Steady and Unsteady Computational Results of Full 2 Dimensional Governing Equations for Annular Internal Condensing Flows Ranjeeth Naik, Soumya Mitra, Amitabh Narain, Nikhil Shankar Acknowledgments: NSF-CBET-1033591 10 th October 2013
Boiling/Condensing Flow Applications Electronics/Data Center Cooling http://www.pgal.com/portfolio/rice-university-data-center Space Based Application Thermal Management Systems and Power Generation Cycles http://spaceflightsystems.grc.nasa.gov Phase change flows with boilers and condensers Lesser space/miniaturization Shear/Pressure driven Higher heat loads Enables phase-change systems of high heat removal and low weight requirements
M in Experimental Observations Traditional and Innovative Vapor Top View IF-HA N-IF Liquid q w (t) Wavy Annular Traditional Condenser Interfacial Wave Motion Heat Flux Meter (HFX) Non-Annular Zone Plug/Slug and Bubbly h = 2 mm Annular to Non-Annular Transition Maps Distance, x Pulsator Innovative Condenser Recirculating Vapor Vapor Acoustic reflectors Vapor Exit Inlet Top View q w (t) HFX Wavy Annular Regime Coleman and Garimella (2003) Enhancements Kivisalu et al., MGST, 2012 and Kivisalu et al., IJHMT, 2013
Need for Predictive/Simulation Capabilities Reliable steady annular flow predictions design innovative boilers/condensers different fluid and thermal boundary conditions Experiments - theory synthesized map of annular to non-annular transition Current Transition Maps insufficient for engineering purpose Enables design/functioning of innovative device operation (G in =?, X in =? X out =?) Annular to Non-Annular Transition Map Mass Flux, G (kg/m 2 s) 500 400 300 200 100 Non- Annular Regime Distance, x Wavy Annular Regime (Discrete and Disperse Waves) x = 0 (inlet) Annular Regime Pulsatile conditions simulation Better understand the experimental heatflux enhancements 0 0 0.2 0.4 0.6 0.8 1 Quality, X Coleman and Garimella (2003)
Problem Description
Problem Description for Internal Condensing Flows X = 0 Annular / Stratified x A Plug / Slug Bubbly DPT 1 X = 40 cm DPT 2 All Liquid Condensing Plate HFX-1 L = 1 m Side view schematic of a shear-driven condensing flow h = 2 mm Liquid Exit Governing Equations Interface Conditions
2- D Simulation Strategy Problem Definition Interface Tracking -4 th order accuracy in time - Unique mix of explicit and implicit methods MATLAB Processing - Data extraction - Interaction between domain through interface conditions COMSOL Liquid Domain -Laminar Flow/ Single- Phase Flow Branch - Heat Transfer in Fluids Vapor Domain - Laminar Flow/ Single- Phase Flow Branch
2- D Simulation Strategy Single Phase Domain Approach COMSOL/MATLAB Platform COMSOL Solve the Individual Domain MATLAB subroutines Data Extraction Interface Tracking Governing Equations Interface Conditions Algorithm
Results (Completed and In-progress)
Consistency Between Completely Different Steady Simulation Tools 0.025 2-D Solution - FORTRAN tool 0.3 2-D solution - FORTRAN tool Non-dimensional film thickness (δ) 0.02 0.015 0.01 0.005 1-D solution 2-D Solution - COMSOL/MATLAB Non- dimensional film thickness, d 0.25 0.2 0.15 0.1 0.05 1-D solution 2-D Solution - COMSOL/MATLAB tool 0 0 5 10 15 20 25 Non-dimensional distance along the channel (x) Gravity Driven R113 U = 0.41 m/s, ΔT = 5 C, h = 0.004 m 0 0 10 20 30 40 50 Non- dimensional distance along the channel, x Shear Driven R113 U = 0.6 m/s, ΔT = 5 C, h = 0.004 m Plot of non-dimensional film thickness along the non-dimensional distance of the channel showing consistency of different codes. Mitra et.al., 2012
X = 0 Steady Code Validation with Experiments (annular regime) Annular / Stratified x A Plug / Slug Bubbly DPT 1 X = 40 cm DPT 2 All Liquid Condensing Plate HFX-1 L = 1 m Side view schematic of a shear-driven condensing flow h = 2 mm Liquid Exit Case q'' W Expt q'' W 2-D @ % Error M x A @ x = 40 x = 40 cm for 2-D x in p in T W A (Expt) (Theory) g/s kpa C W/cm 2 W/cm 2 cm cm Error ± 0.05 ± 0.15 ± 1 ± 25% ± 12 % 1 0.702 99.98 48.6 0.18 0.19 4.1 71 2 0.700 99.99 49.8 0.16 0.14 13.4 90 3 0.700 99.99 50.0 0.15 0.13 11.5 93 4 0.698 99.99 50.7 0.12 0.11 4.2 95 5 1.000 101.07 44.0 0.40 0.40 0.6 57 Ongoing Base Flow Predictions for g y = -g are in Agreement with Experimental Runs
Physics Differences Between Shear and Gravity Driven Steady Condensing Flows Shear Driven Gravity Driven Distance from the condensing surface, (m) Velocity Magnitude (m/s) Distance from the condensing surface, (m) Velocity Magnitude (m/s) y x Distance along the length of the condenser, (m) Horizontal channel g y = - g and g x = 0 Distance along the length of the condenser, (m) Tilted channel, 2 deg g y = - g cos (2 ) and g x = g sin(2 ) Flow Situation
