碩士 / 建國科技大學 / 土木與防災研究所 / 98 / Around the abstract Taiwan main island the surrounding seas, belongs to the island country, therefore the Taiwan main island has many harbors, but the harbor ocean current speed of flow passes in and out very large regarding the ships, based on the ships turnover harbor safe consideration, the ocean current speed of flow simulation to provides the pilot personnel to hold the ship is the indispensable important information. The influence ocean current speed of flow factor is complex, also a nonlinear response, by no means easily by the tradition numerical method or the empirical formula may the precise forecast. Because therefore this research use class nerve network has includes: (1) high speed computation characteristics and so on ability, (2) self-learning capability, (3) high-capacity memory, (4) fault-tolerant ability, use the harbor actual ocean current speed of flow value, by the kind of nerve network analysis method, calculated the harbor ocean current speed of flow, makes the characteristic discussion. This research take the Taiwan Hualian port, the Kaoshiang port, the Keelung Port three harbors does as the object of study, passes through class the Artificial Neural Networks (ANN) input factor integer confirmation, finally designated 4 input factors, respectively be flow to, the wave height, the cycle, the average wind velocity, the survey station material for each hour material, the entire kind of nerve network construction is 4 input factors, a hideaway level, an output level is the ocean current, each has 10 neurons, the weight value is respectively 0.7.
This research mainly take kind of Artificial Neural Networks (ANN) as the main analysis formula, again auxiliary carries on the comparative analysis by the AR(2)pattern, the Taiwan Hualian port, the Kaoshiang port, Keelung Port make the object of study, its comparative analysis distinction target respectively is root-mean-square error (RMS) and correlation coefficient (r), the comparative analysis result demonstrated, its kind of Artificial Neural Networks (ANN) operates correlation coefficient (r) on the Hualian port, the Kaoshiang port, Keelung Port respectively is 0.87, 0.87, the 0.88, AR(2) pattern operates the value is 0.74, 0.74, 0.70, kind of Artificial Neural Networks (ANN) operates the root-mean-square error (RMS) value respectively is 0.21, 0.19, the 0.20, AR(2) pattern operates the value is 0.33, 0.38, 0.37, thus it can be seen Kind of Artificial Neural Networks (ANN) passes through repeatedly intersecting compared to operate, gives the suitable input factor and the neuron, to harbor ocean current speed of flow calculation value, commonly used AR(2) pattern for accurate, also saves the operation time, and erroneous production.
| Identifer | oai:union.ndltd.org:TW/098CTU05653008 |
| Date | January 2010 |
| Creators | 蕭智聰 |
| Contributors | 黃清和 |
| Source Sets | National Digital Library of Theses and Dissertations in Taiwan |
| Language | zh-TW |
| Detected Language | English |
| Type | 學位論文 ; thesis |
| Format | 66 |
Page generated in 0.0115 seconds