基于Hilbert-Huang变换的仿真模型排序评估方法

针对输出为非平稳速变数据的仿真模型验证及排序评估问题,提出了一种基于Hilbert-Huang变换的仿真模型排序评估方法。首先采用经验模态分解对仿真和参考输出的非平稳速变数据进行分解得到趋势项和平稳项;然后从距离和形状两方面刻画趋势项误差特征并进行一致性分析,采用基于Hilbert变换的谱密度相容性检验方法对平稳项进行一致性分析,同时为满足相容性检验的条件,证明了基于Hilbert变换的平滑样本谱密度服从χ²分布;最后引入灰色关联分析综合3类一致性分析结果实现仿真模型的排序评估。通过实例应用,验证了方法的有效性。

Ranking evaluation of simulation models based on Hilbert-Huang transform

To address the validation and ranking evaluation of simulation models whose output are non-stationary and fast variational, a ranking evaluation method based on Hilbert-Huang transform is presented. The simulation output and experimental observations with the non stationary and fast variational data are decomposed into trend items and stationary items respectively. The consistency of trend items is analyzed by comparing the difference in distance and shape and the agreement extent of stationary items is obtained by means of the power spectral density analysis based on Hilbert transform. For satisfying the condition of spectral analysis, the smoothing sample spectral density of stationary items based on Hilbert transform is demonstrated to obey the chi-square distribution. Next the three consistency results are integrated depending on grey relational analysis to accomplish ranking evaluation of simulation models. Finally, the effectiveness of this method is verified by an example.