Search

There have been impressive advances in the physical and mathematical modelling of complex physiological systems in the last few decades, with the potential to revolutionise personalised healthcare with patient-specific evidence-based diagnosis, risk assessment and treatment decision support using digital twins. However, practical progress and genuine clinical impact hinge on successful model calibration, parameter estimation and uncertainty quantification, which calls for novel innovative adaptions and methodological extensions of contemporary state-of-the-art inference techniques from Statistics and Machine Learning. In the present study, we focus on two computational fluid-dynamics (CFD) models of the blood systemic and pulmonary circulation. We discuss state-of-the-art emulation techniques based on deep learning and Gaussian processes, which are coupled with established inference techniques based on greedy optimisation, simulated annealing, Markov Chain Monte Carlo, History Matching and rejection sampling for computationally fast inference of unknown parameters of the CFD models from blood flow and pressure data. The inference task was set as a competitive challenge which the participants had to conduct within a limited time frame representative of clinical requirements. The performance of the methods was assessed independently and objectively by the challenge organisers, based on a ground truth that was unknown to the method developers. Our results indicate that for the systemic challenge, in which an idealised case of noise-free data was considered, the relative deviation from the ground-truth in parameter space ranges from 10−5% (highest-performing method) to 3% (lowest-performing method). For the pulmonary challenge, for which noisy data was generated, the performance ranges from 0.9% to 7% deviation for the parameter posterior mean, and from 35% to 570% deviation for the parameter posterior variance.

The verification of whether small-scale turbulence is isotropic remains a grand challenge. The difficulty arises because the presence of small-scale anisotropy is tied to the dissipation tensor, whose components require the full three-dimensional information of the flow field in both high spatial and temporal resolution, a condition rarely satisfied in turbulence experiments, especially during field scale measurement of atmospheric turbulence. To circumvent this issue, an intermittency-anisotropy framework is proposed through which we successfully extract the features of small-scale anisotropy from single-point measurements of turbulent time series by exploiting the properties of small-scale intermittency. Specifically, this framework quantifies anisotropy by studying the contrasting effects of burstlike activities on the scalewise production of turbulence kinetic energy between the horizontal and vertical directions. The veracity of this approach is tested by applying it over a range of datasets covering an unprecedented range in the Reynolds numbers (Re≈103-106), sampling frequencies (10 kHz to 10 Hz), surface conditions (aerodynamically smooth surfaces to typical grasslands to forest canopies), and flow types (channel flows, boundary-layer flows, atmospheric flows, and flows over forest canopies). For these diverse datasets, the findings indicate that the effects of small-scale anisotropy persists up to the integral scales of the streamwise velocity fluctuations and there exists a universal relationship to predict this anisotropy from the two-component state of the Reynolds stress tensor. This relationship is important towards the development of next-generation closure models of wall turbulence by incorporating the effects of anisotropy at smaller scales of the flow.

The types of constraints encountered in black-box simulation-based optimization problems differ significantly from those addressed in nonlinear programming. We introduce a characterization of constraints to address this situation. We provide formal definitions for several constraint classes and present illustrative examples in the context of the resulting taxonomy. This taxonomy, denoted KARQ, is useful for modeling and problem formulation, as well as optimization software development and deployment. It can also be used as the basis for a dialog with practitioners in moving problems to increasingly solvable branches of optimization.

