A　soft-sensor　method,　based　on　the　recurrent　radial　basis　function　neural　network　(RRBFNN),　was　proposed　in　this　paper　to　solve　the　problem　of　the　permeability　measurement　of　membrane　bio-reactor　(MBR).　First,　the　data　was　collected　from　a　real　wastewater　treatment　process　in　Beijing　and　the　partial　least　squares　(PLS)　technique　was　utilized　to　select　the　variables　which　have　the　largest　correlation　with　the　permeability.　Then,　the　soft-sensor　model　was　developed　to　predict　the　permeability　via　RRBFNN.　Meanwhile,　a　fast　gradient　descent　method　was　used　to　adjust　the　parameters　of　RRBFNN.　Finally,　this　soft-sensor　method　was　applied　to　the　real　wastewater　treatment　process.　The　results　show　that　the　proposed　soft-sensor　method　can　predict　the　permeability　of　MBR　with　high　accuracy.　©　2017,　Editorial　Department　of　Journal　of　Beijing　University　of　Technology.　All　right　reserved.