Preprocessing Method for Vibration Signal Data

Vibration signal is a primary type of data in structural health monitoring. Modal frequencies, damping, and other dynamic characteristics of structures can be identified from vibration signals. However, vibration signals collected from structural health monitoring systems often contain noise and anomalies, which seriously affect the identification of structural dynamic characteristic indicators. In order to improve the accuracy of dynamic characteristic indicator identification, this paper proposes a structural vibration signal preprocessing method based on the characteristics of structural vibration signals. Firstly, outlier values are annotated by using boxplot, and missing values are interpolated using cubic spline interpolation. The original signal is decomposed using variational mode decomposition (VMD) based on the center frequency method, and effective intrinsic mode function (IMF) components are selected through kurtosis value filtering, followed by reconstruction of the selected components. The preprocessing effect of the proposed method is examined through two cases: simulated signals and engineering measured signals. The results show that signals processed by the proposed method exhibit the best signal-to-noise ratio, and modal frequencies can be accurately identified. Therefore, the proposed method effectively cleans vibration signals, which is beneficial for improving the accuracy of dynamic characteristic indicator identification.