This　paper　presents　a　numerical　study　on　the　yield　resistance　and　ultimate　resistance　of　steel　beam　with　random　corrosion　damage.　The　corrosion　is　considered　by　introducing　random　cylindrical　pits　to　the　intact　steel　beam,　in　which　the　thickness　of　the　beam,　both　at　the　section　and　along　the　length　of　the　beam　is　reduced.　Altogether　240　corroded　beams　are　numerically　studied,　wherein　the　effects　of　corrosion　ratio,　corrosion　diameter,　corrosion　depth　and　corrosion　location　on　the　yield　resistance　and　ultimate　resistance　of　steel　beam　are　considered.　It　is　found　that　for　global　corrosion　scenario,　when　the　corrosion　ratio　is　4.76%,　9.26%,　13.37%,　18.25%,　22.36%,　and　25.76%,　due　to　the　random　nature　of　corrosion,　the　reduction　factors　of　ultimate　resistance　of　steel　beam　range　from　0.9-0.92,　0.81-0.86,　0.77-0.8,　0.71-0.76,　0.64-0.67,　and　0.58-0.64,　respectively.　Moreover,　when　corrosion　distributes　only　at　the　bottom　flange　or　top　flange　of　steel　beam,　the　adverse　effects　of　bottom　flange　corrosion　and　top　flange　corrosion　on　the　resistance　of　steel　beam　are　the　same.　The　corrosion　diameter　and　corrosion　depth　have　limited　effects　on　the　resistance　of　steel　beam.　The　relationship　between　the　reduction　factors　for　the　yield　resistance　and　ultimate　resistance　with　the　corrosion　ratio　of　the　beam　is　proposed.　It　is　found　that　for　the　global　corrosion　case,　the　reduction　factors　of　the　yield　resistance　and　ultimate　resistance　of　the　beam　are　linearly　and　negatively　correlated　with　the　corrosion　ratio.　For　each　10%　increase　in　the　corrosion　ratio,　the　reduction　factor　of　yield　resistance　and　ultimate　resistance　decrease　by　18.9%　and　15.1%,　respectively.　The　resistances　of　the　corroded　beam　with　random　corrosion　pits　and　uniform　corrosion　are　also　compared.　The　results　suggest　that　by　using　uniform　corrosion　model,　the　resistance　of　the　corroded　steel　beam　will　be　significantly　overestimated.　When　the　corrosion　ratio　is　about　14%,　the　overestimation　on　the　ultimate　resistance　of　steel　beam　by　using　uniform　corrosion　model　is　more　than　10%.