Machine learning acceleration for nonlinear solvers applied to multiphase porous media flow

Silva, Vinicius L. S.; Salinas, P.; Jackson, Matthew D.; Pain, Christopher C.

doi:10.1016/j.cma.2021.113989

Cited by 12 publications

(3 citation statements)

References 23 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…2021) or continuous two-fluid approaches using level-set (Ambekar, Mondal & Buwa 2021; Jettestuen, Friis & Helland 2021) or Cahn–Hilliard models (Yang & Kim 2021) with improved algorithms making use of machine learning (see, for instance, Silva et al. (2021)).…”

Section: Introductionmentioning

confidence: 99%

A macroscopic model for immiscible two-phase flow in porous media

Lasseux

Valdés‐Parada

2022

J. Fluid Mech.

View full text Add to dashboard Cite

This work provides the derivation of a closed macroscopic model for immiscible two-phase, incompressible, Newtonian and isothermal creeping steady flow in a rigid and homogeneous porous medium without considering three-phase contact. The mass and momentum upscaled equations are obtained from the pore-scale Stokes equations, adopting a two-domain approach where the two fluid phases are separated by an interface. The average mass equations result from using the classical volume averaging method. A Green's formula and the adjoint Green's function velocity pair problems are used to obtain the pore-scale velocity solutions that are averaged to obtain the upscaled momentum balance equations. The macroscopic model is based on the assumptions of scale separation and the existence of a periodic representative elementary volume allowing a local description as usually postulated for upscaling. The macroscopic momentum equation in each phase includes the generalized Darcy-like dominant and viscous coupling terms and, importantly, an additional compensation term that accounts for surface tension effects to momentum transfer that is, otherwise, incompletely captured by the Darcy terms. This interfacial term, as well as the dominant and viscous coupling permeability tensors, can be predicted from the solutions of two associated closure problems that coincide with those reported in the literature. The relevance of the compensation term and the upscaled model validity are assessed by comparisons with direct numerical simulations in a model two-dimensional periodic structure. Upscaled model predictions are found to be in excellent agreement with direct numerical simulations.

show abstract

Section: Introductionmentioning

confidence: 99%

A macroscopic model for immiscible two-phase flow in porous media

Lasseux

Valdés‐Parada

2022

J. Fluid Mech.

View full text Add to dashboard Cite

show abstract

“…The dependence on labeled data constrains the practical application of deep learning techniques in petroleum engineering contexts, necessitating a concerted effort to resolve this challenge. Addressing or accelerating the complex task without relying on labeled data assumes paramount significance (Silva et al, 2021), holding the potential to revolutionize solution methodologies within the field (Kochkov et al, 2021). While some endeavors have been undertaken to leverage deep learning for solving PDEs in the absence of labeled data, this remains a formidable undertaking, especially in the case of non-stationary and highly nonlinear PDE systems.…”

Section: Physics Driven Artificial Neural Network Pde Solvermentioning

confidence: 99%

Artificial intelligence methods for oil and gas reservoir development: Current progresses and perspectives

Xue,

Li,

Dou

2023

Adv. Geo-Energy Res.

View full text Add to dashboard Cite

Artificial neural networks have been widely applied in reservoir engineering. As a powerful tool, it changes the way to find solutions in reservoir simulation profoundly. Deep learning networks exhibit robust learning capabilities, enabling them not only to detect patterns in data, but also uncover underlying physical principles, incorporate prior knowledge of physics, and solve complex partial differential equations. This work presents the latest research advancements in the field of petroleum reservoir engineering, covering three key research directions based on artificial neural networks: data-driven methods, physics driven artificial neural network partial differential equation solver, and data and physics jointly driven methods. In addition, a wide range of neural network architectures are reviewed, including fully connected neural networks, convolutional neural networks, recurrent neural networks, and so on. The basic principles of these methods and their limitations in practical applications are also outlined. The future trends of artificial intelligence methods for oil and gas reservoir development are further discussed. The large language models are the most advanced neural networks so far, it is expected to be applied in reservoir simulation to predict the development performance.

show abstract

“…High-fidelity DNS of flows and devices is still computationally prohibitive and thereby impractical for most engineering applications. Therefore, attempts have been made to accelerate simulations aiming to solve this long-standing daunting problem. , One may consider what algorithms can be used to accelerate simulations, and how these algorithms can be performed to accelerate simulations with adequate accuracy while not compromising generalization ability. Fortunately, ML can be applied to accelerate real-time simulations, and such a kind of major effort may be categorized into the following types (except for hardware acceleration):…”

Section: Current Status and Challengesmentioning

confidence: 99%

Review of Machine Learning for Hydrodynamics, Transport, and Reactions in Multiphase Flows and Reactors

Zhu

Chen

Ouyang

et al. 2022

Ind. Eng. Chem. Res.

116

View full text Add to dashboard Cite

Artificial intelligence (AI), machine learning (ML), and data science are leading to a promising transformative paradigm. ML, especially deep learning and physics-informed ML, is a valuable toolkit that complements incomplete domain-specific knowledge in conventional experimental and computational methods. ML can provide flexible techniques to facilitate the conceptual development of new robust predictive models for multiphase flows and reactors by finding hidden pattern/information/mechanism in a data set. Due to such emergence, we thereby comprehensively survey, explore, analyze, and discuss key advancements of recent ML applications to hydrodynamics, heat and mass transfer, and reactions in single-phase and multiphase flow systems from different aspects: (1) development of multiphase closure models of drag force, turbulence stresses and heat/mass transfer to improve the accuracy and efficiency of typical CFD simulations; (2) image reconstruction, regime identification, key parameter predictions, and optimization of multiphase flow and transport fields; (3) reaction kinetics modeling (e.g., predictions of reaction networks, kinetic parameters, and species production) and reaction condition optimization. These sections also discuss and analyze the key advantages and weakness of ML for solving the problems in the domain of multiphase flows and reactors. Finally, we summarize the under-solving challenges and opportunities in order to identify future directions that would be useful for the research community. Future development and study of multiphase flows and reactors are envisaged to be accelerated by ML and data science.

show abstract

Machine learning acceleration for nonlinear solvers applied to multiphase porous media flow

Cited by 12 publications

References 23 publications

A macroscopic model for immiscible two-phase flow in porous media

A macroscopic model for immiscible two-phase flow in porous media

Artificial intelligence methods for oil and gas reservoir development: Current progresses and perspectives

Review of Machine Learning for Hydrodynamics, Transport, and Reactions in Multiphase Flows and Reactors

Contact Info

Product

Resources

About