NGA high-pass filters remove important real signals; simple tilt correction is preferable when predicting collapse.

Becky Roh, Kenny Buyco, & Thomas H. Heaton

Published August 8, 2017, SCEC Contribution #7434, 2017 SCEC Annual Meeting Poster #255

Horizontal accelerometers record a combination of horizontal ground acceleration and gravitational changes from ground tilt. Traditionally, the tilt effect is removed in data processing to attempt to find the actual ground motion.

NGA uses zero-phase high-pass Butterworth filters to develop corrected records. When applied to the displacement, this type of filter removes long periods that result from tilting, but at the cost of creating a non-causal precursor and removing the static offsets in displacement that naturally occur close to faulting during an earthquake. We question if it is beneficial to use zero-phase high-pass filters to remove long-period accelerations resulting from changes in the ground tilt if they remove static offsets that are crucial in near-source responses.

We examine strong motion data from the 1999 Mw7.6 Chi-Chi, 2015 Mw7.8 Nepal, 2016 Mw7.0 Kumamoto, and 2016 Mw7.8 Kaikoura earthquakes. First, we derive a tilt time history to explain the linear trends found in velocity. In comparison to the peak ground acceleration of these ground motions, the acceleration from tilt is comparatively small, ranging from 0.00027 to 3.42 degrees. After removing the pre-event mean from the acceleration, we apply a tilt correction by removing the linear trend from the velocity. Double integration of the acceleration records after this correction typically gives a stable representation of the actual displacement in the earthquake. We pad the acceleration time series with zeros and apply a zero-phase high-pass 4th-order Butterworth filter at corner periods ranging from 10 to 60 seconds.

High-pass filters remove potentially important components of the real signal. Furthermore, response spectral analysis itself is already a form of filter. Therefore, when generating response spectra to analyze performance of structures, raw or tilt corrected records may be more appropriate than filtered ones.

To investigate the different building responses under raw, tilt corrected, and filtered ground motions, incremental dynamic analysis (IDA) is performed on four representative steel moment frame building models with heights ranging from 3 to 55 stories. For each ground motion, IDA identifies the collapse limit for each building model. Various heights of steel moment frame models are chosen to evaluate if filtering effects become more pronounced as building height increases.

Citation
Roh, B., Buyco, K., & Heaton, T. H. (2017, 08). NGA high-pass filters remove important real signals; simple tilt correction is preferable when predicting collapse.. Poster Presentation at 2017 SCEC Annual Meeting.

Related Projects & Working Groups
Ground Motions