0 valutazioniIl 0% ha trovato utile questo documento (0 voti)
104 visualizzazioni33 pagine
This document provides an overview of discrete phase modeling in computational fluid dynamics. It discusses particle tracking methods including stochastic tracking to model turbulent dispersion. Key aspects covered include the discrete phase model, coupling between phases, particle types and properties, injection definitions, solution strategies, and applications to particle trajectories, mixing, and mass transfer. Visualization techniques like sample planes and Poincare plots are also introduced.
This document provides an overview of discrete phase modeling in computational fluid dynamics. It discusses particle tracking methods including stochastic tracking to model turbulent dispersion. Key aspects covered include the discrete phase model, coupling between phases, particle types and properties, injection definitions, solution strategies, and applications to particle trajectories, mixing, and mass transfer. Visualization techniques like sample planes and Poincare plots are also introduced.
This document provides an overview of discrete phase modeling in computational fluid dynamics. It discusses particle tracking methods including stochastic tracking to model turbulent dispersion. Key aspects covered include the discrete phase model, coupling between phases, particle types and properties, injection definitions, solution strategies, and applications to particle trajectories, mixing, and mass transfer. Visualization techniques like sample planes and Poincare plots are also introduced.
Applied Computational Fluid Dynamics Instructor: Andr a!!er " Andr a!!er #$%%$-$%%&' " Fluent Inc( #$%%$' 2 Discrete phase modeling ) Particle trac!ing( ) *teady +s( unsteady( ) Coupled +s( uncoupled( ) Ad+antages and limitations( ) ,ime stepping( ) Discreti-ation( Particle trajectories in a spray dryer Particle trajectories in a cyclone 3 Discrete phase model ) ,ra.ectories o/ particles0droplets are computed in a Lagrangian /rame( 1 23change #couple' heat4 mass4 and momentum 5ith 2ulerian /rame gas phase( ) Discrete phase volume /raction should pre/era6ly 6e less than 1%7( 1 Mass loading can 6e large #81%%7'( 1 9o particle-particle interaction or 6rea! up( ) ,ur6ulent dispersion modeled 6y: 1 *tochastic trac!ing( 1 Particle cloud model( ) Model particle separation4 spray drying4 li:uid /uel or coal com6ustion4 etc( continuous phase flow field calculation particle trajectory calculation update continuous phase source terms 4 DPM theory ,ra.ectory is calculated 6y integrating the particle /orce 6alance e:uation: typical continuous phase control volume mass, momentum and heat exchange ( ) p i p p i p i i D p i F g u u F dt du / / ) + + = drag force is a function of the relative velocity !dditional forces" Pressure gradient #hermophoretic $otating reference frame %rownian motion &affman lift 'ther user defined) (ravity force ) Coupling 6et5een phases ) ;ne-5ay coupling: 1 Fluid phase in/luences particulate phase +ia drag and tur6ulence( 1 Particulate phase has no in/luence on the gas phase( ) ,5o-5ay coupling: 1 Fluid phase in/luences particulate phase +ia drag and tur6ulence( 1 Particulate phase in/luences /luid phase +ia source terms o/ mass4 momentum4 and energy( 1 23amples include: ) Inert particle heating and cooling( ) Droplet e+aporation( ) Droplet 6oiling( ) De+olatili-ation( ) *ur/ace com6ustion( * ) Particle types are inert4 droplet and com6usting particle( Particle types + # e m p e r a t u r e particle time ,nert heating law -apori.ation law %oiling law # / # v # injection <eat and mass trans/er to a droplet 0 volatile fraction flashes to vapor 1scape $eflect #rap Particle-5all interaction ) Particle 6oundary conditions at 5alls4 inlets4 and outlets: ) For particle re/lection4 a restitution coe//icient e is speci/ied: t t t n n n v v e component: Tangential v v e component: Normal , 1 , 2 , 1 , 2 = = 2 Particle /ates ) =2scaped> tra.ectories are those that terminate at a /lo5 6oundary /or 5hich the =escape> condition is set( ) =Incomplete> tra.ectories are those that 5ere terminated 5hen the ma3imum allo5ed num6er o/ time steps 5as e3ceeded( ) =,rapped> tra.ectories are those that terminate at a /lo5 6oundary 5here the =trap> condition has 6een set( ) =2+aporated> tra.ectories include those