DONATE

IAHR Document Library


« Back to Library Homepage « Proceedings of the 26th IAHR International Symposium on Ice ...

Dimensionality Reduction of Multivariate Sensor Data for Estimation of Ice Load on a Real Ship

Author(s): Hyobeom Heo; Jaeuk Choi; Hyeon Sik Choi; Eun-Jin Oh; Seunghwan Park

Linked Author(s):

Keywords: Sea Ice; Ships and Navigation in Ice

Abstract: Recently, with the increase in the use of the Arctic Route, it is required to review the structural design standards for ships operating in polar seas and to manage safety facilities. Accordingly, maritime research institutes measures various navigation information through sensors to estimate ice load, which is a major variable in ship design, in real ship tests. In particular, since strain gauge sensors are used for estimating the local ice load that causes structural damage to the hull plate, many sensors are attached to increase the accuracy of the local ice load estimation. However, missing values frequently occur due to sensor failure during the real ship test, which leads to an error in the estimated value. Therefore, this study performs strain sensor signal clustering to impute the missing value of sensor used for local ice load estimation. Clustering refers to dividing objects with similar properties into a number of groups. As a result of imputing missing values between sensors within the same cluster, it was confirmed that considering the signal similarity between sensors affects the performance of imputing missing values. Through this, the number of sensors required for local ice load estimation can be reduced by utilizing dimensionality reduction, and reliable estimation of ice load is possible even in the presence of missing values.

DOI:

Year: 2022

Copyright © 2024 International Association for Hydro-Environment Engineering and Research. All rights reserved. | Terms and Conditions