Filters








14 Hits in 6.2 sec

Physics-informed neural network simulation of multiphase poroelasticity using stress-split sequential training [article]

Ehsan Haghighat and Danial Amini and Ruben Juanes
2021 arXiv   pre-print
Additionally, we propose a sequential training approach based on the stress-split algorithms of poromechanics.  ...  Physics-informed neural networks (PINNs) have received significant attention as a unified framework for forward, inverse, and surrogate modeling of problems governed by partial differential equations (  ...  Physics-Informed Neural Networks We approximate the solution variables of the poromechanics problem using deep neural networks.  ... 
arXiv:2110.03049v1 fatcat:zeon4b7wtbau7iabfjpo6ghz5a

Physics-informed neural network solution of thermo-hydro-mechanical (THM) processes in porous media [article]

Danial Amini, Ehsan Haghighat, Ruben Juanes
2022 arXiv   pre-print
Physics-Informed Neural Networks (PINNs) have received increased interest for forward, inverse, and surrogate modeling of problems described by partial differential equations (PDE).  ...  To address these fundamental issues, we: (1)~rewrite the THM governing equations in dimensionless form that is best suited for deep-learning algorithms; (2)~propose a sequential training strategy that  ...  Here, we follow our earlier work on the use of sequential-stress-split.  ... 
arXiv:2203.01514v1 fatcat:ji2qqfkwsjhwhajy5hzoktxdei

Inverse Modeling of Viscoelasticity Materials using Physics Constrained Learning [article]

Kailai Xu, Alexandre M. Tartakovsky, Jeff Burghardt, Eric Darve
2020 arXiv   pre-print
However, inputs and outputs of the neural networks are not directly observable, and therefore common training techniques with input-output pairs for the neural networks are inapplicable.  ...  the context of multi-physics interactions.  ...  This form is expensive to train due to the sequential dependence ofσ n but retains history as contextual information.  ... 
arXiv:2005.04384v1 fatcat:hzjxkhatvrcbfno2rsidhiypy4

Modeling Multiphase Flow Through and Around Multiscale Deformable Porous Materials [article]

Francisco J. Carrillo
2021 arXiv   pre-print
Volume-of-Fluid equations in solid-free-regions and Biot's Poroelasticity Theory in porous regions.  ...  Unlike existing multiscale multiphase solvers, it can match analytical predictions of capillary, relative permeability, and gravitational effects at both the pore and Darcy scales.  ...  This neural network was able to achieve 0.96 and 0.94 labeling accuracy and precision, respectively.  ... 
arXiv:2109.09880v1 fatcat:h4rgdptzwza6nex7w7lfkmiglm

Inverse modeling of nonisothermal multiphase poromechanics using physics-informed neural networks [article]

Danial Amini, Ehsan Haghighat, Ruben Juanes
2022 arXiv   pre-print
We propose a solution strategy for parameter identification in multiphase thermo-hydro-mechanical (THM) processes in porous media using physics-informed neural networks (PINNs).  ...  We report the excellent performance of the proposed sequential PINN-THM inverse solver, thus paving the way for the application of PINNs to inverse modeling of complex nonlinear multiphysics problems.  ...  neural network architectures, and to simulate the outcomes of physical laws.  ... 
arXiv:2209.03276v1 fatcat:on5axoyksbhqpg3baqvcd4dfdq

A multiscale LBM–TPM–PFM approach for modeling of multiphase fluid flow in fractured porous media

Mohamad Chaaban, Yousef Heider, Bernd Markert
2022 International journal for numerical and analytical methods in geomechanics (2022). doi:10.1002/nag.3423  
The outcomes of the numerical model proved the reliability of the multiscale model to simulate multiphasic fluid flow through fractured porous media.  ...  In addition, an anisotropic, phase-field-dependent intrinsic permeability tensor for the fractured porous domains is formulated, which relies on the single-and multiphasic LBM flow simulations.  ...  For further studies, we intend to incorporate machine learning (ML) through deep neural networks (DNN) to create alternative ML-based constitutive models that are trained with extracted data of the microscopic  ... 
doi:10.18154/rwth-2022-07200 fatcat:dzhm6hjdkfbv5n6plh3zuilany

From Fluid Flow to Coupled Processes in Fractured Rock: Recent Advances and New Frontiers

H. S. Viswanathan, J. Ajo‐Franklin, J. T. Birkholzer, J. W. Carey, Y. Guglielmi, J. D. Hyman, S. Karra, L. J. Pyrak‐Nolte, H. Rajaram, G. Srinivasan, D. M. Tartakovsky
2022 Reviews of Geophysics  
Some of the greatest challenges currently facing humanity have roots in the Earth and Energy Sciences.  ...  Approaches such as deep neural networks can be used to train on these input/output quantities.  ...  Immediate research and development investments areas are in using deep neural networks to build towards T-H-M-C emulators and using unsupervised methods to understand the complex coupling of various physical  ... 
doi:10.1029/2021rg000744 fatcat:tjw2o4obqzc5dcg5kcrql57yta

