Schlumberger Technology Corporation

Lingchen Zhu

Sandip Bose

Smaine Zeroug

Tuanfeng Zhang

Peter Tilke

Marie Emeline Cecile LeFranc

Marie Etchebes

Michael Lis

Remy Sabathier

Platforms and methods for generating 3D stratigraphic models are proved wherein the platform utilizes a Motion and Content decomposed GAN that is trained using actual data or generated data. The generation of the stratigraphic models can also include using Wasserstein training loss for improved model diversity.

Rapid and realistic three-dimensional stratigraphic model generator conditioned on reference well log data

Neural network systems and related machine learning methods for geological modeling are provided that employ an improved generative adversarial network including a generator neural network and a discriminator neural network. The generator neural network is trained to map a combination of a noise vector and a category code vector as input to a simulated image of geological facies. The discriminator neural network is trained to map at least one image of geological facies provided as input to corresponding probability that the at least one image of geological facies provided as input is a training image of geological facies or a simulated image of geological facies produced by the generator neural network.

Generating geological facies models with fidelity to the diversity and statistics of training images using improved generative adversarial networks

A process mimicking forward modeler with deposition and erosion at each specific geological time step. The 3D derived properties are high resolution depositional environments and rock properties that are used to generate multiscale labelled synthetic data. These synthetic data can range from 1D logs such as grain size, gamma ray, density, and velocity, to 3D synthetic seismic, and are used directly as training data for various AIML applications.

Stratigraphic forward modeling platform and methods of use

Methods and systems are provided that characterize and evaluate well integrity of a cased well using unsupervised machine learning of acoustic data. Sonic waveform data for acoustic signals received by the receiver array of a sonic logging tool is collected and processed to determine a high-dimensional representation of the sonic waveform data. The high-dimensional representation is input to an unsupervised machine learning system to determine a low-dimensional representation of the sonic waveform data. A clustering method is applied to the low-dimensional representation to identify a set of clusters therein. At least one well integrity property of the depth interval of the cased well is determined based on the set of clusters. In embodiments, the at least one well integrity property can characterize cement condition in an annulus of the cased well as a function of azimuth and depth and can be used to evaluate cement integrity in a depth interval of the cased well.

Methods for characterizing and evaluating well integrity using unsupervised machine learning of acoustic data

A method, computer program product, and computing system are provided for receiving sonic data associated with an inner casing of a well. Predicted ultrasonic data associated with an outer casing of the well may be generated based upon, at least in part, a nonlinear regression model and the received sonic data associated with the inner casing of the well.

Growing community of inventors

Medford, MA, United States of America

Lingchen Zhu