Partial differential equations (PDEs) are ubiquitous in natural science and engineering problems. Traditional discrete methods for solving PDEs usually time-consuming labor-intensive due to the need tedious mesh generation numerical iterations. Recently, deep neural networks have shown new promise cost-effective surrogate modeling because of their universal function approximation abilities. In this paper, we borrow idea from physics-informed (PINNs) propose an improved data-free model,...

We build surrogate models for dynamic 3D subsurface single-phase flow problems with multiple vertical producing wells. The model provides efficient pressure estimation of the entire formation at any timestep given a stochastic permeability field, arbitrary well locations and penetration lengths, matrix as inputs. production rate or bottom hole can then be determined based on Peaceman's formula. original modeling task is transformed into an image-to-image regression problem using...

Breast cancer is considered as the most prevalent cancer. Using ultrasound images a momentous clinical diagnosis method to locate breast tumors. However, accurate segmentation of tumors remains an open problem due artifacts, low contrast, and complicated tumor shapes in images. To address this issue, we proposed boundary-oriented network (BO-Net) for boosting The BO-Net boosts performance from two perspectives. Firstly, module (BOM) was designed capture weak boundaries by learning additional...

Abstract Clay minerals are abundant in shale, characterized by a lamellar structure and dimensions smaller than micron, giving rise to nanometer‐scale pore sizes large specific surface area. They commonly associated with water. However, the spatial distribution of unsaturated water clay is not very well understood, which significantly affects subsequent shale gas flow capacity. Wettability heterogeneity presence hydrocarbons further complicates clay. In this study, we use 3‐D lattice...

Traditional CNNs struggle with SAR and optical image fusion cloud removal due to noise, feature space differences random distribution. This often leads blurred results less texture information. paper proposes a synergistic convolution-transformer method (SCT-CR), which is based on specially designed convolution module that enables the of imagery. The proposed network employs transformer in high-dimensional section better perceive contextual information achieve intelligent extraction global...

In order to determine the matching relationship of polymer molecular weight and reservoir permeability in ASP (alkaline/surfactant/polymer) flooding, a number core flooding experiments with different weights are performed. Two types curves for between pressure difference injection pore volume multiples obtained. One describes characteristics plugging; other well. The cyclotron radius throat used describe permeability. results indicate that when ratio (rh) (rp) is greater than 7, system...

Summary Identification of vuggy intervals and understanding their connectivity are critical for predicting carbonate reservoir performance. Although core samples conventional well logs have been traditionally used to classify facies, this process is labor intensive often suffers from data inadequacies. Recently, convolutional neural network (CNN) algorithms approached human-level performance at multiimage classification identification tasks. In study, CNNs were trained identify facies a in...

Abstract This paper provides a validation of Digital Rocks workflow for computing relative permeability from micro-CT images rock using Lattice Boltzmann simulation. The has the potential to dramatically reduce time and cost measuring by experimental techniques. Additionally, this utilizes smaller volumes than traditional laboratory techniques enabling possibility side-wall cores further reduction in costs avoiding costly coring operations. approach taken is construct computational grids...

Summary Diagenetic effects in carbonate rocks can enhance or occlude depositional pore space. Reliable identification of porosity-enhancing diagenetic features (e.g., vugs and fractures) is essential for petrophysical characterization reservoir properties porosity permeability), construction geological models, reserve estimation, production forecasting. Challenges remain characterizing these from well logs as they are often mixed with variations mineral fluid concentrations. Herein, we...

The fraudulent website image is a vital information carrier for telecom fraud. efficient and precise recognition of images critical to combating dealing with websites. Current research on websites mainly carried out at the level feature extraction similarity study, which have such disadvantages as difficulty in obtaining data, insufficient analysis, single identification types. This study develops model based entropy method leader decision Inception-v3 transfer learning address these...

Lung segmentation on CT images is a crucial step for computer-aided diagnosis system of lung diseases. The existing deep learning based methods are less efficient to segment lungs clinical images, especially that the boundaries not accurate enough due complex pulmonary opacities in practical clinics. In this paper, we propose boundary-guided network (BG-Net) address problem. It contains two auxiliary branches seperately and extract boundaries, an aggregation branch efficiently exploits...

Summary Determination of the optimal well placement strategy in oil or gas fields is crucial for economic reservoir development. The optimization process, however, can be computationally intensive as a result potentially high-dimensional search space and expensive numerical simulation. In this study, machine-learning-based surrogate models are constructed efficient alternatives to simulators accelerate process. A V-Net neural network architecture used with features skip connections, 3D...