For　overcoming　a　lot　of　difficulties　in　condition　monitoring　due　to　the　structural　complexity　and　multiple　excitation　of　air　compressor,　a　fault　detection　method　based　AE　(acoustic　emission)　signal　correlation　dimension　is　investigated.　Based　on　the　detailed　analysis　of　key　parameters　selection　in　calculating　correlation　dimension,　the　embedded　dimension　and　time　delay　are　respectively　computed　by　singular　value　analysis　and　improved　average　displacement　method.　The　typical　faults　of　air　compressor,　including　suction　valve　loosening,　spring　deformation,　spring　failure,　valve　deformation　and　valve　failure,　are　simulated　in　filed.　The　AE　signals　in　various　conditions　are　obtained　to　calculate　the　corresponding　correlation　dimension.　The　results　show　the　proposed　method　can　effectively　identify　the　faults　in　air　compressor.