인공지능을 이용한 연안 소형어선의 초기 제원 추정에 관한 연구

Abstract

In designing a ship, a lot of information is required to build a new ship. The information can be roughly estimated based on the parent ship. Even if there is data on parent ships, however, it is difficult to establish the specifications of a new design ship with only distributed information, and plus it takes longer time to do in various ways. Initial design information affects the entire design process such as basic design or detailed design, and it can have a great influence on the entire duration of ship construction. This paper studied estimating specific dimension at the initial design stage.

Regression analysis through existing graphs or excel is also useful, but the core of this study is to find the appropriate value for the situation by Deep-learning. And if it is applied to practice through programming afterwards, work efficiency can be expected. First, the data for the main specifications of small fishing boats were collected. Then the data have been analyzed through artificial intelligence, which is rapidly growing in recent years. A regression analysis was performed by incorporating a large amount of data into machine learning, and by obtaining the specifications of the ship closest to the parent ship at a high speed when a specific dimension was given. When the specific dimensions are taken, the work inefficiencies which occur in the early stages of the design will be reduced.

As a result through the deep learning, we have get a value close to our target value, and we were able to obtain reliable results with an error rate of 2.79% for breadth and 1.09% for depth.