Neural Networks: An Overview - Neural Networks: Methodology and Applications - page 73

Fig. 1.47. The structure of a neural network for the prediction of ozone concentra-
tion, 24 hours ahead
1.4.11.2 Modeling the Rainfall-Water Height Relation in an Urban
Catchment
The Direction de l'Eau et de l'Assainssement has developed a sophisticated
system for measuring water heights in the sewers of an area of the suburbs
of Paris, and has performed systematic measurements of rainfalls and of the
corresponding water heights. The objective is to optimize the sewer network
and to anticipate serious problems that are likely to arise in the case of severe
rains. Hence the reliability of the water height sensors in the sewers is crucial
for the reliability of the whole system: therefore, the automatic detection of
faults in the water height sensors is mandatory [Roussel 2001].
Neural networks can be accurate models of nonlinear phenomena, which
makes them useful tools for fault detection: if an accurate, real-time model
of normal operation of a process is available, the observation of a statistically