A review of techniques, advances and outstanding issues in numerical modelling for rock mechanics and rock engineering

L. Jing
2003 International Journal of Rock Mechanics And Mining Sciences  
The physical processes and the equations characterizing the coupled behaviour are included in Section 4, with an illustrative example and discussion on the likely future development of coupled models.  ...  The review begins by explaining the special nature of rock masses and the consequential difficulties when attempting to model their inherent characteristics of discontinuousness, anisotropy, inhomogeneity  ...  Hudson who contributed substantially to this review, especially the first two sections and the section about the neural networks, and insisted on removing his name as the co-author.  ... 
doi:10.1016/s1365-1609(03)00013-3 fatcat:fjbcyvbyobhflawsaajzav27ym

Numerical Stabilization of the Melt Front for Laser Beam Cutting [chapter]

Torsten Adolph, Willi Schönauer, Markus Niessen, Wolfgang Schulz
2010 Numerical Mathematics and Advanced Applications 2009  
A saturation function is considered accounting for converging stress-strain curves in cyclic tension tests at fixed load levels.  ...  We use non-conforming Cruzeix Raviart elements for velocity, piecewise constant elements for pressure, and linear Lagrange elements for concentration.  ...  It is based on gathering simulation statistics and training nonlinear regression to approximate simulation complexity metric.  ... 
doi:10.1007/978-3-642-11795-4_6 fatcat:nx4nvuxaxfbcdjknopny53ck5e

Scattering by two spheres: Theory and experiment

Irina Bjo/rno/, Leif Bjo/rno/
1998 Journal of the Acoustical Society of America  
Neural networks were trained to map from the settings of the articulatory synthesizer to the positions of pellets over limited sets of vowels.  ...  To accomplish this, a sequential neural network architecture is used in a simplified dynamical system in which the network's outputs define the activations of gestures within a set of vocal tract constrictions  ...  by the activity of neurons at the corresponding place in the neural array.  ... 
doi:10.1121/1.421626 fatcat:qf2u35fztjfgtgredwhbg53zwm

Ray tracing in a turbulent, shallow‐water channel

Christian Bjerrum‐Niese, René Lützen, Leif Bjo/rno/
1998 Journal of the Acoustical Society of America  
Neural networks were trained to map from the settings of the articulatory synthesizer to the positions of pellets over limited sets of vowels.  ...  To accomplish this, a sequential neural network architecture is used in a simplified dynamical system in which the network's outputs define the activations of gestures within a set of vocal tract constrictions  ...  by the activity of neurons at the corresponding place in the neural array.  ... 
doi:10.1121/1.422455 fatcat:vqypn7trzberzap3bkznkoddcu

Dispersion of axially symmetric waves in fluid‐filled cylindrical shells

X. L. Bao, H. Überall, P. K. Raju, A. C. Ahyi, I. K. Bjo/rno/, L. Bjo/rno/
2000 Journal of the Acoustical Society of America  
A computer model and a physical model of a typical classroom were constructed to further study classroom acoustic situations.  ...  Although designed as a school for the deaf, the approaches are useful for designing any educational classroom environment.  ...  This model was then used to characterize materials in situ, using stored data and artificial neural networks. Close agreement between theoretical and practical measurements has been demonstrated.  ... 
doi:10.1121/1.429206 fatcat:idaehfayzbegvcibxznzcamimy

GEO'98 Abstracts

1998 GeoArabia  
James received his BSc and MSc degrees from the University of Pittsburgh and a PhD degree from the University of Illinois.  ...  His current interests include the application of risk analysis to petroleum exploration and the development of geochemical techniques to allocate commingled production streams to specific sources.  ...  We show that footprint artifacts observed after DMO are mostly a spurious effect of the spatial interpolation scheme conventionally used.  ... 
doi:10.2113/geoarabia030138 fatcat:ahntcmscbngvfatriktkydvtxa

Extended modelling of the multiphasic human brain tissue with application to drug-infusion processes [article]

Arndt Wagner, Universität Stuttgart, Universität Stuttgart
2014
It is obvious that an adequate theoretical modelling of the brain allows a simulation of the occurring biomechanical effects under certain circumstances.  ...  The brain is the most significant and complex organ of human beings and plays a key role as the control centre of the nervous system.  ...  DTI provides a well established and meaningful tool for the detection of the micro-structural information of the neural axons.  ... 
doi:10.18419/opus-531 fatcat:nqife7qc7nbrlf4sosvoju4ska