Due　to　low　sensitivity　in　existing　High-Voltage　Direct　Current　(HVDC)　fault　detection　methods　and　difficulty　in　identifying　high-resistance　grounding　faults,　this　paper　presents　two　signal-terminal　HVDC　transmission　system　fault　detection　methods　based　on　machine　learning.　The　waveform　of　the　fault　voltage　collected　by　the　rectifier　side　detection　device　is　directly　used　as　the　input　data　of　K-Nearest　Neighbor　(KNN)　and　Support　Vector　Machine　(SVM),　eliminating　the　cumbersome　process　of　fault　signal　processing.　Training　is　performed　in　various　fault　areas　and　fault　types.　Then　fault　areas　will　be　detected　by　the　trained　KNN　and　SVM　models.　A　±　500　kv　HVDC　transmission　line　model　was　built　by　electromagnetic　transient　simulation　software　PSCAD/EMTDC　to　simulate　and　compare　different　fault　areas　and　fault　types.　Testing　results　show　that　the　proposed　method　can　reliably　and　accurately　detect　faults　with　a　resistance　up　to　$1000　Ω.　©　2018　IEEE.