A neural network approach for on-line fault detection of nitrogen sensors in alternated active sludge treatment plants

In this paper, an effective strategy for fault detection of nitrogen sensors in alternated active sludge treatment plants is proposed and tested on a simulated set-up. It is based on two predictive neural networks, which are trained using a historical set of data collected during fault-free operation of a wastewater treatment plant and their ability to predict reduced (ammonium) and oxidized (nitrates and nitrites) nitrogen is tested. The neural networks are also characterized by good generalization ability and robustness with respect to the influent variability with time and weather conditions. Then, simulations have been carried out imposing different kinds of fault on both sensors, as isolated spikes, abrupt bias and increased noise. Processing of residuals, based on the difference between measured concentration values and neural networks predictions, allows a quick revealing of the fault as well as the isolation of the corrupted sensor.