system for monitoring and diagnosis of anaerobic wastewater treatment plants

AbstractIn this paper, an expert system (ES) developed for the monitoring and diagnosis of anaerobic wastewater treatmentplants (A WT), is presented. The system was evaluated in a hybrid pilot plant of 1.1 m3 located in an industrialenvironment for the treatment of wastewaters from a fibreboard production factory. The reactor is a hybrid USBF,combining an upflow anaerobic sludge blanket (UASB) in the lower part and an upflow anaerobic filter (UAF) at thetop. ~ 2002 Elsevier Science Ltd. AII rights reserved. -- Introduction biological, operational and hydrodynamic points ofview, their supervision being a task requiring expertise.The complexity of the process makes the supervisionand control of A WT difficult and therefore, differentprocedures based on artificial intelligence (fuzzy logic,expert systems, neural networks, etc.) have been appliedin order to control the operation. Expert systems (ES)are computer programs applied to problem solving areas(narrow domains) where expertise, especially of heuristicnature, is available (rules of thumb). A proper ES musthave the capacity to represent time-dependent knowl-edge, the ability to process incomplete or inaccurateknowledge and the capacity to combine simple ruleswith model based rules. When applied to A WT manage-ment, the long operational time required in anaerobicprocesses (hours or days) represents an advantage toensure real time processing. The major difficulty increating an ES is the generation, management andhierarchisation of knowledge concerning the processDuring the last 10 years, attention has been paid tothe study and development of monitoring, diagnosisando more generally, control systems for wastewatertreatment plants [8,9,10) based on the concept of ES andexamples of anaerobic processes can be found inliterature. The earliest works in this area are related tothe aerobic treatment of municipal wastewaters withactivated sludge processes. Serra et al. [11] designed aAnaerobic wastewater treatment (A WT) is a provenuseful technology for the treatment of highly pollutedwastewaters [1,2] and diluted wastewaters [3]. However,the control applied to most of the anaerobic processeshas been restricted to environmental variables, such aspH, temperature, etc. [4], using conventional controllerssuch as on-off and proportional-integral-derivative(PID) controllers. Nevertheless, control strategies ap-plied to A WT plants could be based on the monitoringof a number of process indicators which give comple-mentary information [5,6]. The most reliable informa-tion to control the process is the result of combining theavailable on-line data, off-line measurements anddetailed knowledge of the process [7].Control of the process depends directly on the type ofreactor, since optimal operation conditions are differentdepending on the technology used. Most of theanaerobic reactors (UAF -upflow anaerobic filter,UASB-upflow anaerobic sludge blanket, EGSB-ex-panded granular sludge bed; FB-fluidised bed, etc.) areefficient technologies, but they are complex from the