SimTech Publications

Abstract

Following recent technological developments in the artificial intelligence space, artificial agents are increasingly taking over organizational tasks typically reserved for humans. Studies have shown that humans respond differently to this, with some being appreciative of their advice (algorithm appreciation), others being averse toward them (algorithm aversion), and others still fully relinquishing control to artificial agents without adequate oversight (automation bias). Using systems thinking, we analyze the existing literature on these phenomena and develop a conceptual model that provides an underlying structural explanation for their emergence. In doing so, we create a powerful visual tool that can be used to ground discussions about the impact artificial agents have on organizations and humans within them.

Abstract

Developing machine learning-based interatomic potentials from ab initio electronic structure methods remains a challenging task for computational chemistry and materials science. This work studies the capability of transfer learning, in particular discriminative fine-tuning, for efficiently generating chemically accurate interatomic neural network potentials on organic molecules from the MD17 and ANI data sets. We show that pre-training the network parameters on data obtained from density functional calculations considerably improves the sample efficiency of models trained on more accurate ab initio data. Additionally, we show that fine-tuning with energy labels alone can suffice to obtain accurate atomic forces and run large-scale atomistic simulations, provided a well-designed fine-tuning data set. We also investigate possible limitations of transfer learning, especially regarding the design and size of the pre-training and fine-tuning data sets. Finally, we provide GM-NN potentials pre-trained and fine-tuned on the ANI-1x and ANI-1ccx data sets, which can easily be fine-tuned on and applied to organic molecules.

Abstract

Virtual and augmented reality technology grants new possibilities to the movement tutorial. On the one hand, VR/ AR visualize the movement instructions in the 3D world, avoiding misperceptions caused by the missing angle and depth information when using traditional 2D screens. With the tracking device in the VR/ AR system, on the other hand, the users receive precise real-time feedback and even correction guidance during training to improve their performance. So far, we have run a series of user studies to propose the design implications for MR-based motion guidance in terms of user perspective, visual encoding, and motion features. And I would like to continue my research on 1) memorability of motion guidance and 2) realistic user scenarios like physiotherapy.

Abstract

Using classical density functional theory, we investigate the influence of solvent on the structure and ionic screening of electrolytes under slit confinement and in contact with a reservoir. We consider a symmetric electrolyte with implicit and explicit solvent models and find that spatially resolving solvent molecules is essential for the ion structure at confining walls, excess ion adsorption, and the pressure exerted on the walls. Despite this, we observe only moderate differences in the period of oscillations of the pressure with the slit width and virtually coinciding decay lengths as functions of the scaling variable $\sigma_ion/łambda_D$, where $\sigma_ion$ is the ion diameter and $łambda_D$ the Debye length. Moreover, in the electrostatic-dominated regime, this scaling behavior is practically independent of the relative permittivity and its dependence on the ion concentration. In contrast, the crossover to the hard-core-dominated regime depends sensitively on all three factors.

Abstract

As social virtual reality (VR) becomes more popular, avatars are being designed with realistic behaviors incorporating non-verbal cues like eye contact. However, perceiving eye contact during a conversation can be challenging for people with visual impairments. VR presents an opportunity to display eye contact cues in alternative ways, making them perceivable for people with visual impairments. We performed an exploratory study to gain initial insights on designing eye contact cues for people with visual impairments, including a focus group for a deeper understanding of the topic. We implemented eye contact cues via visual, auditory, and tactile sensory modalities in VR and tested these approaches with eleven participants with visual impairments and collected qualitative feedback. The results show that visual cues indicating the gaze direction were preferred, but auditory and tactile cues were also prevalent as they do not superimpose additional visual information.

Abstract

Neuromuscular control loops feature substantial communication delays, but mammals run robustly even in the most adverse conditions. In vivo experiments and computer simulation results suggest that muscles’ preflex—an immediate mechanical response to a perturbation—could be the critical contributor. Muscle preflexes act within a few milliseconds, an order of magnitude faster than neural reflexes. Their short-lasting action makes mechanical preflexes hard to quantify in vivo. Muscle models, on the other hand, require further improvement of their prediction accuracy during the non-standard conditions of perturbed locomotion. Our study aims to quantify the mechanical work done by muscles during the preflex phase (preflex work) and test their mechanical force modulation. We performed in vitro experiments with biological muscle fibers under physiological boundary conditions, which we determined in computer simulations of perturbed hopping. Our findings show that muscles initially resist impacts with a stereotypical stiffness response—identified as short-range stiffness—regardless of the exact perturbation condition. We then observe a velocity adaptation to the force related to the amount of perturbation similar to a damping response. The main contributor to the preflex work modulation is not the change in force due to a change in fiber stretch velocity (fiber damping characteristics) but the change in magnitude of the stretch due to the leg dynamics in the perturbed conditions. Our results confirm previous findings that muscle stiffness is activity-dependent and show that also damping characteristics are activity-dependent. These results indicate that neural control could tune the preflex properties of muscles in expectation of ground conditions leading to previously inexplicable neuromuscular adaptation speeds.

