Saudi Arabian Oil Company

Exxonmobil Upstream Research Company

Weichang Li

Sebastian Csutak

Max Deffenbaugh

Damian Pablo San Roman Alerigi

Chicheng Xu

Gregory D Ham

Yang Zhao

Angelo Sampaolo

Yong Ma

Tao Lin

Olanrewaju A Abudu

Sameeh Issa Batarseh

Muhammad Arsalan

Shouxiang Mark Ma

David J Jacobi

Frode Hveding

Daniele Colombo

Murtadha J AlTammar

Mohamed Larbi Zeghlache

Shannon Lee Eichmann

Khalid Mohammed Alruwaili

Dominique Gillard

Ganglin Chen

Roy Matthew Burnstad

Tiffany Dawn McAlpin

Ernesto Sandoval-Curiel

Xiaoxia Xu

Osman Hamid

Yaser Abdullah Alzayer

Ke Chen

Yuchen Jin

Ali Almadan

Lei Fu

Abdulla Al Sarraf

A method for a physics-driven deep learning-based inversion coupled to fluid flow simulators may include obtaining measured data for a subsurface region, obtaining prior subsurface data for the subsurface region, and obtaining a physics-driven standard regularized joint inversion for at least two model parameters. The method may further include obtaining a case-based deep learning inversion characterized by a contracting path and an expansive path. The method may further include forming the physics-driven deep learning inversion with the physics-driven standard regularized joint inversion, the case-based deep learning inversion, and a coupling operator based on a penalty function. The method may further include forming a feedback loop between the physics-driven standard regularized joint inversion and the case-based deep learning inversion for re-training the case-based deep learning inversion. The method may further include generating an inversion solution for reservoir monitoring.

Physics-driven deep learning inversion coupled to fluid flow simulators

A computer-implemented method for automated wellhead monitoring using imaging and computer vision is described. The method includes establishing a baseline for a wellhead using image dimension and perspective calibration. The method also includes constructing at least one computer vision model using images or video from a wellhead database and inputting unseen wellhead images to the trained at least one computer vision model. Additionally, the method includes extracting a wellhead shape and geometric information of the wellhead in the unseen wellhead images and estimating wellhead displacement and growth based on the extracted images.

Automated wellhead monitoring using imaging and computer vision

A computer-implemented method for generating training samples via 3D modeling for machine learning-based greenhouse gas emission detection is described. The method includes generating at least one 3D model corresponding to a real-world facility and at least one 3D model corresponding to gas plume emissions and building a virtual environment comprising a 3D field based on the at least one 3D model corresponding to the real-world facility and the at least one 3D model corresponding to gas plume emissions. The method also includes capturing images or video of the 3D field via simulation of the virtual environment and assigning labels to objects in the images or videos.

Generating training samples via 3D modeling for greenhouse gas emission detection

Some implementations of the present disclosure provide a method that includes: irradiating a target surface with a process beam during a drilling process; in response to irradiating with the process beam, receiving a signal beam that contains light scattered from the target surface as well as light radiating from the target surface; splitting the signal beam into a first portion on a polarization arm and a second portion on a non-polarization arm; performing, on the polarization arm, a first plurality of polarization-dependent intensity and spectrum measurements of the first portion; performing, on the non-polarization arm, a second plurality of intensity and spectrum measurements of the second portion; and based on applying one or more machine learning techniques to at least portions of (i) the first plurality of polarization-dependent intensity and spectrum measurements, and (ii) the second plurality of intensity and spectrum measurements, determining a classification of the target surface.

Real-time multimodal radiometry for subsurface characterization during high-power laser operations

A method may include obtaining core image data regarding a geological region of interest. The method may further include obtaining well log data regarding the geological region of interest from one or more wells. The method may further include determining a sliding window that corresponds to a predetermined window size. The method may further include determining various quantitative image attributes using the core image data, the well log data, and the sliding window. The quantitative image attributes may be determined in a continuous manner by moving the sliding window along the core image data. The method may further include generating predicted rock data for the geological region of interest using the quantitative image attributes, a machine-learning algorithm, and a machine-learning model.

Method and system for image-based reservoir property estimation using machine learning

A computer-implemented method includes: accessing a first database holding information encoding a set of labels that specify a condition of at least one of: a surface pipe, or an underground enclosure that runs at a plurality of depth locations; accessing a second database holding a plurality of inspection logs that record measurement data of the surface pipe or underground enclosure; based on, at least in part, the labeling information and the plurality of inspection logs, training a deep learning model configured to classify, into the set of labels, the condition of the surface pipe or underground enclosure when presented with the inspection logs; applying the deep learning model to one or more newly received inspection logs containing measurement data of a new surface pipe or a new underground enclosure; and subsequently classifying, into the set of labels, the condition of the new surface pipe or the new underground enclosure.

System and method using machine learning for automated casing and/or tubing inspection log interpretation

A system and methods for quantitative hydraulic fracturing surveillance from fiber optic sensing using machine learning is described herein. An exemplary method provides capturing distributed acoustic sensing (DAS) data, distributed temperature sensing (DTS) data, and microseismic data over monitored stages. Operation states and variables at a respective stage are predicted, based on, at least in part, the DAS data, DTS data, or microseismic data. At least one event associated with the predicted operation states and variables is localized at the respective stage.

