Cluster Analysis of Earthquake Ground-Motion Records and Characteristic Period of Seismic Response Spectrum

Classification of earthquake ground-motion records is carried out in this article using K-means cluster technique. Two kinds of cluster attributes are introduced, which are augmented by the frequency band of energy concentration and by the magnitude distance of earthquake ground motions. The effectiveness of cluster analysis is testified through the investigation of 7692 earthquake ground-motion records. It is revealed that the clustered ground-motion records show significant group discrepancies in spectral properties. A new seismic response spectrum is then suggested taking account of the influences of magnitude, propagation distance, and site class. An optimization criterion is introduced to identify the characteristic period of the suggested seismic response spectrum, through a comparative study between the mean response spectrum of the clustered ground-motion accelerations and the seismic response spectrum in provisions. It provides a comprehensive argument for characterizing the spectral properties of ground motions not only with respect to specific site but also with respect to specific magnitude and propagation distance. Numerical investigations indicate that the cluster analysis technique has the benefit of dealing with massive ground-motion records, whereby the suggested seismic response spectrum and associated characteristic periods are applicable of integrated design of seismic structures.