Abstract

The Chemical Master Equation (CME) is a set of linear differential equations that describes the evolution of the probability distribution on all possible configurations of a (bio-)chemical reaction system. Since the number of configurations and therefore the dimension of the CME rapidly increases with the number of molecules, its applicability is restricted to small systems. A widely applied remedy for this challenge is moment-based approaches which consider the evolution of the first few moments of the distribution as summary statistics for the complete distribution. Here, we investigate the performance of two moment-estimation methods for reaction systems whose equilibrium distributions encounter fat-tailedness and do not possess statistical moments.We show that estimation via stochastic simulation algorithm (SSA) trajectories lose consistency over time and estimated moment values span a wide range of values even for large sample sizes. In comparison, the method of moments returns smooth moment estimates but is not able to indicate the non-existence of the allegedly predicted moments. We furthermore analyze the negative effect of a CME solution’s fat-tailedness on SSA run times and explain inherent difficulties. While moment-estimation techniques are a commonly applied tool in the simulation of (bio-)chemical reaction networks, we conclude that they should be used with care, as neither the system definition nor the moment-estimation techniques themselves reliably indicate the potential fat-tailedness of the CME’s solution.

Abstract

Facilitated water permeation through narrow biological channels is fundamental for all forms of life. Despite its significance in health and disease as well as for biotechnological applications, the energetics of water permeation are still elusive. Gibbs free energy of activation is composed of an enthalpic and an entropic component. Whereas the enthalpic contribution is readily accessible via temperature dependent water permeability measurements, estimation of the entropic contribution requires information on the temperature dependence of the rate of water permeation. Here, we estimate, by means of accurate activation energy measurements of water permeation through Aquaporin-1 and by determining the accurate single channel permeability, the entropic barrier of water permeation through a narrow biological channel. Thereby the calculated value for \$\$\backslashtriangle \S\^\\\\backslashddagger\\ \\$\$þinspace=þinspace2.01þinspace±þinspace0.82þinspaceJ/(mol·K) links the activation energy of 3.75þinspace±þinspace0.16þinspacekcal/mol with its efficient water conduction rate of ˜1010 water molecules/second. This is a first step in understanding the energetic contributions in various biological and artificial channels exhibiting vastly different pore geometries.

Abstract

Multi-Agent Reinforcement Learning (MARL) is a promising candidate for realizing efficient control of microscopic particles, of which micro-robots are a subset. However, the microscopic particles' environment presents unique challenges, such as Brownian motion at sufficiently small length-scales. In this work, we explore the role of temperature in the emergence and efficacy of strategies in MARL systems using particle-based Langevin molecular dynamics simulations as a realistic representation of micro-scale environments. To this end, we perform experiments on two different multi-agent tasks in microscopic environments at different temperatures, detecting the source of a concentration gradient and rotation of a rod. We find that at higher temperatures, the RL agents identify new strategies for achieving these tasks, highlighting the importance of understanding this regime and providing insight into optimal training strategies for bridging the generalization gap between simulation and reality. We also introduce a novel Python package for studying microscopic agents using reinforcement learning (RL) to accompany our results.

Abstract

Particle-Based (PB) simulations, including Molecular Dynamics (MD), provide access to system observables that are not easily available experimentally. However, in most cases, PB data needs to be processed after a simulation to extract these observables. One of the main challenges in post-processing PB simulations is managing the large amounts of data typically generated without incurring memory or computational capacity limitations. In this work, we introduce the post-processing tool: MDSuite. This software, developed in Python, combines state-of-the-art computing technologies such as TensorFlow, with modern data management tools such as HDF5 and SQL for a fast, scalable, and accurate PB data processing engine. This package, built around the principles of FAIR data, provides a memory safe, parallelized, and GPU accelerated environment for the analysis of particle simulations. The software currently offers 17 calculators for the computation of properties including diffusion coefficients, thermal conductivity, viscosity, radial distribution functions, coordination numbers, and more. Further, the object-oriented framework allows for the rapid implementation of new calculators or file-readers for different simulation software. The Python front-end provides a familiar interface for many users in the scientific community and a mild learning curve for the inexperienced. Future developments will include the introduction of more analysis associated with ab-initio methods, colloidal/macroscopic particle methods, and extension to experimental data.

Abstract

A theory of neural networks (NNs) built upon collective variables would provide scientists with the tools to better understand the learning process at every stage. In this work, we introduce two such variables, the entropy and the trace of the empirical neural tangent kernel (NTK) built on the training data passed to the model. We empirically analyze the NN performance in the context of these variables and find that there exists correlation between the starting entropy, the trace of the NTK, and the generalization of the model computed after training is complete. This framework is then applied to the problem of optimal data selection for the training of NNs. To this end, random network distillation (RND) is used as a means of selecting training data which is then compared with random selection of data. It is shown that not only does RND select data-sets capable of outperforming random selection, but that the collective variables associated with the RND data-sets are larger than those of the randomly selected sets. The results of this investigation provide a stable ground from which the selection of data for NN training can be driven by this phenomenological framework.

