tilt Angle (Degrees) Time (seconds) V 1 - Distance along the condenser (Non dimensional) V2 - Film thickness (Non dimensional) For small inclinations (~ 10 ) of the condensing plate, the flow becomes gravity driven in mm-scale condensing flows 17

18 Summary An algorithm for successful and accurate computational simulations of steady and unsteady condensing flows has been presented. The results from the computational tool using COMSOL are in good agreement with the 2-D computational code based on SIMPLER Algorithm and a completely independent quasi 1-D tool. Relevant results from the reported computational tool developed here are shown to be in agreement with the experimental results for the inclined channel flow experiments. Differences between gravity dominated flow and shear driven flows are discussed. 18

19 Thank You Questions? 19

20 Vertical Flows for mm-mm Scale Condensers A condenser s gravity-sensitivity can be minimized by using suitable array s of mm-scale ducts. This makes body force effects small relative to shear forces at a pressure penalty. mm-scale condenser mm-scale condenser F body F shear 1g F body F shear F body > F shear F body << F shear 20

21 Condensing Flows in μm Scale Ducts are Shear/Pressure Driven Gravity parameter G p (ρ 22 g x D h3 ) / μ 2 2 is reduced by D h μm scale, and D h < D cr, the flow becomes shear driven for a range of gravity values and for a given average inlet speed, T, and working fluid Σ Re in 3000 A Shear Driven Flow Zone B A representative curve for D is reduced at a certain fixed inlet mass flow rate, pressure and temperature difference ΔT 2000 C 1000 D Transition into shear driven flow zone at D = D cr G p = Fr -1 x* Re 2 G p Fr -1 2 x*re in (ρ 22 g x D inh3 ) / μ

22 Effect of Hydraulic Diameter on the Nature of Flow As D h μm scale, and D h < D cr, the flow becomes shear driven and gravity-insensitive 7.0E E-05 film 1 mm radius for 0g film 0.1 mm radius for 0g film 1mm radius for 1g film 0.1 mm radius for 1g Case A Film thickness, (m) 5.0E E E E E-05 mm-scale D < D cr Same as δ PS Large Diameter Solution Same solution changes for very 1g and slowly 0g for D < Dfor cr D < D cr Case B Case C Case D 0.0E Length along axial direction, x (m) 22

23 Effect of Hydraulic Diameter on the Nature of Flow As D h μm scale, and D h < D cr, the flow becomes shear driven and is accompanied with significant rise in pressure drop 1.4 Pressure Drop Across the Condenser Pressure Drop Across the Condenser, P (P in - P exit ), kpa mm radius 1g 1 mm radius 0g 0.5 mm radius 1g 0.5 mm radius 0g Ongoing research investigations for condensing flows in μm-scale ducts will account for: Significant T s (p i ) variation over the flow Significant vapor density variation Significant surface tension effects Inlet Mass Flow Rate, M in (g/s) 23

24 Validation of Computational Results and Experiments The condensing flow simulation results presented earlier are based on accurate computational simulations that have been quantitatively verified by experiments for gravity driven flows. Flow Regimes in Internal Condensing Flows Gravity Driven Flows (D h > 1 mm) Mostly Annular (Rabas et. al. [2000], Narain et. al. [2009] [2010]*) Shear Driven Flows Annular for: Horizontal mm scale partially condensing flows Results are consistent with Cheng et. al. [2005], Garimella et. al. [1999] (Complex Morphology) (?) Horizontal Tube (> mm-scale) (?) 24

25 Sensitivity of Boiling/Evaporating Flows to Gravity Vector For flow inside boilers/evaporators, the effects of g vector changes are expected to be less dramatic as compared to flow inside condensers. This is because thermal boundary conditions on the heater surface primarily couple with inlet mass flow rate values - which causes body forces to have a secondary influence on heat transfer rates. Courtesy: Incropera et.al [Textbook]. The sensitivity of flow boilers need to be ascertained (research is needed). We do not know what flow boiling information exists with regard to g-sensitivity of Fairchild Corporation s existing aircraft designs. Our forthcoming boiler experiments require that, for air force needs, the boiler be placed on a suitable shaker. 25

26 FUNDAMENTAL RESULTS ON ELLIPTIC-SENSITIVITY IN THE PRESENCE OF FLOW FLUCTUATIONS Theoretical and experimental results on Elliptic-sensitivity are presented for condensers. Analogous experimental results for boilers are expected within a year. 26

27 Shear/Pressure driven condensing (boiling?) flows exhibit a key phenomenon due to fundamentally different behavior compared to gravity driven flows. This is marked on our transition map for annular internal condensing flows Nature of steady equations: quasi-parabolic Nature of unsteady equations: parabolic with ellipticsensitivity Nature of steady equations: quasiparabolic Nature of unsteady equations: in between parabolic and parabolic with elliptic-sensitivity 2 Σ 2 : x** x 0.7 Nature of steady equations: parabolic Nature of unsteady equations: parabolic Shear driven zone Σ 1 Transition between Gravity driven to shear driven flow 2 Σ 1 : x* x 0.7 Σ 2 1 Σ 2 : x** 0 Re in 1 Σ 1 : x* 0 Gravity driven dominated zone zone G p Fr -1 x*re in 2 G p Fr -1 x*re in2 (ρ 22 g x D h3 ) / μ 2 2 x 27

