Upload
others
View
3
Download
0
Embed Size (px)
Citation preview
Modeling Effort on Chamber Clearingfor IFE Liquid Chambers at UCLA
Presented by:
P. Calderoni
Town Meeting on IFE Liquid Wall Chamber DynamicsLivermore, CAMay 5-6, 2003
Outline
This presentation will address twocomponents of our modeling efforts onchamber clearing:
- Vapor Condensation
- Droplet clearing in a pressure decay field
Vapor Condensation Modeling - ApproachCouple UCB model for condensation at a liquid / vapor
interface (based on Schrage kinetic theory) withTsunami calculations in 2-D volume
Apply “enhanced” Tsunami to simulate flibe vaporcondensation experiments, maintaining all assumptions
implicit in Tsunami and condensation model
Compare with experimentsEvaluate effect of measured interface conditions:
traces of non condensable gasesvapor density dropping into transitions regime
accumulation of less volatile BeF2
Generalize liquid / vapor interface model:add diffusional layer at the interface for non condensable gases
add velocity and temperature slip at the interfaceadd diffusional layer at the interface for BeF2
Gas dynamics modeling in IFE liquid chambers
nd 221π
λ =
Gas dynamics regime characterized by Knudsen number:
Considering:Hard-sphere modeldiameter d = 4 ALowest density inHYLIFE chamber =3X1013 #/cm3
Upper limit ofmean free path
in HYLIFE:4.7 cm
LKn λ
= mean free path:
characteristic length
Molecular collisions and collision with system boundaries are equallyimportant
Kn < 0.01 Continuum Regime
0.01 < Kn < 0.1 Slip regime
Mean free path small compared to system - molecularcollision dominate - gas approximated as continuosmedium
Gas approximated as continuos several mean free paths away fromadjoining medium - Kinetic theory near interfaces to account forboth molecular collisions and collision with system boundaries
0.1 < Kn < 3 Transition regime
Kn > 3 Free Molecules RegimeMolecular collisions infrequent - rarefied gas kinetic theory applies
Condensation model assumptions
2-D gas dynamics calculation assumptions:
Two dimensional geometry
Gas phase is a continuum
Gas state changes are isoentropic everywhere exceptat shock waves, which are treated as discontinuities
Liquid structures are rigid - liquid inertia in the time scales of interestprevents structures from moving or deforming - no work is transferredfrom gas to liquid
Gas viscosity is negligible - viscous time scale L2/ν >> dynamic time scale L/c
Inside the volume gas is adiabatic - conduction time scale L2/α >>dynamic time scale L/c - radiative losses assumed to be negligible
Flibe is an ideal gas law with constant γ - fitted EOS corrected fordissociation and ionization not effective in the considered range
Condensation model assumptions
1-D liquid / vapor interface assumptions:
Vapor condenses only on liquid surfaces present as initial condition -no droplet nucleation in the volume
Liquid layers are semi-infinite slabs - thermal diffusion length (α ∆t)1/2
<< surface curvature - initial layer T is uniform - liquid T away from theinterface remains constant - droplet spray cooling not considered
Heat and mass transfer at the interface only in the normal direction - interfacevelocity due to mass addition is neglected because of mass continuity
Heat transfer in the liquid layer by conduction in the normal direction -a convection term due to condensing flux introduced in the energy eq
Liquid surface is always in thermodynamic equilibrium with the vapor - highmass transfer rates during initial transient neglected - continuum assumption
Recombination and chemical diffusion effects for flibe are fast - vaporchemical composition is fixed by initial conditions
Vapor composition is uniform in the volume and at the interface
Interface kinetic theory accommodation coefficients (sticking andevaporation) are assumed to be = 1
Interface condensation: Schrage theory
( )[ ] dwdvduwvuuTkm
Tkm
NdN
bb
wvu
++−
−
= 222
0
2/3,,
2exp
2π
The effect of condensation on the molecular motion is to impose anet flux in the direction normal to the interface:
Integrating over v, w and positive (toward) u:
+
+
−
=+
2/1
0
2/1
020
2/1
0 21
22exp
