一种聚类优化的传感器布置方法研究

Optimal sensor placement by using clustering method

复杂结构试验/理论振型的匹配是模态参数型修正过程的重要内容。以测量并获取用于模态参数识别的最佳信息为目标,实现布置在结构上有限数量的传感器能有效避免信息冗余,提出了一种聚类优化的传感器布置方法。根据结构模态中各自由度振型的动力相似性,应用k-means聚类算法对自由度进行自动集结并分类。采用有效独立法分别从各聚类自由度中搜索出模态分辨率最高的传感器位置作为实际的测量位置。最后通过一个悬臂梁、一个悬臂薄板的数值分析和一个旋转滤光轮组件的模态测试试验对该优化布置方法进行验证。分析结果表明,这种方法能有效选出独立敏感性测点,并且具有较高的搜索效率。

The match of experimental/theoretical models of complex structures is an important part in modal parameter modification. To measure and obtain the best information, a clustering optimal method for sensor placement was proposed. The method can effectively avoid the information redundancy when using a limited number of sensors. According to the dynamic similarity of the mode shape values in important modes, the DOFs were auto-clustered by using A：-means clustering algorithm. The effective independent method was used to search out the sensor locations with the highest modal resolution from each cluster. Finally, the numerical analyses on a cantilever beam and a cantilever plate and the modal test of a filter wheel assembly were carried out to verify the optimization method. The results show that the method can effectively select the independent sensitive locations with a much higher search efficiency.