Computing flame dynamics using massively parallel computers to span scales from the atomistic to the industrial

C o m p u tin g F la m e D y n a m ic s u s in g m a s s iv e ly p a ra lle l c o m p u te r s to s p a n s c a le s fro m th e a to m is tic to th e in d u s tria l Philip S m ith - p r o f e s s o r & ch a ir C h em ica l E n gin eerin g The U n iversity o f Utah F la m e D y n a m ic s example: glass furnace 1 1 P o o l F ire s o b je c tiv e : h e a t f u x to e m b e d d e d o b je c ts (5 0 0 fps) laboratory 0.3m pan fire 5.00E-06 4.50E-06 c 0 4.00E-06 y3 0 3.50E-06 £ 3.00E-06 Q 2.50E-06 2.00E-06 g 1.50E-06 0 0 1.00E-06 W 5.00E-07 0.00E+00 - Batch Continuous position (mm) 0 0 1000 2000 3000 4000 5000 Time, s L a rg e E d d y a s s iv e ly S im u la tio n s & P a ra lle l C o m p u te r s - pool:....... 10m heptane ^ computer:........... LANL (nirvana) 1000processors resolution:............ 300s ^ compute time.... 3 6 hrs real time:.....8 sec. fire a© 3003 X Ks \V X 1503 number of processors volume ren dered tem perature 2 L E S N u m e r ic s - D is c re tiz a tio n : ( s t a g g e r e d m e s h ) time discretization: 2nd&3rdorder strong stability preserving (SSP) Runge-Kutta methods » spatial velocity discretization: ° non-dissipative 2ndorder central differencing ° conservative for momentum & kinetic energy » spati al scalar d iscretization: weighted higher-order Essentially NonOscillatory (WENO) scheme o uniformly high order accurate but somewhat dissipative (less dissipative than ENO) 3 J P 8 s u rro g a te G a s c h e m . k in e tic m e c h a n is m ^ „ . _ . ^ J P 8 S u r r o g a t e Fuel (m ol %): ^ G a s M e c h a n ism : a. 190 Species a 912 Reactions 1% Benzene <*• 26% Toluene * 5% Methyl Cyclo Hexane (MCH) » 3% Iso-Octane (i-C8) <*• 65% n-Dodecane (n-C12) Stoichiometric mixture fraction: 0.0639 <f soot formation process M o le c u la r S c a le s o x id a tio n 50 ms • w * M o le c u la r D y n a m ic s "■ limited to time steps < vibrational period K in e tic M o n te C a rlo a time step determined by kinetics ~ms ^ A p p r o a c h to b r id g e t h e tim e s c a l e s Fractal clusters a g g l o m e r a t i o n Fr; 10-30 nm c a r b o n iz a tio n 1 §8o < G ro w th b y s u rfa c e r e a c t io n a n d c o a g u la tio n P a r tic le in c e p tio n Diameter = 1-2 nm P o ly m e r iz a t io n 10 ms P A H f o r m a tio n 1 ms 2‘'3 4 R e a c t i o n M o d e l: a n ifo ld lib r a r ie s r e a c ti o n m o d e l p a r a m e t e r iz a tio n & lib ra ry g e n e r a ti o n u s e DNS to e v a l u a te e f f e c tiv e n e s s o f r e a c ti o n m o d e l f o r d e s c r ib i n g s t a t e s p a c e v a r ia b le s S G S tu rb u le n c e /m ix in g m o d e l: O D T (o n e -d im e n s io n a l tu rb u le n c e ) ODT represents a notional "line o f sight" through a fully resolved flow field. in collaboration with Alan Kerstein &Rod Schmidt, SNL 5 6 V e rific a tio n & V a lid a tio n S c h e m a follow in g O berkam pf, S a n d ia N atio n a l L ab Verification: The process of determining that a model implementation