Abstract
Abstract This paper presents an approach for creating a digital seismic twin of a building using a dense network of low cost sensors and intelligent analysis of vibration time series. Vibration data recorded at different levels of the structure are processed using statistical features describing the intensity and variability of motion. Anomalous behavior is detected using a machine learning method trained on normal operating conditions. The results show that normal structural states form a compact group in feature space, while abnormal and transitional states are located outside this group. The proposed approach enables early detection of weak dynamic changes and demonstrates potential for real time monitoring and improved safety of urban structures. In addition, the system architecture supports continuous data acquisition and scalable processing, allowing integration into existing monitoring infrastructures. The proposed framework can be applied to various types of buildings and urban environments, providing a cost-effective solution for structural health assessment. Future work includes incorporating advanced deep learning models and validating the approach on larger datasets under diverse environmental conditions.