Schlumberger Technology Corporation

Tao Zhao

Aria Abubakar

Hiren Maniar

Xiaoli Chen

Anisha Kaul

Chunpeng Zhao

Vanessa Simoes

A method includes receiving seismic data and an initial velocity model, generating a first seismic image based at least in part on the seismic data and the initial velocity model, training a machine learning model to predict salt masks based at least in part on seismic images, merging the initial velocity model and the first salt mask to generate a first modified velocity model, generating an updated velocity model based at least in part on the first modified velocity model, generating a second seismic image based at least in part on the updated velocity model, predicting a second salt mask based at least in part on the second seismic image and the updated velocity model, using the trained machine learning model, and merging the updated velocity model and the second salt mask to generate a second modified velocity model.

Automatic salt geometry detection in a subsurface volume

A method includes receiving geophysical data representative of a geophysical structure; providing the geophysical data as one or more input data to a neural network; training the neural network to reconstruct the geophysical structure that was received and provide one or more uncertainty metrics for one or more features of the geophysical structure that is reconstructed; reconstructing, using the neural network that has been trained, the geophysical structure; and determining, using the neural network that has been trained, the one or more uncertainty metrics by implementing a second drop out condition on the one or more nodes of the one or more hidden layers of the neural network. The training is performed at least partially by implementing a first drop out condition on one or more nodes of one or more hidden layers of the neural network to randomly set an output of the one or more nodes to zero.

Uncertainty analysis for neural networks

A method includes receiving first training well logs, generating second training well logs by injecting one or more different types of systematic errors, random errors, or both into at least a portion of the first training well logs, training a machine learning model to correct well logs by configuring the machine learning model to reduce a dissimilarity between at least a portion of the first and second training well logs, receiving one or more implementation well logs, and generating one or more corrected well logs by correcting at least a portion of the one or more implementation well logs using the machine learning model that was trained.

Growing community of inventors

Houston, TX, United States of America

Tao Zhao