Two-dimensional (2D) turbulence features an inverse energy cascade that produces large-scale flow structures such as large-scale vortices (LSVs) and unidirectional jets. We investigate the dynamics of such large-scale structures using extensive direct numerical simulations (DNS) of stochastically forced, viscously damped 2D turbulence within a periodic rectangular (Cartesian) domain [0,Lx]×[0,Ly]. LSVs form and dominate the system when the domain aspect ratio δ=Lx/Ly≈1, while unidirectional jets predominate at δ≳1.1. At intermediate values of δ, both structures are metastable, and fluctuation-induced transitions between LSVs and jets are observed. Based on large-scale energy balance in the condensate, we derive and verify predictions for the dependence of the total kinetic energy and the flow polarity on the nondimensional control parameters. We further collect detailed statistics on the lifetimes of LSVs and jets from DNS runs of up to 10738 viscous diffusive time units in length. The distribution of the lifetimes is consistent with that of a memoryless Poisson process. The data are compatible with an exponential dependence of the mean lifetime on the aspect ratio δ. In addition, the mean lifetimes depend sensitively on the Reynolds number Re: As Re increases, the energy gap between LSV (lower energy) and jet states (higher energy) arising from anisotropic dissipation increases, leading to an increase in lifetimes that is approximately exponential in Re for both LSVs and jets. Similarly, as the ratio of the forcing scale to the domain size increases, the transition rates increase sharply, confirming earlier findings. We investigate the transition dynamics in terms of kinetic energy, flow polarity, modal amplitude, and 2D phase-space diagrams, revealing that the transitions occur in two stages: In the initial stage, an efficient redistribution of kinetic energy by nonlinear triadic interactions facilitates a rapid transition from LSVs to jets and vice versa. In the second stage, the kinetic energy of the newly formed structure slowly adjusts to its associated (higher or lower) equilibrium value on a longer, viscous timescale, leading to a time delay that results in hysteretic transition behavior. Fluctuation-induced transitions may also occur between different numbers of jets. Our findings shed new light on the dynamics of coherent large-scale structures in anisotropic turbulence.

A mathematical model and numerical framework are presented for computing multi-physics multi-layer coating flow dynamics, with applications to the leveling of multi-layer paint films. The algorithm combines finite difference level set methods and high-order accurate sharp-interface implicit mesh discontinuous Galerkin methods to capture a complex set of multi-physics, incorporating Marangoni-driven multi-phase interfacial flow and the transport, mixing, and evaporation of multiple dissolved species. In particular, we develop several numerical methods for this multi-physics problem, including: high-order local discontinuous Galerkin methods for Poisson problems with Robin boundary conditions on implicitly-defined domains, to capture solvent evaporation; finite difference surface gradient methods, to robustly and accurately incorporate Marangoni stresses; and a coupled multi-physics time stepping approach, to incorporate all the different solvers at play including quasi-Newtonian fluid flow. The framework is applicable to an arbitrary number of layers and dissolved species; here, we apply it in a variety of settings, including multi-solvent evaporative paint dynamics, the flow and leveling of multi-layer automobile paint coatings in both 2D and 3D, and an examination of interfacial turbulence within a multi-layer matter cascade. Our results reproduce several phenomena observed in experiment, such as the formation of Marangoni plumes and Bénard cells. We also use the model to study the impact of long-wave deformational surface modes on immersed interfaces as well as the emergence of the final multi-layer film profile.

Residual trapping is an important process that affects the efficiency of cyclic storage and withdrawal and in-situ production of hydrogen in geological media. In this study, we have conducted pore-scale modeling to investigate the effects of pore geometry and injection rate on the occurrence and efficiency of residual trapping via dead-end bypassing. We begin our theoretical and numerical analyses using a single rectangular pore to understand the key controls in bypassing. We further investigated two factors affecting bypassing: (a) a continuous cycle of injection-extraction of H2, and (b) variable pore geometry. Based on our pore-scale simulations, we found that: (a) a higher pore height/width ratio (h/w) and a higher injection rate cause more residual trapping, which is unfavorable for withdrawal of H2; (b) the trapping percentage increases with the h/w first and then decreases after h/w reaches 0.5; (c) and a converging-shaped pore can result in less trapping volume. Based on a theoretical comparison of the residual trapping behavior of H2 and CO2, we discuss the mechanisms that are applicable to CO2 residual trapping and the possibility of developing engineering controls of H2 storage and production.

