单水滴致噪声平均能量密度谱的神经网络模型

摘要

The average energy spectral density of underwater noise for a given single waterdrop is the basic parameter of rain noise theory proposed by Medwin. The neural networks model is constructed to calculate the average energy noise spectral density for a single waterdrop by noise spectral levels caused by a swarm of waterdrops with confirmed drop-size distribution. It is proved to be effective by the comparison of the calculated data with the laboratorial data and the characters of the MSE through three different training models.%单水滴击水致水下噪声平均能量密度谱是Medwin雨致噪声理论中的基本参量.基于人工神经网络算法,建立了利用一定粒度分布的水滴群及其击水后的噪声谱反演单水滴致水下噪声能量密度谱的模型.通过分析三种训练模式,得到反演值与真实值的对比及误差指标MSE的起伏性和收敛速度,验证了利用人工神经网络算法研究雨致噪声的有效性.