Unsteady Simulation Capability - Wave Resolution Distance from the condensing surface, y (m) x 10-4 2.3 2.2 2.1 2 1.9 1.8 1.7 1.6 1.5 1.4 1.3 Inlet Vapor Speed = 2.53 m/s, ΔT = 13.1 C Film Thickness Steady Film Initial Distrubance t = 0.056 s t = 0.104 s 0.05 0.1 0.15 0.2 Distance along the length of the condenser, x (m) Plot of film thickness along the length of channel for condensing flow
Unsteady Simulation Capability - Wave Resolution Inlet Vapor Speed = 2.53 m/s, ΔT = 13.1 C x 10-6 7 6 5 t = 0.008 s t = 0.04 s t = 0.056 s t = 0.072 s t = 0.088 s t = 0.096 s t = 0.104 s Amplitude 4 3 2 1 0 100 200 300 400 500 600 Wave number (1/m) Plot of Fast Fourier Transform as a function of wave number identifies critical wave-number
Unsteady Simulation Capability Interfacial Mass Flux Resolution Inlet Velocity,U 2 Vapor ṁ VK = -ρ 2 (v 2 i v s i ).n^ Interface ṁ Energy = 1/h fg. k 1 [ T 1 / n] i 1 Liquid ṁ LK = -ρ 1 (v 1 i v s i ).n^ Condensing Surface Interfacial mass flux (kg/m 2 s) ṁ VK - Based on kinematic constraints on the interfacial values of vapor velocity fields ṁ LK - Based on kinematic constraints on the interfacial values of liquid velocity fields ṁ Energy - Based on based on net energy transfer constraint
Interfacial Mass Flux, (kg/m 2 s) Unsteady Simulation Capability Interfacial Mass Flux Resolution 0.06 0.055 0.05 0.045 Time = 0.036 s Mṁ Liq LK Mṁ Vap VK Mṁ Energy 0.12 0.14 0.16 0.18 0.2 0.22 0.24 Distance along the length of the condenser, x (m) Plot of unsteady interfacial mass flux along the length of the condenser showing convergence of the interfacial variables.
Conclusions Developed Fundamental 2-D steady/unsteady predictive tools for annular flow condensation (and flow boiling not discussed). With regard to convergence and satisfaction of the interfacial conditions in the presence of waves, it shows unsurpassed accuracy (relative to other methods). Developed Engineering 1-D approx. tools for annular condensing and boiling flows. Validated the scientific tool by comparison with the experimental data (MTU). Suitable integration of simulations and experiments will aid in building of next generation thermal management systems involving phase change.
Thank You. Questions?
Heat Transfer Enhancements for Annular Flows Effects of externally imposed pressure-difference or inlet mass flow rate pulsations Kivisalu et al., MGST, 2012 and Kivisalu et al., IJHMT, 2013 Back
Simulation Tools Engineering 1-Dimensional tool (IJHMT, 2012) Annular flow condensation (Assisted Dr. Soumya Mitra) Annular flow boiling (Current Ph. D. Work) Inlet Pressure: 200 kpainlet Pressure: 200 kpa Total Inlet Mass Flow Rate: Total 4.51 Inlet g/s Mass Flow Rate: 4.51 g/s Re-Circulating Flow Rate: Re-Circulating 3.38 g/s Flow Rate: 3.38 Mg/s M V-e-C Temperature total Difference: Temperature 46 o C Difference: 46 o C O(Δp) = p exit p in 4.8 kpa O(Δp) = p exit p in 4.8 kpa h M V-e-C h M V-in-B h Inlet Pressure: 157 kpa Inlet Pressure: 157 kpa Liquid Inlet Mass Flow Liquid Rate: Inlet 1.13 Mass g/s Flow Rate: 1.13 g/s Re-Circulating M V-in-B Flow Rate: Re-Circulating 4.21 g/s Flow Rate: 4.21 g/s Temperature Difference: Temperature 36 o C Difference: 36 o C O(Δp) = p in p exit 10.0 O(Δp) kpa= p in p exit 10.0 kpa M total X X y M L-e-C M L-e-C M L-in-B M L-in-B X X y y q x q w condenser 50 W/cm 2 w condenser 50 W/cm 2 x q x q w boiler 50 W/cm 2 w boiler 50 W/cm 2 (without enhancement) (without enhancement) (without enhancement) (without enhancement) Plot of film thickness along the length of channel for condensing and boiling flow
Computational Approach Iterative solution strategy At discrete number of spatial locations, make an initial guess of interface variables - {d, τ i, p i, T Li, u Vi, v Vi, T Vi } for the steady problem. For the unsteady problem, start with known or specified values of these variables at t = 0. Governing Equations Interface Conditions
Computational Approach Iterative solution strategy Liquid domain calculations Vapor u L i v L i p 1 i p 2 i d Liquid Solve liquid domain by a finite-element method on COMSOL, using stress boundary conditions {i.e. tangential stress (shear) and normal stress (pressure) specified}, and saturation temperature conditions at the interface. Post-process the solution to obtain {u Li, v Li, T Li }.
