Seismic Motion Detection and Classification Methodology for Buildings Using DFT and SVM

Abstract

Analyzing the acceleration signals of buildings during a seismic event helps us to identify the vibrational intensity of the structure. This research proposes a methodology to recognize the level of the vibrational intensity of a building during an earthquake with a few seconds before its maximum vibration using the P wave response. The proposed methodology is based on the Discrete Fourier Transform (DFT) and Support Vector Machine (SVM). This methodology results in an alert level that is classified according to its vibrational intensity (low, moderate and high). Each building could have different alerts in the event of an earthquake since may have different natural frequencies. A prototype implemented with a Raspberry Pi V4 B embedded system, two acceleration sensors called MPU6050 based on MEMS and a Wi-Fi antenna, mainly, is used. Also, filters were used to attenuate noise. The STA/LTA algorithm is used to compare the detection time. The results show that it is convenient to use the methodology for two main reasons. Firstly, this uses one tenth of the samples needed in the STA/LTA algorithm for detection and with the same efficiency. Secondly, the classification of the alert level in the building has a correlation greater than 0.8 with the PGA of the seismic signal that occurred in its structure.