Determining the added value of surface distributed acoustic sensors in sparse geophone arrays using transfer learning with a convolutional neural network
Samir Jreij, Whitney Trainor‐Guitton, James Simmons · 2018
Distributed sensors have widely been used in boreholes and their added value is apparent in these environments. Surface acquisitions with distributed sensors have not been quite as successful due to the limited understanding of the types of waves that the instrument records. This paper discusses experiments to identify if there is any added value to using distributed acoustic sensors with sparse geophone arrays in 2-D surface acquisition. The results qualitatively show that 2-D surface DAS arrays are able to recover migrated images similar to sparse, multi-component geophone arrays. Quantitative analysis was also performed using transfer learning in a convolutional neural network. The quantitative analysis shows that adding distributed sensors for this experiment only helped in decreasing false negatives and increasing the true negatives in identifying reflectors. This paper provides the framework for future quantitative analysis in the geophysics field using machine learning. Presentation Date: Thursday, October 18, 2018 Start Time: 8:30:00 AM Location: 212A (Anaheim Convention Center) Presentation Type: Oral