28 Elliptic Sensitivity for Shear Driven Internal Condensing Flows (Consider Partially Condensing Annular/Stratified Flows) Exit Condition Control M V-e or p e In the above thought experiment, one asks whether the exit condensate flow rate ( ) (or natural pressure difference Δp) can be changed to achieve multiple quasi-steady solutions (not necessarily annular/stratified). In other words: do these flows exhibit elliptic-sensitivity (i.e. do these flows listen to both upstream and downstream conditions)? - Yes! Because net mean energy into the control volume can be changed by a change in the interface energy transfer (associated with interface location and mass transfer). Clearly, the above different Δp impositions are impossible for single-phase flows or adiabatic twophase flows (with zero interfacial mass transfer) because, the information only travels downstream (i.e. they are parabolic flows), and energy flow across the interface being zero 28

29 Basic Results on the Special Nature of Condensing Flows M in 1g Vapor (x,t) Liquid P 1 P M V-e (t) M L-e (t) Condensing Surface For any internal condensing flow (shear or gravity driven), there is a unique steady (termed natural ) annular/stratified (or film condensation) solution/realization which can be realized when the set up allows the flow to seek its own exit condition. However one can actively impose different steady or quasi-steady exit conditions - other than the natural one - for purely shear driven or mixed flows. This typically leads to other time dependent or quasi-steady solutions which may cause the flow regime boundaries (from annular stratified to non-annular (plug, slug, etc.) flows) to shift. 29

30 Test-Section and Schematic of the Observed Flow Annular/Stratified flows Plug/Slug flows Bubbly flows Vapor exit (closed for full condensation) Vapor Inlet h = 2 mm Steel slab HFX - 1 HFX - 2 Liquid exit x Bubbly L = 1 m HFX 1, HFX 2: These are heat flux meters which have thermo-electric coolers underneath them. 30

31 Experimental Proof of Elliptic-Sensitivity P Na Imposition of different P P Na P imposed Mean exit pressure: fixed Mean inlet mass flow rate: fixed 31

32 Experimental Proof of Elliptic-Sensitivity P Na Imposition of different P P Na P imposed Non-natural imposition of P results in a change in average heat flux and corresponding change in interface location For this representative case, the 90 Pa change in Δp results in approximately 38 % enhancement in the average heat flux. 32

33 Experimental Proof of Elliptic-Sensitivity T V Natural-1 Natural-2 R condensation T SH R slab T SL R coolant T Res T w (x, t) & q w (x, t) response exhibits transient behavior because cooling method does not hold T w (x, t) steady. The increase in heat transfer rates have to negotiate the resistance which leads to the thermal transients. 33

34 Theoretical/Computational Proof of Elliptic-Sensitivity Input: Imposition of different M L-e (t) M in 1g Condensing Surface Vapor (x,t) Liquid P1 P M V-e (t) M L-e (t) Non-dimensional Liquid Exit Flow Rate M Continuous 'Off-Natural' Control On-Off Control for an 'Off-Natural' Mean Continuous Control at 'Natural' Exit Flow Condition L-e-c 1 (t) M L-e-c2 (t) ML-e-Natural Non-dimensional Time t* Mean Value of the On-Off Control Strictly steady solution Quasi-steady/periodic imposition of fluctuations 34

35 Control through Liquid Mass Flow Rate at the Exit Non-dimensional Film Thickness δ Continuous 'Off-Natural' Control On-Off Control for an 'Off-Natural' Mean Control at 'Natural' Exit Flow Condition 9.00E E-02 At all times t > E E E E-02 At time t < t* 3.00E E E E E E E E E E+01 Non-dimensional Distance, x For all times annular/stratified solutions exist for these special controls at or near natural value. Solution After certain becomes time t* nonannular t*), no at annular time t > t* stratified (t > solution exists for all offnatural constant steady controls. This is further substantiated by the instability result for a constant steady control case with mean at an offnatural value. 35

36 Dynamic Stability at Natural and Instability for Continuous Off-Natural Control Initial disturbance at t = 0 Initial disturbance at t = 0 Disturbance dies at t = 125 Disturbance grows at t = 125 t* Steady Imposition at Natural is Stable Off-Natural Steady Imposition is Unstable But Off-Natural Quasi-Steady Imposition is Robust 36

37 Energy Response to Exit Condition Imposition Consider: Non-Dimensional Net Mechanical Energy into the Condenser (Partial Condensation versus time Non Dimensional Net Mechanical Energy in the Condenser Controllability Through Liquid Exit Mass Flow Rate Constant Steady Control at 'Natural' Exit Condition Constant Steady Control at an 'Off-Natural' Mean On-Off Control at an 'Off-Natural' Mean Mean energy value for Energy Response to constant steady control at on-off control "natural" value of exit liquid mass flow rate Energy Response to on-off control with the mean higher than "natrual" value Energy Response to constant steady control at an"offnatural" value of exit liquid mass flow rate Non Dimensional Time t Natural Mean steady off-natural value is achieved for an on-off control in the vicinity of the natural. Typically negative slopes are associated with the constant steady off-natural control. For constant steady control of exit liquid mass flow rate at off-natural value, energy keeps piling with a non-zero positive slope or draining inside the domain. This leads, eventually, to a situation where However annular/stratified for on-off control solutions with the do mean not of exist the control after near a the certain natural transition value, mean time. energy These annular/stratified in the condenser flows also are settles unstable down and to a only steady their value transition near behavior the steady for natural negligible value to non-zero and, initial therefore, disturbances nearby quasi-steady can be computationally annular/stratified studied solutions with exist the and help thus of flow the current regime simulation transition technique. is avoided. 37

