Abstract: Inordertoreducethecalculationcapacityoffaultdiagnosisalgorithmsandoptimizethealgorithmparameters,anoptimalfaultdiagnosisalgorithmbasedonreducedorderobserverwasproposed.Firstly,areducedorderobserverwasdesignedtoestimatethefaultwithoutcomplexmatrixblockingoperationbydecouplingthestatesofsystemandstatesoffaultbasedontheprincipleoforthogonality.Secondly,theproblemofdeterminingtheobservergainwastranslatedtotheproblemofdesigningoptimalcontrollerforthedualsystemoffaultdiagnosiserrorsystem,basedonwhichtheminimumvarianceestimationoffaultwasimplemented.Thirdly,afaulttolerantcontrollerwasproposedtocompensatethefaultinfluencebyutilizingtheresultofdiagnosis,soastosegregatethesystemwithfaults.Finally,anumericalsimulationwascarriedouttoverifytheeffectivenessoftheproposedmethod.Theresultshowsthatthedesigneddiagnosisalgorithmcanfastconvergetotherealvalueoffaultin7secondsforatwoordersystem.

Key words: reducedorderobserver, optimalgain, principleofduality, faultcompensation