Abstract

In multiphase flows, the evolution of fluid-fluid interfaces is of interest in many applications. In addition to fluid dynamic forces governing the flow in the entire volume, surface tension determines droplet interfaces. Here, the analysis of interface kinematics can help in the investigation of interface deformation and the identification of potential breakups. To this end, we developed a visualization technique using metric and shape tensors to analyze interface stretching and bending. For interface stretching, we employ the eigenpairs of the metric tensor defined for the deformation rate of the fluid surface. For interface bending, we present a technique that locally captures the interface curvature change in terms of a shape tensor, extracting its principal directions and curvatures. We then visualize interface deformation by combining both representations into a novel glyph design. We apply our method to study multiphase flow simulations with particular emphasis on interface effects. These include the interplay between fluid dynamics and surface tension forces leading to breakup processes following droplet collisions, as well as droplet-droplet interactions of different fluids where Marangoni convection along the surface is explicitly taken into account.

Abstract

We developed a new approach comprised of different visualizations for the comparative spatio-temporal analysis of displacement processes in porous media. We aim to analyze and compare ensemble datasets from experiments to gain insight into the influence of different parameters on fluid flow. To capture the displacement of a defending fluid by an invading fluid, we first condense an input image series to a single time map. From this map, we generate a spatio-temporal flow graph covering the whole process. This graph is further simplified to only reflect topological changes in the movement of the invading fluid. Our interactive tools allow the visual analysis of these processes by visualizing the graph structure and the context of the experimental setup, as well as by providing charts for multiple metrics. We apply our approach to analyze and compare ensemble datasets jointly with domain experts, where we vary either fluid properties or the solid structure of the porous medium. We finally report the generated insights from the domain experts and discuss our contribution's advantages, generality, and limitations.

Abstract

Pharmaceutical tablet disintegration is a critical process for dissolving and enabling the absorption of the drug substance into the blood stream. The tablet disintegration process consists of multiple connected and interdependent mechanisms: liquid penetration, swelling, dissolution, and break-up. One key dependence is the dynamic change of the pore space in a tablet caused by the swelling of particles while the tablet takes up liquid. This study analysed the changes in the pore structure during disintegration by coupling the discrete element method (DEM) with a single-particle swelling model and experimental liquid penetration data from terahertz-pulsed imaging (TPI). The coupled model is demonstrated and validated for pure microcrystalline cellulose (MCC) tablets across three porosities (10, 15, and 22%) and MCC with three different concentrations of croscarmellose sodium (CCS) (2, 5, and 8% w/w). The model was validated using experimental tablet swelling from TPI. The model captured the difference in the swelling behaviour of tablets with different porosities and formulations well. Both the experimental and modelling results showed that the swelling was lowest (i.e., time to reach the maximum normalised swelling capacity) for tablets with the highest CCS concentration, cCCS = 8%. The simulations revealed that this was caused by the closure of the pores in both the wetted volume and dry volume of the tablet. The closure of the pores hinders the liquid from accessing other particles and slows down the overall swelling process. This study provides new insights into the changes in the pore space during disintegration, which is crucial to better understand the impact of porosity and formulations on the performance of tablets.

Abstract

This article presents guitARhero, an Augmented Reality application for interactively teaching guitar playing to beginners through responsive visualizations overlaid on the guitar neck. We support two types of visual guidance, a highlighting of the frets that need to be pressed and a 3D hand overlay, as well as two display scenarios, one using a desktop magic mirror and one using a video see-through head-mounted display. We conducted a user study with 20 participants to evaluate how well users could follow instructions presented with different guidance and display combinations and compare these to a baseline where users had to follow video instructions. Our study highlights the trade-off between the provided information and visual clarity affecting the user’s ability to interpret and follow instructions for fine-grained tasks. We show that the perceived usefulness of instruction integration into an HMD view highly depends on the hardware capabilities and instruction details.

Abstract

In this work, an approach for strongly temperature-dependent thermoelastic homogenization is presented. It is based on computational homogenization paired with reduced order models (ROMs) that allow for full temperature dependence of material parameters in all phases. In order to keep the model accurate and computationally efficient at the same time, we suggest the use of different ROMs at few discrete temperatures. Then, for intermediate temperatures, we derive an energy optimal basis emerging from the available ones. The resulting reduced homogenization problem can be solved in real time. Unlike classical homogenization where only the effective behavior, i.e., the effective stiffness and the effective thermal expansion, of the microscopic reference volume element are of interest, our ROM delivers also accurate full-field reconstructions of all mechanical fields within the microstructure. We show that the proposed method referred to as optimal field interpolation is computationally as efficient as simplistic linear interpolation. However, our method yields an accuracy that matches direct numerical simulation in many cases, i.e., very accurate real-time predictions are achieved. Additionally, we propose a greedy sampling procedure yielding a minimal number of direct numerical simulations as inputs (two to six discrete temperatures are used over a range of around 1000 K). Further, we pick up a black box machine-learned model as an alternative route and show its limitations in view of the limited amount of training data. Using our new method to generate an abundance of data, we demonstrate that a highly accurate tabular interpolator can be gained easily.