38 Energy Response to Exit Condition Imposition Consider: Mean Non-Dimensional Net Mechanical Energy into the Condenser (Partial Condensation) for Different Quasi-Steady Realizations Assembled Results on Mechanical Energy in the Condenser Control-Volume for "Natural" and Nearby On-Off Controls Non-dimensional Liquid Exit Flow Rate 0.2 Continuous 'Off-Natural' Control On-Off Control for an 'Off-Natural' Mean Continuous Control at 'Natural' Exit Flow Condition Non-dimensional Time Non-Dimensional Energy Recived or Supplied by Controller Natural Lower Condensation Limited Range of Annular Flows Realized through On-Off Control Natural Higher Condensation Annular flows realized through on-off control in this range correspond to the limited steady energy band associated with this control. Non-Dimensional Liquid Mass Flow Rate at the Exit For periodic steady-in-the-mean realizations in the vicinity of strictly steady natural realization indicated above nearby quasi-steady solutions exist. Therefore, in the presence of fluctuations, PID control of both the mean inlet and the exit pressures become feasible. 38

39 Energy Response to Exit Condition Imposition Consider: Mean Overall Interface Energy Transfer Rates (Non-Dimensional) for Different Quasi-Steady Realizations Assembled Results on Interfacial Energy Transfer for "Natural" and Nearby Quasi-steady Prescriptions Mean Non-DimensionalValues of Energy Spent Across the Interface Natural Lower Condensation Higher Condensation Natural Mean Non-Dimensional Liquid Mass Flow Rate at the Exit 39

40 EFFECT OF ELLIPTIC-SENSITIVITY ON SYSTEM PERFORMANCE (ONGOING AND FUTURE RESEARCH) 40

41 Implications of Elliptic-Sensitivity on System Level Repeatability T sink QC T source < T sink QB For the same steady Q,, T source, and T sink ; and a transient load history shown B M below, there could be significant drifts in boiler temperature for a given load history due to elliptic-sensitivity associated with the two-phase components. That is, performance of boiler at Time I may not be same as that at Time I. 41

42 Research Needs We need to learn what aspects of our facility can be altered/upgraded to address issues of interest to ongoing Air Force-Fairchild Corporation s CRADA. We would like to collaborate and learn from you. We can do this by doing experiments and simulations that are not covered by NSF (3 year grant on mm-scale condensing flows) and NASA (1 year grant on flow boiling) grants. The possibilities are: i. Do a mm-scale tubular boiling experimental research of Air Force s interest. ii. Collaborate with Air Force on putting our mm-scale boilers/condensers on a suitable shaker that model different g-force history segments of your interest. For this, we may need a miniaturization of our facility as well as change in our working fluid (from FC-72 to fluids of your interest). The new equipment can be developed at AFRL or MTU. However suitable shaker experiments can be performed by additions to the existing MTU facility. iii. Provide simulation support for annular regime condensing/boiling flows under different g-force histories (e.g. on shakers, aircraft g-force history, etc.) iv. Learn your planned system and flow control details to see how elliptic/parabolic sensitivity issues discovered (and being developed) by us for flow condensation and flow boiling may be of assistance to you. 42

43 Conclusions 1. Novel transition maps for steady annular/stratified condensing flows yield a proper subdivision of the parameter space into gravity, shear, and mixed driven flow zones. This is of help in a-priori estimation of effects of changes in steady gravity levels. 2. For mm-scale ducts, the steady flow computational results as obtained from the 1-D/2- D solution techniques have been validated by comparisons with vertical tube experiments. Similar computational results and associated horizontal channel condensing flow experimental results are being synthesized. 3. Elliptic-sensitivity results for condensing flows (results for boiling flows are expected) were established both theoretically and experimentally. Its significance for controlling thermal transients and ensuring system level repeatability was discussed. 43

44 Thank You Questions? 44

45 Experiments Using a Shaker to Assess Gravity-Sensitivity +5g +3g Small enough duration to minimize impact on flow g Z g Z -5g -3g t 1 t 2 time t 1 t 2 time Sample g-force vector history for a particular aircraft maneuver in a vertical plane Representative g (t) profile for experiments Can the response for any duration [t 1, t 2 ] in an actual flight trajectory be assessed by mounting the boiler/condenser on a shaker with a periodic acceleration as shown above? Back 45

46 External Flows Available Knowledge/Literature Fundamental Laminar/Laminar Solutions: Internal Flows Stagnant Vapor 1g Nusselt Problem [1914], Narain et. al. [2007] 1g or 0g Koh Problem [1961], Narain et. al. [2010]* 1g or 0 g (Narain et. al. [2004], [2009], [2010]*) Experimental Investigations: Correlation for average heat transfer coefficients: (Cavallini et. al. [1974], Shah et. al.[1979], etc.) Flow regime visualization (Garimella et. al. [1999], Cheng et. al. [2005]), Flow regime maps (Carey [1992]) Gravity Driven Flows (D h > 1 mm) Mostly Annular (Rabas et. al. [2000], Narain et. al. [2009] [2010]*) Internal Flows Shear Driven Flows Annular for: Horizontal mm scale partially condensing flows Small μm scale channel and cylinder with surface tension effects under self-selected exit conditions (Narain et. al. [2010]*). More commonly: Complex Morphology (Cheng et. al. [2005], Garimela et. al. [1999]), Narain et al. [2010]* (?) Horizontal Tube (> mm-scale) (?) 46