We provide the first rigorous study of the finite-size error in the simplest and representative coupled cluster theory, namely the coupled cluster doubles (CCD) theory, for gapped periodic systems. Given exact Hartree-Fock orbitals and their corresponding orbital energies, we demonstrate that the correlation energy obtained from the approximate CCD method, after a finite number of fixed-point iterations over the amplitude equation, exhibits a finite-size error scaling as O(Nk−[Formula presented.]). Here Nk is the number of discretization points in the Brillouin zone and characterizes the system size. Under additional assumptions ensuring the convergence of the fixed-point iterations, we demonstrate that the CCD correlation energy also exhibits a finite-size error scaling as O(Nk−[Formula presented.]). Our analysis shows that the dominant error lies in the coupled cluster amplitude calculation, and the convergence of the finite-size error in energy calculation can be boosted to O(Nk−1) with accurate amplitudes. This also provides the first proof of the scaling of the finite-size error in the third order Møller-Plesset perturbation theory (MP3) for periodic systems.

Correction to: Transport in Porous Media (2024) 151:55–84https://doi.org/10.1007/s11242-023-02033-0. There are three numbers in Table 2 of the original paper that were incorrect. Specfically, the value of the density of hydrogen (H2) for the DB model and the values of density and viscosity of H2 for the TOUGH2 model listed in Table 2 of the original paper were incorrect. (Table presented.) Properties of the H2-water upconing system for comparison against the DB model. Property DB model Used for TOUGH2 Gas cap thickness, total reservoir thickness, and radial extent (outer radius) of the reservoir Infinite, infinite, infinite 50 m, 100 m (with open boundary at bottom), 100 m (open boundary condition) Porosity (ϕ) 0.10 0.10 Permeability (kH) 1.0 × 10−12 m2 1.0 × 10−12 m2 Permeability (kV) 1.0 × 10−12 m2 1.0 × 10−12 m2 Relative permeability (krel) Not applicable Linear with Slr = 0.99 Distance from well to H2-water interface (d) 10 m 10 m Extraction rate of rate of H2 (Qm) − 5.5 kg s−1 − 5.5 kg s−1 Density of water 996 kg m−3 996 kg m−3 Density of H2 7.32 kg m−3 7.87 kg m−3 Viscosity of water 6.54 × 10−4 Pa s 5.11 × 10−4 Pa s Viscosity of H2 9.31 × 10−6 Pa s 9.53 × 10−6 Pa s A corrected Table 2 is shown below. The erroneous values in Table 2 were not used in any of the modeling and simulation. Accurate values for density and viscosity in the modeling and simulation come from CoolProp for the DB model and from EOS7CH for the TOUGH2 simulations.

Pulmonary hypertension is a cardiovascular disorder manifested by elevated mean arterial blood pressure (>20 mmHg) together with vessel wall stiffening and thickening due to alterations in collagen, elastin, and smooth muscle cells. Hypoxia-induced (type 3) pulmonary hypertension can be studied in animals exposed to a low oxygen environment for prolonged time periods leading to biomechanical alterations in vessel wall structure. This study introduces a novel approach to formulating a reduced order nonlinear elastic structural wall model for a large pulmonary artery. The model relating blood pressure and area is calibrated using ex vivo measurements of vessel diameter and wall thickness changes, under controlled pressure conditions, in left pulmonary arteries isolated from control and hypertensive mice. A two-layer, hyperelastic, and anisotropic model incorporating residual stresses is formulated using the Holzapfel-Gasser-Ogden model. Complex relations predicting vessel area and wall thickness with increasing blood pressure are derived and calibrated using the data. Sensitivity analysis, parameter estimation, subset selection, and physical plausibility arguments are used to systematically reduce the 16-parameter model to one in which a much smaller subset of identifiable parameters is estimated via solution of an inverse problem. Our final reduced one layer model includes a single set of three elastic moduli. Estimated ranges of these parameters demonstrate that nonlinear stiffening is dominated by elastin in the control animals and by collagen in the hypertensive animals. The pressure-area relation developed in this novel manner has potential impact on one-dimensional fluids network models of vessel wall remodeling in the presence of cardiovascular disease.

