平稳海浪频谱分析方法的比较研究

发布时间：2019-06-03 14:18

【摘要】：海浪谱是描述海浪过程的基本特征参数。利用谱描述海浪并获得海浪要素具有重要的理论和实用意义。海浪观测与分析、海浪预报、海洋环境研究、造船与水港工程等都需要结合海浪谱进行,海浪谱是目前海浪研究的中心问题之一。由于海浪本身的复杂性,小部分海浪谱是由观测到的波要素连同某些假定推导而来,大部分海浪谱则是利用定点波面记录通过特殊谱分析方法得到。国内外已有的诸多对谱分析方法的研究都是对采样特征以及谱估计参量利用控制变量法进行单独讨论,并没有考虑二者之间可能存在的关系,且由于之前的研究讨论一般都是基于现场实测海浪数据,因此也未能分析目标靶谱对谱估计结果可能造成的影响。基于此,本文进行了相关的研究计算。本文通过对具有不同特征目标靶谱,在不同的采样特征条件下由线性波浪叠加的方法生成的多组随机波面应用不同的谱分析方法进行谱分析计算,并用估计精度衡量了估计谱结果相对于靶谱的估计效果,得到了不同参量条件下对应的估计精度曲线,计算分析了不同特征量对FTA法,Welch法,AR模型法谱分析的影响。结果表明影响谱估计效果的因素不只有谱估计方法的参量以及采样特征量,所要研究的目标海浪谱的特征同样在一定程度上决定着能否得到有效的估计谱结果。有可能出现FTA法的最大推移乘积数,Welch法的分段数,以及AR模型的阶数无论如何变化,都得不到满意的谱估计结果的情况。而在各特征参量选值合适的情况下,三种方法都可以取得满意的结果。其中,目标靶谱谱宽对Welch法以及AR模型法的精度曲线分布影响较大,对FTA法估计效果影响较小。谱峰尖度则对三种谱估计方法均有影响,其中AR模型法对谱峰尖度比较大的目标靶谱的估计稳定性较好,而FTA法和Welch法则只在谱峰尖度较小时估计效果较好。采样时长对Welch法以及AR模型法的稳定性影响并不显著,且采样时长与最优段数与模型阶数之间还存在一定的数值关系。采样时长的减小会导致FTA谱估计的稳定性下降。采样频率在满足奈奎斯特采样频率的前提下,对FTA法与Welch法没有显著影响,而AR模型法的最优阶数与采样频率之间存在某种数值关系。
[Abstract]:Wave spectrum is the basic characteristic parameter to describe the wave process. It is of great theoretical and practical significance to use spectrum to describe waves and obtain wave elements. Wave observation and analysis, wave prediction, marine environment research, shipbuilding and water port engineering all need to be combined with wave spectrum. Wave spectrum is one of the central problems in wave research at present. Because of the complexity of the wave itself, a small part of the wave spectrum is derived from the observed wave elements together with some assumptions, and most of the wave spectrum is obtained by the special spectral analysis method by using the fixed-point wavefront records. Many existing studies on spectral analysis methods at home and abroad discuss the sampling characteristics and spectral estimation parameters separately by using the control variable method, and do not consider the possible relationship between the two methods. Because the previous research and discussion are generally based on the measured wave data, it is not possible to analyze the possible influence of the target spectrum on the spectral estimation results. Based on this, the related research and calculation are carried out in this paper. In this paper, the spectral analysis of multiple groups of random wavefront generated by linear wave superposition method under different sampling characteristics is carried out by using different spectral analysis methods for the target spectrum with different characteristics. The estimation effect of the estimated spectrum results relative to the target spectrum is measured by the estimation accuracy, and the corresponding estimation accuracy curves under different parameters are obtained. The effects of different characteristic quantities on the spectral analysis of FTA method, Welch method and AR model method are calculated and analyzed. The results show that the factors that affect the spectral estimation effect are not only the parameters of the spectral estimation method and the sampling characteristics, but also the characteristics of the target sea wave spectrum to a certain extent determine whether the effective estimation spectrum results can be obtained. It is possible that the maximum moving product of FTA method, the piecewise number of Welch method and the order of AR model can not get satisfactory spectral estimation results. Under the condition that each characteristic parameter is suitable, all three methods can obtain satisfactory results. Among them, the spectral width of the target has a great influence on the accuracy curve distribution of the Welch method and the AR model method, but has little effect on the estimation effect of the FTA method. The spectral peak cusp has an influence on all three spectral estimation methods, among which the AR model method has better stability in the estimation of the target spectrum with large spectral peak cusp, while the FTA method and Welch rule have better estimation effect only when the spectral peak cusp is small. The sampling time has no significant effect on the stability of Welch method and AR model method, and there is still a certain numerical relationship between the sampling time and the optimal number of segments and the order of the model. The decrease of sampling time will lead to the decrease of the stability of FTA spectrum estimation. On the premise of satisfying the Nyquist sampling frequency, the sampling frequency has no significant effect on the FTA method and the Welch method, but there is a numerical relationship between the optimal order of the AR model method and the sampling frequency.
【学位授予单位】：大连理工大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：P731.22

【参考文献】