A Basic Study on the Determination of Optimum Ship Route Using Neural Network

本研究利用神经网络这一模式识别的有力工具，提出了一种船舶气象导航的新方法。提出的天气路由神经网络由三层组成，即输入层、隐层和输出层。将航行前5天和后5天覆盖北太平洋的网格点上的5天平均500 hPa高度输入神经网络，利用分析的波浪资料模拟集装箱船在从旧金山弗朗西斯科到东京的不同航线上的航行，并计算这些航线的通过时间，从而产生教师信号。然后，基于通过时间计算每条路线的分数，以便在0.1至0.9的范围内。最高分0.9和最低分0.1分别分配给最短时间路线和最长时间路线。将1978-1983年5个冬季500 hPa高度场的连续两个5天平均值反复输入到网络中进行学习，并修改隐层和输出层各单元的权值和阈值，使网络的输出信号与教学信号相吻合。学习完成后，得到了一组新的连续两个5天平均500 hPa高度场，（1989-1991）的数据输入到网络中，验证了网络的有效性，结果表明，训练后的神经网络的输出信号与目标信号，即模拟计算出的路线得分相吻合，对于大多数旅行来说都很好。总之，建议的天气路由神经网络可以提供最佳或次最佳路线的大部分航程给定的准确连续两个5天平均500 hPa高度模式。

In this research, a new method of ship weather routing was developed using the neural network which is known as a powerful tool of pattern recognition. The proposed weather routing neural network consists of three layrs, i.e.input, hidden and output layrs. The 5-day mean 500hPa heights on the grid points covering the North Pacific Ocean for the first 5 days and the latter 5 days during the voyage were input to the neural network.The teacher signals were produced by simulating the navigation of a container ship on the various routes from San Francisco to Tokyo using the analyzed wave data and calculating the passage times of these routes. A score of each route was then computed based on the passage time so as to range from 0.1 to 0.9. The highest score 0.9 and the lowest score 0.1 were allocated to the minimum time route and the maximum time route, respectively. These scores were used as the teacher signals.To perform the learning of the proposed neural network, many successive two 5-day mean 500hPa height patterns during 5 winter seasons (1978-1983) were input to the network repeatedly, and the weights and threshold of each unit of the hidden and output layrs were modified so as to let the output signals from the network coincide with the teaching signals. After the completion of the learning, a new set of successive two 5-day mean 500hPa height patterns in the different winter seasons (1989-1991) were input to the network to verify the effectiveness of the network.As a result, the output signals of a trained neural network coincided with the target signals, i.e.the scores of the routes calculated by the simulations, very well for most of the voyages. In conclusion, the proposed weather routing neural network could provide the optimum or sub-optimum routes for most of the voyages given the accurate successive two 5-day mean 500hPa height patterns.