Variations in the Natural Frequencies of Millikan Library Caused by Weather and Small Earthquakes

Abstract

In 2001, the Southern California Seismic Network (a part of the California Integrated Seismic Network) installed a 3-component Episensor force-balance accelerometer on the top floor of Millikan Library (station name MIK). This accelerometer is recorded by a 24-bit Quanterra data logger, and the dynamic range of the system is about 10 to the 7th. This means that ambient vibrations of the building can be recorded with a signal to noise ratio exceeding 100. Data from this system is recorded in a similar fashion to other more conventional seismographs that are installed in the ground throughout southern California. In particular, data is stored at 80 sps for any local earthquakes large enough to record on the seismic network, and data is also stored continuously after anti-alias filtering and resampling to 20 sps. The authors analyze a near-continuous three-year-long recording of the motion of the top floor of Millikan Library, which is a 9-story concrete building (a combination of shear walls and moment-resisting frames) built in 1966. The fundamental mode natural frequencies of the building (about 1.7 Hz N-S, 1.2 Hz E-W, and 2.5 Hz torsion) can be clearly observed from even short stretches of ambient vibration data. Using this data the authors have documented that changes in natural frequency are unambiguously correlated with the following factors. 1) There are diurnal variations in all natural frequencies on the order of 1-2%. 2) Heavy rain increases the E-W and torsional frequencies by up to 3%. The frequencies typically return to the pre-rain levels within about 5 days. 3) High temperatures (40 ° C) raise all natural frequencies by 1-2%. 4) Strong winds (50 kph) decrease natural frequencies. 5) Construction of non-structural office partitions on floors 3, 4, and 5 in 2003 increased natural frequencies, with the E-W frequency increasing by 5%. In addition to studying ambient vibrations, the authors also studied the apparent natural frequencies in a number of small to moderate earthquakes in southern California. They obtained the apparent natural frequencies by deconvolving the ground motions at a site 100 meters from the building (SCSN station GSA) from the Millikan roof records. They observed that the natural frequency dropped significantly for even small motions. For example, the E-W apparent frequency dropped from 1.18 Hz (ambient) to 1.06 Hz during a M 5.4 earthquake at a distance of 119 km (peak ground acceleration of 0.018 g). It returned to pre-earthquake levels within several minutes.