A　feature　extraction　method　for　gear　faults　based　on　Fractional　Fourier　Transform　(FRFT)　is　proposed.　In　this　method,　Fractional　Fourier　Transform　of　the　gear　vibration　signals　is　carried　out　by　using　transform　order　as　the　variable.　Peak　points　are　found　through　two-dimensional　search　according　to　threshold　in　the　fractional　Fourier　plane　formed　by　signal　energy,　where　the　optimal　order　is　determined.　The　construction　of　a　narrow　band　pass　filter　could　be　used　to　separate　the　useful　components　of　gear　vibration　signal　from　the　background　noise.　The　wavelet　packet　decomposition　and　reconstruction　is　then　applied　to　the　useful　components　extracted　from　the　gear　vibration　signal　to　obtain　the　feature　signal　of　frequency　bands.　And　it　is　lined　up　as　gear　fault　feature　to　each　frequency　band　feature　signal　of　the　highest　decomposition　layer　after　wavelet　packet　decomposition.　Experimental　results　show　that　the　gear　fault　feature　extraction　method　based　on　FRFT　proposed　in　this　paper　is　able　to　availably　suppress　background　noise　and　highlight　the　target　component.　The　wavelet　packet　frequency　band　energy　can　accurately　reflect　the　gear　meshing　transmission　information,　which　can　effectively　extract　the　gear　fault　features.　©　2017,　Chinese　Mechanical　Engineering　Society.　All　right　reserved.