Based　on　wavelet　packet　decomposition　and　conditions　of　the　support　hector　kernel　function,　a　non-linear　wavelet　basis　is　introduced　to　construct　the　kernel　function　of　support　vector　machine　(SVM),　and　a　tighten　wavelet　support　vector　machine　(WSVM),　which　has　strong　generalization　ability　is　obtained.　Wavelet　packet　decomposition　is　applied　to　the　structural　response　signals　under　ambient　vibration,　feature　vectors　are　obtained　by　feature　extraction　according　to　wavelet　energy　spectrum.　The　feature　vectors　are　used　for　training　and　classification　as　the　inputs　of　the　tighten　SVM.　Hence,　a　new　structural　damage　detection　method　called　complete　WSVM　is　established.　This　method　is　used　to　a　single-layer　spherical　lattice　dome　for　damage　detection.　The　structural　damage　location　and　degree　can　be　detected　and　classified,　and　the　result　is　highly　accurate.　This　approach　has　some　advantages,　such　as　engineering　oriented,　low　cost　and　convenient.　Copyright　(c)　2005　John　Wiley　&　Sons,　Ltd.