Parallel Computation for Natural Convection in Cavities

l’arallcl computation for thmnal coavcctive flows iII cavit ics with adiabatic horizontal boundarim and driven by diff’mmtial ll~ati~]g of I,llc two vertical end walls, is investigated usiug the lnt,c] l’araso]i, 1 rl tc] ‘J1OUC.lISi,OIIC 1 )clta, and Chay ‘J’31). A lmrallcl co~nput atioll code has lMXII ill]l)lelncllt,cd by usil)g a finite-diffcmlcc method with a II)ultigrid elliptic solver aIId a l)ufort-l~rankcl SCIICIIIC. ‘1’lIe domai]l dcx:o]r]])osil,ioll techniques arc discussed ill detail. ‘J’hc pa rallcl code is l)ulnclically stable, colnl)ut,ational]y efficient, and pc)rta blc to various lmallcl architmt,urcs which sul)lx)rt cil,hcr I’VM or NTX libraries for COIIIIIIU llicatliolls. l~illally, ]iumcrica] results for various l{aylcigll lIu It Il)(Irs aIId l’ralldtl numbers arc presented. 1. IN’1’ROl~lJCl’ION Cknlvcctivc ]notions d r i v e n by latcra] tcmpcratur( gradimlts jll cavjtics arc iIIl})orta IIt j]] II IalIy areas of il)tcmst in jndust, ry a~ld jl) IIat, urc. A~)l)lications jncludc llca,tj I)g al)d vcl)tilat,joll coIItrol in bui]dinp, dcsjgll al)d cons t ruc t ion , cooling systcll)s for IIuclcar reactors in tlIc nuc]car industry, solal-ciIcrgy collectors iI) tl)c l)owcr jl)dustry alId some otlIcr areas. l)UC to tlIc wide IaItgc of al)l)licatjons, studies of IIatural convection flow and l)cat tralls[cr IIavc been vigorously pursued for InaTIy years. A typical Ino(lcl of co]lvcctioll” drivml by a latcra] thcrlnal gradicl)t, consis ts of a two dilncllsjonal Iwctallgular cavity wjtl) tl)c two vcrtica] cnd walls lIcld at d i [[crcIIt col)stallt, tc~nj)craturcs. III order to dctcrlnil)c ilIc {low st,ruciurc a]ld lIcai, trausfcr across cavities wit,l I diflcrcl)t physjcal propcrti(x, lIulncrous al)alylical, cxpcrjIncntal al)d compuhtiollal tccl)l)iqucs have bccII used [1] [2] [4] [5] . ‘1’l)c })rcsclltj s tudy wjll focus olI IIul))crical silnulat,io]l on large or sl~lall as~)cct ratios 1, cavjty flows (Ii >>1, or 1,<<1 ) wit], va r ious l)randt] nu,nlms a])d IUiylcigll nu,nbcrs by S(x!l’al parallc] Systclns [3] [7] [8] [9].