betaFeedback

WorksPhysics and Astronomy

Variational learning of energy-stable macroscopic flow equations from microscopic laws

Abstract

We present a generalized variational approach for constructing energy-stable macroscopic partial differential equations (PDEs) directly from microscopic physical laws. Rather than fitting the PDE evolution for prescribed variables from macroscopic trajectories, we construct an extendable variational structure by adjoining auxiliary variables to seek autonomous reduced dynamics in an enlarged state space. By defining these variables as micro–macro encoders with explicit microscopic counterparts, we cast the learning problem as the sample-wise joint identification of the reduced state representation and the variational ingredients from time-discrete data, enabling systematic law-level transfer across scales. The resulting PDEs preserve frame indifference, physical constraints, and energy stability by construction, and admit a fluctuation–dissipation-consistent stochastic extension. Numerical results for the non-Newtonian hydrodynamics of polymeric fluids show that the learned models capture the viscoelastic responses induced by nonlinear and heterogeneous molecular interactions beyond empirical closures.

Introduction

Multiscale modeling seeks tractable macroscopic descriptions of collective behavior emerging from microscopic dynamics [1]. When clear scale separation exists, asymptotic reduction can often guide the choice of resolved variables and derive their effective dynamics. Micro–macro frameworks such as the heterogeneous multiscale method [2] and the equation-free approach [3] exploit such separation by using short microscopic simulations to infer macroscopic dynamics on the fly. However, for systems without clear scale separation, the situation is fundamentally less settled. While macroscopic reduced models are often formulated as autonomous partial differential equations (PDEs) for selected field variables, such as density, velocity, temperature, or other structural variables, the prescribed set of variables may not admit such an autonomous closure. Classical moment approaches [4, 5] enlarge the resolved state with higher-order moments, while the adequacy of the prescribed moment set remains problem dependent. For such systems, an adequate set of resolved variables and closure relations are generally unknown a priori, and transferring the microscopic physical laws to a structurally admissible reduced description remains a central challenge.

Recent data-driven approaches, including sparse-regression [6, 7, 8, 9] and neural network (NN)-based [10] methods, have advanced the inference of reduced dynamics from spatiotemporal observations. Notably, machine learning (ML)-based moment closures have enabled kinetic-to-hydrodynamic reduction while preserving conservation laws [11]. Beyond closure inference, macroscopic models must also preserve the underlying structure, such as frame indifference, thermodynamic admissibility, and energy stability. Rooted in Onsager’s reciprocal relations [12, 13], variational and thermodynamic formulations [14, 15, 16, 17, 18, 19] encode the coupling of reversible transport and irreversible dissipation, yielding macroscopic models with energy stability by construction. Recent ML approaches [20, 21, 22, 23, 24, 25, 26, 27] identify the corresponding variational ingredients from macroscopic evolution data. Despite these advances, two fundamental difficulties remain for multiscale modeling. Existing learning formulations are generally based on pre-selected resolved variables [28]. This assumes that the chosen variables admit autonomous dynamics, which may not be guaranteed. Moreover, the reduced dynamics is often inferred from trajectories generated by microscopic simulations. Such trajectory-based learning accesses microscopic laws only through their projected macroscopic responses, rather than directly exploiting their explicit analytical form. Systematic law-level transfer in data-driven multiscale closure with physical constraints remains largely open.

In this work, we develop a generalized variational learning framework for constructing energy-stable macroscopic PDE models. The central novelty is the joint identification of the reduced state representation and its variational closure by directly exploiting the microscopic laws, rather than fitting the evolution in PDE form for prescribed variables from macroscopic trajectories. The framework generalizes an extendable variational structure [16, 17, 19], adjoining a set of auxiliary field variables with explicit microscopic counterparts, to adaptively seek closed and structurally admissible reduced dynamics in an enlarged state space. Crucially, the auxiliary variables serve as physically interpretable micro–macro encoders, whose macroscopic evolution follows analytically from the microscopic dynamics. This analytical correspondence enables the microscopic laws to be directly transferred to the macroscopic closure. Consequently, the reduced state representation and variational ingredients are identified sample-wise from time-discrete data and parameterized by symmetry-preserving NNs. Upon variation, the learned structure yields autonomous PDEs satisfying physical and symmetry constraints, and enables stable numerical discretization and long-time simulation.

