Characterizing onboard noise in ships: Insights from statistical, machine learning and advanced noise index analyses

Understanding the specific characteristics of onboard noise can significantly benefit human activity in different kinds of ships. Here, the perceived noise can significantly change from space to space in terms of both levels and frequency content, depending on location onboard in respect to the primary sources and on propagation paths. Moreover, secondary sources as ventilation systems contribute to generate a complex soundscape for each ship. In this work, noise spectra reported in the literature were collected and analyzed to obtain an overview of the noise characteristics in different spaces onboard military and commercial vessels, i.e. accommodation, navigation bridge, and other workspaces. A prevalence of high noise levels in the different spaces that can be over the commonly accepted limits was found, characterized by higher levels at low and high frequencies. Different noise indexes not yet adopted in this sector were tested on the data to derive more information about onboard noise annoyance. Supervised and unsupervised machine learning algorithms were also applied to qualitatively classify and study the collected spectra. The analysis shows the presence of different spectral typologies, i.e. balanced spectral shapes (quasi-linear decay of the noise level per octave band with no dominant frequency components) or unbalanced ones.