47 Internal Flows Gravity Driven Flows (D h > 1 mm) Mostly Annular (Narain et. al. [2009] [2010]*) Back 47

48 Shear Driven Flows Back 48

49 Research Tools that are Employed for Reported Results and Planned Research Computational Simulation Tool Experimental Flow Loop Facility Experiments Fluid: FC72 Vertical Tube: D h = 6 mm 2 G 90 kg/m 2 -s Horizontal Channel: D h = 2 mm 2 G 200 kg/m 2 -s Newly Invented Film Thickness Sensor (Narain et. al., JHT, 2010) 49

50 First Principles Underlying Flow Physics and Computational Problem Continuum governing equations (Mass, momentum, and energy for each differential element in the interior of the two phases) Interface conditions (on the unknown interface these are restrictions imposed by: kinematics, mass transfer, momentum transfer, energy transfer, and thermodynamics) Vapor P m Wall, Far field or Line of symmetry Boundary Liquid P m Heat released Methodology Other conditions Wall conditions Conditions at infinity (if any) Initial conditions (t = 0) Inlet conditions Exit conditions (need?) Special features Sharp interface Single-phase solutions interact through interface conditions Interface condition is used for interface tracking Height function with adaptive grid (current) Level-set function (planned) 50

51 Experimental Facility for Internal Condensing Flows Test Section Auxiliary Condenser Rotameter Coriolis Flow Meter Data Acquisition Inlet Flow Valve Displacement Pump 1 L/V Separator Displacement Pump 2 Vacuum Pump Evaporator Water Pump 15 51

52 Experimental Test-Section Vertical Tube Horizontal Channel 52

53 Combined Experimental Facility Picture 53

54 Comparison of Results Obtained by 1-D and 2-D Solution Techniques for Annular/Stratified Flows Gravity Driven Flow in mm Scale Vertical Ducts Shear Driven Flow in 0g, Horizontal, and μm Scale Ducts Non-dimensional film thickness (δ) D code 1-D code Non- dimensional film thickness, d D solution 1 D solution Non-dimensional distance along the channel (x) Non- dimensional distance along the channel, x The 2-D and 1-D prediction for other flow variables (interfacial velocity, pressure, etc.) exhibit similar good agreements for different flow conditions and tube geometries as well. Within their own regimes, they also agree with experiments. Both 1g and 0g flows are stable. Note: (i) gravity driven smooth flows become wavy for Re d > 30, but they remain annular/stratified. (ii) Shear driven & 0g flows though stable (as shown) are not always experimentally realized except under controlled conditions. 54

55 COMPUTATIONAL RESULTS ON COMPARISONS BETWEEN GRAVITY DRIVEN FLOWS AND SHEAR DRIVEN FLOWS 55

56 Transition Between Gravity Driven and Shear Driven Flows Parameters affecting the flows: {x, Re in, G p Fr -1 x*re in2, Fr -1 y = 0, Ja/Pr 1, ρ 2 / ρ 1, μ 2 /μ 1 } Method of Cooling: T w (x) = Constant 2 Σ 2 : x** x 0.7 Shear driven zone Σ 1 Transition between Gravity driven to shear driven flow 2 Σ 1 : x* x 0.7 Σ 2 (within 4% of P.S.) 1 Σ 2 : x** 0 Re in 1 Σ 1 : x* 0 Gravity driven zone (within 4% of Nusselt) G p Fr -1 x*re in2 (ρ 22 g x D h3 ) / μ 2 2 x Transition map in {x, Re in, Fr -1 x} space for chosen {Ja/Pr 1, ρ 2 / ρ 1, μ 2 /μ 1 } = {0.004, , } 56

57 Transition Between Gravity Driven and Shear Driven Flows In the Re in and G p plane for a constant Ja/Pr 1, ρ 2 / ρ 1, μ 2 /μ 1 Transition from Gravity Driven to Shear Driven Flow 1 Σ 1 : x* Σ 2 2 Σ 1 : x* x 0.7 Transition between gravity dominated flow and shear driven flow zone 2 Σ 2 : x** x 0.7 Σ 2 Arbitrarily chosen limit for G p Re in Shear Driven Flow Zone 1 Σ 2 : x** 0 Correlation presented for this region Gravity Dominated Flow Zone G p = Fr -1 x* Re in 2 57

58 Projection of the Transition Map in Re in -G p Plane and Correlations for Gravity Driven and Shear Driven Flows Error bar x 1 Σ 1 : x* Re in Σ 2 Shear Driven Flow Zone Transition from Gravity Driven to Shear Driven Flow 2 Σ 1 : x* x 0.7 For 0g flows, x 0.35 (Ja /Pr ) (ρ /ρ ) Transition between gravity dominated flow and shear driven flow zone 2 Σ 2 : x** x 0.7 Gravity Dominated Flow Zone G p = Fr -1 x* Re in 2 Σ 2 1 Σ 2 : x** * * * δ ps(x) Rein *(μ 2/μ 1) Re (ρ /ρ ) (μ /μ ) * in * * * -1 (Ja 1/Pr 1) (Fr x ) Correlation presented for this region G p Fr -1 x*re in 2 (ρ 22 g x D h3 ) / μ 2 2 For gravity driven flows, 1 4k μ ΔT x 1 1 1/4 δ Nu ( x ) [ ] Yc g ρ 1 (ρ1 ρ 2) hfg [4 (Ja/Pr ) (x/g )] 1 p 0g and other correlations (see paper) are for parameter space given by the following: 0 x x A < x FC or 0 x x Re in Ja/Pr E-4 ρ 2 /ρ µ 2 / µ Fr -1 x 0.01 Nu x = (h x * L c )/k 1 = 1/d q (x) = h x * T 1/4 58

