Advances in smart infrastructure produces a natural demand of system identification techniques for structural health and performance monitoring that can be scaled to regions and large asset inventories. Conventional approaches require sensors to be installed, often in long-term deployments, on the monitored infrastructure systems, which is a costly undertaking when thousands of systems (e.g., bridges) need to be monitored. This paper presents a novel mode shape identification method for bridges that uses data collected from moving vehicles as input—a paradigm that can overcome limitations associated with conventional approaches. The method consists of two steps. First, the data collected from moving measurement points are mapped to virtual fixed points to generate a sparse matrix. Then, a “soft-imputing” technique is employed to fill the sparse matrix. Finally, a singular value decomposition is applied to extract the mode shapes of the bridge. Experiments with synthetic, yet realistic, data are conducted to verify the method. The sensitivity of the proposed approach to different factors, including the number of vehicles, car speed, road roughness, and measurement errors, are also investigated. The results show that the proposed method is capable of identifying the mode shapes of the bridge accurately.

中文翻译：

---

通过非参数稀疏矩阵完成使用以交通速度行驶的移动车辆的桥梁模态形状识别

智能基础设施的进步产生了对用于结构健康和性能监控的系统识别技术的自然需求，这些技术可以扩展到区域和大型资产库存。传统方法需要在受监控的基础设施系统上安装传感器，通常是长期部署，当需要监控数千个系统（例如，桥梁）时，这是一项成本高昂的工作。本文提出了一种新的桥梁振型识别方法，该方法使用从移动车​​辆收集的数据作为输入——一种可以克服与传统方法相关的局限性的范例。该方法由两个步骤组成。首先，将从移动测量点收集的数据映射到虚拟固定点以生成稀疏矩阵。然后，使用“软输入”技术来填充稀疏矩阵。最后，应用奇异值分解来提取桥梁的模态振型。使用合成但真实的数据进行实验以验证该方法。还研究了所提出的方法对不同因素的敏感性，包括车辆数量、车速、道路粗糙度和测量误差。结果表明，所提出的方法能够准确识别桥梁的模态振型。也被调查。结果表明，所提出的方法能够准确识别桥梁的模态振型。也被调查。结果表明，所提出的方法能够准确识别桥梁的模态振型。