Inthispaper,byusingthewell-knownhigh-gainobserverdesign,anupdatelawforthegainandanadaptiveestimationofparameters,anewmethodoffaultdiagnosisforaclassofnonlinearsystemsispresented.Withoutresorttoanytransformationfortheparameters,theestimationerrorsofthestatesandtheparametersareguaranteedtobegloballyexponentiallyconvergentbyapersistentexcitationcondition.Comparedtotheexistingresults,itcanbeappliedtononlinearsystemswithnonlineartermsadmittinganincrementalratedependingonthemeasuredoutput.Acasestudyfurtherverifiesthevalidityoftheproposedresearch.