The　warning　of　membrane　fouling　is　of　great　important　to　maintain　the　stable　operation　of　membrane　bioreactor　(MBR).　However,　traditional　methods　are　so　error-prone　that　probably　do　not　acquire　reliable　solutions　of　membrane　fouling　due　to　its　uncertainties.　To　overcome　this　problem,　an　intelligent　warning　method　is　proposed　to　monitor　the　status　of　MBR　in　this　paper.　The　main　advantages　in　this　paper　are　as　follows.　First,　an　identification　method,　based　on　robust　deep　neural　network　(RDNN),　is　developed　to　diagnose　the　different　types　of　membrane　fouling.　Second,　a　decision-making　method,　based　on　the　restricted　Boltzmann　machine　(RBM),　is　designed　to　distinguish　the　operational　suggestion.　Third,　an　intelligent　warning　system,　based　on　the　above　two　methods　and　some　sensors,　is　developed　to　mitigate　the　membrane　fouling　in　real　wastewater　treatment　plants.　Finally,　the　simulation　and　experimental　results　demonstrate　the　proposed　warning　method　can　obtain　the　higher　identification　accuracy　of　membrane　fouling　than　other　methods.