＇Energy-fault＇　method　is　introduced　for　faults　warning　of　ventilators,　which　is　based　on　wavelet　package　analysis　and　BP　neural　network.　Character　vectors　which　reflect　different　faults　state　of　ventilators　are　extracted　from　different　frequency　segments　with　the　technology　of　wavelet　package　analysis,　and　taking　them　into　BP　neural　network　model　which　is　trained　with　character　vectors　of　typical　faults　sample.　The　faults　states　of　ventilators　are　identified　with　the　BP　neural　network　model.　The　results　of　research　show　that　this　kind　of　faults　diagnosis　technology　is　an　effective　way　to　implement　faults　warning.　©2007　IEEE.