tra.ectories along 5hich the particles 5ere e+aporated 5ithin the domain( ) =A6orted> tra.ectories are those that /ail to complete due to numerical0round-o// reasons( I/ there are many a6orted particles4 try to redo the calculation 5ith a modi/ied length scale and0or di//erent initial conditions( 11 ) 2ach in.ection is trac!ed repeatedly in order to generate a statistically meaning/ul sampling( ) Mass /lo5 rates and e3change source terms /or each in.ection are di+ided e:ually among the multiple stochastic trac!s( ) ,ur6ulent /luctuations in the /lo5 /ield are represented 6y de/ining an instantaneous /luid +elocity: ) 5here is deri+ed /rom the local tur6ulence parameters: ) and is a normally distri6uted random num6er( i i i u u u 3 + = i u3 3 2 3 k i u =
,ur6ulence: discrete random 5al! trac!ing
12 ) *tochastic trac!ing turned on( ) Fi+e trac!s per in.ection point( ) Adds random tur6ulent dispersion to each trac!( ) ,rac!s that start in the same point are all di//erent( ) *tochastic trac!ing turned o//( ) ;ne trac! per in.ection point( ) ?ses steady state +elocities only and ignores e//ect o/ tur6ulence( *tochastic trac!ing e3ample - paper plane 1) In.ection set-up ) In.ections may 6e de/ined as: 1 *ingle: a particle stream is in.ected /rom a single point( 1 @roup: particle streams are in.ected along a line( 1 Cone: #A-D' particle streams are in.ected in a conical pattern( 1 *ur/ace: particle streams are in.ected /rom a sur/ace #one /rom each /ace'( 1 File: particle streams in.ection locations and initial conditions are read in /rom an e3ternal /ile( 1* In.ection de/inition ) 2+ery in.ection de/inition includes: 1 Particle type #inert4 droplet4 or com6usting particle'( 1 Material #/rom data 6ase'( 1 Initial conditions #e3cept 5hen read /rom a /ile'( ) Com6usting particles and droplets re:uire de/inition o/ destination species( ) Com6usting particles may include an e+aporating material( ) ,ur6ulent dispersion may 6e modeled 6y stochastic trac!ing( 1+ *olution strategy: particle trac!ing ) Cell should 6e crossed in a minimum o/ t5o or three particle steps( More is 6etter( ) Ad.ust step length to either a small si-e4 or $% or more steps per cell( ) Ad.ust =Ma3imum 9um6er o/ *teps(> ) ,a!e care /or recirculation -ones( ) <eat and mass trans/er: reduce the step length i/ particle temperature 5ildly /luctuates at high +apori-ation heats( 12 Particle trac!ing in unsteady /lo5s ) 2ach particle ad+anced in time along 5ith the /lo5( ) For coupled /lo5s using implicit time stepping4 su6-iterations /or the particle trac!ing are per/ormed 5ithin each time step( ) For non-coupled /lo5s or coupled /lo5s 5ith e3plicit time stepping4 particles are ad+anced at the end o/ each time step( 24 *ample planes and particle histograms ) ,rac! mean particle tra.ectory as particles pass through sample planes #lines in $D'4 properties #position4 +elocity4 etc(' are 5ritten to /iles( ) ,hese /iles can then 6e read into the histogram plotting tool to plot histograms o/ residence time and distri6utions o/ particle properties( ) ,he particle property mean and standard de+iation are also reported( 21 Poincar plots ) Poincar plots are made 6y placing a dot on a gi+en sur/ace e+ery time a particle tra.ectory hits or crosses that sur/ace( ) <ere it is sho5n /or a /lo5 inside a closed ca+ity 5ith tangentially oscillating 5alls( ) ,he /igure on the le/t sho5s streamlines( ) ,he /igure on the right sho5s a Poincar plot /or the top sur/ace( ) ,his method can 6e used to +isuali-e /lo5 structures( Aref and Naschie. Chaos applied to fluid mixing. Page 187. 1995. 22 Leap/rogging +orte3 rings ) ,5o ideal coa3ial +orte3 rings 5ith the same sense o/ rotation 5ill leap/rog each other( ) ,he /or5ard +orte3 increases in diameter and slo5s do5n( ,he rear5ard +orte3 shrin!s and speeds up( ) ;nce the +ortices traded places4 the process repeats( ) ,he photographs on the le/t are e3perimental +isuali-ations using smo!e rings4 and the /igures on the right are Poincar plots /rom a CFD simulation sho5ing the same structures( Aref and Naschie. Chaos applied to fluid mixing. Page 33 187. 1995. 