摘要
本文提出了一种非正态分布随机序列建模和模拟的新方法。该方法的基本思想是将ARMA模型产生的正态分布的相关序列直接通过Johnson分布转换式得到需要的非正态分布序列,而转换前后的序列的自相关函数基本一致。该方法克服了Davies方法中存在的一些问题,尤其适用于二维非正态分布序列的建模和模拟。实际加工表面轮廓的模拟结果和给定偏斜度和驼峰度的二维时序模型的模拟结果证实了该方法的有效性。
In this paper a new simulation method for random series with non-normal distribution is presented The basic idea of this method is that the desired non-normal series are directly obtained from normal series which is produced from ARMA model by using Johnson' Distribution Transformation, and the correlation function of series before transformation is very similar as that after transformation. Some limitations existing in the Davies' method are overcame in this method and it especially suited the modelling and simulating of two dimension .(2-D) non-normal field. The results of simulation of real surface profiles and 2-D field with given skewness and kurtosis have proved that this method is useful.
基金
国家自然科学基金6872031的一部份
关键词
时间序列
模拟
非正态分布
建模
time series, simulation, non-normal distribution
