Automated modal identification and tracking of high-rise buildings are challenging due to complex ambient excitations, closely spaced modes, nonlinear and time-varying modal parameters, and demands for high efficiency and accuracy, yet critical for structural health monitoring (SHM) to detect potential damage and ensure structural safety. This study proposes an automated intelligent modal identification and tracking (AIMIT) method for high-rise buildings without prior information. The AIMIT framework consists of four main steps: (1) a deep learning-based model is developed to automatically extract two fundamental structural indices from dynamic response data; (2) a Monte Carlo-based stochastic subspace identification method is employed to generate a large set of modal candidates; (3) an iterative screening strategy is proposed to eliminate spurious modes and enable robust automatic modal identification; and (4) an adaptive tracking algorithm is applied for continuous modal tracking. Numerical validation demonstrates the high efficiency (processing 1-hour data in just 2.4 seconds) and accuracy (errors for damping ratio < 30%) of the proposed method. Field applications across multiple high-rise buildings further confirm its general applicability. The proposed AIMIT approach significantly enhances SHM performance in high-rise buildings and holds promise for extension to other civil structures such as bridges and long-span systems.

中文翻译：

---

无需先验信息的高层建筑的自动智能模态识别和跟踪

由于复杂的环境激励、紧密间隔的模态、非线性和时变模态参数以及对高效率和准确性的要求，高层建筑的自动模态识别和跟踪具有挑战性，但对于结构健康监测 （SHM） 检测潜在损坏和确保结构安全至关重要。本研究提出了一种无先验信息的高层建筑自动智能模态识别和跟踪 （AIMIT） 方法。AIMIT 框架包括四个主要步骤：（1） 开发基于深度学习的模型，从动态响应数据中自动提取两个基本结构指标;（2） 采用基于蒙特卡洛的随机子空间识别方法来生成大量模态候选者;（3） 提出了一种迭代筛选策略来消除杂散模式并实现稳健的自动模态识别;（4） 自适应跟踪算法应用于连续模态跟踪。数值验证证明了所提出的方法的高效率 （仅需 2.4 秒处理 1 小时数据） 和准确性 （阻尼比误差 < 30%）。多座高层建筑的现场应用进一步证实了其普遍适用性。拟议的 AIMIT 方法显著提高了高层建筑的 SHM 性能，并有望扩展到其他土木结构，如桥梁和大跨度系统。