Abstract

Glaucoma as an eye disease influences the optic nerve, resulting in progressive vision loss and, thus, blindness. For this disease, the most important risk factor is high intraocular pressure. Therefore, it is important to accurately measure the intraocular pressure. The present work aimes to present a mathematical description of a capacitive pressure sensor based on a Micro-Electro-Mechanical-Systems (MEMS) to measure intraocular pressure (IOP). The relatively high working bias voltage of MEMS capacitive pressure sensors restricts their potential applications as implantable sensors. Hence, Polydimethylsiloxane (PDMS) is employed as a porous elastomeric substance between the deformable and fixed electrodes of the capacitor. With a low young modulus and a higher dielectric constant, it reduces the sensor's working bias voltage. The PDMS's permittivity and young modulus are a function of the porosity volume fraction based on displacement in terms of a power law with fractional power constant. The dynamic equation of the microplate's transversal motion is used in the developed model, taking mid-plane stre-tching into account along with the generated force owing to the PDMS film squeezing. To decompose the governing nonlinear equation, a weak formulation is used with appropriate basis functions, thus integrating the attained ordinary differential equations over time. The sensor response to static pressure and step-wise alteration of the applied pressure is examined by dynamic and static analysis. The results of pull-in voltage reveal that using the PDMS as a dielectric causes a considerable reduction. Additionally, the effect of the PDMS elasticity on the capacitance and displacement was assessed along with the effects of the geometrical parameters on the sensor response.

Abstract

The simulation of contaminant transport through the vadose zone enjoys high significance for decision makers and contaminated site planners since the vadose zone can serve as a filter, but many contaminants can be transported from this region to aquifers. The intention of this paper is to utilize the extended Theory of Porous Media (eTPM) to develop a ternary model for the simulation of contaminant transport in the vadose zone whose application is subsequently shown via a numerical example. The simulation was conducted for 140 days, during which the contamination source was removed after 25 days. The results indicate that the contaminant reached the water table after 76 days. The concentration of the contaminant reaching the groundwater was 17% less than that of the contaminant source.

Abstract

Pretrained model parameters and pregenerated evaluation data for our visual analysis system for scene-graph-based visual question answering (https://doi.org/10.18419/darus-3589).

Abstract

Source code of our visual analysis system to explore scene-graph-based visual question answering.This approach is built on top of the state-of-the-art GraphVQA frameworkwhich was trained on the GQA dataset. Instructions on how to use our system can be found in the README.

Abstract

Scene-graph-based Visual Question Answering (VQA) has emerged as a burgeoning field in Deep Learning research, with a growing demand for robust and interpretable VQA systems. In this paper, we present a novel visual analysis approach that addresses two critical objectives in VQA: identifying and correcting prediction issues and providing insights into model decision-making processes through visualizing internal information. Our approach builds on the GraphVQA framework, which uses graph neural networks to process scene graphs representing images and which was trained on the widely-used GQA dataset. Our analysis tool aims at users familiar with the basics of graph-based VQA. By leveraging query-based scene analysis and visualization of crucial internal states, we are able to detect and pinpoint reasons for inaccurate predictions, facilitating model refinement and dataset curation. Identifying expressive internal states is a challenge. Through rigorous computer-based evaluations and presentation of a use case, we demonstrate the effectiveness of our analysis tool and model state visualization.

Abstract

Affective Tutoring Systems (ATS) detect and mitigate critical emotional learner states with the aim of providing individualized support. In tutoring systems for safety-critical work environments, students are trained to achieve and maintain high performance, therefore an ATS should be capable of identifying critical emotional states hindering performance. Interindividual differences in the emotion-performance-relationship can be considered by using the ARC categorization system. The present contribution aims at developing a questionnaire-based method of classifying new learners to the categories. To that end, we investigated differences in personality traits between the different categories. In an airspace surveillance task, we measured performance, emotional valence, emotional arousal, and personality traits in N = 50 subjects. Results showed that a positive valence-performance-relationship, compared to a negative valence-performance-relationship, is associated with higher Neuroticism, lower Conscientiousness, and lower Openness to experience. There were no significant differences in the traits Agreeableness and Extraversion. Based on these results, a future ATS for safety-critical work environments could classify new learners in the ARCs using self-report data and thus dispense with physiological sensors. Thereby, user state diagnosis and evaluation for high performance is possible, setting the ground for an ATS adapting to critical emotional learner states.

Abstract

Water ingestion in gas turbines is used in industrial applications to improve the thermal efficiency by cooling the air before and throughout compression. However, this also leads to interactions between the liquid droplets and the compressor parts, which cause a faster degradation of the structure. The current work addresses the numerical investigation of the atomization process at the trailing edge of a compressor blade as there have only been experimental investigations in literature considering the ambient conditions in a gas turbine compressor. Direct numerical simulations (DNS) are carried out using the multiphase flow solver Free Surface 3D (FS3D). Therefore, a numerical setup has been developed to model the trailing edge as a thin plate corresponding to experiments performed at ITLR 22. Four different cases have been performed to take the experimentally observed atomization processes into account. Additionally, the dependence of the simulation results on the grid resolution and with it the limits to reproduce the experimental findings has been investigated in a grid study. The results show that the developed numerical setup works well and the different atomization processes are reproduced qualitatively, with best results for very high grid resolutions.