This paper introduces a methodology designed to augment the inverse design optimization process in scenarios constrained by limited compute, through the strategic synergy of multi-fidelity evaluations, machine learning models, and optimization algorithms. The proposed methodology is analyzed on two distinct engineering inverse design problems: airfoil inverse design and the scalar field reconstruction problem. It leverages a machine learning model trained with low-fidelity simulation data, in each optimization cycle, thereby proficiently predicting a target variable and discerning whether a high-fidelity simulation is necessitated, which notably conserves computational resources. Additionally, the machine learning model is strategically deployed prior to optimization to compress the design space boundaries, thereby further accelerating convergence toward the optimal solution. The methodology has been employed to enhance two optimization algorithms, namely Differential Evolution and Particle Swarm Optimization. Comparative analyses illustrate performance improvements across both algorithms. Notably, this method is adaptable across any inverse design application, facilitating a synergy between a representative low-fidelity ML model, and high-fidelity simulation, and can be seamlessly applied across any variety of population-based optimization algorithms.

In turbulent flows, energy production is associated with highly organized structures, known as coherent structures. Since these structures are three dimensional, their detection remains challenging in the most common situation in experiments, when single-point temporal measurements are considered. While previous research on coherent structure detection from time series employs a thresholding approach, either in spectral or temporal domain, the thresholds are ad hoc and vary significantly from one study to another. To circumvent this issue, we introduce the level-crossing method and show how specific features of a turbulent time series associated with coherent structures can be objectively identified, without assigning a priori any arbitrary threshold. By using two wall-bounded turbulence time-series datasets (at a Reynolds number of 104), we successfully extract through level-crossing analysis the impacts of coherent structures on turbulent dynamics and therefore open an alternative avenue in experimental turbulence research. By utilizing this framework further, we discover a metric, characterized by a statistical asymmetry between the peaks and troughs of a turbulent signal, to quantify inner-outer interaction in wall turbulence. Most importantly, through phase-randomized surrogate data modeling, we demonstrate that the level-crossing statistics are quite sensitive to the nonlinear dependencies in a turbulent signal. Physically, this finding implies that the large-scale coherent structures modulate the near-wall turbulent dynamics through a nonlinear interaction associated with low-speed streaks, a mechanism not identifiable from spectral analysis alone. Moreover, a connection is established between extreme value statistics and level-crossing analysis, thereby allowing additional possibilities to study extreme events in other dynamical systems.

Abstract: This work presents a multi-resolution physics-informed recurrent neural network (MR PI-RNN), for simultaneous prediction of musculoskeletal (MSK) motion and parameter identification of the MSK systems. The MSK application was selected as the model problem due to its challenging nature in mapping the high-frequency surface electromyography (sEMG) signals to the low-frequency body joint motion controlled by the MSK and muscle contraction dynamics. The proposed method utilizes the fast wavelet transform to decompose the mixed frequency input sEMG and output joint motion signals into nested multi-resolution signals. The prediction model is subsequently trained on coarser-scale input–output signals using a gated recurrent unit (GRU), and then the trained parameters are transferred to the next level of training with finer-scale signals. These training processes are repeated recursively under a transfer-learning fashion until the full-scale training (i.e., with unfiltered signals) is achieved, while satisfying the underlying dynamic equilibrium. Numerical examples on recorded subject data demonstrate the effectiveness of the proposed framework in generating a physics-informed forward-dynamics surrogate, which yields higher accuracy in motion predictions of elbow flexion–extension of an MSK system compared to the case with single-scale training. The framework is also capable of identifying muscle parameters that are physiologically consistent with the subject’s kinematics data.

