b"PageNo.5 Herewehavetakena100'roofsectionandappliedaloadcausing theroofingmembranetostretch1-1/2%toalengthof101-1/2'. Becauseoftheelongationpropertiesofthisbuilt-upmembrane,the membraneitself didnotfail.However,astheloadreleasedand thedeckreturnedtonorma l ,theroofingmembranedidnotreturn toitsnormallengthandroofbucklesweretheresult.Thesebuckles willridthemselvesofthetarsandbitumenandprematureroof failureswillresult. Tosummarize,allbuilt-usstemswillerforrnifthere isnobuildingordeckmov~ment.Whereorcesareconcentrated andresultincracks,theforceexertedbyexpansionandcontraction isgreaterthanthestrengthorelongationabilityofanybuilt-up roofsystem.Becauseofbuildinanddeckmovementbuilt-uroof systemsshouldnotbefastenedir~ctlytote deck. Illustration#6verysimplyconveysthe~roblemofexpansion andcontractioninrelationtothebuilt-uproof : Thecartoonshowstheroofmembraneor125lb.weaklingingrips withthe33,000lb.structuralmuscle.Asthis33,000lb.muscle flexes,abuilt-uproofwilldooneoftwothings:Itwillelongate andreturn,formingbucklesandprematurerooffailures,orit will stretchandsplittheroofingmembraneandcauseimmediateroof failures . ReferbacktoIllustration#1,thegraphshowingthe comparativestrengthofthe17built-uproofsamplesthatwere cutfromroofsthathadnotfailed,andwerebroughttothe1960, convention.Thesebuilt-uproofshadanaverageof103lbs.tensile strength.Theylastedfrom20to54years;onereason:theydid1 nothavetofightany33,000lb.muscles.Inthepast54yearst Mese roofshavegonethroughsometerrificthermalshocks.Yet,they ' didnotfail . I \\"