accurately represents the developer's conceptual description of the model and the solution to the model. Validation: The process of determining the degree to which a model is an accurate representation of the real world from the perspective of the intended uses of the model. F ire V & V Verification P y ra m id C o m p le te System M etal c o n ta in e r w ith HMX in a JP -8 po o l fire 1 0 m JP -8 po o l fire S u b syste m C ases / 1 m M ethane b u o y a n t fla m e i 2 m JP -8 pool fire M etal c o n ta in e r in a fla m e x B e n c h m a rk C ases B u o ya n cy d riv e n fla m e s (lik e TN F d a ta ) T u rb u le n t flo w & h e a t tra n s fe r p a s t c o m p le x g e o m e try C o u p le d P ro b le m s 7 F ir e V e r ific a tio n O b je c tiv e : to r e d u c e - Tools: ^ d is c r e tiz a tio n e r r o r 4 spatial, temporal, and boundary conditions la c k o f ite r a tiv e co n v erg en ce a lg o rith m ic e r r o r s a Scalability, Portability p e rfo rm an c e in e ffic ie n c y » scalability, % of peak performance, & portability ^ c o m p u te r p r o g r a m m in g e r r o r s ^ a n a ly tic s o lu tio n s ^ m e th o d o f m a n u fa c tu re d s o lu tio n s ^ b e n c h m a r k s o lu tio n s ^ g rid c o n v e r g e n c e s tu d ie s ^ s in g le a r id e r r o r e stim a to rs F ire V & V P y ra m id C o m p le te System M etal c o n ta in e r w ith HMX in a JP -8 po o l fire 1 0 m JP -8 po o l fire S u b syste m C ases / 1 m M ethane b u o y a n t fla m e i x 2 m JP -8 pool fire M etal c o n ta in e r in a fla m e -- I B e n c h m a rk C ases B u o ya n cy d riv e n fla m e s (lik e TN F d a ta ) T u rb u le n t flo w & h e a t tra n s fe r p a s t c o m p le x g e o m e try C o u p le d P ro b le m s 8 V a lid a tio n o f C u t-C e ll M e th o d Flow Interaction with a Cylinder or Sphere (2D and 3D) Inclined Channel Flow P.G. Tucker and Z. Pan, 2000 etal., etal., J.-H. Chen 1995; Ye 1999; Kirkpatrick 2003; D. Kim and H. Choi, 2002 Concentric and Non-concentric Journal Bearing Flow Laminar Mixing Tank 1111 A. Souvaliotis I win M. Abid et al., 1994; A.D. Harvey et al., 1996; Y. Xu and G. McGrath, 1996; V.V. Ranade, 1997 etal., 1995; J.M. Ottino 1989 (book) F ire V & V et al., P y ra m id C o m p le te System M etal c o n ta in e r w ith HMX in a JP -8 po o l fire 1 0 m JP -8 po o l fire S u b syste m C ases 1 m M ethane b u o y a n t fla m e 2 m JP -8 pool fire M etal c o n ta in e r in a fla m e B e n c h m a rk C ases B u o ya n cy d riv e n fla m e s (lik e TN F d a ta ) T u rb u le n t flo w & h e a t tra n s fe r p a s t c o m p le x g e o m e try C o u p le d P ro b le m s 9 D N S a p r io r i e v a l u a t i o n • ^ c o m b u s tio n r e a c tio n o f m o d e ls o.a -----f,X ideal ■- SLFM ...... f ideal - ■ EQ = 0.4| CD E o 0.2 10 10 id 2 1 7. S ) 10 73.5 mol% n-dodecane 5.5 mol% i-octane 10 mol% MCH 11 mol% aromatic V alid atio n o f G a s - P h a s e C h e m ic a l K inetic M e c h a n ism fo r S u r r o g a te M ixture 2.5 20 04 1.5 1.0 05 2 ‘■{C <20 CH*103 60 30 40 10 20 0.5 05 1.0 0.1 0.2 0.3 0.4 0.1 0.2 