Based　on　complementary　strategies,　a　new　AI　method　was　put　forward　to　solve　the　fault　diagnosis　of　the　multilevel　reciprocating　compressor.　Using　rough　set　theory　to　enrich　the　data,　then　exploiting　fuzzy　neural　network(FNN)　to　perform　diagnosing,　the　diagnosing　rules　could　be　obtained　from　the　network.　Additionally,　a　new　FNN　approach　method　and　a　criterion　about　the　data　enriching　were　presented.　The　correctness　and　the　effectiveness　were　showed　in　the　experiment.