59 EXPERIMETNAL VERIFICATIONS FOR GRAVITY DRIVEN FLOWS 59

60 Experimental Data Obtained for Fully Condensing Flows Range of operating conditions and properties for the experimental data Ja 1 /Pr Re in Gravity Parameter μ 2 /μ ρ 2 / ρ 1 60

61 Comparisons Between Theory and Experiments for Partially Condensing Flows (Annular/Stratified Regime) Given T 1g is predicted and measured where: Z e = / Annular Flow Regime Verification Video Percentage agreement within ± 2% Percentage agreement within ± 3%

62 Measured p x FC Comparisons Between Theory and Experiments for Fully Condensing Flows (Annular/Stratified Regime) 1g Given : and T x FC is predicted and measured p is predicted and measured There is a good agreement between Δp p in p exit obtained both from experiments and predicted by quasi 1-D computational theory Run M in Tw T sat ΔT p in p xp-3 p xp-6 p exit Δp Δp_comp (g/s) ( C) ( C) ( C) (kpa) (kpa) (kpa) (kpa) (kpa) (kpa) No. ±0.03 ±0.03 ±0.05 ±1 ±0.15 ±1 or ± 0.2 ±0.03 ±0.03 ± The agreement is within ± 12% For shear driven cases, more detailed comparisons between theory and experiments is expected from a better instrumented horizontal rectangular testsection (forthcoming paper). 62

63 Comparisons Between Theory (1-D No Waves) and Experiments for Fully and Partially Condensing Flows HTC DSR (W/m 2 K) % 30% more than 30% 0% ±15 % ±30 % 15% partial condensation Heat Transfer Coefficient Re d > 800 over a significant section HTC exp. (W/m 2 K) Better agreement with simulations with waves and turbulent effects are possible Back 63

64 Experimental Result Showing the Deviation from Laminar/Laminar Flow and Onset of Turbulence Near the Interface Discrepancy vs. Re in /(Ja/Pr 1 ) Discrepancy (%) % 30% >30% Re in /(Ja/Pr 1 )

65 Sample Comparisons of Experiments with Various Correlations for Partially Condensing Flows Comparison of Heat Transfer Values from Different Correlations 30% 15% 15% % Axis Title h_l (Simu) Nusselt Shah Dobson-Chato Cavallini 0% ±15 % ±30 % Axis Title

66 Recommendation: Use Physics Based Sub-Categories Parameters affecting the flows: {x, Re in, G p ((ρ 22 g x D h3 ) / μ 22 ), Ja/Pr 1, ρ 2 / ρ 1, μ 2 /μ 1 } Experimental correlations often replace: {x, Ja/Pr 1, Re in } by Re d and distance x by local value of vapor quality X or Z Effectively, for common refrigerants, the parameters (Re in, G p, Re d ) impose the following restrictions: Small: if Re in < Re cr (x, G p, Ja/Pr 1 ) 50,000 Laminar Vapor model - OK Re in Large: if Re in > Re cr (x, G p, Ja/Pr 1 ) 50,000 Vapor Turbulence becomes important Small: if G p is small (?) (mm-scale or g x = 0) Shear Driven Flows G p Large: if G p is large (?) (mm-scale or moderate g x ) Gravity Driven Flows Small: if Re d < Re dcr (x, G p, Ja/Pr 1 ) 1,000 Laminar Condensate Re d Large: if Re d > Re dcr (x, G p, Ja/Pr 1 ) 1,000 Turbulent Condensate 66

67 Annular Flow Correlations Practices Method of Cooling: T w (x) = Constant Correlations Low G p Moderate G p High G p Low Re d High Re d Low Re in High Re in Proposals based on this paper Cavallini [1974] Shah [1979] Dobson & Chato [1998] Azer et.al. [1971] Travis et.al. [1973] Soliman et.al. [1968] 67

68 FUNDAMENTAL RESULTS ON EXIT-CONDITION SENSITIVITY AS A PART OF BOUNDARY DATA, FLOW CONTROLLABILITY, AND FLOW REALIZABILITY 68

69 Assume Lam/Lam Flows & Look for Annular Flows Transition Between Gravity Driven and Shear Driven Flows Parameters affecting the flows: {x, Re in, G p ((ρ 22 g x D h3 ) / μ 22 ), Fr -1 y = 0, Ja/Pr 1, ρ 2 / ρ 1, μ 2 /μ 1 } Method of Cooling: T w (x) = constant Parabolic Flows Elliptic Flows Mixed Flows 2 Σ 2 : x** x 0.7 Shear driven zone Re in 1 Σ 1 : x* 0 Σ 1 Transition between Gravity driven to shear driven flow 2 Σ 1 : x* x Σ 2 : x** 0 Gravity driven zone Σ 2 Experiments show robust film flows possible and exit conditions can not be imposed Experiments show stable film flows only under controlled exit conditions G p Fr -1 x*re in 2 (ρ 22 g x D h3 ) / μ 2 2 Chosen {Ja/Pr 1, ρ 2 / ρ 1, μ 2 /μ 1 } = {0.004, , } x The above maps takes the goals of Chen, Gerner, and Tien [1986] significantly forward. 69

