The　non-linear,　time-varying,　and　uncertain　characteristics　of　municipal　solid-waste　incineration　(MSWI)　process　cause　the　difficulty　of　recognizing　the　combustion　condition　of　MSWI.　Currently,　engineers　judge　the　operating　conditions　in　MSWI　plants　using　a　video　of　the　combustion　flame　inside　the　incinerator,　but　this　method　does　not　help　maintain　stable　operating　conditions.　Thus,　this　paper　proposes　a　convolutional　neural　network-based　recognition　method　for　controlling　combustion　conditions　in　MSWI　processes.　First,　the　flame　image　is　pre-processed　by　resizing,　and　its　features　are　extracted　using　the　convolutional　neural　network.　Thereafter,　the　classification　of　combustion　conditions　is　realized　from　the　Softmax　layer.　Moreover,　the　proposed　method　is　validated　with　experimental　results　based　on　flame　image　data　of　an　MSWI　plant　in　Beijing.