基于遗传BP神经网络的海底沉积物声速预报
发布时间：2018-08-22 17:49
【摘要】：随着海洋地质学等学科的发展,以及海洋工程和海洋开发的需要,海底沉积物声学特性研究具有重要的现实意义,并受到越来越广泛的重视。海底沉积物通常被认为是一种固液双相介质,其结构和物理性质直接决定了声波在其中的传播速度,是声波传播的物理基础。构建明确、统一的海底沉积物声速与物理参数模型,对于开展声速反演、地声模型建立、工程实践等方面的研究都具有重要的意义。国内外的研究学者对纵波声速与沉积物物理参数之间的相关关系进行了大量实际调查工作,建立了适用于不同海域沉积物的声速与物理参数之间的经验公式。这些经验公式的建立在一定程度上揭示了两者之间的相互关系,但由于经验公式大多采用简单的回归拟合得到,再加上海洋沉积环境的多样性及复杂性,在进行声速预报时,存在回归误差过大、适用范围有限、缺乏物理意义等问题。针对这些问题,本文将在已有BP神经网络预测的基础上,运用遗传算法优化其初始权值和阈值的方法,构建出基于含水量、孔隙度的声速预报模型进行声速预报。同时,将南沙海域采集得到的海底沉积物样品分为两部分,随机抽取120组涵盖陆架、陆坡、海槽等地貌单元的样品作为训练数据,另外剩余6组作为测试数据。经试验对比后发现,在对本区域进行声速预报时,宜采用遗传算法优化的BP神经网络,其要优于传统的单参数、双参数回归拟合预报方法和国内外其他学者所得到的经验公式。此种预报方法具有一定的科学依据和广泛的应用前景,可在今后为建立明确、统一的声速预报模型提供参考。
[Abstract]:With the development of marine geology and the need of marine engineering and marine development, the study of acoustic characteristics of seabed sediments has important practical significance and has been paid more and more attention. Seafloor sediments are generally considered as a solid-liquid biphasic medium, whose structure and physical properties directly determine the velocity of sound wave propagation in which, is the physical basis of acoustic wave propagation. The establishment of a clear and unified model of acoustic velocity and physical parameters of seabed sediment is of great significance for the research of acoustic velocity inversion, the establishment of a geoacoustic model, and engineering practice. Researchers at home and abroad have carried out a lot of practical investigations on the correlation between longitudinal wave velocity and sediment physical parameters, and established empirical formulas between sound velocity and physical parameters suitable for sediment in different sea areas. The establishment of these empirical formulas reveals the relationship between them to some extent. However, because most of the empirical formulas are obtained by simple regression fitting, coupled with the diversity and complexity of the marine sedimentary environment, in the prediction of sound velocity, There are some problems such as too large error of regression, limited scope of application and lack of physical meaning. Aiming at these problems, based on the existing BP neural network prediction, this paper uses genetic algorithm to optimize its initial weight and threshold value, and constructs a sound velocity prediction model based on water content and porosity to predict sound velocity. At the same time, the samples collected from Nansha sea area were divided into two parts. 120 groups of geomorphologic units including shelf, slope and trough were randomly selected as training data, and the remaining 6 groups were used as test data. It is found that the BP neural network, which is optimized by genetic algorithm, is superior to the traditional regression forecasting method with single parameter, double parameter and the empirical formula obtained by other scholars at home and abroad in the prediction of sound velocity in this region. This forecasting method has certain scientific basis and wide application prospect, which can be used as reference for the establishment of a clear and unified sound velocity prediction model in the future.
【学位授予单位】：中国科学院研究生院(海洋研究所)
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：P733.2;P714
本文编号：2197853
[Abstract]:With the development of marine geology and the need of marine engineering and marine development, the study of acoustic characteristics of seabed sediments has important practical significance and has been paid more and more attention. Seafloor sediments are generally considered as a solid-liquid biphasic medium, whose structure and physical properties directly determine the velocity of sound wave propagation in which, is the physical basis of acoustic wave propagation. The establishment of a clear and unified model of acoustic velocity and physical parameters of seabed sediment is of great significance for the research of acoustic velocity inversion, the establishment of a geoacoustic model, and engineering practice. Researchers at home and abroad have carried out a lot of practical investigations on the correlation between longitudinal wave velocity and sediment physical parameters, and established empirical formulas between sound velocity and physical parameters suitable for sediment in different sea areas. The establishment of these empirical formulas reveals the relationship between them to some extent. However, because most of the empirical formulas are obtained by simple regression fitting, coupled with the diversity and complexity of the marine sedimentary environment, in the prediction of sound velocity, There are some problems such as too large error of regression, limited scope of application and lack of physical meaning. Aiming at these problems, based on the existing BP neural network prediction, this paper uses genetic algorithm to optimize its initial weight and threshold value, and constructs a sound velocity prediction model based on water content and porosity to predict sound velocity. At the same time, the samples collected from Nansha sea area were divided into two parts. 120 groups of geomorphologic units including shelf, slope and trough were randomly selected as training data, and the remaining 6 groups were used as test data. It is found that the BP neural network, which is optimized by genetic algorithm, is superior to the traditional regression forecasting method with single parameter, double parameter and the empirical formula obtained by other scholars at home and abroad in the prediction of sound velocity in this region. This forecasting method has certain scientific basis and wide application prospect, which can be used as reference for the establishment of a clear and unified sound velocity prediction model in the future.
【学位授予单位】：中国科学院研究生院(海洋研究所)
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：P733.2;P714
【相似文献】
相关期刊论文 前2条
1 罗忠辉;卢博;;基于主元分析技术的海底沉积物声速预报方程[J];热带海洋学报;2009年03期
2 ;[J];;年期
相关硕士学位论文 前1条
1 陈文景;基于遗传BP神经网络的海底沉积物声速预报[D];中国科学院研究生院(海洋研究所);2016年
，本文编号：2197853
本文链接：https://www.wllwen.com/kejilunwen/haiyang/2197853.html