2 Tkmuerf
Tkmu
Tkmu
mTkn
bbb
bN π
πφ
Re-writing:
( )[ ] serfseTR
pG s ++
= −+ 1
22
2/1
0
ππµ where ( )
2/12/1
2/10
2/1
00 222
=
=
=
γγγ
µ MRTu
TRus
RTp ρ=Finally, the net flux across theinterface:
2/12/1
21
21
−
Γ=
lslse
vsvsck RT
pfRT
pfGππ
Modeling the interface and liquid layer
Energy equation
with bc
0),( ≥=∞ ∞ tTtT l
Two equations to couple liquid and vapor properties at the interface:
mass balance 000
≥=== +− ==tGuuG
xlslsvsvsxρρ
energy balance conduction in the vaporneglected for short diffusionlength - neglect radiation
Vapor stagnationenthalpy:
Where for flibe
[ ]refvsatplpgreffgvsatfg TpTccThpTh −−+= )()()())((
−
×= Tsatp
10054407.910
760101300
0),(
0
0 ≥∂
∂−=
=
txtxTkhG
xlvs
[ ] [ ]2
)()(2
0 vsvssatfgvssatvspgvs
upThpTTch ++−=
0),(
0int ≥
∂∂
−=′′ txtxTkq l
In the liquid layer:
∞≤≤∂
∂=
∂∂
+∂
∂ xxtxT
xtxTu
ttxT
ls 0),(),(),(2
2
α
Convection term added toaccount for condensing massacross the interface - ulsevaluated from G
Initial non-equilibrium conditionsIn early stages of condensation the contact of highly superheatedvapor with the cold surface causes high mass transfer rates atthe interface - the effect (suction) is to increase the vaporvelocity that is evaluated by Tsunami
Schrage theory fails to account for high mass transfer ratesbecause of the surface equilibrium assumption - velocityassociated with mass flux predicted by the Schrage eq can behigher than physical limitations associated with super sonicchoking effect
Gas dynamics limiting flux:
cuMcMMcG vv <
−
++−
−−
−
++−
−
+=−+
− 11
12
max 12
11
12
12
11
12 γ
γγ
γγγ
γρ
γγγ
γρ
cuuG vvv >= ρmax
),(min maxvk GGG =Correction:
Numerical iteration scheme
( )
xkcG
TxkTchG
Tl
plk
ll
lplvsk
ls
∆+
∆++
=∞
0
2/12/1
21
21
−
Γ=
lslse
vsvsck RT
pfRT
pfGππ
Vapor (p, T) at the interface are given by Tsunami (as well as the gas dynamic limiting flux)
Equilibrium assumption reduces unknown liquid properties to one, surface temperature:
Condensation gives the second eq to solve for T and G:
)( 1−= nls
n TGG )( nls
nls GTT = ε≤
−−
−
1
1
nls
nls
nls
TTT
))((
1
nlowlsls
nhighls
nl
nlowls
TGTT
TT
−−
−−
=
=
2)( n
lowlsnlsnewn
ls
nls
newnhighls
TTT
TT
−−
−−
+=
=
[ ] [ ]11111
1112
1 2)(
−−
−−−
−−+
− −∆∆
−+−∆∆
+= ni
ni
l
nni
ni
ni
ni
ni TT
xtGTTT
xtTT
ρα
1
2)(2
−
∆
+∆
<∆x
Gx
tl
n
ρα
In the liquid layer, using an upwind scheme for the condensation case:
Iteration step:
Newton-Raphson averaging method:
or
2)( n
highlsnlsnewn
ls
nls
newnlowls
TTT
TT
−−
−−
−=
=
Introducing condensation effect in Tsunami
The condensation module evaluates G at each step - the mass flux condition must now beused at Tsunami boundary cells interface instead of the usual adiabatic condition adoptedat cells interfaces in the volume
Tsunami numerical scheme requires computation of mass (continuity), momentum (Riemannsolver across the discontinuity) and energy (adiabatic assumption) fluxes at the edge ofeach cell
Mass flux is G
Energy flux from same interface balance -written in Tsunami terms:
+
−=
2)1(
2vs
vs
vs upGfluxEnergyργ
γ
( )( )TvssvsvsR
TvsvsvsL
puuU
puU
,,
,,
+−=
=
ρ
ρ ( )TpuU **** ,,ρ= **uG ρ= ερ<
−=∆
max
**
vnorm G
GuG
vss
Guρ20 =
0)1(
0)1(11
11
<∆+=
>∆−=−−
−−
GifGuu
GifGuuknorm
ks
ks
knorm
ks
ks
Momentum flux determined by mirror node, introducing suction velocity:
Iteration scheme:
Numerical domaingeometry
Uniform grid: 10 x 100 cells - 0.5 x 0.5 cm each
4.1
0331.0
==
=
v
p
CC
molkgm
γ
0
00 TR
Wp⋅⋅
=ρ10
0 −⋅
=γTRe
Background gas is flibe, considered as anideal gas with:
Initial background conditions specified as:
Injected gas considered by Tsunami as DEBRIS,initially available in a 3 x 3 cell volume V0
Initial superheated vapor conditions specified astotal injected mass [kg] and total initial energy [J]
Code runs - parameters case study
Tsunami BCrecovered:
us = 0
us = 2 uvs
Top and bottom boundaryare impermeable
Boundary conditions:
Open interface
Impermeable surface
Condensation
Parametric study for:
Sticking coefficient at the interface
Liquid layer thickness
Initial liquid temperature (constant at solidwall interface)
Reference case:
fc = 1
0.5 mm
600 C
Liquid initial temperature
Liquid layer thickness
Liquid initial temperature
