Investigation of harbor oscillations originated from the vessel-induced bores using methods of autoregressive model and Mahalanobis distance. (1st July 2020)

• Record Type: Journal Article
• Authors: Shao, Dong; Xing, Yun; Zheng, Zirui; Jiang, Gang
• Is Part Of: Ocean engineering. Volume 207(2020)
• Publication Date: 2020-07-01
• Subjects: Harbor oscillations -- Vector autoregressive model -- Statistical pattern recognition -- Numerical experiments; Ocean engineering -- Periodicals; Ocean engineering; Periodicals; 620.4162
• DOI: 10.1016/j.oceaneng.2020.107385
• Languages: English
• ISSNs: 0029-8018

Abstract

Abstract: This paper presents a study for investigating and identifying features of oscillation patterns induced by vessel motions in the context of statistical pattern recognition on the basis of time-domain oscillation series in addition to traditional frequency-domain analysis. Numerical simulations are carried out for a semi-enclosed harbor of variable depth with different vessel motions as external agitation sources and the oscillation responses are measured at multiple locations along the backwall in the longshore direction and along one sidewall in the offshore direction. Vector autoregressive type models are proposed and applied on the recorded oscillation time series. From clustering of the selected vector autoregressive model coefficients fitted to the oscillation responses, both the influences of a changing bathymetry and those of various vessel motions can be recognized, although some of them are insensitive to frequency-domain analysis and cannot be revealed with the amplitude spectra. With Mahalanobis distances calculated from the spatially distributed measurements at observation points within the harbor, an insight into the internal structure and energy distribution of the vessel-induced oscillations can be achieved by extracting the probability density distributions and the occurrence or absence of certain oscillation components or modes can be analyzed with the proposed pattern recognition approach. Highlights: A sensitive statistical pattern recognition method is presented for harbor oscillation feature identification. The vessel-induced oscillations are studied both in frequency domain and with the proposed pattern recognition procedure. Influences of bathymetry and vessel motions are recognized by the statistical variations in the vector autoregressive model. The proposed method provides an insight into the internal structure and energy distribution of the harbor oscillations.