Prediction of structural response is necessary for evaluating condition and quantifying vulnerability of structural systems exposed to seismic loads. Traditional modeling techniques for infrastructure systems such as finite elements are typically limited by inherent modeling assumptions as well as the prohibitive computational effort required for analysis. This necessitates the development of surrogate models that serve as a basis for predicting structural response. Deconvolution interferometry is a viable data-driven approach for such a task that uses single component sensor data to generate a set of impulse response functions for a structure of interest that constitutes the required surrogate model of the structure. The resulting surrogate model aids in both dynamic characterization as well as for accurate response prediction. However, it is limited to cases where motions in various degrees of freedom of a structure can be decoupled. This decoupling requires dense sensor deployment as well as prior knowledge about the structure’s geometry. To overcome this limitation, in this paper we propose a multi-component deconvolution seismic interferometry approach to develop a surrogate model for response prediction for cases with sparse sensor deployment and limited information about the structure of interest. The resulting model incorporates various sources of uncertainties namely measurement noise, effects of variations of temperature and humidity, and human activity induced vibrations by predicting a probabilistic structural response. We demonstrate the efficacy of the proposed algorithm by applying it to field monitoring data collected from structures with sparse sensor deployment in the Groningen region of the Netherlands for a period of approximately 10 months on average.

结构响应的预测对于评估条件和量化承受地震荷载的结构系统的脆弱性是必要的。用于基础结构系统（例如有限元）的传统建模技术通常受固有建模假设以及分析所需的过高计算量的限制。这就需要开发替代模型，以作为预测结构响应的基础。反卷积干涉测量法是一种可行的数据驱动方法，可用于使用单个组件传感器数据为目标结构生成一组脉冲响应函数的脉冲响应函数，该脉冲响应函数构成了所需的结构替代模型。生成的替代模型有助于动态表征以及准确的响应预测。然而，它限于可以将结构的各种自由度的运动解耦的情况。这种解耦需要密集的传感器部署以及有关结构几何形状的先验知识。为了克服这一局限性，在本文中，我们提出了一种多分量反褶积地震干涉法，以开发一种替代模型，用于传感器部署稀疏且有关结构信息有限的情况下的响应预测。生成的模型包含各种不确定性来源，即测量噪声，温度和湿度变化的影响以及人类活动通过预测概率结构响应而引起的振动。