Automatic microseismic-event detection via supervised machine learning

Abstract

Microseismic methods are crucial for real-time monitoring of the hydraulic fracturing dynamic status. However, unlike the active-source seismic events, the microseismic events usually have very low signal-to-noise ratio (SNR), which makes its processing challenging. To overcome the noise issue of the weak microseismic events, we propose a novel method for microseismic event detection based on the support vector machine (SVM) classification with a Gaussian kernel. For the preprocessing, fix-sized segmentation with a length of 2* wavelength is used to divide the data into segments. 123 features in total, which are used as input data to train the SVM model, have been extracted. These features include 63 1D time/spectral-domain features, and 60 2D texture features. Afterwards, we use a combination of univariate feature selection and random-forest-based recursive feature elimination for feature selection. This feature selection strategy not only finds the best features but also decides the number of features that are needed for the best accuracy. Regarding the essential training process, a C-SVM model, where coefficient C is used to control the tolerance of error item, is considered and a cross-validation (CV) process is implemented for automatic parameter setting. In the end, a group of synthetic and real microseismic data with different levels of complexity show the effectiveness of the proposed method.