Computational Approach Iterative solution strategy Vapor domain calculations Vapor u v i d v v i Liquid Using the liquid domain solution, compute u i V from continuity of tangential velocity, v i V from interfacial mass flux equality m VK m Energy and T i V using saturation temperature conditions at the interface. Using the computed {u Vi, v Vi, T Vi } on the current location of interface d, solve the vapor domain by the finite element method on COMSOL. Post-process the solution to obtain new values of tangential and normal stresses. For this, use momentum-balance condition at the interface and the computed values of vapor domain interfacial stresses.
Computational Approach Interface Tracking Update δ on moving grid which remains fixed over a time interval [t, t+δt] of interest. The interface is tracked through the reduced form of m LK m Energy given as: δ(x,0) δ δ δ u(x,t) v(x,t) t x δ(0,t) 0 steady (x) or other prescriptions
Computational Approach Interface Tracking EXPLICIT MARCHING: The evolution equation wave equation (1 st order hyperbolic PDE) We predict a location of interface at t = t* + Δt Map the existing/current solution to the new domain. Obtain a new predicted solution. Back IMPLICIT MARCHING Predict new interface with 4 th order accuracy in time with the help of its well defined characteristics equation. Map the current/existing solution on to the new domain implied by the new interface location (corrected location or the value for the time t*+dt). Repeat above steps for convergence and march in time.
Accuracy of 2-D Computational Tool The accuracy of the 2-D solution is ensured through satisfaction of the following: the convergence of the flow variables in the interior of each fluid domain satisfaction of all the interface conditions, grid independence of each domain and the flow problem 0.07 0.06 Interfacial mass flux, (kg/m 2 s) 0.05 0.04 0.03 0.02 0.01 0.00 0 0.05 0.1 0.15 0.2 0.25 Distance along the condenser, m Plot of interfacial mass flux along the length of the condenser showing convergence of the interfacial variables (Mitra (2012)).
Grid Independence Distance from the condensing surface (y), m Grid independence of Vapor domain 2.0E-03 1.5E-03 1.0E-03 5.0E-04 0.0E+00 0 0.2 0.4 0.6 0.8 Vapor velocity profile (m/s) No Refinement Refinement level 1 Refinement level 2 Refinement level 3 Refinement level 4 Distance from the condensing surface (y), m Grid independence of Liquid domain 7.E-05 No Refinement 6.E-05 Refinement level 1 5.E-05 Refinement level 2 4.E-05 3.E-05 2.E-05 1.E-05 0.E+00 0 0.02 0.04 0.06 0.08 Condensate velocity profile (m/s)
Grid Independence Vapor domain is typically solved with refinement level of 3 or 4 Liquid domain is typically solved with refinement level of 2 or 3 Back
Existing Simulation Methodologies Single domain solution approach Coarse and band approach to track interface extremely dense grid needed Unable to satisfy the mass flux transfer criteria as a result Level Set Methods Implicit Method Tracking VOF methods Marker Cell Approach
Refrigerant : FC - 72 Channel height = 2 mm Inlet mass flow rate = 0.4 g/s Temperature difference =17.45 C Flow Specifications Back
Assumptions Negligible interfacial thermal resistance Equilibrium thermodynamics on either side of the interface are assumed to hold. No non-equilibrium thermodynamic model for the interfacial mass-flux is used to obtain a solution.
Limitations of FORTRAN Code and Benefits of New Code Limitations Smaller domain problems Issues with meshing algorithm and noise resolution Slower convergence Inability to simulate non-annular regimes Benefits of COMSOL/MATLAB Code Simulate non-annular regimes with some modification Well developed single phase solvers Vapor compressibility effects
Our Hypothesis: Physics of Dramatic Enhancements δ(x*): Time-averaged thickness at x=x* M v (x,t) V x 2 L x 1 δ(x,t) M L (x,t) x δ** δ* x=x* x δ ads Adsorbed layer δ ads 100-200 nm Adsorbed layer interacts and destabilizes the micro layer causing wave troughs to stick Time of dwell/sticking is significant compared to externally imposed pulsation s time-scale Time averages of film thickness is significantly smaller for the pulsatile cases