0.3 0.4 0.2 0.3 0.4 height above burner, cm Data reference: D oute', C., Delfau, J.L., Akrich, R. and Vovelle, C. (1995) Chem ical structure o f atm ospheric pressure prem ixed n-decane and kerosene flam es Combust. Sci. and Tech. 106(4-6): 327. F ire V & V P y ra m id C o m p le te System M etal c o n ta in e r w ith HMX in a JP -8 po o l fire 1 0 m JP -8 po o l fire S u b syste m C ases / 1 m M ethane b u o y a n t fla m e i x 2 m JP -8 pool fire M etal c o n ta in e r in a fla m e -- I B e n c h m a rk C ases B u o ya n cy d riv e n fla m e s (lik e TN F d a ta ) T u rb u le n t flo w & h e a t tra n s fe r p a s t c o m p le x g e o m e try C o u p le d P ro b le m s 11 O D T S G S V a l id a t i o n : C B C d a t a : d e c a y in g tu r b u le n c e F ire V & V P y r a m id C o m p le te System M etal c o n ta in e r w ith HMX in a JP -8 po o l fire S u b syste m C ases 1 m M ethane b u o y a n t fla m e 2 m JP -8 pool fire M etal c o n ta in e r in a fla m e 12 T e m p e r a tu r e , s o o t v o lu m e f r a c tio n s a lo n g o + 0 A □ 500 th e a x ia l h e ig h t Tem perature (Expt) Tem perature (10) Tem perature (25) S oot Vol. F raction (Expt) S oot Vol. F raction (10) S oot Vol. F raction (25) 1.0E -06 6.0E -07 750 a ♦ 500 4.0 E -0 7 2.0E -07 250 ; 0 F r a c tio n ❖ Soot Volume ❖ 000 O El T e m p e ra tu re (K ) 250 0.0E+00 0 0.5 1 1.5 2 z,m F ire V & V P y r a m id C o m p le te System M etal c o n ta in e r w ith HMX in a JP -8 po o l fire 1 0 m JP -8 po o l fire S u b syste m C ases / 1 m M ethane b u o y a n t fla m e i x 2 m JP -8 pool fire M etal c o n ta in e r in a fla m e 1 B e n c h m a rk C ases B u o ya n cy d riv e n fla m e s (lik e TN F d a ta ) T u rb u le n t flo w & h e a t tra n s fe r p a s t c o m p le x g e o m e try C o u p le d P ro b le m s 13 H M X in a F ir e - T i m e t o fir e & c a n is te r c o m p u ta tio n E x p lo sio n : rVo Conp«nwn iw CatWMT Tapand Baton ♦ ■ 4 • +- All et al, 99: 0.75 atmair All et al, 99: 1atmair Vilynov and Zarko, 89 Strakovski (1989) Lengelle (1985) Atwood (1988) C-SAFE (1999) O C-SAFE (2001) X Flux to HMX Surface Flux to Steel Surface Cookoff I 10 Heat Flux (W/cm2) Complete System Prediction/Minimization of NO x emissions from glass furnaces Combustion Space IFRF Glass Furnace r Glass Melt Single Port Model T wo Port Model TNF Flame Laminar CFD Turbulence Model Mixing Model Reaction Model NOx Model Radiation Model Soot Model 14 C o n c lu s io n ;:‘ # V V & V - L E S w ith g e n e r a l i z a b l e s g s m o d e lin g a b s tr a c tio n (surrogate) Fuel a few species Molecular Scale feasible chemical kinetic mechanism Reaction Model continuum micro -reduce degrees of freedom (resolved field vars) -molecular scale transport -radiation props. Mixing Model mesoscale model (subgrid hetero. due to turbulence) -nonlinear -filtering/ averaging 100's of species + / P o o l F ire in C r o s s w i n d h e p ta n e p o o l - 5 m /s c r o s s w in d v P j 1 8 .5 mLLNL(frost) computer: 490p resolution:.......... 2003 ^ compute time ....90 hrs fire time:........... 12 sec. 15