A NEURAL-NETWORK ARCHITECTURE FOR NOISE PREDICTION

This paper presents application of neural networks (NNs) to the proble m of prediction of noise caused by urban traffic. The most representat ive physical variable quantifying noise emissions ir the equivalent so und pressure level Up to now it has been identified on the basis of se mi-empirical models, typically regression analysis, which generally do not provide very accurate approximations of the trend followed by sou nd pressure level. The authors have attempted to overcome this difficu lty by adopting a neural approach based on a BackPropagation Network ( BPN). Results obtained by the comparison of the BPN approach with thos e provided by selected relationships found in relevant literature, sho w how good is the approach proposed. The neural solution to the proble m has shown the necessity, in certain phases, of a set of acoustic mea surements which is as free as possible of error. The complexity of err or identification by means of classical approaches has led the authors to explore the possibility of a neural solution to this problem as we ll. The authors therefore propose use of a neural architecture made up of two cascading levels. At the first level a supervised classifying network, the learning vector quantization (LVQ) network,filters the da ta discarding all the wrong measurements, while at the second level th e BPN predicts the sound pressure level.