Understanding　the　hydraulic　cleaning　mechanism　of　the　membrane　fouled　with　activated　sludge　(AS)　and　establishment　of　the　flux　predictive　model　can　effectively　improve　the　cleaning　efficiency　of　the　membrane　in　MBR.　In　this　study,　an　instantaneous　flux　prediction　model　was　established　by　considering　the　cascade　mechanisms　(irreversible　swelling,　reversible　dissolution　and　re-deposition)　for　the　gel　and　dynamic　equilibrium　mechanisms　(direct　reversible　dissolution　and　re-deposition)　for　the　cake　in　the　hydraulic　cleaning　process.　In　addition,　the　XDLVO　theory　was　used　to　calculate　the　variation　of　interaction　energy　between　AS　and　membrane　surface　in　the　swelling　process　of　AS.　The　result　showed　that　the　proposed　model　was　fully　adequate　for　the　200　kDa　PAN　membrane　fouled　with　AS　suspension　(R-2=0.9987)　or　yeast　suspension　(R-2=0.9958)　in　the　hydraulic　cleaning　process.　The　swelling　of　AS　can　significantly　reduce　the　attractive　interaction　energy　and　improve　the　membrane　cleaning　efficiency.　Moreover,　the　predictive　model　showed　better　accuracy　(sigma　=　0.39%)　than　other　models　(Kong＇s　model　(sigma　=　0.71%)　and　Cabero＇s　model　(sigma　=　0.88%)).　In　addition,　the　proposed　model　also　exhibited　excellent　performance　for　the　multiple　fouling　and　cleaning　processes　(R-2＞0.9860)　and　for　other　membranes　(100　kDa　PES　membrane　(R-2=0.9995),　200　kDa　PVDF　membrane　(R-2=0.9976)　and　0.1　mu　m　PAN　membrane　(R-2=0.9999))　fouled　with　AS　suspension.