Enhancing the accuracy of physics-informed neural network surrogates in flash calculations using sparse grid guidance

Article Energy & Fuels

PTFlash : A vectorized and parallel deep learning framework for two-phase flash calculation

Jingang Qu et al.

Summary: This paper introduces a fast and parallel framework, PTFlash, which uses vectorized algorithms and neural networks to accelerate phase equilibrium calculations, greatly reducing computation time while maintaining high precision.

FUEL (2023)

添加到收藏夹

Article Green & Sustainable Science & Technology

An integrated model with stable numerical methods for fractured underground gas storage

Wendi Xue et al.

Summary: Underground gas storage formed from fractured depleted oil/gas reservoirs has potential for storing large quantities of natural gas. To address the absence of a numerical model, we propose an integrated model that considers the wellbore, reservoir, and gas properties. This model incorporates a gas flow direction factor, transient heat transfer models, and accurate gas property models. We analyze stability, efficiency, and global mass conservation of the reservoir model and propose a new source term equation and a pressure correction method for stable and efficient solution. Simulation results show improved efficiency and strict global mass conservation.

JOURNAL OF CLEANER PRODUCTION (2023)

添加到收藏夹

Article Mechanics

Removing the performance bottleneck of pressure-temperature flash calculations during both the online and offline stages by using physics-informed neural networks

Yuanqing Wu et al.

Summary: Pressure-temperature (PT) flash calculations are a performance bottleneck of compositional-flow simulations. The computing burden of PT flash calculations is shifted from the online stage to the offline stage with the sparse grid surrogate, achieving great acceleration. Physics-informed neural networks remove the computing burden of PT flash calculations in the offline stage by not carrying out the heavy-burden routines. Numerical experiments validate the correctness and applicability of this approach. To the best of our knowledge, this is the first work to remove the performance bottleneck of PT flash calculations during both the online and offline stages of compositional-flow simulations.

PHYSICS OF FLUIDS (2023)

添加到收藏夹

Article Computer Science, Interdisciplinary Applications

Fully implicit two-phase VT-flash compositional flow simulation enhanced by multilayer nonlinear elimination

Yiteng Li et al.

Summary: The newly developed two-phase VT-flash compositional flow algorithm with multilayer nonlinear elimination significantly improves robustness and efficiency by implicitly removing locally large nonlinearities and restoring large step length for Newton iterations. The algorithm utilizes adaptive time stepping control and a modified shadow region method to enhance computational efficiency, demonstrating increased number of Newton iterations but significantly enlarged timestep size under the multilayer nonlinear elimination method.

JOURNAL OF COMPUTATIONAL PHYSICS (2022)

添加到收藏夹

Article Multidisciplinary Sciences

A field-based general framework to simulate fluids in parallel and the framework's application to a matrix acidization simulation

Yuanqing Wu et al.

Summary: Researchers have found that common operations in fluid simulations can be abstracted into a general framework based on field operations, allowing for parallelization. This framework simplifies and unifies operations in fluid simulations, improving efficiency.

PLOS ONE (2022)

添加到收藏夹

Article Engineering, Multidisciplinary

Physics informed neural networks for continuum micromechanics

Alexander Henkes et al.

Summary: Physics informed neural networks are a method used in applied mathematics and engineering to solve partial differential equations. However, due to their global approximation approach, they face challenges in displaying localized effects and strong nonlinear solution fields. To overcome these issues, researchers have studied adaptive training strategies and domain decomposition.

COMPUTER METHODS IN APPLIED MECHANICS AND ENGINEERING (2022)

添加到收藏夹

Review Engineering, Petroleum

Compositional Reservoir Simulation: A Review

Larry C. Young

Summary: This article introduces the uses of compositional reservoir models, with a focus on isothermal compositional models based on an equation of state. The article describes the differential-algebraic equation system and the physical nature of the parameters, comparing model formulations and computational features. The article highlights the importance of correct physics and chemistry understanding in numerical calculations and identifies areas for further research.

SPE JOURNAL (2022)

添加到收藏夹

Article Thermodynamics

A generalized machine learning-assisted phase-equilibrium calculation model for shale reservoirs

Fangxuan Chen et al.

Summary: This paper proposes a novel ML-assisted framework for phase equilibrium calculations in shale reservoirs. By utilizing machine learning techniques, the computation time needed for nano-scale phase equilibrium calculations can be significantly reduced while maintaining high accuracy. The framework can be compiled into a reservoir simulator to accelerate flash calculation.

FLUID PHASE EQUILIBRIA (2022)

添加到收藏夹

Article Engineering, Multidisciplinary

A mixed formulation for physics-informed neural networks as a potential solver for engineering problems in heterogeneous domains: Comparison with finite element method

Shahed Rezaei et al.

Summary: Physics Informed Neural Networks (PINNs) can find solutions to boundary value problems by minimizing a loss function that incorporates governing equations, initial and boundary conditions. This study proposes an improved method that uses the spatial gradient of the primary variable as an output and applies the strong form of the equation as a physical constraint. By comparing with finite element methods, it is shown that this approach has advantages, and the potential of combining PINN with physical FE simulations for designing composite materials is discussed.

COMPUTER METHODS IN APPLIED MECHANICS AND ENGINEERING (2022)

添加到收藏夹

Article Computer Science, Interdisciplinary Applications

Efficient training of physics-informed neural networks via importance sampling

Mohammad Amin Nabian et al.

Summary: PINNs are deep neural networks trained to compute the response of systems governed by PDEs using automatic differentiation. Although successful, they still need improvements in computational efficiency, which is why this paper studies the performance of an importance sampling approach for efficient training of PINNs.

COMPUTER-AIDED CIVIL AND INFRASTRUCTURE ENGINEERING (2021)

添加到收藏夹

Article Computer Science, Interdisciplinary Applications

An adaptive high-order piecewise polynomial based sparse grid collocation method with applications

Zhanjing Tao et al.

Summary: This paper introduces an adaptive sparse grid collocation method onto arbitrary order piecewise polynomial space, including Lagrange and Hermite interpolation methods. Error estimates are provided, and numerical results are used to compare different collocation schemes in function interpolation, integration, and uncertainty quantification benchmark problems.

JOURNAL OF COMPUTATIONAL PHYSICS (2021)

添加到收藏夹

Article Thermodynamics

Physics-Informed Neural Networks for Heat Transfer Problems

Shengze Cai et al.

Summary: Physics-informed neural networks (PINNs) have gained popularity in engineering fields for their effectiveness in solving realistic problems with noisy data and partially missing physics. Through automatic differentiation to evaluate differential operators and defining a multitask learning problem, PINNs have been applied to various heat transfer problems, bridging the gap between computational and experimental heat transfer.

JOURNAL OF HEAT TRANSFER-TRANSACTIONS OF THE ASME (2021)

添加到收藏夹

Article Engineering, Multidisciplinary