Nonparametric　methods　for　assessing　fragility　do　not　require　distributional　assumptions　and　accurately　capture　the　inherent　characteristics　of　structural　fragility.　Achieving　efficient　and　accurate　fragility　estimation　is　crucial　for　the　practical　application　of　nonparametric　methods.　This　study　proposed　an　optimal　selection　method　for　ground　motion　records　based　on　hierarchical　clustering　for　nonparametric　seismic　fragility　estimation,　using　the　mean　period　of　ground　motion　as　the　clustering　feature.　Four　reinforced　concrete　frame　structure　models　with　varying　numbers　of　storeys　were　developed.　A　total　of　1724　real　horizontal　ground　motion　records　were　employed　to　constitute　the　overall　sample　set.　The　fragility　curves　of　the　structures　were　derived　using　the　Latin　hypercube　sampling　method　and　the　hierarchical　clustering　method,　respectively.　The　results　demonstrate　that　using　the　mean　period　as　the　clustering　feature　is　both　reasonable　and　effective　for　fragility　estimation.　For　8storey　and　10-storey　structures,　the　proposed　method　accurately　estimated　the　overall　fragility　using　only　6.09　%　of　the　total　sample　size.　Furthermore,　when　both　methods　utilized　the　same　quantity　of　samples　(less　than　10　%　of　the　total　sample　size),　the　proposed　method　yielded　fragility　estimates　that　were　more　closely　aligned　with　the　overall　fragility　than　Latin　Hypercube　Sampling.　This　study　provides　new　insights　into　the　development　of　nonparametric　methods　for　seismic　fragility　estimation,　facilitating　the　efficient　and　accurate　construction　of　seismic　fragility　curves　for　structures.