We illustrate the framework through the non-Newtonian hydrodynamics of polymeric fluids, a canonical multiscale problem where macroscopic stress and transport depend on microscopic molecular interactions [29, 30, 31]. The hydrodynamic equations contain an unclosed polymer stress coupled to constitutive equations for suitable structural variables. Classical models [32] typically prescribe low-dimensional moments of the polymer configuration distribution, such as end-to-end conformation tensors, together with empirical forms for stress, objective tensor derivatives, and relaxation dynamics. While these models capture qualitative viscoelastic responses, their molecular fidelity may deteriorate with nonlinear or heterogeneous molecular interactions. Recent data-driven approaches [33, 34] have learned constitutive models from rheological data and scattering-based microstructural trajectories. Within the present framework, we introduce auxiliary variables as generalized conformation tensors, interpreted as nonlinear features of the polymer configuration distribution with explicit microscopic counterparts. These tensors and variational ingredients are identified from time-discrete samples using the microscopic laws. The constructed variational structure generates the polymer stress and constitutive dynamics jointly, rather than fitting them separately at the PDE level as in [35, 36]. The resulting model captures molecular interaction effects beyond empirical closures, while preserving frame indifference, energy stability, and the positive definiteness of the conformation tensors, and further enables a stochastic extension with a consistent fluctuation–dissipation relation. Although demonstrated for polymeric fluids, the framework extends in principle to other multiscale problems, including kinetic-to-hydrodynamic reduction [11].

Preliminary

Microscopic description

We consider a polymer solution consisting of NmN_m molecules, with each polymer represented by a bead–spring model containing NN particles. Let q∈RN×3\mathbf{q} \in\mathbb{R}^{N \times3} denote the configuration of a molecule and let r=Kq\mathbf{r} = \mathbf{K}\mathbf{q} collect its NbN_b bond vectors, where rn∈R3×1\mathbf{r}_n \in\mathbb{R}^{3 \times1} is the nn–th bond vector and K∈RNb×N\mathbf{K} \in\mathbb{R}^{N_b \times N} is the connectivity matrix determined by the molecular structure. The intramolecular potential of the polymer system is

Vm=∑m=1NmV(r(m)),V(r)=∑n=1NbVb(∣rn∣),(1)V_m = \sum_{m=1}^{N_m} V\left(\mathbf{r}^{(m)}\right), \qquad V(\mathbf{r}) = \sum_{n=1}^{N_b} V_b\left(\lvert\mathbf{r}_n\rvert\right), \tag*{(1)}

where VbV_b denotes the bond potential and the superscript indexes the polymer molecules. In principle, the molecular architecture K\mathbf{K} and bond interactions VbV_b, together with the coupling to the surrounding solvent and thermal fluctuations, determine the microscopic dynamics of the polymer solution. In practice, directly resolving all polymer and solvent degrees of freedom over hydrodynamic scales could become computationally prohibitive. This motivates the macroscopic reduced model based on various empirical closures [29, 37].

Macroscopic description

At the continuum level, the incompressible hydrodynamic equations take the form

∇⋅u=0\nabla\cdot\mathbf{u} = 0
ρ(∂u∂t+u⋅∇u)=−∇p+∇⋅(τs+τp)+g,(2)\rho\left(\frac{\partial\mathbf{u}}{\partial t} + \mathbf{u}\cdot\nabla\mathbf{u}\right) = -\nabla p + \nabla\cdot\left(\boldsymbol{\tau}_s+\boldsymbol{\tau}_p\right) + \mathbf{g}, \tag*{(2)}

where ρ\rho, u\mathbf{u}, pp, and g\mathbf{g} denote the fluid density, velocity, pressure, and external body force, respectively. τs\boldsymbol{\tau}_s is the solvent stress tensor and takes the closed Newtonian form τs=η(∇u+∇uT)\boldsymbol{\tau}_s = \eta\left(\nabla\mathbf{u}+\nabla\mathbf{u}^{\mathsf{T}}\right) with shear viscosity η\eta. However, a closure gap arises from the polymer stress τp\boldsymbol{\tau}_p. In principle, τp\boldsymbol{\tau}_p can be determined from the microscopic interactions in Eq. (1). Let f(r,x,t)f(\mathbf{r},\mathbf{x},t) denote the local probability density function (PDF) of the polymer configuration, and let ⟨⋅⟩x,t\langle\cdot\rangle_{\mathbf{x},t} denote the corresponding ensemble average with respect to ff. For notational simplicity, we suppress the subscript (x,t)(\mathbf{x},t) hereafter. τp\boldsymbol{\tau}_p takes the microscopic form

