用人工神经网络技术表达船舶型线的算法

应用人工神经网络技术表达船舶型线，针对多元函数逼近问题，分析了影响ＢＰ算法的诸多因素。尝试了逐层学习算法，算法中的优化方法选用共轭梯度法，学习率的调整采用一维法求取最佳值的方法，编辑算结构表明，其收敛速度比普通ＢＰ算法快约一个数量级，在此基础上，以一般３．６万ｔ散货船的型线为例，完成了船舶型线的人工神经元网络表达。

Algorithm Improvement Using ANN Technology to Express Hull Lines

Applying technology of artificial neural networks (ANN) to express hull lines, aiming at the approximation problem of a continuous function of n variables, factors affecting the effect of BP algorithm were analyzed and a new improving method was practiced. The newly obtained method adopts the algorithm of optimizing hidden layer output (OHLO) in which the optimization method is Fletcher Reeves conjugation gradient method and the adjustment method of the learning rate is one dimensional searching method. The result shows that it can speed the convergence of the learning in a measure. In this paper, a 36 000 t bulk carrier was taken as an example to express the hull lines using this newly obtained method. The conclusion shows that using the ANN method to express the hull lines is viable and effective.