Instantaneous Modal Frequency Identification of Bridge Cables under the Influence of Abnormal Monitoring Data Using SSTFT and M-DBSCAN Models

As a crucial element of bridge structures, the identification of cable modal frequencies holds significant importance. The dynamic characteristics of cables are directly associated with their tension, and precisely estimating cable tension is fundamental for evaluating the safety and performance of bridges. However, in practical monitoring, sensor malfunctions and data transmission errors often contaminate the monitoring datasets with various types of abnormal data. It can severely disrupt modal parameter identification and lead to inaccurate estimations of cable tension. This paper presents an instantaneous modal frequency identification framework utilizing the synchrosqueezing short-time Fourier transform (SSTFT) and the modified density-based spatial clustering of applications with noise model (M-DBSCAN). SSTFT enables high-precision time-frequency analysis of dynamic monitoring data by dynamically adjusting the energy distribution of frequency components. Meanwhile, M-DBSCAN can effectively detect and eliminate incorrect frequency identification results caused by abnormal data. The accuracy and effectiveness of the proposed framework are verified using the acceleration monitoring data of the cables of Qinglinwan Bridge. The results show that after removing abnormal data, the negative correlation between modal frequency and ambient temperature is accurately captured, laying a foundation for further analysis of high-frequency fluctuation features. Moreover, this framework integrates abnormal data detection directly into the frequency identification process, reducing computational costs and enhancing the reliability of identification results.