A　fault　set　and　a　symptom　set　were　established　in　order　to　exactly　judge　and　to　quickly　dispose　in　turbine　startup　of　a　power　plant.　There　are　ten　typical　faults　in　the　fault　set　and　sixteen　fault　symptoms　in　the　symptom　set.　In　consideration　of　the　various　kinds　of　change　directions　and　ranges　of　the　fault　symptom　parameters,　the　fuzzy　disposal　of　nine　degrees　is　put　forward　to　build　a　set　of　typical　fault-character-sample　mode.　A　neural　network　model　for　fault　diagnosis　was　obtained　by　fuzzy　theory　and　radial　basis　function,　and　it　was　validated　by　using　evaluator.　It　shows　that　the　fuzzy　fault　disposal　and　the　swiftness　of　training　constringency　are　very　satisfied　in　turbine　startup　of　this　power　plant.　Copyright　©　2005　by　ASME.