Modal parameters provide an insight into the dynamical properties of structures. In the time–frequency domain–based methods, time–frequency ridges contain crucial information on the characteristics of multicomponent signals, and manually extracting time–frequency ridges is a huge burden, especially when long-time time-varying modal parameters are focused on. In this study, time–frequency ridge extraction is converted into a multi-objective optimization problem, and a new hybrid method of multi-objective particle swarm optimization and k-means clustering is proposed to solve such a multi-objective optimization problem. In the hybrid method, the particle swarm is partitioned into sub-swarms by k-means clustering, and the sub-swarms are used to search new solutions for updating a finite-sized external archive, which is used as the exclusive centroids of the k-means clustering. Simultaneously, the finite-sized external archive serves as global best positions of sub-swarms. Both simulated and experimental cases are applied to validate the hybrid method. With the aid of the hybrid method, the influence of varying temperatures on modal parameters of a column beam is experimentally analyzed in detail.

中文翻译：

---

多目标粒子群和k-means聚类的混合方法及其在时频域模态参数估计中的应用

模态参数可洞悉结构的动力学特性。在基于时频域的方法中，时频脊包含有关多分量信号特性的重要信息，手动提取时频脊是一个巨大的负担，尤其是在长期使用随时间变化的模态参数时。该研究将时频脊线提取转化为一个多目标优化问题，并提出了一种新的多目标粒子群优化与k均值聚类的混合方法来解决这种多目标优化问题。在混合方法中，粒子群通过k均值聚类划分为多个子群，这些子群用于搜索新的解决方案以更新有限大小的外部档案，用作k均值聚类的排它质心。同时，有限大小的外部存档将成为子群的全局最佳位置。仿真和实验案例均被用于验证混合方法。借助混合方法，通过实验详细分析了温度变化对柱梁模态参数的影响。