A　bench-scale　anoxic-oxic　(A/O)　system　was　used　to　treat　a　synthetic　starch　wastewater　for　1　year.　The　objectives　were　to　improve　nitrogen　removal　and　reduce　operating　costs　by　using　on-line　sensors　DO,　pH　and　ORP.　It　is　found　that　the　DO　in　the　first　aerobic　zone　could　be　used　to　estimate　influent　ammonia　load　under　constant　aeration.　The　pH　profiles　and　DO　profiles　along　the　length　of　aerobic　zones　give　good　indication　of　the　nitrification,　and　the　pH　profiles　in　the　aerobic　zones　and　anoxic　zones　could　be　classified　into　two　types:　descending　and　rising　type.　Further,　ORP　value　at　the　end　of　aerobic　zone　showed　a　good　correlation　with　the　effluent　ammonia　and　nitrate　nitrogen　concentration,　good　effluent　quality　corresponded　with　the　range　of　40-60　mV　of　ORP　in　this　study.　ORP　was　also　correlated　well　with　nitrate　concentration　at　the　end　of　anoxic　zone,　and　the　maximum　denitrification　capacity　could　be　achieved　when　ORP　was　maintained　at　about　-90　mV.　A　fuzzy　inference　system　for　on-line　aeration,　nitrate　recirculation　flow　and　external　carbon　dosage　control　is　developed　on　the　basis　of　these　observations　and　is　implemented　on　the　bench-scale　system.　Promising　control　performance　is　achieved.　(c)　2006　Elsevier　B.V.　All　rights　reserved.