Real-time fault detection and isolation in biological wastewater treatment plants

摘要

Automatic fault detection is becoming increasingly important in wastewater treatment plantnoperation, given the stringent treatment standards and the need to protect the investment costsnfrom the potential damage of an unchecked fault propagating through the plant. This paperndescribes the development of a real-time Fault Detection and Isolation (FDI) system based onnan adaptive Principal Component Analysis (PCA) algorithm, used to compare the current plantnoperation with a correct performance model based on a reference data set and the output ofnthree ion-specific sensors (Hach-Lange gmbh, Du¨ sseldorf, Germany): two Nitrataxw NOx UVnsensors, in the denitrification tank and downstream of the oxidation tanks, where an Amtaxwnammonium-N sensor was also installed. The algorithm was initially developed in the Matlabnenvironment and then ported into the LabView 8.20 (National Instruments, Austin, TX, USA)nplatform for real-time operation using a compact Field Point w, a Programmable AutomationnController by National Instruments. The FDI was tested with a large set of operational datanwith 1 min sampling time from August 2007 through May 2008 from a full-scale plant. Afterndescribing the real-time version of the PCA algorithm, this was tested with nine months ofnoperational data which were sequentially processes by the algorithm in order to simulate annon-line operation. The FDI performance was assessed by organizing the sequential data in twondiffering moving windows: a short-horizon window to test the response to single malfunctionsnand a longer time-horizon to simulate multiple unrepaired failures. In both cases the algorithmnperformance was very satisfactory, with a 100% failure detection in the short window case,nwhich decreased to 84% in the long window setting. The short-window performance wasnvery effective in isolating sensor failures and short duration disturbances such as spikes,nwhereas the long term horizon provided accurate detection of long-term drifts and provednrobust enough to allow for some delay in failure recovery. The system robustness is based onnthe use of multiple statistics which proved instrumental in discriminating among the variousncauses of malfunctioning.