Water Temperature Prediction and Gyroscope Signal Denoising using Deep Learning Technology

Title
Water Temperature Prediction and Gyroscope Signal Denoising using Deep Learning Technology
Author(s)
김민규; 양현; 김종화
KIOST Author(s)
Kim, Min Kyu(김민규)Yang, Hyun(양현)
Publication Year
2019-12-12
Abstract
As from January 1, 2020, the International Maritime Organization (IMO) will enforce strong regulations limiting sulfur content of ship fuel oil from 3.5% to 0.5% to reduce air pollutants. It is important to limit sulfur content of ship fuel oil to reduce air pollutants, but it is also important to reduce unnecessary energy waste during ship operation. In order to do this, the ship needs to maintain the designated route correctly. To maintain the sea route, a ship used autopilot system composed of controller such as PD type, Fuzzy PID type, etc. These type controllers have excellent performance on the assumption that there is no noise. However, in a real environment, measurement noise caused from gyroscope is applied to autopilot system, which degrades the performance of controller. In order to solve this problem, Kalman Filter, which is widely used for state estimation, is applied, but this also cannot completely eliminate noise. In this study, therefore, the denoising method to reduce effect of noise is proposed by combining Kalman Filter and Multi-Layer Perceptron (MLP) which is a kind of artificial neural network. Since motions of a ship are divided into the forward direction and the rotation motions, Kalman Filter is applied in case of forward direction motion and MLP is applied in case of rotation motion.
URI
https://sciwatch.kiost.ac.kr/handle/2020.kiost/21024
Bibliographic Citation
The 7th Asian/16th Korea-Japan Workshop on Ocean Color, pp.P-08, 2019
Publisher
Burapha University
Type
Conference
Language
English
Publisher
Burapha University
Related Researcher
Files in This Item:
There are no files associated with this item.

qrcode

Items in ScienceWatch@KIOST are protected by copyright, with all rights reserved, unless otherwise indicated.

Browse