Natural Hazards Research (Dec 2021)
Novel method to pick P waves based on a grey rational analysis
Abstract
For monitoring the earthquake ground motion, it is essential to rapidly and accurately detecting P waves' arrival in the seismic signal. To detect the seismic events and pick phases' arrival, some traditional methods are based on the statistical theory, such as the short-term average and long-term average ratio (STA/LTA) method, the modified energy ratio (MER) method and the Akaike information criteria (AIC) method. However, when the seismic event occurred, the ground motion signal no longer obeys the normal distribution, and the probability distribution of the seismic signals is unknown or uncertainty. Due to these problems, the traditional methods are limited to pick P waves. A new P waves' picking method is proposed based on grey relational analysis. The method depends on the grey relational coefficients, which are calculated by a known template and the seismic signal. The grey relational coefficients represent the similarity between a chosen template and the seismic signal. The grey relational coefficients' changes represent the P waves' arrival. Therefore, by calculating the grey relational coefficients, the P waves can be detected with a given confidence level. Through the sample analysis of Wenchuan Ms.8.0 Earthquake's main-shock and its aftershocks, by the grey relational analysis (GRA) method the P waves' picking mean error is 0.02s and its variance error is 0.053s respectively, which performs better than the mean error (0.09s) and variance error (0.18s) by the AIC method. To verify the robustness of the GRA method some random noises were added to the original seismic data, and the P waves' picking were performed by both the GAR method and the AIC method. Then the variance error of the GRA method increased to 0.10s, which is also less than the variance error (0.22s) by the AIC method. The simulation results shows that the GRA method not only provides the better accuracy than the AIC method, but also is suitable for the continuous seismic signals. The GRA method is based on the non-statistical theory and will be a beneficial attempt for the EEW system.