Abstract

 The Derjaguin--Landau--Verwey--Overbeek (DLVO) theory, introduced more than 70 years ago, is a hallmark of colloidal particle modeling. For highly charged particles in the dilute regime, it is often supplemented by Alexander's prescription (Alexander et al. in J Chem Phys 80:5776, 1984) for using a renormalized charge. Here, we solve the problem of the interaction between two charged colloids at finite ionic strength, including dielectric mismatch effects, using an efficient numerical scheme to solve the nonlinear Poisson--Boltzmann (NPB) equation with unknown boundary conditions. Our results perfectly match the analytical predictions for the renormalized charge by Trizac and coworkers (Aubouy et al. in J Phys A 36:5835, 2003). Moreover, they allow us to reinterpret previous molecular dynamics (MD) simulation results by Kreer et al. (Phys Rev E 74:021401, 2006), rendering them now in agreement with the expected behavior. We furthermore find that the influence of polarization becomes important only when the Debye layers overlap significantly.

Abstract

We expose in a tutorial fashion the mechanisms which underlie the synthesis of optimization algorithms based on dynamic integral quadratic constraints. We reveal how these tools from robust control allow to design accelerated gradient descent algorithms with optimal guaranteed convergence rates by solving small-sized convex semi-definite programs. It is shown that this extends to the design of extremum controllers, with the goal to regulate the output of a general linear closed-loop system to the minimum of an objective function. Numerical experiments illustrate that we can not only recover gradient decent and the triple momentum variant of Nesterov's accelerated first order algorithm, but also automatically synthesize optimal algorithms even if the gradient information is passed through non-trivial dynamics, such as time-delays.

Abstract

We propose novel time-domain dynamic integral quadratic constraints with a terminal cost for exponentially weighted slope-restricted gradients of not necessarily convex functions. This extends recent results for subdifferentials of convex function and their link to so-called O'Shea-Zames-Falb multipliers. The benefit of merging time-domain and frequency-domain techniques is demonstrated for linear saturated systems.

Abstract

The benefits of augmented reality (AR) have been demonstrated in both medicine and fitness, while its application in areas where these two fields overlap has been barely explored. We argue that AR opens up new opportunities to interact with, understand and share personal health data. To this end, we developed an app prototype that uses a Snapchat-like face filter to visualize personal health data from a fitness tracker in AR. We tested this prototype in two pilot studies and found that AR does have potential in this type of application. We suggest that AR cannot replace the current interfaces of smartwatches and mobile apps, but it can pick up where current technology falls short in creating intrinsic motivation and personal health awareness. We also provide ideas for future work in this direction.

Abstract

This dataset includes the system matrices in velocity formulation and a parallel coordinates plot of the sensitivity analysis from the paper titled "Port-Hamiltonian Fluid-Structure Interaction Modeling and Structure-Preserving Model Order Reduction of a Classical Guitar". The system matrices can be used to perform further system reformulations according to the paper or model reductions for the basis generation variants mentioned in the paper. The parallel coordinates plot is an interactive way to visualize the multitude of numerical experiments. For this purpose, the desired combination possibilities can simply be marked. The color of the path is a measure of the relative error due to the model reduction. The plot is in .html file format and can be opened with any browser. For visualisation purposes all error values >1 were set to 1.

Abstract

This dataset provides three tables: 1) Regime map data: The tabulated data represent the material properties, the impact conditions, the separation mechanism as well as the non-dimensional quantities of all experiments and simulations relevant for the associated publication. 2) The tabulated data of the points corresponding to the figures in the associated publication which were also highlighted in the regime maps. 3) A list of performed simulations which provides the relevant setup data.

Abstract

Tabulated data of the setups and outcomes of the immiscible liquid and single liquid droplet collisions for the preprint J. Potyka and K. Schulte: Ä VOF Method for Three Dimensional Direct Numerical Simulations of Immiscible Droplet Collisions". 2023.

Abstract

Supplemental material for the poster "Visual-Explainable AI: The Use Case of Language Models" published at the International Conference on Data-Integrated Simulation Science 2023.Collection of videos and images showing our interactive visualization systems for exploring language models: - Text classification (https://github.com/MunzT/hiddenStatesVis) - Neural machine translation (https://github.com/MunzT/NMTVis)

