Neural Network Approach for Automatic Classification of Ship-Radiated Noise

Abstract

Tonal signals from ship-radiated noise contain important information for a ship's design, maintenance, operating condition diagnosis, and machinery monitoring. The signals mainly consist of two components–the speed-dependent component and the speed-independent component. Therefore, some procedures are required to detect and classify the tonal signals from ship-radiated noise. We apply two neural network approaches for the detection of a tonal signal by peak extraction, and the classification of the tonal signal by pattern recognition in some numerical experiments on a simulation signal and ship-radiated noise obtained from a real ship.