τp(x,t)=np⟨∑n=1Nbrn⊗∇rnV(r)⟩−npNbkBTI,(3)\boldsymbol{\tau}_p(\mathbf{x},t) = n_p\left\langle\sum_{n=1}^{N_b}\mathbf{r}_n\otimes\nabla_{\mathbf{r}_n}V(\mathbf{r})\right\rangle- n_pN_bk_BT\mathbf{I}, \tag*{(3)}

where npn_p is the polymer number density and is assumed spatially homogeneous as a constant. To close the macroscopic model in Eq. (2), a common approach is to approximate τp\boldsymbol{\tau}_p in terms of a pre-selected conformation tensor c\mathbf{c} along with its constitutive dynamics, i.e.,

τp=G(c)DcDt=H(c).(4)\boldsymbol{\tau}_p = \mathbf{G}(\mathbf{c}) \qquad\frac{\mathcal{D}\mathbf{c}}{\mathcal{D}t} = \mathbf{H}(\mathbf{c}). \tag*{(4)}

For simple dumbbell-shaped molecules with Nb=1N_b = 1, a standard choice is the conformation tensor c=⟨rrT⟩\mathbf{c} = \langle\mathbf{r}\mathbf{r}^{\mathsf{T}}\rangle. G\mathbf{G} and H\mathbf{H} represent the stress and reduced dynamics. Their constitutive forms are often determined by different (e.g., linear or mean-field) approximations while preserving the symmetry constraints, such as the Hookean [38], Giesekus [39], FENE-P [40], and FENE-S [41]. Notably, DcDt\frac{\mathcal{D}\mathbf{c}}{\mathcal{D}t} denotes an objective tensor derivative, which also needs to satisfy the frame-indifference constraint. This constraint does not uniquely determine its form; common choices include upper-convected [38] and co-rotational [42].

Despite the broad applications, Eq. (4) is based on a prescribed reduced state and specific constitutive assumptions. For general molecular structures and interactions, neither an adequate reduced state nor its constitutive structure is known a priori. The problem addressed here is to jointly identify the reduced state and an extendable variational structure directly from the microscopic laws, by which the stress, objective transport, and reduced dynamics are generated rather than fitting the individual PDE forms for pre-selected variables.

Learnable extendable variational non-Newtonian hydrodynamic model

Extendable admissible variational structure

We consider the open isothermal incompressible hydrodynamic model in Eq. (2) and embed it in an enlarged state space. Let χ=(M,c1,…,cNc)\boldsymbol{\chi}=(\boldsymbol{M},\boldsymbol{c}_1,\ldots,\boldsymbol{c}_{N_c}) represent a set of macroscopic field variables where M=ρu\boldsymbol{M}=\rho\boldsymbol{u} is the fluid momentum with ∇⋅M=0\nabla\cdot\boldsymbol{M}=0 and all functional derivatives with respect to M\boldsymbol{M} are constrained variations on the divergence-free subspace. ci∈S+3\boldsymbol{c}_i\in\mathbb{S}_{+}^{3} is a set of generalized second-order conformation tensors representing learned nonlinear features of the polymer configuration PDF f(r,x,t)f(\boldsymbol{r},\boldsymbol{x},t). Their microscopic definitions will be identified jointly with the variational closure below.

To construct the reduced dynamics, we represent the evolution of a general observable F[χ]F[\boldsymbol{\chi}] as the coupling of a reversible and an irreversible process taking a generalized energy variational form [16, 17, 19], i.e.,

F˙={F,H}+(F,H)−,(5)\dot{F}=\{F,\mathcal{H}\}+(F,\mathcal{H})_{-}, \tag*{(5)}

where {⋅,⋅}\{\cdot,\cdot\} is the skew-symmetric Poisson bracket that generates the reversible transport and (⋅,⋅)−(\cdot,\cdot)_{-} is a non-positive dissipative bracket that governs the irreversible energy dissipation and molecular relaxation, i.e.,

