Prediction of Stress on Ship Structures using Ship Measurement Data and Machine Learning

Abstract

In the maritime field, ship measurement projects have been conducted to examine the condition of ships in service, and various data have been accumulated. From the viewpoint of ensuring the safety of ships, it is significant to get the history of the stress generated on the ship structures. However, even though it is necessary to analyze ship measurement data together with the natural conditions which affected ships greatly, it’s difficult to accurately grasp such conditions, and the variation in the data will be large. On the other hand, in the field of machine learning NGBoost has been proposed as an algorithm for probabilistic prediction. Therefore, in this study, we performed prediction of stress on ship structure from the statistical values of the ship measurement data by using NGBoost. Study results show that prediction of stress is possible even for data with large variations.