Page 76 - konishipaper

wasasurprise，b~cause theheadoftheliquidconvectionpreceded theflame'sIeadingedgebyaproximately1cm[Ito，Masuda，and Saito，191]，whichisentirelydiferent仕omAkita'sresult[1972J. To make sure of the ac curacy of the data， we repeated the experimentsseveraltimesandfoundthereproducibilitytobewithin 土 90%. Thus，wepropos巴danewinechanism:theflamespreadin theuIUformspread回 gionisgovemedbytheIiquidconvection[Ito， Masuda，andSaito，191]. However，atthattimewedidnotrealy understandwhyAkita'sshadowgraftresultsweresodiferentfrom ourmresuIts.
Later，HowardRos'sgroupatNASALewisResearchCenter apIiedRainbowSchIierenDeflectometry(RSD)tomeasureliquid convection generated by the spreading flame [MiI Ier and Ros s， 192]. TheyfoundlitleIiquidconvectionaheadoftheflame's leadingedgeintheuniformspreadregion. TheirRSDdatawere similartoAkita'sshadowgraphsanddisagredwithourHIresults. Overthepastthreyears，bothNASAandourgrouphavetriedto understandthereasonsforthedisagrement，butfailedtodosountil a'series of p釘 ametric experiments was completed. In those experiments，theefectsofsixdiferentparameterswereinvestigated: sensitivityofbothRSDandHIsysterns，impurityofpropanol(if any)，therelativehumidityoftheair，theambientairtemperaωre， severaldife:ienttypesofignitionmethod，andfinalydiferenttray widths. W efoundthatthe白rstfivepぽ ametershadlit tleinfluenceon thedi5agremen.t However，whenwecheckedthee百ectoftray width(0，5. 1，and2cm)ontheliquidconvection，theleastsusp回 ted parameter，asurprisingresultemerged: the0.5cmwidetrayhada largeIiquidconvectionintheuniformspreadregion，whilethetrays
白atwere1cmand2cmwidehada出inlayerofliquidconvection. Itwas50thinandsmalthatonlyaverysensitivenieasurement techniquecoulddetecti.t Theresultforthe.2・crntrayw部 very similartotheNASA'sresultforthe2・cmtray. A closerlo okattheir RSDresultalsorevealedasmalandthinlayerofIiquidconvection thatwasalwaysaheadoftheflame'sleadingedge. Comparisonof thesefindingsledustotheagrementthatintheuniformspread regionthereisalwaysasmalliquidconvectionindicatingthatthe majQrheattransferisbythatmode. Italsobecamec1eartha.tthe 5hadowgraph [Akita，1972] did not detect the liquid convection becausethesizeoftheIiquidconvect
'fheCurrentProblem NASA'smicrogravitytestresultsshowedthatpulsatingspread
didnotocurineithershaloword田 ppolexperiments[Ros， 194];insteadtheflamewasextinguished，indicatingtheimportance ofthebuoyancye能 ctinthegasphaseonthemechanismof pulsation. SchiIer and Sirignano，using numerical calculations [192]，predictedtheexi5tenceofaverysmalcirculationinthegas phaseaheadoftheflame'sleadingedge. Theysugestedthatthe gas~phase circulation，whichisuniqueinthepulsatingspreadand apearswiththeliquid~phase cirω，lation，mayplayanimportantrole
inflamepulsation[192，196]. 百lereareLDVflowmeasurement 142