论文部分内容阅读
人工神经网络作为一个具有高度非线性映射能力的计算模型,在工程中具有广泛的应用前景。在数值预测方面,它不需要预选确定样本的数学模型,仅通过学习样本数据即可以进行预测。文章介绍了BP神经网络,并针对实际应用中收敛速度慢,平台效应等问题对网络进行了改进并优化,详尽地给出了改进的三层BP神经网络数值预测算法。为测试该算法,选用了著名的XOR(异或)问题和和一个高度非线性的0-1矩阵预测问题对其进行了验证。计算结果表明文中算法能给出令人满意的精度。最后结合船舶与海洋工程的两个实际问题,探讨了利用改进的BP神经网络进行数值预测的方法和应该注意的问题,并给出了一些有益的建议。实践表明,文中给出的改进的BP神经网络数值预测算法值得在船舶与海洋工程中加以应用并推广。
Artificial neural network, as a computational model with highly nonlinear mapping ability, has a wide range of application prospects in engineering. In the numerical prediction, it does not need to pre-select the mathematical model to determine the sample, but only by learning the sample data can be predicted. In this paper, the BP neural network is introduced, and the network is improved and optimized for the problems of slow convergence and platform effect in practical applications. The improved three-layer BP neural network numerical prediction algorithm is given in detail. To test this algorithm, we selected the famous XOR (XOR) problem and a highly nonlinear 0-1 matrix prediction problem to verify it. The calculation results show that the proposed algorithm can give satisfactory accuracy. Finally, combining the two practical problems of ship and ocean engineering, the methods of using BP neural network to predict the value and the problems that should be noticed are discussed, and some useful suggestions are given. Practice shows that the improved BP neural network numerical prediction algorithm given in this paper deserves to be applied and popularized in marine and offshore engineering.