{F,G}=−{G,F}(F,F)−≤0.(6)\{F,G\}=-\{G,F\}\qquad(F,F)_{-}\leq0. \tag*{(6)}

H\mathcal{H} is the free energy functional consisting of the kinetic energy and the polymer entropy, i.e.,

H[χ]=∫Ω[∣M∣22ρ−s(c1,…,cNc)]dx,(7)\mathcal{H}[\boldsymbol{\chi}]=\int_{\Omega}\left[\frac{|\boldsymbol{M}|^{2}}{2\rho}-s(\boldsymbol{c}_1,\ldots,\boldsymbol{c}_{N_c})\right]\mathrm{d}x, \tag*{(7)}

where ss is the macroscopic entropy density and −s-s the configuration free energy density with the detailed form to be learned. By imposing Eq. (6), the macroscopic model in Eq. (5) ensures the free energy dissipation dHdt≤0\frac{\mathrm{d}\mathcal{H}}{\mathrm{d}t}\leq0 in the absence of external body force. An external body force g\boldsymbol{g} is treated as prescribed in the momentum balance without modifying the variational structures.

For two general observables F[χ]F[\boldsymbol{\chi}] and G[χ]G[\boldsymbol{\chi}], the Poisson bracket takes the extendable form

{F,G}=∫ΩM⋅[(GM⋅∇)FM−(FM⋅∇)GM]dx+∑i=1Nc∫Ω[Fi:TGMci−Gi:TFMci]dx,(8)\{F,G\}=\int_{\Omega}\boldsymbol{M}\cdot\left[(G_{\boldsymbol{M}}\cdot\nabla)F_{\boldsymbol{M}}-(F_{\boldsymbol{M}}\cdot\nabla)G_{\boldsymbol{M}}\right]\mathrm{d}x+\sum_{i=1}^{N_c}\int_{\Omega}\left[F_i:\mathcal{T}_{G_{\boldsymbol{M}}}\boldsymbol{c}_i-G_i:\mathcal{T}_{F_{\boldsymbol{M}}}\boldsymbol{c}_i\right]\mathrm{d}x, \tag*{(8)}

where FM:=δF/δMF_{\boldsymbol{M}}:=\delta F/\delta\boldsymbol{M} under constraint ∇⋅M=0\nabla\cdot\boldsymbol{M}=0, Fi:=δF/δciF_i:=\delta F/\delta\boldsymbol{c}_i and Tvc:=−v⋅∇c+(∇v)c+c(∇v)T\mathcal{T}_{\boldsymbol{v}}\boldsymbol{c}:=-\boldsymbol{v}\cdot\nabla\boldsymbol{c}+(\nabla\boldsymbol{v})\boldsymbol{c}+\boldsymbol{c}(\nabla\boldsymbol{v})^{\mathsf{T}}. By Ref. [43], Eq. (8) satisfies the Jacobi identity and forms the incompressible hydrodynamic transport.

To model the constitutive dynamics, we construct the non-positive dissipative bracket consisting of both symmetric and skew-symmetric parts. The symmetric part governs the free energy dissipation and takes the extendable form

(F,G)−s=−∫Ωη2(DFM):(DGM) dx−∑i,j=1Nc∫ΩFi:Mij:Gj dx,(9)(F,G)_{-}^{s}=-\int_{\Omega}\frac{\eta}{2}(DF_M):(DG_M)\,\mathrm{d}x-\sum_{i,j=1}^{N_c}\int_{\Omega}F_i:M_{ij}:G_j\,\mathrm{d}x, \tag*{(9)}

where Dv:=∇v+∇vTDv:=\nabla v+\nabla v^{\mathsf{T}}. The first part represents the dissipation due to the Newtonian stress with viscosity η\eta. The second part represents the polymer molecule relaxation, where Mij(c1,⋯ ,cNc)∈R3×3×3×3M_{ij}(c_1,\cdots,c_{N_c})\in\mathbb{R}^{3\times3\times3\times3} represents the conformation-dependent friction matrices coupling among the individual features. The block operator M=[Mij]i,j=1Nc\mathcal{M}=[M_{ij}]_{i,j=1}^{N_c} is self-adjoint, frame-indifferent, with M⪰0\mathcal{M}\succeq0. Its detailed form will be learned.