Abstract: This work presents an approach for automating the discretization and approximation procedures in constructing digital representations of composites from micro-CT images featuring intricate microstructures. The proposed method is guided by the Support Vector Machine (SVM) classification, offering an effective approach for discretizing microstructural images. An SVM soft margin training process is introduced as a classification of heterogeneous material points, and image segmentation is accomplished by identifying support vectors through a local regularized optimization problem. In addition, an Interface-Modified Reproducing Kernel Particle Method (IM-RKPM) is proposed for appropriate approximations of weak discontinuities across material interfaces. The proposed method modifies the smooth kernel functions with a regularized Heaviside function concerning the material interfaces to alleviate Gibb's oscillations. This IM-RKPM is formulated without introducing duplicated degrees of freedom associated with the interface nodes commonly needed in the conventional treatments of weak discontinuities in the meshfree methods. Moreover, IM-RKPM can be implemented with various domain integration techniques, such as Stabilized Conforming Nodal Integration (SCNI). The extension of the proposed method to 3-dimension is straightforward, and the effectiveness of the proposed method is validated through the image-based modeling of polymer-ceramic composite microstructures.

We present a higher-order finite volume method for solving elliptic PDEs with jump conditions on interfaces embedded in a 2D Cartesian grid. Second, fourth, and sixth order accuracy is demonstrated on a variety of tests including problems with high-contrast and spatially varying coefficients, large discontinuities in the source term, and complex interface geometries. We include a generalized truncation error analysis based on cell-centered Taylor series expansions, which then define stencils in terms of local discrete solution data and geometric information. In the process, we develop a simple method based on Green's theorem for computing exact geometric moments directly from an implicit function definition of the embedded interface. This approach produces stencils with a simple bilinear representation, where spatially-varying coefficients and jump conditions can be easily included and finite volume conservation can be enforced.

This research computationally investigates the complex dynamic stall phenomena of a cross-flow turbine blade utilizing modal analysis to identify pertinent events within the cycle. The blade rotation perpendicular to the freestream generates a curved relative flow, a nonsinusoidal variation of relative flow speed and angle of attack, and the necessity of traveling through its own wake. These complexities have challenged traditional predictors of dynamic stall such as pitch rate, pitching moment, or relative angle of attack. To investigate these phenomena, aerodynamic loads and flow fields on the blade from large-eddy simulations are examined across two tip speed ratios. Proper orthogonal decomposition of the velocity fields is employed to analyze the spatiotemporal evolution of the dominant flow features. The modes' time development coefficients reveal a stronger representation of the flow at the higher rotation rate, capturing the trend of relative flow velocity magnitude and lift generation on the blade, along with critical events such as vortex formation and detachment. Additionally, mean power generation is enhanced by 40% by applying a nonconstant rotation rate (intracycle control or angular velocity control). The flow fields, supported by corresponding changes in the modal analysis, demonstrate that a delayed stall behavior is responsible for the additional power extraction. Finally, flow curvature, history effects, and induced flow are identified as significant factors that modify the dynamic stall onset and resulting force and moment curves as compared to nonrotating pitching or plunging foils.

Abstract: Model reduction methods usually focus on the error performance analysis; however, in presence of uncertainties, it is important to analyze the robustness properties of the error in model reduction as well. This problem is particularly relevant for engineered biological systems that need to function in a largely unknown and uncertain environment. We give robustness guarantees for structured model reduction of linear and nonlinear dynamical systems under parametric uncertainties. We consider a model reduction problem where the states in the reduced model are a strict subset of the states of the full model, and the dynamics for all of the other states are collapsed to zero (similar to quasi‐steady‐state approximation). We show two approaches to compute a robustness guarantee metric for any such model reduction—a direct linear analysis method for linear dynamics and a sensitivity analysis based approach that also works for nonlinear dynamics. Using the robustness guarantees with an error metric and an input‐output mapping metric, we propose an automated model reduction method to determine the best possible reduced model for a given detailed system model. We apply our method for the (1) design space exploration of a gene expression system that leads to a new mathematical model that accounts for the limited resources in the system and (2) model reduction of a population control circuit in bacterial cells.

