一种基于识别概率直方图的随机减量改进技术

An improved random decrement technique based on identification probability histogram

在采用随机减量技术进行工作模态分析过程中,触发条件和参考通道通常需要人为指定。受实际环境和测量误差等的影响,单一触发条件和特定参考通道导致模态辨识结果具有一定的偶然性。结合基于数据挖掘的识别概率直方图方法,对传统随机减量技术进行了改进。在遍历参考通道和多次触发取值的情况下获得多个结构模态辨识结果,并根据频率容差和模态置信度容差准则进行一致性聚类,继而绘制相应的识别概率直方图,最终选取识别概率大的聚类结果作为结构模态结果。通过结构算例和模型实验,显示了基于识别概率直方图的随机减量改进技术的有效性和抗噪性。

During the operational modal identification of structures with the Random Decrement Technique,the trigger condition and reference channel are chosen artificially.Subjected to the operational environment and the measurement errors,the identification results will exhibit uncertainties for the case of single trigger condition and specific reference channel.To avoid these problems,an identification-probability histogram (IpHist) method by using the data mining technique is proposed to improve the traditional RDT method.Firstly,a series of modal data is identified with different trigger conditions and reference channels.Then,all the alternative modes are clustered into several categories by using the criteria of frequency tolerance and MAC tolerance,and the identification probability of each category is obtained from the corresponding identification-probability histogram.Finally,the clustered modes with large identification probability are chosen to be the structural modes.By taking a 3DOF numerical model and an actual truss model,the modal identification results by RDT based on IpHist are shown and compared with considerating the measurement noise.