On Long Numerical Simulations at Extreme Seas
On Long Numerical Simulations at Extreme Seas
复制标题
极端海洋的长时间数值模拟
DOI：
--
复制
发表时间：
2005
期刊：
影响因子：
--
通讯作者：
V. Belenky
中科院分区：
文献类型：
--
作者：
V. Belenky
本文着重讨论了在从具有恒定频率步长的谱重建波记录时所遇到的已知的“统计自重复”效应。为这样的记录计算的相关函数显示高值的重复模式。这意味着在过程的某些时间段之间存在不切实际的强概率依赖性。即使直接从频谱计算自相关函数而不实际重建波记录，这种影响仍然存在。尝试使用可变频率步进来解决该效应会导致误差的扩散。它显示为一系列不切实际的高值，而不是离散的峰值。作者以前的工作也发现，该效应的存在或不存在强烈地依赖于数值积分的方法。本文的结论是，这种影响的本质是一个数值误差所造成的高度振荡特性的被积函数在拉普拉斯变换（频谱和自相关函数之间的关系）为大的时间值。本文还讨论了这一结论的实际意义。
The paper is focused on the previously known effect of “statistical self-repetition” encountered when a wave record is reconstructed from spectrum with a constant frequency step. A correlation function calculated for such a record shows a repetitive pattern of high values. It means an unrealistically strong probabilistic dependence between certain time sections of the process. The effect remains even if the autocorrelation function is calculated directly from spectrum without actual reconstruction of the wave record. An attempt to work the effect around using a variable frequency step leads to a spread of the error. It appears as a series of unrealistically high values instead of discrete peaks. The previous work of the author also found that the presence or absence of the effect strongly depends on the method of numerical integration. The paper concludes that the nature of this effect is a numerical error caused by the highly oscillatory character of an integrand in Laplace transform (relation between spectrum and autocorrelation function) for large values of time. The paper also discusses the practical implications of this conclusion.