We develop and solve a constrained optimization model to identify an integrable optics rapid-cycling synchrotron lattice design that performs well in several capacities. Our model encodes the design criteria into 78 linear and nonlinear constraints, as well as a single nonsmooth objective, where the objective and some constraints are defined from the output of Synergia, an accelerator simulator. We detail the difficulties of optimizing within the 32-dimensional, simulation-constrained decision space and establish that the space is nonempty. We use a derivative-free manifold sampling algorithm to account for structured nondifferentiability in the objective function. Our numerical results quantify the dependence of approximate solutions on constraint parameters and the effect of the form of objective function.

Subclinical leaflet thrombosis (SLT) is a potentially serious complication of aortic valve replacement with a bioprosthetic valve in which blood clots form on the replacement valve. SLT is associated with increased risk of transient ischemic attacks and strokes and can progress to clinical leaflet thrombosis. SLT following aortic valve replacement also may be related to subsequent structural valve deterioration, which can impair the durability of the valve replacement. Because of the difficulty in clinical imaging of SLT, models are needed to determine the mechanisms of SLT and could eventually predict which patients will develop SLT. To this end, we develop methods to simulate leaflet thrombosis that combine fluid-structure interaction and a simplified thrombosis model that allows for deposition along the moving leaflets. Additionally, this model can be adapted to model deposition or absorption along other moving boundaries. We present convergence results and quantify the model's ability to realize changes in valve opening and pressures. These new approaches are an important advancement in our tools for modeling thrombosis because they incorporate both adhesion to the surface of the moving leaflets and feedback to the fluid-structure interaction.

A new adaptive clipping-and-redistribution method is presented which provides bounds-preservation for multidimensional interpolation in the context of high-order finite-volume discretizations with adaptive mesh refinement (AMR). The underlying finite-volume method (FVM) for the computational fluid dynamics applications is fourth-order accurate for smooth solutions and utilizes AMR for computational efficiency in solving multiscale problems involving turbulence and combustion. High-order interpolation between different AMR levels is required. However, this operation often leads to numerical issues because combustion species must have physical bounds preserved. The present study overcomes two major challenges in the development of the high-order interpolation method. First, the method needs to be bound-preserving near extrema or discontinuities to prevent the emergence of unphysical oscillations while maintaining fourth-order accuracy in smooth flows. Second, the method needs to satisfy the conservation requirement in multiple dimensions, particularly in the context of curvilinear coordinate transformations. Additionally, the method is designed to be localized and computationally inexpensive. The new interpolation scheme is demonstrated by solving reacting flows, which are extremely sensitive to unphysical overshoots in conserved quantities. The test problems are shock-induced (Formula presented.) - (Formula presented.) combustion and a (Formula presented.) -air flame in a practical bluff-body combustor. Results show the method prevents new extrema near discontinuities while maintaining high-order accuracy in smooth regions. In particular, the method is extremely beneficial for combustion with stiff chemistry. With the proposed new method, even if flame fronts cross AMR interfaces or new grids are created in the vicinity of the flame, solution stability is retained.

Turbulent signals are known to exhibit burstlike activities, which affect the turbulence statistics at both large and small scales of the flow. In our study, we pursue this problem from the perspective of an event-based framework, where bursting events are studied across multiple scales in terms of both their size and duration. To illustrate our method and assess any dependence on the Reynolds number (Re), we use two data sets: from the Melbourne wind tunnel (Re≈14750) and from SLTEST, an atmospheric surface layer experiment (Re≈106). We show that an index, namely, the "burstiness index,"can be used successfully to describe the multiscale nature of turbulent bursting while accounting for the small-scale intermittency effects. With this index, we demonstrate that irrespective of Re, the presence of large amplitude fluctuations in the instantaneous velocity variance and momentum flux signals are governed by the coherent structures in the flow. For small-scale turbulence, a Re dependence is noted while studying the scalewise evolution of the burstiness features of second-order streamwise velocity increments ((Δu)2). Specific to the wind-tunnel data set, the burstiness index of the (Δu)2 signal displays a strong dependence on height and decreases as the scales increase with the maximum being obtained at scales comparable to the dissipative structures. However, such features are nearly absent in the atmospheric flows. To conclude, this research paves a way to evaluate the effect of bursts on the turbulence statistics at any specified scale of the flow.