23 Massi+e particle trac!ing ) Massi+e particle trac!ing re/ers to analyses 5here tens o/ thousands to millions o/ particles are trac!ed to +isuali-e /lo5s or to deri+e statistics o/ the /lo5 /ield( ) ,5o e3amples: 1 A mi3ing tan! 5ith three impellers( 1 A mi3ing tan! 5ith /our impellers( 1 oth animations sho5 the motion o/ more than $B! particles( 1 It can 6e seen that one large /lo5 loop /orms in the three impeller system4 and t5o /lo5 loops /orm in the /our impeller system( 24 ,hree impellers Animation courtesy of ightnin !nc. 2) Four impellers Animation courtesy of ightnin !nc. 2* Mi3ing +essel - +elocity +ectors ) *maller diameter impeller #&%7 o/ +essel diameter'( ) Impeller .et e3tends to the +essel 6ottom( ) Larger diameter impeller #5%7 o/ +essel diameter'( ) Impeller .et 6ends o// to the 5all and the /lo5 direction at the 6ottom is re+ersed( 2+ Mi3ing +essel - trac!ing o/ sand particles ) *maller diameter impeller( ) ,he sand is dispersed throughout the 5hole +essel 5ith a small dead spot in the center o/ the 6ottom( ) Larger diameter impeller( ) Due to the re+ersed /lo5 pattern at the 6ottom4 sand does not get suspended throughout the 5hole +essel( 20 Cenics static mi3er ) A static mi3er is a de+ice in 5hich /luids are mi3ed4 6ut has no mo+ing parts( ) ,he mi3ing elements induce a /lo5 pattern that results in the material 6eing stretched and /olded to /orm e+er smaller structures( ) Mi3ing can 6e analy-ed 6y loo!ing at species concentration or 6y loo!ing at particle paths( ) CP? time: 1 1BB$( 1%%4%%% cells( *un *parc II 5or!station( ;ne 5ee!( 1 1BBA( A5%4%%% cells( Cray CB% supercomputer( ;+ernight( 1 $%%1( A5%4%%% cells( ,5o processor ?ni3 5or!station( A% minutes( 1 $%%1( ,5o-million cells( ,5o processor ?ni3 5or!station( Fi+e hours( 22 Lattice-olt-mann Method ) Calculations 6y Dos Der!sen4 Del/t ?ni+ersity4 $%%A( ) ?n6a//led stirred tan! e:uipped 5ith a Eushton tur6ine( Cross *ection Fessel Gall 34 Lattice-olt-mann Method ) Calculations 6y Dos Der!sen4 Del/t ?ni+ersity4 $%%A( ) ?n6a//led stirred tan! e:uipped 5ith /our impellers( 31 Mi3ing mechanism ) Laminar mi3ing( ) CFD simulation( ) *i3 elements( ) 2ach element splits4 stretches and /olds the /luid parcels( ) 2+ery t5o elements the /luid is mo+ed inside-out( 32 Particle trac!ing animations 33 *pecies mi3ing animation 34 Huestion: are particle tra.ectories closedI ) 9ot in tur6ulent /lo5s( ) Fiscous4 periodic /lo5s may ha+e periodic points( ,hese are points 5here the particle returns to its initial position a/ter a certain amount o/ time( ) rou5erJs /i3ed-point theorem: Under a continuous mapping f : S S of an n- dimensional simplex into itself there exists at least one point x S such that f (x)x! ) Application to particle trac!s: 1 In +iscous periodic /lo5s in closed4 simply connected domains there 5ill al5ays 6e at least one periodic point 5here a particle returned to its original location( 1 In other situations4 there is no guarantee that there is any closed tra.ectory4 and there may 6e none at all( 3) Huestion: ho5 /ast do particles separateI ) I/ 5e place t5o particles in/initesimally close together4 5ill they stay together4 or separateI ) ,he separation distance is go+erned 6y the Lyapuno+ e3ponent o/ the /lo54 5hich states that the particles 5ill separate e3ponentially as a /unction o/ time t: ) ,he higher the Lyapuno+ e3ponent4 the more chaotic the /lo5 and the more stretching occurs( ) Lyapuno+ e3ponents can ha+e any +alue4 most o/ the time 6et5een % and 1%4 and usually 6et5een %(5 and 1( t e t
) 4 ) = 3* Particle trac!ing accuracy ) ,here are three types o/ errors: discreti-ation4 time integration4 and round-o//( ) Eesearch has sho5n that in regular laminar /lo5s the error in the particle location increases as t"4 and in chaotic /lo5s almost e3ponentially( ) 2rrors tend to align 5ith the direction o/ the streamlines in most /lo5s( ) As a result4 e+en though errors multiply rapidly #e(g( %(17 error /or $%4%%% steps is 1(%%1 $%4%%% K &(L2L'4 :ualitati+e /eatures o/ the /lo5 as sho5n 6y the de/ormation o/ material lines can 6e properly reproduced( ut the length o/ the material lines may 6e o/ 6y as much as 1%%7( ) ;+erall4 particle trac!ing4 5hen properly done4 is less di//usi+e than sol+ing /or species transport4 6ut numerical di//usion does e3ist( 3+ *ummary ) 2asy-to-use model( ) Clear and simple physics( ) Eestricted to +olume /ractions M 1% 7( ) Particle trac!ing can 6e used /or a +ariety o/ purposes: 1 Fisuali-ation( 1 Eesidence time calculations( 1 Com6ustion( 1 Chemical reaction( 1 Drying( 1 Particle /ormation processes(