Abstract

Modeling reservoir sedimentation is particularly challenging due to the simultaneous simulation of shallow shores, tributary deltas, and deep waters. The shallow upstream parts of reservoirs, where deltaic avulsion and erosion processes occur, compete with the validity of modeling assumptions used to simulate the deposition of fine sediments in deep waters. We investigate how complex numerical models can be calibrated to accurately predict reservoir sedimentation in the presence of competing model simplifications and identify the importance of calibration parameters for prioritization in measurement campaigns. This study applies Bayesian calibration, a supervised learning technique using surrogate-assisted Bayesian inversion with a Gaussian Process Emulator to calibrate a two-dimensional (2d) hydro-morphodynamic model for simulating sedimentation processes in a reservoir in Albania. Four calibration parameters were fitted to obtain the statistically best possible simulation of bed level changes between 2016 and 2019 through two differently constraining data scenarios. One scenario included measurements from the entire upstream half of the reservoir. Another scenario only included measurements in the geospatially valid range of the numerical model. Model accuracy parameters, Bayesian model evidence, and the variability of the four calibration parameters indicate that Bayesian calibration only converges toward physically meaningful parameter combinations when the calibration nodes are in the valid range of the numerical model. The Bayesian approach also allowed for a comparison of multiple parameters and found that the dry bulk density of the deposited sediments is the most important factor for calibration.

Abstract

In this paper, we discuss how to synthesize stabilizing Model Predictive Control (MPC) algorithms based on convexly parameterized Integral Quadratic Constraints (IQCs), with the aid of general multipliers. Specifically, we consider Lur’e systems subject to sector-bounded and slope-restricted nonlinearities. As the main novelty, we introduce point-wise IQCs with storage in order to accordingly generate the MPC terminal ingredients, thus enabling closed-loop stability, strict dissipativity with regard to the nonlinear feedback, and recursive feasibility of the optimization. Specifically, we consider formulations involving both static and dynamic multipliers, and provide corresponding algorithms for the synthesis procedures. The major benefit of the proposed approach resides in the flexibility of the IQC framework, which is capable to deal with many classes of uncertainties and nonlinearities. Moreover, for the considered class of nonlinearities, our method yields larger regions of attraction of the synthesized predictive controllers (with reduced conservatism) if compared to the standard approach to deal with sector constraints from the literature.

Abstract

We present a novel physics-informed system identification method to construct a passive linear time-invariant system. In more detail, for a given quadratic energy functional, measurements of the input, state, and output of a system in the time domain, we find a realization that approximates the data well while guaranteeing that the energy functional satisfies a dissipation inequality. To this end, we use the framework of port-Hamiltonian (pH) systems and modify the dynamic mode decomposition, respectively, operator inference, to be feasible for continuous-time pH systems. We propose an iterative numerical method to solve the corresponding least-squares minimization problem. We construct an effective initialization of the algorithm by studying the least-squares problem in a weighted norm, for which we present the analytical minimum-norm solution. The efficiency of the proposed method is demonstrated with several numerical examples.

Abstract

Understanding traffic scenes requires considering heterogeneous information about dynamic agents and the static infrastructure. In this work we propose SCENE, a methodology to encode diverse traffic scenes in heterogeneous graphs and to reason about these graphs using a heterogeneous Graph Neural Network encoder and task-specific decoders. The heterogeneous graphs, whose structures are defined by an ontology, consist of different nodes with type-specific node features and different relations with type-specific edge features. In order to exploit all the information given by these graphs, we propose to use cascaded layers of graph convolution. The result is an encoding of the scene. Task-specific decoders can be applied to predict desired attributes of the scene. Extensive evaluation on two diverse binary node classification tasks show the main strength of this methodology: despite being generic, it even manages to outperform task-specific baselines. The further application of our methodology to the task of node classification in various knowledge graphs shows its transferability to other domains.

Abstract

The force developed by actively lengthened muscle depends on different structures across different scales of lengthening. For small perturbations, the active response of muscle is well captured by a linear-time-invariant (LTI) system: a stiff spring in parallel with a light damper. The force response of muscle to longer stretches is better represented by a compliant spring that can fix its end when activated. Experimental work has shown that the stiffness and damping (impedance) of muscle in response to small perturbations is of fundamental importance to motor learning and mechanical stability, while the huge forces developed during long active stretches are critical for simulating and predicting injury. Outside of motor learning and injury, muscle is actively lengthened as a part of nearly all terrestrial locomotion. Despite the functional importance of impedance and active lengthening, no single muscle model has all of these mechanical properties. In this work, we present the viscoelastic-crossbridge active-titin (VEXAT) model that can replicate the response of muscle to length changes great and small. To evaluate the VEXAT model, we compare its response to biological muscle by simulating experiments that measure the impedance of muscle, and the forces developed during long active stretches. In addition, we have also compared the responses of the VEXAT model to a popular Hill-type muscle model. The VEXAT model more accurately captures the impedance of biological muscle and its responses to long active stretches than a Hill-type model and can still reproduce the force-velocity and force-length relations of muscle. While the comparison between the VEXAT model and biological muscle is favorable, there are some phenomena that can be improved: the low frequency phase response of the model, and a mechanism to support passive force enhancement.Competing Interest StatementThe authors have declared no competing interest.