70 Exit Condition Issue for Shear Driven Internal Condensing Flows (Consider Partially Condensing Annular/Stratified Flows) Exit Condition Control M V-e or p e In the above thought experiment, one asks whether the exit condensate flow rate ( ) (or equivalent exit pressure) can be used to control the flow and achieve multiple quasisteady solutions (not necessarily annular/stratified). In other words: are these flows elliptic (i.e. do these flows listen to both upstream and downstream conditions)? - Yes! Clearly, the above control is impossible for single-phase flows or adiabatic two-phase flows (with zero interfacial mass transfer) because, the information only travels downstream (i.e. they are parabolic flows). Related Issues/Questions: What is the nature of the steady governing equations? Are they parabolic (as in single-phase or air-water flows)? Are there significant differences between gravity-driven and shear-driven flows? 70

71 The basic result on ellipticity is: Summary of Results Requires exit boundary conditions in general and responds to them. But, in the absence of external constraints, parabolic boundary conditions (i.e. inlet and wall boundary conditions) suffice for determining the natural unconstrained steady solution not just the annular/stratified type but inclusive of other regimes (plug/slug, bubbly, etc.) Experimental proof for stable and repeatable natural partially condensing shear driven flows Experimental proof for stable and repeatable natural fully condensing shear driven flows Imposition of exit conditions- allowed for shear driven flows changes the liquid/vapor morphology or interface locations and hence significantly changes heat transfer coefficient (or thermal resistance R condensation for condensing flows). This fact is being experimentally proven. 71

72 Differences between Shear Driven Flows in 0g Channel and Horizontal Channel Film Thickness - 0 gravity Film Thickness - Transverse gravity Non-dimensional Film Thickness, δ Unlike 0g flows, at these downstream locations no "natural" exit conditions exist for maintaining annular/stratified flows. Reasons: See intersecting charactersitics on the next slide Non-dimensional Distance, x Annular/Stratified flow is not possible after x > x* (Ranjeeth et. al, 2010) 72

73 Shear Driven Flow in 0g Channel vs. horizontal Channel Study of Characteristics for flow in a 0g and horizontal channel Time Non-annular/stratified flow morphology x Annular/Stratified flow is not possible after x > x* (Ranjeeth et. al, 2010) Back 73

74 Shear Driven Flows in a Horizontal Channel Inlet Mass Flow Rate (g/s) and Delta Temperature** ( C) Partial Condensation Channel Test Section Inlet and Wall Conditions Time (1 iteration 1 second) Temperature Difference Inlet Mass Flow Rate Inlet Absolute Pressure * Measured at 10 cm downstream test section. ** Temperature difference between calculated T sat (p in ) and T wall average Inlet Pressure* (kpa) 74

75 Shear Driven Flows in a Horizontal Channel 1.4 Partial Condensation Channel Test Section Output Data Exit Mass Flow Rates (g/s) Exit Vapor Mass Flow Rate Exit Liquid Mass Flow Rate Pressure Difference * Pressure difference between measuring points located at 10 cm and 90 cm downstream test section Pressure Difference* (kpa) Time (1 iteration 1 second) Morphology Video Back 75

76 Comparisons Between Theory and Experiments for Fully Condensing Flows (Annular/Stratified Regime) For a fully condensing flow (with x FC < L), Δp p in p exit is obtained both from experiments and quasi 1-D computational theory Run M in Tw T sat ΔT p in p xp-3 p xp-6 p exit Δp Δp_comp (g/s) ( C) ( C) ( C) (kpa) (kpa) (kpa) (kpa) (kpa) (kpa) No. ±0.03 ±0.03 ±0.05 ±1 ±0.15 ±1 or ± 0.2 ±0.03 ±0.03 ± The agreement is within ± 12% More detailed comparisons between theory and experiments is expected from a better instrumented horizontal rectangular test-section (forthcoming). 76

77 Conclusions 1. The novel proposed transition maps for annular/stratified flows show a proper subdivision of the parameter space into gravity, shear, and mixed flow zones. 2. For mm-scale range, the results for both gravity and shear driven flows as obtained from the 1-D solution technique were validated by successful comparisons with 2-D results as well as relevant experimental results. 3. The unique annular/stratified steady solutions define an exit condition which is termed as natural. For shear driven flows, unless natural exit condition is specified or is accessible, there could be other quasi-steady/unsteady realization of the governing unsteady equations because of inherent exit condition sensitivity. This leads to more complex non-annular flow morphologies. 77

78 Generalized Summary Steady Governing Equations: Are They Parabolic or Elliptic? Thought Experiment: Exit Condition p e1 p e2 78