Because of its simplicity and the ability to produce a stable, slow-propagating crack, the Double-Torsion (DT) method has been used widely for investigating the critical and subcritical propagation of a slow-propagating tensile (mode-I) crack. However, to determine the complex relationship between the crack velocity vc vs. the strain energy release rate G (or the stress intensity factor K) from laboratory measurements, several corrections must be made to account for the impact of sample and crack geometry. Particularly, DT test typically produces a crack with a curved edge profile instead of a straight line, causing the local vc and G vary along the crack front. The experimentally measured vc and G data merely reflect collective, averaged behavior of the crack. This makes inversion for the intrinsic, “true” crack growth kinetics necessary, based upon the knowledge of the crack geometry. Simple and effective correction methods have been proposed and validated for the slow, chemical-reaction-controlled part (Region I) of the vc−G curve. However, reliable methods for the highly nonlinear, transport-dominated part (Region II) and its sudden transition to the dynamic propagation part (Region III) are still lacking. In this paper, we propose a method for determining the intrinsic vc−G relationship across all three Regions based upon DT test data, using a simple model function and its numerical inversion. The performance of this approach is examined and demonstrated using both synthetic and laboratory data for subcritical crack growth in soda lime glass.

Simulating the dynamics of a nonequilibrium quantum many-body system by computing the two-time Green's function associated with such a system is computationally challenging. However, we are often interested in the time-diagonal of such a Green's function or time-dependent physical observables that are functions of one time. In this paper, we discuss the possibility of using dynamic mode decomposition (DMD), a data-driven model order reduction technique, to characterize one-time observables associated with the nonequilibrium dynamics using snapshots computed within a small time window. The DMD method allows us to efficiently predict long time dynamics from a limited number of trajectory samples. We demonstrate the effectiveness of DMD on a model two-band system. We show that, in the equilibrium limit, the DMD analysis yields results that are consistent with those produced from a linear response analysis. In the nonequilibrium case, the extrapolated dynamics produced by DMD is more accurate than a special Fourier extrapolation scheme presented in this paper. We point out a potential pitfall of the standard DMD method caused by insufficient spatial/momentum resolution of the discretization scheme. We show how this problem can be overcome by using a variant of the DMD method known as higher order DMD.

We develop a stochastic optimization framework to identify governing equations in multi-physics systems. The proposed approach discovers partial differential equations (PDEs) by combining user’s prior knowledge of the underlying physics of a target system and its observed data. The technique relies on evolutionary processes to randomly generate PDEs and stochastically optimize their structure and coefficients to the data by exploring the infinite model space. Furthermore, the method captures the spatiotemporal dynamics of physical system by direct evaluation of the candidate PDEs under physical constraints. To achieve significant computational speedup, the proposed stochastic optimization method relies on a series of novel modifications. These consist of the incorporation of a multi-purpose loss function into the parallel fitness test and bloat control techniques into the evolutionary processes. As such, these innovations lead to a significant improvement of the effectiveness and computational efficiency in identifying PDEs from given data. We demonstrate the applicability of methodology in two illustrative examples: the nonlinear Burgers’ equation and the linear/nonlinear advection-dispersion equation. The exact PDEs are successfully identified, even with significant data noise, and captures the underlying physics of the target system. Through series of simulations, we assess the accuracy, robustness, and limitations of the proposed approach. In particular, we show the impact of key dimensionless groups (that accounts for the competition between various physical phenomena) in controlling the accuracy of the identification process. This work shows that developed identification method is a promising effective and robust gray box modeling tool for identifying PDEs.