Abstract

We consider Gaussian subordinated Lévy fields (GSLFs) that arise by subordinating Lévy processes with positive transformations of Gaussian random fields on some spatial domain. The resulting random fields are distributionally flexible and have in general discontinuous sample paths. Theoretical investigations of the random fields include pointwise distributions, possible approximations and their covariance function. As an application, a random elliptic PDE is considered, where the constructed random fields occur in the diffusion coefficient. Further, we present various numerical examples to illustrate our theoretical findings.

Abstract

The projection lemma (often also referred to as the elimination lemma) is one of the most powerful and useful tools in the context of linear matrix inequalities for system analysis and control. In its traditional formulation, the projection lemma only applies to strict inequalities, however, in many applications we naturally encounter non-strict inequalities. As such, we present, in this note, a non-strict projection lemma that generalizes both its original strict formulation as well as an earlier non-strict version. We demonstrate several applications of our result in robust linear-matrix-inequality-based marginal stability analysis and stabilization, a matrix S-lemma, which is useful in (direct) data-driven control applications, and matrix dilation.

Abstract

We discuss the modelling framework of port-Hamiltonian descriptor systems and their use in numerical simulation and control. The structure is ideal for automated network-based modelling since it is invariant under power-conserving interconnection, congruence transformations and Galerkin projection. Moreover, stability and passivity properties are easily shown. Condensed forms under orthogonal transformations present easy analysis tools for existence, uniqueness, regularity and numerical methods to check these properties. After recalling the concepts for general linear and nonlinear descriptor systems, we demonstrate that many difficulties that arise in general descriptor systems can be easily overcome within the port-Hamiltonian framework. The properties of port-Hamiltonian descriptor systems are analysed, and time discretization and numerical linear algebra techniques are discussed. Structure-preserving regularization procedures for descriptor systems are presented to make them suitable for simulation and control. Model reduction techniques that preserve the structure and stabilization and optimal control techniques are discussed. The properties of port-Hamiltonian descriptor systems and their use in modelling simulation and control methods are illustrated with several examples from different physical domains. The survey concludes with open problems and research topics that deserve further attention.

Abstract

Nowadays, active human body models are becoming essential tools for the development of integrated occupant safety systems. However, their broad application in industry and research is limited due to the complexity of incorporated muscle controllers, the long simulation runtime, and the non-regular use of physiological motor control approaches. The purpose of this study is to address the challenges in all indicated directions by implementing a muscle controller with several physiologically inspired control strategies into an open-source extended Hill-type muscle model formulated as LS-DYNA user-defined umat41 subroutine written in the Fortran programming language. This results in increased usability, runtime performance and physiological accuracy compared to the standard muscle material existing in LS-DYNA. The proposed controller code is verified with extensive experimental data that include findings for arm muscles, the cervical spine region, and the whole body. Selected verification experiments cover three different muscle activation situations: (1) passive state, (2) open-loop and closed-loop muscle activation, and (3) reflexive behaviour. Two whole body finite element models, the 50th percentile female VIVA OpenHBM and the 50th percentile male THUMS v5, are used for simulations, complemented by the simplified arm model extracted from the 50th percentile male THUMS v3. The obtained results are evaluated additionally with the CORrelation and Analysis methodology and the mean squared error method, showing good to excellent biofidelity and sufficient agreement with the experimental data. It was shown additionally how the integrated controller allows simplified mimicking of the movements for similar musculoskeletal models using the parameters transfer method. Furthermore, the Hill-type muscle model presented in this paper shows better kinematic behaviour even in the passive case compared to the existing one in LS-DYNA due to its improved damping and elastic properties. These findings provide a solid evidence base motivating the application of the enhanced muscle material with the internal controller in future studies with Active Human Body Models under different loading conditions.

Abstract

Physics-informed neural networks~(PINNs) leverage data and knowledge about a problem. They provide a nonnumerical pathway to solving partial differential equations by expressing the field solution as an artificial neural network. This approach has been applied successfully to various types of differential equations. A major area of research on PINNs is the application to coupled partial differential equations in particular, and a general breakthrough is still lacking. In coupled equations, the optimization operates in a critical conflict between boundary conditions and the underlying equations, which often requires either many iterations or complex schemes to avoid trivial solutions and to achieve convergence. We provide empirical evidence for the mitigation of bad initial conditioning in PINNs for solving one-dimensional consolidation problems of porous media through the introduction of affine transformations after the classical output layer of artificial neural network architectures, effectively accelerating the training process. These affine physics-informed neural networks~(AfPINNs) then produce nontrivial and accurate field solutions even in parameter spaces with diverging orders of magnitude. On average, AfPINNs show the ability to improve the Formula: see text relative error by Formula: see text after 25,000 epochs for a one-dimensional consolidation problem based on Biot's theory, and an average improvement by Formula: see text with a transfer approach to the theory of porous media.

Abstract

Microrobots for, e.g., biomedical applications, need to be equipped with motility strategies that enable them to navigate through complex environments. Inspired by biological microorganisms we re-create motility patterns such as run-and-reverse, run-and-tumble, or run-reverse-flick applied to active rodlike particles in silico. We investigate their capability to efficiently explore disordered porous environments with various porosities and mean pore sizes ranging down to the scale of the active particle. By calculating the effective diffusivity for the different patterns, we can predict the optimal one for each porous sample geometry. We find that providing the agent with very basic sensing and decision-making capabilities yields a motility pattern outperforming the biologically inspired patterns for all investigated porous samples.

