G. Cammarata
A. Fichera
Istituto di Macchine, Facolta di Ingegneria, Universita di Catania---Italy, Viale A. Doria 6, 95125 Catania, Italy
S. Graziani
L. Marletta
Universita di Catania, 95125 Catania, Italy
A lot of research work has been done for linking the noise emissions to the most relevant traffic parameters. So far, correlations are derived from multiple regression analysis under the hypothesis that modeling noise pollution in urban areas is a linear problem. Such an assumption can be overcome if more general approaches are used. This paper reports on the results obtained by the authors in modeling the traffic noise by means of neural network (NN) and fuzzy logic (FL). The models allow the prediction of the continuous equivalent level as a function of some typical parameters such as the traffic intensity (number of vehicles per hour), type of vehicles (cars, trucks, motorcycles), and geometrical features of the road (width of the lane, height of buildings). The database used for the NN and FL approaches refer to some typical towns of middle Europe, thus the resulting models should be valid for medium size cities. Comparison of the results with the measured data show better predictions than those based on more classical methods.