Monitoring Fracture Saturation With Internal Seismic Sources and Twin Neural Networks

Seismic coda-wave analysis is a well-developed method for detecting subtle physical changes in complex media by measuring arrival times the late-arriving energy from multiply scattered or reflected waves. However, challenge arises when waves are not sufficiently separated time direct arrivals to provide clear coda wave train. Additional complications monitoring fracture systems arise signals originate unsynchronized internal sources, such as natural induced seismicity, acoustic emission, transportable intra-fracture sources (chattering dust), that generate uncontrolled vary time, amplitude and frequency content. Here, we use twin neural network (TNN also known Siamese network) dimensionality reduction analyze chattering dust classify fluid saturation state of synthetic system. The TNN with shared weights generates low-dimensional representation data input minimizing contrastive loss, serving multiclass classifier accurately classifies whether multiple fractures system fully saturated partially saturated, change has occurred different These results show information buried unresolved codas can be extracted using machine learning monitor evolution caused chemical processes even fields overlap.