Line detection techniques to pinpoint slamming impulses in time-frequency images of hull acceleration measurements

Abstract

The detection and localization of slamming is investigated using a hull monitoring system on a polar vessel. The system comprises an accelerometer array, resulting in time histories in which slamming impacts must be detected. Proposed methods exploit the attribute that slamming impacts cause broadband excitation which result in line features in short-time Fourier transform spectrograms and Morlet wavelet scalogram images. The proposed methods are benchmarked against the MATLAB findpeaks function which requires a pre-defined detection threshold and falsely identifies several slams if consecutive peaks of the free decay exceed the threshold. An evaluation on simulated data investigates artefacts such as an unpredictable frequency of incidence, varying impulse magnitude and ever-changing noise floor. The likely wave impact site is pinpointed at the sensor which first displays an impulsive peak. Preliminary insights are extracted from full-scale experiments on the stationary vessel in stable wave conditions and a systematically varied relative heading. The Morlet wavelet transform delivers the most promising slamming detection capabilities and concurs that stern slamming is prevalent in following and stern-quartering seas. A threshold is beneficial to consistently identify significant slams despite varying ocean conditions.