In addition, we introduce a skew-symmetric part

(F,G)−a=∑i=1Nc∫Ω[Fi:(∇GM:Ei)−Gi:(∇FM:Ei)]dx,(10)(F,G)_{-}^{a}=\sum_{i=1}^{N_c}\int_{\Omega}\left[F_i:(\nabla G_M:\mathcal{E}_i)-G_i:(\nabla F_M:\mathcal{E}_i)\right]\mathrm{d}x, \tag*{(10)}

References

  1. [1]P. W. Anderson. More is different. Science, 177(4047):393–396, 1972.pubmed.ncbi.nlm.nih.gov/17796623
  2. [2]Weinan E and Bjorn Engquist. The heterogenous multiscale methods. Commun. Math. Sci., 1(1):87–132, 03 2003.
  3. [3]Ioannis G. Kevrekidis, C. William Gear, James M. Hyman, Panagiotis G Kevrekidid, Olof Runborg, and Constantinos Theodoropoulos. Equation-free, coarse-grained multiscale computation: Enabling macroscopic simulators to perform system-level analysis. Commun. Math. Sci., 1(4):715–762, 12 2003.
  4. [4]Harold Grad. On the kinetic theory of rarefied gases. Communications on pure and applied mathematics, 2(4):331–407, 1949.DOI
  5. [5]C. David Livermore. Moment closure hierarchies for kinetic theories. Journal of Statistical Physics, 83(5-6):1021–1065, 1996.
  6. [6]Hayden Schaeffer, Russel Caflisch, Cory D. Hauck, and Stanley Osher. Sparse dynamics for partial differential equations. Proceedings of the National Academy of Sciences, 110(17):6634–6639, 2013.DOI
  7. [7]Steven L. Brunton, Joshua L. Proctor, and J. Nathan Kutz. Discovering governing equations from data by sparse identification of nonlinear dynamical systems. Proceedings of the National Academy of Sciences, 113(15):3932–3937, 2016.DOI
  8. [8]Samuel H. Rudy, Steven L. Brunton, Joshua L. Proctor, and J. Nathan Kutz. Data-driven discovery of partial differential equations. Science Advances, 3(4):e1600214, 2017.
  9. [9]Daniel A. Messenger and David M. Bortz. Weak SINDy for partial differential equations. Journal of Computational Physics, 443:110525, 2021. doi: 10.1016/j.jcp.2021.110525.DOI
  10. [10]Zichao Long, Yiping Lu, Xianzhong Ma, and Bin Dong. PDE-net: Learning PDEs from data. In Proceedings of the 35th International Conference on Machine Learning, volume 80 of Proceedings of Machine Learning Research, pages 3208–3216. PMLR, 2018.arxiv.org/abs/1710.09668
  11. [11]Jiequn Han, Chao Ma, Zheng Ma, and Weinan E. Uniformly accurate machine learning-based hydrodynamic models for kinetic equations. Proceedings of the National Academy of Sciences, 116(44):21983–21991, 2019.DOI
  12. [12]Lars Onsager. Reciprocal relations in irreversible processes. I. Phys. Rev., 37:405–426, 1931.DOI
  13. [13]Lars Onsager. Reciprocal relations in irreversible processes. II. Phys. Rev., 38:2265–2279, Dec 1931.DOI
  14. [14]Philip J Morrison. A paradigm for joined hamiltonian and dissipative systems. Physica D: Nonlinear Phenomena, 18(1-3):410–419, 1986.DOI
  15. [15]Miroslav Grmela and Hans Christian Öttinger. Dynamics and thermodynamics of complex fluids. I. development of a general formalism. Physical Review E, 56(6):6620–6632, 1997.DOI
  16. [16]Hans Christian Öttinger. Beyond Equilibrium Thermodynamics. Wiley, 2005. ISBN 978-0-471-72791-0.DOI
  17. [17]Antony N Beris and Brian J Edwards. Thermodynamics of flowing systems: with internal microstructure. Number 36. Oxford University Press on Demand, 1994.DOI
  18. [18]Fang-Hua Lin, Chun Liu, and Ping Zhang. On hydrodynamics of viscoelastic fluids. Communications on Pure and Applied Mathematics, 58(11):1437–1471, 2005. doi: 10.1002/cpa.20074.DOI
  19. [19]Mi-Ho Giga, Arkadz Kirshtein, and Chun Liu. Variational Modeling and Complex Fluids. In Yoshikazu Giga and Antonín Novotný, editors, Handbook of Mathematical Analysis in Mechanics of Viscous Fluids, pages 73–113. Springer International Publishing, 2018.