Quantitative hydraulic fracturing surveillance from fiber optic sensing using machine learning

A method to perform a field operation with well log repeatability verification is disclosed. The method includes generating, by repeatedly performing well logging of a wellbore penetrating a subterranean formation in a field, a set of well log data files each comprising a plurality of data channels, each data channel comprising a series of measurement data records representing a downhole property along a depth in the wellbore, analyzing, by a computer processor, a main log and a repeat log of the set of well log data files to determine a repeatability measure of the set of well log data files, presenting, using a graphical user interface, the repeatability measure to a user, and facilitating, based on a user input in response to presenting the repeatability measure, the field operation.

System and method for well log repeatability verification

A computer-implemented method for constrained multi-phase virtual flow metering and forecasting is described. The method includes predicting instantaneous flow rates and forecasting future target flow rates and well dynamics. The method includes constructing a virtual sensing model trained using forecasted target flow rates and well dynamics. The method includes building a constrained forecasting model by combining unconstrained flow forecasting models, well dynamics models, and virtual sensing models, wherein the constrained forecasting model forecasts multi-phase flow rates.

Model-constrained multi-phase virtual flow metering and forecasting with machine learning

A method for mapping fiber optic distributed acoustic sensing (DAS) measurements to particle motion involves obtaining, from a fiber optic DAS system in a wellbore, a first set of DAS data associated with a first seismic wave; obtaining, from a discrete seismic receiver in the wellbore, measured particle motion data associated with the first seismic wave; generating training data from the first set of DAS data and the measured particle motion data; training a machine learning model using the training data; obtaining a second set of DAS data associated with a second seismic wave; and determining a predicted particle motion in response to the second seismic wave using the machine learning model applied to the second set of DAS data.

Method and system for mapping fiber optic distributed acoustic sensing measurements to particle motion

A computer-implemented method may include obtaining well logs data pertaining to a well of interest. The method may further include training a physics-constrained machine learning (PCML) model using the obtained well logs data as inputs. The method may further include outputting one or more sonic logs and mechanical properties of interest determined by using the trained PCML model and the obtained well logs data for the well of interest. The method may further include updating the determined sonic logs and mechanical properties of interest based on a breakout model and field breakout data for the well of interest. The method may further include outputting the final sonic logs for the well of interest. The method may further include determining one or more mechanical properties for well planning based on the final sonic logs for the well of interest.

Method and system for correcting and predicting sonic well logs using physics-constrained machine learning

Systems, apparatuses, and computer-implemented methods are provided for the sensing and prediction of properties of source rock. Disclosed here is a method of predicting the maturity of a source rock that includes obtaining a plurality of data of a sample source rock from a plurality of data acquisition devices placed in vicinity of the sample source rock and analyzing the received data using a predictive correlation to determine maturity of the sample source rock. The predictive correlation is generated by applying a machine learning model to correlate the plurality of data acquired from a plurality of representative source rocks with a plurality of properties of the plurality of representative source rocks.

Collaborative sensing and prediction of source rock properties

Seismic survey data is received, indexed into index sets, and each index set partitioned into data blocks. For each particular data block of a particular index set, the particular data block is sliced into frequency slices. For each particular frequency slice of the particular data block, the particular frequency slice is processed to remove random and erratic noise by: forming a Hankel matrix from the particular frequency slice: determining an optimal rank for the Hankel matrix, determining a clean signal and erratic noise from the ranked Hankel matrix, and returning the clean signal and erratic noise for the particular frequency slice. A clean signal is assembled from the index sets.

Automated system and methods for adaptive robust denoising of large-scale seismic data sets

Seismic shot gather data is received from a computer data store for processing. The received seismic shot gather data is separated into downgoing and upgoing wavefields, a time-frequency-wavenumber (t-f-k) three-dimensional (3D) data cube comprising multiple time-frequency (t-f) slices is formed. The downgoing wavefields are wavelet transformed from a time (t) domain to a t-f domain and the upgoing wavefields are wavelet transformed from the t domain to the t-f domain. A wavelet cross-correlation is performed between the downgoing wavefields in the t-f domain and the upgoing wavefields in a t-f-k domain to generate wavelet cross-correlated data. Soft-threshold filtering if performed for each t-f slice of the t-f-k 3D data cube. An inverse wavelet transform is performed to bring wavelet cross-correlated data from the t-f-k domain to a time-receiver (t-x) domain. All seismic shots of the received seismic shot gather data are looped over and the wavelet cross-correlated data is stacked as a virtual source gather.

Use of wavelet cross-correlation for virtual source denoising

Received shot gathers are sorted to a common receiver gather. A target three-dimensional (3D) amplitude spectrum of seismic wavefield direct arrivals is computed from synthetic data. A 3D amplitude spectrum of seismic wavefield direct arrivals in field data is computed for each receiver. A matched filter is calculated from the 3D amplitude spectrum of field data to target response and applied to downgoing seismic wavefields separated from the common receiver gather to generate filtered downgoing seismic wavefields. Time-dependent smoothing of the filtered downgoing seismic wavefields is performed to generate smoothed downgoing seismic wavefields. A cross-correlation is calculated between upgoing seismic wavefields separated from the common receiver gather and the smoothed downgoing seismic wavefields.

Growing community of inventors

Katy, TX, United States of America

Weichang Li