碩士 / 國防大學中正理工學院 / 造船工程研究所 / 95 / Computing of ship stability is an important issue that the designer must investigate thoroughly for ship safety. Depending on the type of ship, designers define the parameters of ship stability based on the method they trust. However, these methods are mostly established in the form of approximate expressions or empirical rules. Therefore, during the stage of ship design, designers must consider the purposes of ships, and make choice of stability parameters according to the ship owner’s demand.
This study is divided into three parts. First of all, ship geometry data of various types of ship are collected and transformed into dimensionless variables. Secondly, the types of ship are classified by using perception and back-propagation network. Then, the sea keeping quality of various ships from the neural network results are discussed. Besides linear classification problem, nonlinear classification problem can be solved by using back-propagation network with transfer functions to transfer the original data into a linear problem. Finally, the vertical center of gravity（KG）and the height of transverse metacenter above keel（KM）values are calculated from 80 sets of ship geometry data by using both linear regression and neural network method. Using the relations of KG and KM , metacentric height（GM）can be evaluated. Comparing with the actual KG and KM values, the estimation errors of both the regression analysis and back-propagation network method can be obtained. The result shows that the errors which are trained and predicted by back-propagation network method are, more accurate than result from regression method.
| Identifer | oai:union.ndltd.org:TW/095CCIT0345004 |
| Date | January 2007 |
| Creators | Jing-Tzai Tu, 涂進財 |
| Contributors | Tzeng-Yuan Heh, 賀增原 |
| Source Sets | National Digital Library of Theses and Dissertations in Taiwan |
| Language | zh-TW |
| Detected Language | English |
| Type | 學位論文 ; thesis |
| Format | 92 |
Page generated in 0.0126 seconds