Research progress of baseline correction for near-field strong motion accelerogram

The baseline drift is often found in strong motion accelerograms, especially in the near-field accelerogram. The small baseline drift in accelerogram will lead to linear drift in the velocity time history obtained through single integration and parabolic drift in the displacement time history obtained through double integration, which are unreasonable non-physical characteristics. Low frequency errors and tilting or rotation of the ground are the main factors causing baseline drift in strong motion accelerograms. This paper introduces the types of low frequency error and analyzes the effects of low frequency errors and ground tilt or rotation on the velocity and displacement time history obtained from the original accelerogram in detail. The principle of baseline correction for near-field strong motion accelerogram is discussed. Considering the baseline drift caused by the low frequency error, we discuss the basic principle of the most widely used high-pass filtering method, and how to select filters. Then we analyze adding zero pads before acausal filtering, cosine to smoothing the transition band of zero pads and accelerogram. The selection criteria of acausal cut-off frequency is investigated. Furthermore, we discuss how to avoid the incompatibility of velocity, displacement time histories, and acceleration response spectra after removing zero pads. Aiming at the baseline drift caused by ground tilt or rotation, the principle of the two-stage baseline correction method is illustrated by Iwan. The research progress and compute methods of the baseline correction methods developed and improved based on the principle of Iwan’s method are also discussed, and the problems of these methods are pointed out. We also illustrate the selection of baseline correction methods of near-field strong motion accelerogram for different magnitude scales. Because there is no available method that can quantify the contributions of different error sources and not available strong motion accelerogram of six components (three translational and three rotational components), most of the baseline correction methods still belong to the experience and the semi-empirical. There is no general baseline correction method applied to all near-field strong motion accelerograms of different earthquakes. With the accumulation of near-field strong motion accelerogram on a global scale, the maturity of extensive data analysis and machine learning, based on the principle of the Iwan method, automatically and quickly identified the reliable beginning times of the strong shaking phase and the post-seismic phase from the massive near-field strong motion accelerograms will be the future development direction.