Reliability　evaluation　of　complex　network　is　one　of　main　topics　in　complex　engineering　systems,　especially　for　Internet　of　Things　(IoT).　The　reliability　of　IoT　partially　depends　on　its　large-scale　network.　Especially,　the　lifetime　distribution　of　large-scale　network　is　critical　for　its　health　management.　However,　the　large　scale　of　the　network　usually　leads　to　an　expensive　simulation　time　cost.　Instead　of　direct　simulation,　we　propose　a　method　to　infer　the　large-scale　network　lifetime　using　small-scale　networks　with　the　universal　scaling　function.　We　first　find　the　scaling　relationships　between　network　lifetime　and　network　size　in　a　network　model　with　failure　coupling　for　two-dimensional　square　lattice　network　and　Cayley　tree　network.　Network　lifetime　with　different　size　can　be　described　by　one　universal　scaling　function.　Then　we　perform　theoretical　analysis　to　derive　the　scaling　relationships.　Finally　we　apply　these　scaling　relationships　to　wireless　sensor　network　with　coupled　failures　in　more　realistic　situation　as　case　study.　From　the　simulation　results　of　smaller-scale　networks,　we　can　infer　the　lifetime　distribution　of　large-scale　networks　based　on　universal　scaling　function.　The　computation　time　and　accuracy　are　compared　with　standard　Monte　Carlo　simulation　which　shows　that　our　method　is　faster　and　accurate.　Our　research　shows　that　the　proposed　method　using　universal　scaling　functions　can　help　us　to　infer　the　lifetime　properties　of　large-scale　networks　with　low　computational　cost.　Our　method　can　help　fast　reliability　evaluations　of　large-scale　complex　networks　with　high　accuracy.　IEEE