Statistical Prediction of Ship Performance Using Onboard Measurement Data

• Published in: Journal of the Japan Society of Naval Architects and Ocean Engineers Vol. 13; pp. 41 - 50
• Main Authors: Minoura, Munehiko, Koike, Kenta
• Format: Journal Article
• Language: English
• Published: Tokyo The Japan Society of Naval Architects and Ocean Engineers 2011; Japan Science and Technology Agency
• Subjects: Estimates; Marine; Mathematical models; Onboard; Probability density functions; Sea states; Seas; Ships; Stochasticity
• ISSN: 1880-3717; 1881-1760
• DOI: 10.2534/jjasnaoe.13.41

It becomes increasingly important that the performance of a ship in actual seas is predicted accurately due to an increasing awareness about global warming and other environmental problems. The ship responses such as the ship motions, the stresses and the added resistances in actual seas vary with sea states and the ship speed. The onboard measurement data has this information,therefore the data is useful. However there is not an established analysis method of the onboard data because the data involves factors which are unexplained by the theoretical calculation. The ship responses which are caused by uncertainty sea states are stochastic variable. Therefore, the statistical analysis is useful in case of the prediction of ship responses with using the onboard data.The paper proposes a statistical method which estimates the correlation between sea states and ship responses for a short time as the conditional probability density function (PDF) by using the onboard measurement data. Two kinds of method are adopted to estimate the conditional PDF of ship speed loss for encounter sea states. On one hand it is parametric method, on the other it is nonparametric method. Using a time-series of sea states and ship speed loss during a one-year voyage obtained by an oceangoing simulation, the expectation and confidence interval of mean ship speed loss are estimated to validate these stochastic models based on each method. We conclude that the stochastic models are available. In particular, the stochastic model based on the nonparametric method offers an accurate prediction of the mean speed loss.