Theobjectiveofthispaperistodevelopaneuralnetwork-basedresidualgeneratortodetectthefaultintheactuatorsforaspecificcommunicationsatelliteinitsattitudecontrolsystem(ACS).First,adynamicmultilayerperceptronnetworkwithdynamicneuronsisused,theseneuronscorrespondtoasecondorderlinearInfiniteImpulseResponse(IIR)filterandanonlinearactivationfunctionwithadjustableparameters.Second,theparametersfromthenetworkareadjustedtominimizeaperformanceindexspecifiedbytheoutputestimatederror,withthegiveninput-outputdatacollectedfromthespecificACS.Then,theproposeddynamicneuralnetworkistrainedandappliedfordetectingthefaultsinjectedtothewheel,whichisthemainactuatorinthenormalmodeforthecommunicationsatellite.Thentheperformanceandcapabilitiesoftheproposednetworkweretestedandcomparedwithaconventionalmodel-basedobserverresidual,showingthedifferencesbetweenthesetwomethods,andindicatingthebenefitoftheproposedalgorithmtoknowtherealstatusofthemomentumwheel.Finally,theapplicationofthemethodsinasatellitegroundstationisdiscussed.