Onsite Early Prediction of PGA Using CNN With Multi-Scale and Multi-Domain P-Waves as Input

Abstract

Although convolutional neural networks (CNN) have been applied successfully to many fields, the onsite earthquake early warning by CNN remains unexplored. This study aims to predict the peak ground acceleration (PGA) of the incoming seismic waves using CNN, which is achieved by analyzing the first 3 s of P-wave data collected from a single site. Because the amplitude of P-wave data of large and small earthquakes can differ, the multi-scale input of P-wave data is proposed in this study in order to let the CNN observe the input data in different scales. Both the time and frequency domains of the P-wave data are combined into multi-domain input, and therefore the CNN can observe the data from different aspects. Since only the maximum absolute acceleration value of the time history of seismic waves is the target output of the CNN, the absolute value of the P-wave time history data is used instead of the raw value. The proposed arrangement of the input data shows its superiority to the one directly inputting the raw P-wave data into the CNN. Moreover, the predicted PGA accuracy using the proposed CNN approach is higher than the one using the support vector regression approach that employed the extracted P-wave features as its input. The proposed CNN approach also shows that the accuracy of the predicted PGA and the alert performances are acceptable based on data from two independent and damaging earthquakes.

Cumulative absolute velocity prediction for earthquake early warning with deep learning

“…EEW systems usually determine the potential earthquake damageability based on the threshold values of PGA or PGV (Fahjan et al, 2011;Hsu & Huang, 2021;Y. M. Wu et al, 2011;Zollo et al, 2010), so it is necessary to assess the effect of predicted CAV by DLcav in estimating earthquake damage.…”

“…EEW systems usually determine the potential earthquake damageability based on the threshold values of PGA or PGV (Fahjan et al., 2011; Hsu & Huang, 2021; Y. M. Wu et al., 2011; Zollo et al., 2010), so it is necessary to assess the effect of predicted CAV by DLcav in estimating earthquake damage. For nuclear power plants, it is proposed that incoming earthquakes would pose damaging threats when the CAV is above 0.3 g s (EPRI, 1988), and this same threshold value is used in this study to determine the damageability of incoming earthquakes.…”

“…The other approach is to predict the PGV and PGA based on the first arriving wave from a single station, which is often used in onsite EEW systems (Böse et al, 2012;Colombelli et al, 2015;Hsu et al, 2013; Y. M. Wu & Kanamori, 2005;Zollo et al, 2010). This second approach does not require earthquake source information and can be more spontaneous than the first one, and many variations have been proposed in this category (Caruso et al, 2017;Hsu et al, 2016Hsu et al, , 2021Liu et al, 2019;Parolai et al, 2017).…”

Peak ground acceleration prediction for on-site earthquake early warning with deep learning

“…Jozinović et al 50 used seismic data from central Italy and successfully predicted PGA using the initial 7–15 s three-component seismic waves from multiple stations as inputs to CNN, indicating that this method has similar errors to the GMPE method developed by Bindi et al 26 . To fully utilize the information of initial P-waves, Hsu et al 51 proposed a method for predicting PGA by inputting P-waves into a CNN after multiscale and multidomain preprocessing and showed that the accuracy of this method exceeded that of the SVR method proposed in 2013 36 . Chiang et al 52 inputted three-component first-arrival seismic waves into a CNN to predict whether the PGA exceeded the threshold in a classified form.…”

Early Peak Ground Acceleration Prediction for On-Site Earthquake Early Warning Using LSTM Neural Network

“…Recently, many advanced methods have been proposed to estimate the magnitude and intensity of the earthquakes based on deep learning approaches. Hsu and Huang et al (2021) attempted to exploit a deep convolutional neural network (CNN) to automatically extract useful features from previously measured P-wave data without losing too much information in the seismic waveforms and successfully predict the PGA of coming earthquakes based on a single-station approach. Similarly, Chiang et al (2022) employed CNN to extract features of the initial P wave measured at a single station and predicted whether the maximum PGA of the coming earthquake of that station was larger than a pre-defined threshold.…”