APPLICATION OF THE SELF-ORGANIZING MAPS FOR DURATION CHARACTERISTICS OF EARTHQUAKE MOTION

Abstract

 The evaluation of earthquake motion is basic problem in the earthquake engineering. In this study, the four-dimensional feature vector representing characteristics of duration of earthquake motion is defined based on the discretized probability mass function form of Husid plot of acceleration time histories. The feature vector is composed of average, standard deviation, skewness and kurtosis corresponding to first raw moment and second to fourth central moment of the PMF, respectively. As a case study, the proposed method was applied to the earthquake motions observed from the 2011 off the Pacific coast of Tohoku Earthquake. It was confirmed that the feature vectors well represented the duration characteristics. Next, the algorithm of SOM (Self-Organizing Maps) was applied to the set of feature vectors. The duration characteristics were visualized in a systematic way by the two-dimensional mapping using SOM.