Schlumberger Technology Corporation

Services Petroliers Schlumberger

Geoquest Systems B.v.

Hiren Maniar

Aria Abubakar

Xiaoli Chen

Haibin Di

Nicolae Moldoveanu

Atul Laxman Katole

Mandar Shrikant Kulkarni

Purnaprajna Raghavendra Mangsuli

Cen Li

Aaron Scollard

Tao Zhao

Sunil Manikani

Anisha Kaul

Vanessa Simoes

Ekansh Verma

Preeti Gupta

Denzil Francis Crasta

Gayatri Novenita

Karan Pathak

A method for digitizing image-based data includes receiving an image file including one or more target objects, generating an intermediate image by removing noise from the image file using a denoising machine learning model, identifying the one or more target objects included in the intermediate image using an object segmentation machine learning model, discretizing the one or more target objects that were identified using the trained object segmentation machine learning model, and storing the one or more target objects that were discretized in a data file, visualizing the one or more target objects, or both.

Raster image digitization using machine learning techniques

A method includes determining a top of salt (TOS) surface in a seismic volume based on a crossline direction of the seismic volume and an inline direction of the seismic volume. The method also includes determining a binary mask based upon the TOS surface. The method also includes sampling seismic data in the seismic volume to obtain a training seismic slice. The method also includes sampling the binary mask to obtain a mask slice. The method also includes selecting a first coordinate in the training seismic slice to produce a first tile. The method also includes selecting a second coordinate in the mask slice to produce a second tile. The method also includes generating or updating a model of the seismic volume based upon the first tile and the second tile.

Cascaded machine-learning workflow for salt seismic interpretation

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 seismic training data comprising a plurality of images each including a plurality of traces, predicting a location of a feature in at least some of the plurality of traces based on a location of an amplitude peak therein, applying labels to the locations, classifying pixels of the plurality of images as representing the feature or not representing the feature, using a semantic segmentation model, adjusting the labels based on the classification of the pixels, training, using the adjusted labels and the seismic training data, a machine-learning model to identify the feature, and identifying the feature in a different seismic data set using the trained machine-learning model.

Feature detection in seismic data

A machine-implemented method, at least one non-transitory computer-readable medium storing instructions, and a computing system are provided for attenuating noise. A computing system receives a seismic image and generates a first image using a first neural network configured to identify low-frequency ground roll in a seismic image, and a second image using a second neural network configured to identify reflections in the seismic image. A combined image is generated by combining the first image and the second image. The first neural network and the second neural network are adjusted to reduce a difference between the combined image and the seismic image using frequency constraint to guide separation of the seismic image into the first image and the second image.

Ground roll attenuation using unsupervised deep learning

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.

Automatic well log correction

A method includes receiving well log data comprising a plurality of well logs, identifying one or more sections of one or more well logs of the plurality of well logs that have substantially complete data, training a reconstruction neural network to reconstruct incomplete well logs based on the one or more sections of the one or more well logs that have substantially complete data, and reconstructing one or more incomplete well logs of the plurality of well logs using the reconstruction neural network.

Unsupervised well log reconstruction and outlier detection

A method for seismic processing includes extracting, using a first machine learning model, one or more seismic features from seismic data representing a subsurface domain, receiving one or more well logs representing one or more subsurface properties in the subsurface domain, and predicting, using a second machine learning model, the one or more subsurface properties in the subsurface domain at a location that does not correspond to an existing well based on the seismic data, the one or more well logs, and the one or more seismic features that were extracted from the seismic data.

Subsurface property estimation in a seismic survey area with sparse well logs

A training log is selected from a plurality of well logs. A log window of a plurality of log windows is selected from the training log. A positive window is generated from the log window. A negative window is selected from the training log. A siamese neural network (SNN) is trained that includes a first self attention neural network (ANN) and a duplicate self attention neural network with the log window, the positive window, and the negative window, to recognize a similarity between the log window and the positive window and to differentiate against the negative window.

Growing community of inventors

Houston, TX, United States of America

Hiren Maniar