  20. [20]Haijun Yu, Xinyuan Tian, Weinan E, and Qianxiao Li. OnsagerNet: Learning stable and interpretable dynamics using a generalized Onsager principle. Physical Review Fluids, 6(11):114402, 2021.DOI
  21. [21]Shenglin Huang, Zegun He, Bryan Chem, and Celia Reina. Variational Onsager neural networks (VONNs): A thermodynamics-based variational learning strategy for non-equilibrium PDEs. Journal of the Mechanics and Physics of Solids, 163:104856, 2022.DOI
  22. [22]Zhen Zhang, Yeonjong Shin, and George Em Karniadakis. GFINNs: GENERIC formalism informed neural networks for deterministic and stochastic dynamical systems. Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences, 380(2229):20210207, 2022. doi: 10.1098/rsta.2021.0207.DOI
  23. [23]Filippo Masi and Ioannis Stefanou. Multiscale modeling of inelastic materials with thermodynamics-based artificial neural networks (TANN). Computer Methods in Applied Mechanics and Engineering, 398:115190, 2022.DOI
  24. [24]Xiaoli Chen, Beatrice W. Soh, Zi-En Ooi, Eleonore Vissol-Gaudin, Haijun Yu, Kostya S. Novoselov, Kedar Hippalgaonkar, and Qianxiao Li. Constructing custom thermodynamics using deep learning. Nature Computational Science, 4(1):66–85, 2023. doi: 10.1038/s43588-023-00581-5.DOI
  25. [25]Jun Sur Richard Park, Siu Wun Cheung, Youngsoo Choi, and Yeonjong Shin. tLaSDI: Thermodynamics-informed latent space dynamics identification. *Computer Methods in Applied Mechanics and Engineering*, 429:117144, 2024. doi: 10.1016/j.cma.2024.117144.DOI
  26. [26]David Nieto Simavilla, Andrea Bonfanti, Imanol García-Beristain, Pep Espa˜nol, and Marco Ellero. Hammering at the entropy: a GENERIC-guided approach to learning polymeric rheological constitutive equations using PINNs. *Journal of Fluid Mechanics*, 1016:A11, 2025. doi: 10.1017/jfm.2025.10325.
  27. [27]Suguru Shiratori, Elham Kiyani, Khemraj Shukla, and George Em Karniadakis. GIMLET: Generalizable and Interpretable Model Learning through Embedded Thermodynamics, 2025.DOI
  28. [28]Shenglin Huang, Zequn He, Nicolas Dirr, Johannes Zimmer, and Celia Reina. Statistical-Physics-Informed Neural Networks (Stat-PINNs): A machine learning strategy for coarse-graining dissipative dynamics. *Journal of the Mechanics and Physics of Solids*, 194:105908, January 2025. doi: 10.1016/j.jmps.2024.105908.DOI
  29. [29]R.G. Owens and T.N. Phillips. *Computational Rheology*. Imperial College Press, 2002. ISBN 9781860941863.
  30. [30]R. B. Bird, C. F. Curtiss, R. C. Armstrong, and O. Hassager. *Dynamics of Polymeric Liquids, Volume 2: Kinetic Theory, 2nd Edition*. Wiley, 2nd edition, 1987.
  31. [31]Weiqing Ren and Weinan E. Heterogeneous multiscale method for the modeling of complex fluids and micro-fluidics. *Journal of Computational Physics*, 204(1):1 – 26, 2005.DOI
  32. [32]Ronald G. Larson. *Constitutive Equations for Polymer Melts and Solutions*. Butterworth-Heinemann Press, 1988.
  33. [33]Kyle R. Lennon, Gareth H. McKinley, and James W. Swan. Scientific machine learning for modeling and simulating complex fluids. *Proceedings of the National Academy of Sciences*, 120 (27):e2304669120, 2023. doi: 10.1073/pnas.2304669120.DOI
  34. [34]Charles D. Young, Patrick T. Corona, Anukta Datta, Matthew E. Helgeson, and Michael D. Graham. Scattering-informed microstructure prediction during lagrangian evolution (SIMPLE)—a data-driven framework for modeling complex fluids in flow. *Rheologica Acta*, 62:587–604, 2023. doi: 10.1007/s00397-023-01412-0.DOI
