Numerical calculation of subsonic jets in crossflow with reduced numerical diffusion

Abstract

A serles of calculatlons are reported for two, expenmentally studled, subsonlc Jet In crossflow geometrles. The parametrlc varlatlon examlned lnvolves the lateral spaclng of a row of Jets. The flrst serles of calculatlons corresponds to a wldely spaced Jet geometry, SID = 4, and the second serles corresponds to closely spaced Jets, SID = 2. The calculatlons are done wlth alternate dlfferenClng schemes to lllustrate the lmpact of numer1cal d1ffus1on. The calculated Jet traJectorles agreed well wlth experlmental data In the w1dely spaced Jet geometry, but not 1n the closely spaced geometry. Introduct10nIn gas turblne englnes the m1xlng of Jets 1n crossflow plays a domlnant role In establlshlng the temperature proflle leavlng the combustor. Th1s temperature proflle, In turn, slgnlflcantly affects the durab1llty of the turb1ne. It 1S the process of tallorlng thlS temperature proflle lnto a pattern acceptable to the turblne that usually consumes the greatest amount of des1gn and development testlng In the evolutlon of a new combustlon system.To reduce thlS deslgn and development t1me (l.e., expense) combUStor deslgners currently use emplr1cal correlat1ons to deslgn rows of Jets In the combustor wall that wlll provlde an acceptable temperature proflle. The dlfflculty arlses due to the llmlted app11cablllty of these correlatlons. Emplrlca1 corre1atlons are constructed from "ldea11zed" flow fle1d experlments. The flow fle1ds In practlca1 combustlon systems exhlblt severe nonunlformltles 1n the ve10clty and temperature fle1ds lnto Wh1Ch the Jets are lnJected. Emplr1ca1 corre1atlons can be used In these sltuatlons only at great rlsk.Wlth the lncreased capabl11tles of current comput1ng systems, a much more promls1ng approach 1S to employ numerlcal flow codes In the deslgn process. Th1S approach also lnvolves some r1sk Slnce current combustor flow codes have not been fully verlf1ed agalnst most of the complex flow fle1d features occurrlng wlthln the ~ombustor. Indeed as noted by Kenworthy et a1., the deflclencles of current computer codes make the pred1ctlon of combustor eXlt temperature profl1es untenable. \Currently there are two maln factors 11mltlng the predlctlve accuracy of combustor flow codes. Flrst, the proper phYS1CS must be represented In the equatlons solved by the numerlca1 code. The actual phYS1CS removed In the large number of mode11ng assumpt10ns may severely restrlct the usefu11ness of the code. Second, the numerlca1 accuracy of these codes must be 1mproved.Current computer codes employ upwlnd dlfferenc1ng WhlCh can lntroduce an apprec1ab1e error 1n the calculated results. ThlS error (or numer1cal dlffuslon) lS frequently of such a large magn1tude that lt swamps or obsecures the turbulence model used 1n the ca1culatlon. Stud1es under the Hot Sectlon Technology (HOST) aerotherma1 mode11ng program ldent1fled thlS error as be1ng one of the key "bottlenecks" to the development of lmp20~ed physlca1 submode1s In combustor flow codes. -Many prev10us numer1ca1 stud1es 2 ,5-7 have examlned Jets In crossf1ow....

Numerical studies of twin‐jet impingement for STOVL flow application

“…Although a two equation k-e turbulence model was employed in these calculations, the use of a first-order accurate scheme for convection and very coarse meshes made any statements on turbulence model accuracy difficult due to the intrusion of numerical diffusion. Several solutions of the Reynolds averaged Navier-Stokes equations have already been computed for both single (Baker et al, 1982;Chien and Schetz, 1975;Harloff and Lytle, 1988;Patankar et al, 1977;Roth, 1987;Sykes et al, 1986) and multiple jets in cross flow (Claus, 1985;Demuren, 1983;Khan et al, 1981). Calculations of a turbulent jet by Patankar et al (1977) exhibit good agreement for the jet centre-line and the rate of decay of the jet velocity but deviate significantly from the experimental velocity profiles along the jet axis after the jet deflects in the reversed flow region.…”

K-epsilon turbulence model assessment with reduced numerical diffusion for coaxial jets

“…These eomputetions have been compared with detailed experimental data obtained using a tvo-component phsse/Doppler technique. Prior studies which deal with some of the aforementioned issues related to the evaluation of turbulence models include the work of Leschziner and Redi (Ref 13,14), Hacliman et al (Ref 15), and Claus (Ref 16).…”

Computational fluid dynamics prediction of intake ingestion relevant to short take-off and vertical landing aircraft

“…Childs and Nixon [27] presented calculations relevant to ground-effect flow using the k±å model and confirmed the gross features of the flow, but little comparison was given with experimental data to enable a quantitative judgement of the calculations. Claus [26] has addressed the question of false diffusion in three-dimensional jet-in-crossflow calculations and found it is necessary to adopt second-order QUICK discretization for the convective terms in order to obtain numerically accurate solutions. The k±å turbulence model with the QUICK numerical scheme was also successfully employed by Marquis [28], who concluded that the turbulent structure of the flow was independent of the velocity ratio (R jet velocity/crossflow velocity) as long as this was relatively high (of order 30).…”