79 The Steady Governing Equations are Neither Elliptic Nor Parabolic The quasi 1-D/ full 2-D code results indicate that the steady gravity driven flows behave (in most situations) almost like a parabolic flow as it has a strong attractor even in the absence of exit condition specification. Nature of Steady Solutions and Phase Portrait Diagram 1g R 1 off curve x = R 1 on S x = 22 Solution Curve S x = d Q 1 off curve x = Q 1 on S x= P 1 on S P 1 off curve x = d P 1 merges with x = P 1 off x = u f u f 79

80 However shear driven (0g or horizontal) internal condensing steady flows behave somewhat like elliptic problems as the steady solutions have weak attractors and can be controlled by imposition of exit conditions. Nature of Steady Solutions and Phase Portrait Diagram Obtained from Quasi 1-D Approach 0g N 1 off curve x = 10.3 M 1 off curve x = 10.3 M 1 on S Solution Curve S d L 1 on S L 1 off curve x = 0.11 u f 80

81 Sensor Principle An identifiable The amount part of green the amount of green fluorescent fluorescent light light collected collected from from the fluorescent dopant material depends depends only on instantaneous only on thickness d for δ a for given a given concentration of the of dopant. the dopant. 200 mm fiber-optic cable Violet illuminating light Violet Laser Light Source Film Thickness Sensor Probe d Liquid film d of 0.5 to 3.0mm Suitable filtering of noise light associated with light source Greenish fluorescent light is collected in the outer cable Suitable filtering of of noise light light 1 Receiving mm-dia fiberbundle cable Suitable Long Pass Filtering Fluorescent Signal Detector (Photomultiplier Fluorescent Signal Tube) Detector (Photomultiplier Tube) Back 81

82 First Principles Underlying Flow Physics and Computational Problem Continuum governing equations (Mass, momentum, and energy for each differential element in the interior of the two phases) Interface conditions (on the unknown interface these are restrictions imposed by: kinematics, mass transfer, momentum transfer, energy transfer, and thermodynamics) Vapor P m Wall, Far field or Line of symmetry Boundary Liquid P m Heat released Methodology Other conditions Wall conditions Conditions at infinity (if any) Initial conditions (t = 0) Inlet conditions Exit conditions (need?) Special features Sharp interface Single-phase solutions interact through interface conditions Interface condition is used for interface tracking Height function with adaptive grid (current) Level-set function (planned) Back 82

83 Comparison of Results Obtained by 1-D and 2-D Solution Techniques for Annular/Stratified Flows Gravity Driven Flow in mm Scale Vertical Ducts Shear Driven Flow in 0g, Horizontal, and μm Scale Ducts Non-dimensional film thickness (δ) D code 1-D code Non- dimensional film thickness, d D solution 1 D solution Non-dimensional distance along the channel (x) Non- dimensional distance along the channel, x The 2-D and 1-D prediction for other flow variables (interfacial velocity, pressure, etc.) exhibit similar good agreements for different flow conditions and tube geometries as well. Within their own regimes, they also agree with experiments. Both 1g and 0g flows are stable. Note: (i) gravity driven smooth flows become wavy for Re d > 30, but they remain annular/stratified. (ii) Shear driven & 0g flows though stable (as shown) are not always experimentally realized except under controlled conditions. Back 83

84 Computational Approach Adaptive computational grids GRID B EMPLOYS xud i LINES yv m2 yv j+1 yv j xud i xud i+1 GRID A LINES FOR CFD I = 2 Interfacecell I = 1 yv 2 xu 2 xu i xu i+1 xu l1 84

85 Computational Approach Iterative solution strategy At discrete number of spatial locations, guess {d, u SLi, v SLi, SLi, u Vi, v Vi, Vi } for the steady problem at t = 0 and, for the unsteady problem (incompressible vapor and unspecified exit condition) at t > 0, for the). Adjust these seven guess functions: {d, u SLi, v SLi, SLi, u Vi, v Vi, Vi } with the help of seven interface conditions. The following steps implement this philosophy by separate singlephase (liquid and vapor domain) calculations with a sharp interface. 85

86 Computational Approach Iterative solution strategy (contd.) Liquid domain calculations Vapor d shifted u SL i v SL i p 1 i p2i d Liquid Ghost liquid over a single cell After fixing {u SLi, v SLi, SLi } on shifted interface (d shifted ), solve liquid domain under shifted interface by a finite-volume (SIMPLER) or a finite-element method. The {u SLi, v SLi, SLi } are adjusted to satisfy tangential stress (shear), normal stress (pressure), and saturation temperature conditions at the interface. 86

87 Computational Approach Iterative solution strategy (contd.) Vapor domain calculations u v i v v i Vapor d Liquid After fixing {u Vi, v Vi, Vi } on interface d, solve vapor domain above interface by the same finite-volume method (SIMPLER). UPDATE the guesses for u Vi, v Vi, and i V with the help of: continuity of tangential velocity, interfacial mass flux equality m, and saturation temperature VK m Energy conditions at the interface. 87

88 Computational Approach However Popular Level-Set Methods Use m LK m Energy V 0 t where, (x,t) = 0 locates the interface with V extended extended v (k T k T ) (1 ) (1 h ) I I fg where,i 1 for liquid and 2 for vapor with subscript I = 1 is for liquid and I = 2 is for vapor In the new COMSOL/MATLAB based approach, we propose to retain our current approach in principle but use the above level-set equation for locating the interface through = 0. This will allow investigation of flow regime transitions from annular/stratified flows to plug/slug flows. 88

