A　method　that　combines　temperature　field　detection,　adaptive　FCM　(Fuzzy　c-means)　clustering　algorithm　and　RBF　(Radial　basis　function　network)　neural　network　model　is　proposed.　This　method　is　used　to　analyze　the　thermal　error　of　the　spindle　reference　point　of　the　tauren　EDM　(Electro-discharge　machining)　machine　tool.　The　thermal　imager　is　used　to　obtain　the　temperature　field　distribution　of　the　machine　tool　while　the　machine　tool　simulates　actual　operating　conditions.　Based　on　this,　the　arrangement　of　temperature　measurement　points　is　determined,　and　the　temperature　data　of　the　corresponding　measurement　points　are　got　by　temperature　sensors.　In　actual　engineering,　too　many　temperature　measurement　points　can　cause　problems　such　as　too　high　cost,　too　much　wiring.　And　normal　processing　can　be　affected.　In　order　to　establish　that　the　thermal　error　prediction　model　of　the　machine　tool　spindle　reference　point　can　meet　the　actual　engineering　needs,　the　adaptive　FCM　clustering　algorithm　is　used　to　optimize　the　temperature　measurement　points.　While　collecting　the　temperatures　of　the　optimized　temperature　measurement　points,　the　displacement　sensors　are　used　to　detect　the　thermal　deformation　data　in　X,　Y,　Z　directions　of　the　spindle　reference　position.　Based　on　the　test　data,　the　RBF　neural　network　thermal　errors　prediction　model　of　the　machine　tool　spindle　reference　point　is　established.　Then,　the　test　results　are　used　to　verify　the　accuracy　of　the　thermal　errors　analysis　model.　The　research　method　in　this　paper　provides　a　system　solution　for　thermal　error　analysis　of　the　tauren　EDM　machine　tool.　And　this　builds　a　foundation　for　real-time　compensation　of　the　machine　tool＇s　thermal　errors.