A rapid detection method of earthquake infrasonic wave based on decision-making tree and the BP neural network

摘要

In this paper, we propose a rapid automatic detection method based on decision-making tree combined with BP neural network for the earthquake infrasonic wave. Three factors of frequency (F), duration period (P) and amplitude (A) of seismic infrasonic waves were selected as the network input parameters in the three-layered BP neural network. A total of 30 different infrasonic waves were tested in this model. The results indicate that the successful decision rates can reach 0.8 with input parameters F, P and A. When using proper thresholds for the input parameters, such as F=0.005 Hz, P=500 s and A=5 Pa, the detection results are very closed to the true input signals, and the infrasonic sources as well as their main characteristics can be effectively recognised and classified rapidly. This new method could provide clues and thoughts for the short-term earthquake infrasonic wave detection.