  35. [35]Huan Lei, Lei Wu, and Weinan E. Machine learning based non-newtonian fluid model with molecular fidelity. *Physics Review E*, 102:043309, 2020.DOI
  36. [36]Lidong Fang, Pei Ge, Lei Zhang, Weinan E, and Huan Lei. Deepn²: A deep learning-based non-newtonian hydrodynamic model. *Journal of Machine Learning*, 1(1):114–140, 2022. doi: 10.4208/jml.220115.DOI
  37. [37]M. Doi, P.A.P.M. Doi, and S.F. Edwards. *The Theory of Polymer Dynamics*. Clarendon Press, 1986. ISBN 9780198519768.
  38. [38]J. G. Oldroyd and Alan Herries Wilson. On the formulation of rheological equations of state. *Proceedings of the Royal Society of London. Series A. Mathematical and Physical Sciences*, 200 (1063):523–541, 1950. doi: 10.1098/rspa.1950.0035.DOI
  39. [39]H. Giesekus. A simple constitutive equation for polymer fluids based on the concept of deformation-dependent tensorial mobility. Journal of Non-Newtonian Fluid Mechanics, 11(1): 69 – 109, 1982. ISSN 0377-0257. doi: https://doi.org/10.1016/0377-0257(82)85016-7.
  40. [40]A. Peterlin. Hydrodynamics of macromolecules in a velocity field with longitudinal gradient. Journal of Polymer Science Part B: Polymer Letters, 4(4):287–291, 1966. doi: 10.1002/pol.1966.110040411.DOI
  41. [41]Peng Yu, Qiang Du, and Chun Liu. From micro to macro dynamics via a new closure approximation to the fene model of polymeric fluids. Multiscale Modeling & Simulation, 3(4):895–917, 2005. doi: 10.1137/030602794.DOI
  42. [42]S. Zaremba. Sur une forme perfectionee de la theorie de la relaxation. Bull. Int. Acad. Sci. Cracovie, pages 594–614, 1903.
  43. [43]Jerrold Marsden and Alan Weinstein. Coadjoint orbits, vortices, and Clebsch variables for incompressible fluids. Physica D: Nonlinear Phenomena, 7(1-3):305–323, May 1983. doi: 10.1016/0167-2789(83)90134-3.DOI
  44. [44]Hans Christian Öttinger and Miroslav Grmela. Dynamics and thermodynamics of complex fluids. II. illustrations of a general formalism. Physical Review E, 56(6):6633–6655, 1997.DOI
  45. [45]Brandon Amos, Lei Xu, and J. Zico Kolter. Input convex neural networks. In Proceedings of the 34th International Conference on Machine Learning, volume 70, pages 146–155. PMLR, 06–11 Aug 2017.arxiv.org/abs/1609.07152
  46. [46]Bruno H. Zimm. Dynamics of polymer molecules in dilute solution: Viscoelasticity, flow birefringence and dielectric loss. The Journal of Chemical Physics, 24(2):269–278, 1956.DOI
  47. [47]Jia Zhao, Qi Wang, and Xiaofeng Yang. Numerical approximations for a phase field dendritic crystal growth model based on the invariant energy quadratization approach. International Journal for Numerical Methods in Engineering, 110(3):279–300, 2017.DOI
  48. [48]Jie Shen, Jie Xu, and Jiang Yang. The scalar auxiliary variable (SAV) approach for gradient flows. Journal of Computational Physics, 353:407–416, 2018.DOI
  49. [49]Jie Shen, Jie Xu, and Jiang Yang. A new class of efficient and robust energy stable schemes for gradient flows. SIAM Review, 61(3):474–506, 2019.arxiv.org/abs/1710.01331
  50. [50]Nusret Balci, Becca Thomases, Michael Renardy, and Charles R. Doering. Symmetric factorization of the conformation tensor in viscoelastic fluid models. Journal of Non-Newtonian Fluid Mechanics, 166(11):546–553, 2011. ISSN 0377-0257. XVIth International Workshop on Numerical Methods for Non-Newtonian Flows.
  51. [51]Raanan Fattal and Raz Kupferman. Time-dependent simulation of viscoelastic flows at high weissenberg number using the log-conformation representation. Journal of Non-Newtonian Fluid Mechanics, 126(1):23–37, 2005. ISSN 0377-0257.DOI
