Investigation of the influence of non-acoustical factors on the estimation of railway induced vibration annoyance using artificial neural networks

Rodolfo Venegas , Eulalia Peris , James Woodcock , Gennaro Sica , Andy Moorhouse , David Waddington

Groundborne vibration can be a source of annoyance in residential areas situated close to railway lines. Using field data from case studies (N=929) comprised of face-to-face interviews and internal measurements of vibration exposure, this paper aims to investigate vibration annoyance as a function of the vibration level, socio-demographic data, and several individual and situational factors. The influence of these variables on vibration annoyance is investigated through a sensitivity analysis performed on several artificial neural network models. Comparisons are also made with analyses performed using conventional regression or classification techniques. The results show that the vibration annoyance prediction is improved when the non-acoustical variables are considered along with the vibration level. The results further suggest that artificial neural networks could provide an alternative way of calculating vibration exposure-response relationships. The data used in this paper were collected within the Defra funded UK study “Human Response to Vibration in Residential Environments” conducted by the University of Salford.

Mots clés: vibration annoyance

Domaines: Acoustique [physics.class-ph]

Rodolfo Venegas, Eulalia Peris, James Woodcock, Gennaro Sica, Andy Moorhouse, et al.. Investigation of the influence of non-acoustical factors on the estimation of railway induced vibration annoyance using artificial neural networks. Acoustics 2012, Apr 2012, Nantes, France. ⟨hal-00810945⟩