Abstract: An improved ARMA identification method based on IEEMD decomposition algorithm was proposed. Firstly, an IEEMD analysis for the measured acceleration signal was carried out, and then the question whether there was a modal aliasing in the obtained intrinsic mode functions (IMFs) was verified by using clustering analysis. Secondly, the fuzzy comprehensive evaluation method was used to calculate the fuzzy similarity coefficients between each IMF and the measured signal so as to select the effective IMF component, and then the principal component analysis and Pareto Diagram method were used to reconstruct the signal of preserved IMFs, so that the effective decomposition of the measured signal and the noise attenuation effect could be achieved. Finally, the reconstructed dynamic signal was regarded as the input of ARMA algorithm to identify the modal parameters. Through analyzing the error percentage between each order’s frequency and truth value by comparative analysis, it is indicated that the vibration signal processed by IEEMD method is used as the input of the ARMA algorithm, so it can get the frequency which is most close to the real values and the error percentage is below 3%. The results verify that the proposed identification method can effectively identify the frequency of the cable-stayed bridge.

Key words: bridge engineering, IEEMD, fuzzy comprehensive evaluation method, principal component analysis, Pareto diagram method, ARMA

摘要： 提出了一种基于IEEMD分解的ARMA改进识别算法。首先对实测加速度信号进行IEEMD分析，之后利用聚类分析检验所得的本征模态函数（IMFs）中是否存在模态混叠；然后采用模糊综合 评价法计算每个IMF与实测信号之间的模糊相似系数，以便选出有效的IMF分量；再利用主成分分析和帕累托图法对保留下来的IMFs进行信号的重构，进而达到对实测信号的有效分解和降噪效 果；最后将重构的动力信号作为ARMA算法的输入，进行模态参数识别。通过对比分析每阶频率与实际值的误差百分比，可知利用IEEMD处理之后的振动信号作为ARMA算法的输入能得到与真实值 最为接近频率值，且误差的百分比都在3%以下，验证了该识别方法能有效的识别到斜拉桥的频率。

关键词: 桥梁工程, IEEMD, 模糊综合评价法, 主成分分析, 帕累托图, ARMA