  52. [52]T. Vaithianathan, Ashish Robert, James G. Brasseur, and Lance R. Collins. An improved algorithm for simulating three-dimensional, viscoelastic turbulence. Journal of Non-Newtonian Fluid Mechanics, 140(1):3–22, 2006. ISSN 0377-0257.DOI
  53. [53]Thomas T. Perkins, Douglas E. Smith, and Steven Chu. Single Polymer Dynamics in an Elongational Flow. Science, 276(5321):2016–2021, 1997. doi: 10.1126/science.276.5321.2016.DOI
  54. [54]Harold R. Warner. Kinetic theory and rheology of dilute suspensions of finitely extendible dumbbells. Industrial & Engineering Chemistry Fundamentals, 11(3):379–387, 1972. doi: 10.1021/i160043a017.DOI
  55. [55]R.B. Bird, P.J. Dotson, and N.L. Johnson. Polymer solution rheology based on a finitely extensible bead spring chain model. Journal of Non-Newtonian Fluid Mechanics, 7(2):213–235, 1980. ISSN 0377-0257.DOI
  56. [56]Richard M. Jendrejack, Juan J. De Pablo, and Michael D. Graham. Stochastic simulations of DNA in flow: Dynamics and the effects of hydrodynamic interactions. The Journal of Chemical Physics, 116(17):7752–7759, 2002. doi: 10.1063/1.1466831.DOI
  57. [57]Jing Tang, Daniel W. Trahan, and Patrick S. Doyle. Coil-stretch transition of dna molecules in slit-like confinement. Macromolecules, 43:3081–3089, 2010. doi: 10.1021/ma902689c.DOI
  58. [58]Charles M. Schroeder, Hazen P. Babcock, Eric S. G. Shaqfeh, and Steven Chu. Observation of Polymer Conformation Hysteresis in Extensional Flow. Science, 301(5639):1515–1519, September 2003. doi: 10.1126/science.1086070.DOI
  59. [59]Chun Liu and Huan Sun. On energetic variational approaches in modeling the nematic liquid crystal flows. Discrete and Continuous Dynamical Systems, 23(1&2):455–475, 2009. doi: 10.3934/dcds.2009.23.455.DOI
  60. [60]M. Gregory Forest, Ruhai Zhou, and Qi Wang. Chaotic boundaries of nematic polymers in mixed shear and extensional flows. Phys. Rev. Lett., 93:088301, Aug 2004. doi: 10.1103/PhysRevLett.93.088301.DOI
  61. [61]Bob Eisenberg, YunKyong Hyon, and Chun Liu. Energy variational analysis of ions in water and channels: Field theory for primitive models of complex ionic fluids. The Journal of Chemical Physics, 133(10):104104, 2010. doi: 10.1063/1.3476262.DOI
  62. [62]P. J. Hoogerbrugge and J. M. V. A. Koelman. Simulating microscopic hydrodynamic phenomena with dissipative particle dynamics. Europhys. Lett., 19(3):155–160, 1992.DOI
  63. [63]P. Español and P. Warren. Statistical mechanics of dissipative particle dynamics. Europhysics Letters, 30(4):191–196, 1995.DOI
  64. [64]R. D. Groot and P. B. Warren. Dissipative particle dynamics: Bridging the gap between atomistic and mesoscopic simulation. Journal of Chemical Physics, 107(11):4423–4435, 1997.DOI
  65. [65]David A. Nicholson and Gregory C. Rutledge. Molecular simulation of flow-enhanced nucleation in n-eicosane melts under steady shear and uniaxial extension. The Journal of Chemical Physics, 145(24):244903, 2016. doi: 10.1063/1.4972894.DOI
  66. [66]Takahiro Murashima, Katsumi Hagita, and Toshihiro Kawakatsu. Elongational viscosity of weakly entangled polymer melt via coarse-grained molecular dynamics simulation. Nihon Reoroji Gakkaishi, 46(5):207–220, 2018. ISSN 0387-1533.DOI
  67. [67]Diederik Kingma and Jimmy Ba. Adam: A method for stochastic optimization. International Conference on Learning Representations (ICLR), 12 2015.

Paper details

vlnc:W_pjw52b8u
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents