In　order　to　solve　the　difficulties　of　rail　detection　(such　as　low　timeliness,　poor　reliability,　and　high　dependence　on　manual　inspection),　this　paper　proposes　a　monitoring　and　early　warning　technology　for　internal　cracks　of　railhead.　Through　the　finite　element　simulations　of　Lamb　wave　and　practical　experiments　of　rail　cutting,　the　data　for　driving　algorithms　can　be　obtained.　On　the　basis　of　using　Shannon　Wavelet　Transform　(SWT)　to　extract　the　first　arrival　wave　and　Hilbert-Huang　Transform　(HHT)　to　analyze　its　time-frequency　properties,　this　paper　presents　an　innovative　triple-threshold　judgment　method.　It　sets　up　three　lines　of　defence　to　decline　the　impact　of　environmental　factors.　Empirical　results　show　that　this　technology　can　effectively　monitor　and　warn　the　internal　cracks　of　railhead　under　a　low　false　alarm　rate.　The　sensitivity　is　2 mm,　which　meets　the　needs　of　practical　engineering　and　makes　up　for　the　gap　in　rail　structure　health　monitoring.　©　Springer　Nature　Singapore　Pte　Ltd.　2020.