We introduce a machine-learning framework named statistics-informed neural network (SINN) for learning stochastic dynamics from data. This new architecture was theoretically inspired by a universal approximation theorem for stochastic systems, which we introduce in this paper, and the projection-operator formalism for stochastic modeling. We devise mechanisms for training the neural network model to reproduce the correct statistical behavior of a target stochastic process. Numerical simulation results demonstrate that a well-trained SINN can reliably approximate both Markovian and non-Markovian stochastic dynamics. We demonstrate the applicability of SINN to coarse-graining problems and the modeling of transition dynamics. Furthermore, we show that the obtained reduced-order model can be trained on temporally coarse-grained data and hence is well suited for rare-event simulations.

The combination of neural networks and quantum Monte Carlo methods has arisen as a promising path forward for highly accurate electronic structure calculations. Previous proposals have combined equivariant neural network layers with a final antisymmetric layer in order to satisfy the antisymmetry requirements of the electronic wavefunction. However, to date it is unclear if one can represent antisymmetric functions of physical interest, and it is difficult to precisely measure the expressiveness of the antisymmetric layer. This work attempts to address this problem by introducing explicitly antisymmetrized universal neural network layers. This approach has a computational cost which increases factorially with respect to the system size, but we are nonetheless able to apply it to small systems to better understand how the structure of the antisymmetric layer affects its performance. We first introduce a generic antisymmetric (GA) neural network layer, which we use to replace the entire antisymmetric layer of the highly accurate ansatz known as the FermiNet. We demonstrate that the resulting FermiNet-GA architecture can yield effectively the exact ground state energy for small atoms and molecules. We then consider a factorized antisymmetric (FA) layer which more directly generalizes the FermiNet by replacing the products of determinants with products of antisymmetrized neural networks. We find, interestingly, that the resulting FermiNet-FA architecture does not significantly outperform the FermiNet. This strongly suggests that the sum of products of antisymmetries is a key limiting aspect of the FermiNet architecture. To explore this further, we investigate a slight modification of the FermiNet, called the full determinant mode, which replaces each product of determinants with a single combined determinant. We find that the full single-determinant FermiNet closes a large part of the gap between the standard single-determinant FermiNet and FermiNet-GA on small atomic and molecular problems. Surprisingly, on the nitrogen molecule at a dissociating bond length of 4.0 Bohr, the full single-determinant FermiNet can outperform the largest standard FermiNet calculation with 64 determinants.

The gas–water interface in Underground Hydrogen Storage (UHS) reservoirs creates the possibility that water will upcone to the well during hydrogen (H2) withdrawal with detrimental impacts. We study the upconing of water to a hydrogen injection/withdrawal (I/W) well using both an analytical solution and numerical simulation. We carried out sensitivity analyses of the engineered properties (e.g., distance of well bottom to gas–water interface, withdrawal rate) and the intrinsic properties (e.g., reservoir permeability, porosity) of an idealized UHS system. Horizontal permeability is the main parameter controlling the height of upconing. Daily I/W cycles to some degree mitigate upconing because injection pushes down the gas–water interface. Sampling-based global sensitivity analyses show clearly that reservoirs with large horizontal permeability are preferred for avoiding upconing. Minimizing withdrawal rate and maximizing either the distance from well to gas–water interface or the length of the perforated well interval are important engineering controls to minimize upconing.

In this paper, the generalized SEIHRDP (susceptible-exposed-infective-hospitalized-recovered-death-insusceptible) fractional-order epidemic model is established with individual migration. Firstly, the global properties of the proposed system are studied. Particularly, the sensitivity of parameters to the basic reproduction number are analyzed both theoretically and numerically. Secondly, according to the real data in India and Brazil, it can all be concluded that the bilinear incidence rate has a better description of COVID-19 transmission. Meanwhile, multi-peak situation is considered in China, and it is shown that the proposed system can better predict the next peak. Finally, taking individual migration between Los Angeles and New York as an example, the spread of COVID-19 between cities can be effectively controlled by limiting individual movement, enhancing nucleic acid testing and reducing individual contact.