Temperature distribution in the 2-D axi-symmetrical numerical domain as a function
of time for the reference case
Temperature distribution - middle cells
Gas temperatures evaluatedby Tsunami fall below theimposed initial backgroundtemperature whenremaining mass is low
Temperature distribution - Top
In the boundary cells atthe top and bottom ofthe chambers thevapor interface T ishigher then Tsunamievaluated temperaturein the inner cells
Heat conduction in the liquid layer
Pressure distribution in the 2-D axi-symmetrical numerical domain as a
function of time for the reference case
Liquid surface equilibriumassumptions not valid fortransient condensation
Ohno fitted equation for flibe:
−
×= Tsatp
10054407.910
760101300
Pressure distribution - middle cells
Ideal gas assumption for flibeoverestimates vapor pressureduring high temperaturesinitial transient
Density distribution in the 2-D axi-symmetrical numerical domain as a
function of time for the reference case
Free molecular regime
Gas dynamics modeling in IFE liquid chambers
01.0≅=L
Kn λ
Direct Simulation MonteCarlo method
HIBALL - Wisconsin (1989)KOYO - Osaka Un. (2002)
solving Boltzmann equation
dvdggtxvftxcftxvftxcfcfF
xnfc
tnf
Ω×−∫∫=
=∂∂⋅+
∂∂⋅+
∂∂
),()],,(),,(),,'(),,'([
)()(
χσ
by decoupling molecularmotion from collisions:
),(*)1(),,()0,,()1(),(*xcFtJtxcF
xcFtDxcF∆+=∆
∆−=
Continuum fluid regime
Hydrodynamic code TsunamiHYLIFE - UCB (1979-2003)
NIF - LLNL, UCB (1994-1996)
solving Euler equation
2
2
21),,(
),,(
0
ueEpuEpuuF
EuUxF
tU
T
T
ρ
ρρρ
ρρρ
+=
++=
=
=∂∂
+∂∂
with the eq of state
),( Epp ρ=
Proposed extension to gas dynamics slip regimeKinetic theory of gas dynamics on a diffused surface in dilute gasconditions:• references include older theoretical studies of Cuette flow conditionsand newer numerical studies with DSMC methods (1988 - 2001)• models are based on imposing a velocity and temperature slip to thegas near the surface to compensate for the difference in the velocitydistribution of the particles approaching and leaving the surface• DSMC simulation show model is valid for Kn < 0.1
Proposed extension for Tsunami is based on Harvie and Fletcher study(2001) that explicitly include the mass flux in the velocity and temperatureslip formulation:
xvUvvi ∂∂
−= λ0
+
−
=ΦGG
xv
fffffUvvcmc
cml ∂
∂
−+Φ−−+
=−)1()1)(1(1
0 λxvv∂∂
Φ−=λ
0
xTUTTi ∂∂′−= λ0
459 −=′ γU
mcct ffff )1( −+=
xT
ffffUTT
ct
tcl ∂
∂
−Φ−+
−Φ−−′=−)1)(1(
)1(20 γ
λ
xTTTl ∂∂
Φ−Φ
=−4.04.1
35.110 λ
Droplet Clearing in a Pressure DecayField
Problem Definition: Droplet clearingrepresents another aspect of thechamber clearing issue. Dropletsproduced from the blast should becleared away before the next shot.
Approach: Start with thedevelopment of an incompressiblecode for analyzing droplet heat andmass transfer with respect to aknown pressure decay
Goal: Ultimately to couple theTsunani code with the developedincompressible free surface heatand mass transfer code for chamberclearing evaluation
Movie: A hot droplet reacting to thecold surrounding environment
Droplet Heat Transfer and Phase Change withTruchas: Preliminary Evaluation
Truchas is a software programdeveloped at LANL to simulatesolidification manufacturingprocesses, most notably metalcasting and welding operations.Include models and algorithmsfor:• Interfacial motion and heat
transfer• Properties varying with
temperature• Phase change• finite volume method• Simulations are fully 3
dimensional onunstructured grids.
Movie: A hot droplet falling down through a coldenvironment
Internal circulation caused by the temperaturedifference is now employed in Truchas code
through Boussinesq approximation
Tgg o∆=∆ αρρBuoyancy force
Melting temp=723KBottom temp=723 kTop temp=523 kInitial temp inside droplet=735 kinitial temp outside droplet=723 k
x
y
0 1 2 3 40
0.5
1
1.5
2
2.5
3
3.5
4
Assumptions & Near Term Goal
The mass evaporated from the droplet intothe surrounding pressure field will bediscarded
Incorporate a time dependent temperatureboundary condition (set at Tsatcorresponding to saturated temperate as afunction of known pressure decay)
Await approval from LLNL to modifyTruchas code