89 Computational Approach Iterative solution strategy (contd.) Our Current Practice is to Update d (by tracking the interface) on an adaptive Eulerian Grid which remains fixed over a time interval [t, t+ t] of interest. Current method uses reduced form of m LK y ( xt, ) 0 m given as: Energy and tracks the interface through the δ(x,0) δ δ δ u(x,t) v(x,t) t x δ(0,t) 0 steady (x) or other prescriptions Back 89

90 Alternative Theory/Computational Results from a Quasi 1-D Semi-Analytical Approach The method uses exact analytical solutions of the underlying 2-D governing equations under thin film approximation. Only the vapor phase momentum and mass balances employ one-dimensional governing equations with an assumed vapor profile. Hence this method is called Quasi 1-D. π(x): nearly uniform cross-sectional pressure d(x): film thickness u f (x): interfacial speed d(x), u f (x), p(x) are non-dimensional For the unknown, y(x) = [d(x), u f (x), p(x), dp/dx (x) ζ(x)] T, the governing equation dy/dx = g(y) is to be solved, subject to the condition y(0) = [d(0), u f (0), p(0), ζ(0)] T or y(ε) = [d(ε), u f (ε), p(ε), ζ(ε)] T for x > ε Salient Features: The problem is singular Problem is neither parabolic (because of the presence of ζ(0 + ) in the y(x)) nor clearly elliptic (since explicitly defined values of ζ(0 + ) is not admissible.) Back 90

91 Nature of Steady Solutions and Phase Portrait Diagram 1g R 1 off curve x = R 1 on S x = 22 Solution Curve S x = d Q 1 off curve x = Q 1 on S x= P 1 on S P 1 off curve x = d P 1 merges with x = P 1 off x = u f Strictly steady solution in 1g behaves like a parabolic solution as it does not need an exit condition specification. u f 91

92 Nature of Steady Solutions and Phase Portrait Diagram 0g N 1 off curve x = 10.3 M 1 off curve x = 10.3 M 1 on S Solution Curve S d L 1 on S L 1 off curve x = 0.11 u f Strictly steady solution in 0g behaves like an elliptic problem with neutral to unstable steady solution that can take different exit conditions. 92

93 Alternative Theory/Computational Results from a Quasi 1-D Semi-Analytical Approach The method uses exact analytical solutions of the underlying 2-D governing equations under thin film approximation. Only the vapor phase momentum and mass balances employ one-dimensional governing equations with an assumed vapor profile. Hence this method is called Quasi 1-D. π(x): nearly uniform cross-sectional pressure d(x): film thickness u f (x): interfacial speed d(x), u f (x), p(x) are non-dimensional For the unknown, y(x) = [d(x), u f (x), p(x), dp/dx (x) ζ(x)] T, the governing equation dy/dx = g(y) is to be solved, subject to the condition y(0) = [d(0), u f (0), p(0), ζ(0)] T or y(ε) = [d(ε), u f (ε), p(ε), ζ(ε)] T for x > ε Salient Features: The problem is singular Problem is neither parabolic (because of the presence of ζ(0 + ) in the y(x)) nor clearly elliptic (since explicitly defined values of ζ(0 + ) is not admissible.) Back 93

94 Transition Between Gravity Driven and Shear Driven Flows Parameters affecting the flows: {x, Re in, Fr -1 x, Fr -1 y = 0, Ja/Pr 1, ρ 2 / ρ 1, μ 2 /μ 1 ) Shear driven zone Gravity driven zone Transition between Gravity driven to shear driven flow 94

95 Transition Between Gravity Driven and Shear Driven Flows In the Re in and Fr -1 x for a constant Ja/Pr 1, ρ 2 / ρ 1, μ 2 /μ 1 Transition from Gravity Driven to Shear Driven Flow Zone B 8000 Gravity Dominated Flow Zone Re in 4000 Zone A Correlation presented for this region Shear Driven Flow Zone Transition between gravity dominated flow and shear driven flow zone Fr -1 x 95

96 Transition Between Gravity Driven and Shear Driven Flows For 0g flows, δ (x) 0.75x 0.35 x (Ja /Pr ) (ρ /ρ ) Rein *(μ 2/μ 1) Rein (ρ 2/ρ 1) (μ 2/μ 1) (Ja /Pr ) * * * ps * * * FC For Gravity driven and mixed flows (shaded purple in the flow regime map) 0.75x x 0.26 (Ja /Pr ) (ρ /ρ ) * * * δ(x) Rein *(μ *(Fr ) 2/μ 1) Re x (ρ /ρ ) (μ /μ ) * in * * * -1 (Ja 1/Pr 1) (Fr x ) FC

97 Exit Condition Issue for Internal Condensing Flows (Consider Partially Condensing Annular/Stratified Flows) Exit Condition Control M V-e or p e In the above thought experiment, one asks whether the exit condensate flow rate ( ) (or equivalent exit pressure) can be used to control the flow and achieve multiple quasisteady solutions (not necessarily annular/stratified). In other words: are these flows elliptic (i.e. do these flows listen to both upstream and downstream conditions)? - Yes! Clearly, the above control is impossible for single-phase flows or adiabatic two-phase flows (with zero interfacial mass transfer) because, the information only travels downstream (i.e. they are parabolic flows). Related Issues/Questions: What is the nature of the steady governing equations? Are they parabolic (as in single-phase or air-water flows)? Are there significant differences between gravity-driven and shear-driven flows? 97