Detection and identification of seismic P-Waves using Artificial Neural Networks

Detection and identification of seismic P-Wave is useful in event location and event detection. This involves an intensive amount of pattern recognition. For the recognition of seismic phases, no probabilistic distribution model performs as well as Artificial Neural Network(ANN). Back Propagation Neural Network (BPNN) was applied for the automatic detection and identification of local and regional seismic P-Waves. For a set of three-component seismic data, four attributes were used as input to the ANN: Degree of Polarization (DOP), Auto Regression Coefficient (ARC), Ratio between Short time average and Long time average (STA/LTA) and Ratio of Vertical power to Total power (RV2T). These four attributes were calculated in the frequency band of 1-8 Hz with a 2 second moving window. The results of preliminary training and testing with a set of various local and regional earthquake recordings show that the ANN achieved 95% correct rate of P-Wave detection and identification. 90% of the P-Waves were detected with a maximum deviation of 0.1 sec from correct manual pick up.