Reciprocating　air　compressor＇s　structure　is　complex,　and　it　has　various　excitation　sources　when　running,　moreover,　there　are　a　few　fault　samples　in　actual　fault　diagnosis,　so　it　is　difficult　to　implement　intelligent　diagnosis.　Support　Vector　Machine　based　on　Statistical　Learning　Theory　just　overcomes　this　deficiency,　and　it　provides　a　new　approach　for　diagnosis　technology　to　develop　into　intelligent　diagnosis.　The　application　of　Support　Vector　Machine　on　fault　diagnosis　for　reciprocating　air　compressor　and　a　concrete　implementation　scheme　are　discussed　in　this　paper.　A　fault　diagnosis　system　for　reciprocating　air　compressor　is　established,　and　the　vibration　signals　of　rolling　bear　in　reciprocating　air　compressor＇s　crankcase　are　simulated　in　a　test-bed.　The　test　result　shows　that　this　system　has　strong　adaptability　for　reciprocating　air　compressor　diagnosis　of　a　few　samples　and　could　recognize　fault　rapidly　and　accurately.