Fiber-optic distributed seismic sensing data generator and its application for training classification nets.

Area of Science:

• Geophysics
• Machine Learning
• Data Science

Background:

• Distributed acoustic sensing (DAS) offers high resolution and sensitivity for seismic measurements.
• Increasing demand for long-distance distributed seismic sensing (DSeiS) has spurred interest in advanced data processing.
• Current machine learning methods for DSeiS require extensive in situ calibration.

Purpose of the Study:

• To develop a geophysics-driven approach for generating synthetic DSeiS data.
• To utilize synthetic data for training DSeiS classification algorithms.
• To enhance the accuracy of machine learning models in seismic data analysis.

Main Methods:

• A novel geophysics-driven methodology was employed to generate synthetic DSeiS data.
• Synthetic data were used to train an artificial neural network classifier.
• The performance of the trained classifier was evaluated using experimental DSeiS records.

Main Results:

• The geophysics-driven approach successfully generated realistic synthetic DSeiS data.
• Training machine learning classifiers with synthetic data significantly improved classification accuracy.
• The validated approach demonstrates the effectiveness of incorporating geophysical models into synthetic data generation.

Conclusions:

• A geophysics-driven method for synthetic DSeiS data generation is effective for training machine learning models.
• This approach reduces the need for laborious in situ calibration experiments.
• The study highlights a promising direction for advancing automated seismic data analysis.