Application of machine learning to microseismic event detection in distributed acoustic sensing data

Abstract

This study presents the first demonstration of the transferability of a convolutional neural network (CNN) trained to detect microseismic events in one fiber-optic distributed acoustic sensing (DAS) data set to other data sets. DAS is being increasingly used for microseismic monitoring in industrial settings, and the dense spatial and temporal sampling provided by these systems produces large data volumes (approximately 650 GB/day for a 2 km long cable sampling at 2000 Hz with a spatial sampling of 1 m), requiring new processing techniques for near-real-time microseismic analysis. We have trained the CNN known as YOLOv3, an object detection algorithm, to detect microseismic events using synthetically generated waveforms with real noise superimposed. The performance of the CNN network is compared to the number of events detected using filtering and amplitude threshold (short-term average/long-term average) detection techniques. In the data set from which the real noise is taken, the network is able to detect >80% of the events identified by manual inspection and 14% more than detected by standard frequency-wavenumber filtering techniques. The false detection rate is approximately 2% or one event every 20 s. In other data sets, with monitoring geometries and conditions previously unseen by the network, >50% of events identified by manual inspection are detected by the CNN.

Distributed Acoustic Sensing (DAS) for Natural Microseismicity Studies: A Case Study From Antarctica

“…VSP methods are now applied in other environments, such as at glaciers (Booth et al., 2020). Recently, DAS has been applied to passive seismic investigations including: the study of tectonic earthquakes (Ajo‐Franklin et al., 2019; Dou et al., 2017; Jousset et al., 2018; Lindsey et al., 2019; Sladen et al., 2019; Wang et al., 2018); ambient noise studies (Ajo‐Franklin et al., 2019; Dou et al., 2017; Martin et al., 2018; Spica et al., 2020; Zeng et al., 2017); and microseismicity in a variety of settings including hydraulic fracture reservoir stimulation (Baird et al., 2020; Karrenbach et al., 2019; Stork et al., 2020; Verdon et al., 2020), geothermal seismicity (Li & Zhan, 2018), and alpine glacier icequakes (Walter et al., 2020).…”

An introduction to distributed training of deep neural networks for segmentation tasks with large seismic data sets

“…A trade-off can occur between input data dimensions and model size where, as opposed to subsampling data, a smaller, simpler model is used. For example, in the microseismic event detection use case both Stork et al (2020) and Consolvo and Thornton (2020) train…”

“…From a single station viewpoint, i.e., where traces are handled independently of one another, recurrent NNs have been shown to be particularly powerful in offering an alternative to the commonly used short-time average, long-time average detection procedure (e.g., Zheng et al, 2018;Birnie and Hansteen, 2020). While from an array point of view, both Stork et al (2020) and Consolvo and Thornton (2020) have illustrated how CNNs can be used for detecting an events arrival within a certain time-space bounding box.…”

Seismic Applications of Downhole DAS

“…Figure 10 shows that event locations are more clustered than those obtained with surface geophones. Machine learning approaches have also been attempted and have shown reasonable performance [ 109 , 110 ]. However, they are highly dependent on the size and quality of training datasets, and substantial effort is needed in generating those.…”