Due　to　its　synthetic　and　complex　characters,　the　combustion　process　with　fixed　air-fuel　ratio　of　shaft　ore-roasting　furnace　is　very　difficult　to　he　controlled　stably,　the　fault　is　appeared　frequently　and　lead　to　the　combustion　efficiency　laigh.　To　deal　with　this　problem,　an　intelligent　control　approach　has　been　developed　for　the　air-fuel　ratio　combination　of　case-based　reasoning　and　neural　network.　The　fault　prediction　model　performs　to　predict　the　typical　fault　with　the　help　of　case-based　reasoning　technology　is　obtained　with　the　working　trend　and　the　fault　cases.　According　to　these,　the　tuning　value　of　air-fuel　ratio　are　given　by　the　algorithm　based　on　neural　network.　The　proposed　method　has　been　successfully　applied　to　the　combustion　process　of　a　shaft　furnace,　with　increase　of　control　accuracy　for　the　combustion　temperature,　reduction　of　gas　consumption　and　the　fault　ratios.