Automatic P-phase picking based on machine learning and AIC algorithm and its application in engineering geological hazards warning

Abstract

Accurate P-phase first arrival time is a premise for improving accuracy of seismic source localizations and achieving hazard warning. Traditional algorithms failed to meet the requirements of high precision and accuracy for microseismic (MS) monitoring in deep geological engineering. In this study, a multi-step method: convolutional neural network combined with K-means and AIC (CNN-KA) for picking up arrival time of P-phases is proposed. Firstly, convolutional neural network (CNN) technique is used to recognize waveforms of rock fractures instead of manual. Secondly, maximum overlapping discrete wavelet transform and multi-resolution analysis are combined to denoise. Finally, a new picker was developed by introducing K-means clustering algorithm, which was used to extract the target time window where the P-phase was located. It compensates for inherent shortcomings of AIC when applied to field data itself. Experiments and engineering applications show that the average absolute error of the proposed method (CNN-KA) is 0.0915s at frequency of 200Hz, which is 86.65% lower than STA/LTA algorithm. Automatic location error of rock fracture MS events is reduced from 37.33m to 10.89m. CNN-KA was able to warn a potential geological hazard in a coal mine of Anhui Province, China. The in-situ mine pressure data validated the validity of CNN-KA. The proposed workflow greatly improves accuracy of P-phases and identification of rock fracturing events in geo-engineering. The computed results can be used further for calculating precise parameters of MS sources and early warning of engineering geohazards.