Fuzzy　neural　network,　which　is　based　on　fuzzy　theory　and　BP　neural　network　plays　an　important　role　in　practical.　But　the　difficulty　is　how　to　construct　its　structure　model.　In　this　paper,　according　to　the　hypostasis　of　hidden　layer,　a　fizzy　neural　network　model　based　on　fuzzy　clustering　is　brought　forward,　in　which　features　of　the　samples　are　extracted　and　output　information　is　synthetically　considered.　And　a　satisfied　result　is　reached　in　the　application　of　oil-gas　recognition　in　Tarim　Oilfield.