To　improve　the　performance　of　case-based　reasoning　in　reverse　osmosis　membrane　fault　diagnosis,　the　Fuzzy　ART　network　was　first　used　to　divide　the　source　case　into　multiple　clusters.　Then,　the　noise　cases　and　the　redundancy　cases　in　each　cluster　were　deleted　and　the　outliers　were　reserved　to　realize　the　maintenance　of　the　case　base.　Finally,　this　maintenance　strategy　and　case-based　reasoning　method　were　applied　to　the　fault　diagnosis　of　the　reverse　osmosis　membrane,　and　the　comparative　experiment　was　carried　out　by　using　the　historical　data　of　the　reverse　osmosis　membrane　fault　in　a　smelter,　which　verified　the　advantages　of　the　improved　CBR　method　in　the　diagnosis　accuracy,　and　the　research　has　certain　practical　application　value.　©　2018,　Editorial　Department　of　Journal　of　Beijing　University　of　Technology.　All　right　reserved.