Abstract

Bacteria often form biofilms in porous environments where an external flow is present, such as soil or filtration systems. To understand the initial stages of biofilm formation, one needs to study the interactions between cells, the fluid and the confining geometries. Here, we present an agent based numerical model for bacteria that takes into account the planktonic stage of motile cells as well as surface attachment and biofilm growth in a lattice Boltzmann fluid. In the planktonic stage we show the importance of the interplay between complex flow and cell motility when determining positions of surface attachment. In the growth stage we show the applicability of our model by investigating how external flow and biofilm stiffness determine qualitative colony morphologies as well as quantitative measurements of, e.g., permeability.

Abstract

The design of biocatalytic reaction systems is highly complex owing to the dependency of the estimated kinetic parameters on the enzyme, the reaction conditions, and the modeling method. Consequently, reproducibility of enzymatic experiments and reusability of enzymatic data are challenging. We developed the XML-based markup language EnzymeML to enable storage and exchange of enzymatic data such as reaction conditions, the time course of the substrate and the product, kinetic parameters and the kinetic model, thus making enzymatic data findable, accessible, interoperable and reusable (FAIR). The feasibility and usefulness of the EnzymeML toolbox is demonstrated in six scenarios, for which data and metadata of different enzymatic reactions are collected and analyzed. EnzymeML serves as a seamless communication channel between experimental platforms, electronic lab notebooks, tools for modeling of enzyme kinetics, publication platforms and enzymatic reaction databases. EnzymeML is open and transparent, and invites the community to contribute. All documents and codes are freely available at https://enzymeml.org.

Abstract

This paper presents an efficient algorithm for the sequential positioning, also called nested dissection, of two planes in an arbitrary polyhedron. Two planar interfaces are positioned such that the first plane truncates a given volume from this arbitrary polyhedron and the next plane truncates a second given volume from the residual polyhedron. This is a relevant task in the numerical simulation of three-phase flows when resorting to the geometric Volume-of-Fluid (VoF) method (Hirt and Nichols, 1981) with a Piecewise Linear Interface Calculation (PLIC) (Youngs, 1982). An efficient algorithm for this task significantly speeds up the three-phase PLIC algorithm. The present study describes a method based on a recursive application of the Gaussian divergence theorem, where the fact that the truncated polyhedron shares multiple faces with the original polyhedron can be exploited to reduce the computational effort. A careful choice of the coordinate system origin for the volume computation allows for successive positioning of two planes without reestablishing polyhedron connectivity. Combined with a highly efficient root finding, this results in a significant performance gain in the reconstruction of the three-phase interface configurations. The performance of the new method is assessed in a series of carefully designed numerical experiments. Compared to a conventional decomposition-based approach, the number of iterations and, thus, of the required truncations was reduced by up to an order of magnitude. The PLIC positioning run-time was reduced by about 90% in our reference implementation. Integrated into the multi-phase flow solver Free Surface 3D (FS3D), an overall performance gain of about 20% was achieved. The reference implementation of the efficient sequential PLIC positioning is available as a Fortran module at https://doi.org/10.18419/darus-2488. Allowing for simple integration into existing numerical schemes, the proposed algorithm is self-contained, requiring no external decomposition libraries.

Abstract

This dataset provides simulation results from a high-fidelity human upper-arm finite element model under varying muscle activations and an example script to load the data.The upper arm model consists of the bones of the radius and ulna for the forearm and the humerus for the upper arm. The elbow joint that connects them is modeled as a simple hinge joint. It contains five muscles that operate this joint, two extensors and three flexors: m. triceps brachii, m. anconeus, m. biceps brachii, m. brachialis, m. brachioradialis. The geometry of all these components is modeled after the Visible Human Male dataset. Each of the muscles is associated with a corresponding muscle activation, which represents the percentage of the maximum possible active stress that the muscle is currently exerting. The provided simulation results themselves contain data on element- and node-level including positional data as well as stress data and node-connectivity information. The saved simulation results constitutes equilibrium states for the given activation levels.

Abstract

Computational homogenization is the gold standard for concurrent multi-scale simulations (e.g., FE2) in scale-bridging applications. Often the simulations are based on experimental and synthetic material microstructures represented by high-resolution 3D image data. The computational complexity of simulations operating on such voxel data is distinct. The inability of voxelized 3D geometries to capture smooth material interfaces accurately, along with the necessity for complexity reduction, has motivated a special local coarse-graining technique called composite voxels (Kabel et al. Comput Methods Appl Mech Eng 294: 168--188, 2015). They condense multiple fine-scale voxels into a single voxel, whose constitutive model is derived from the laminate theory. Our contribution generalizes composite voxels towards composite boxels (ComBo) that are non-equiaxed, a feature that can pay off for materials with a preferred direction such as pseudo-uni-directional fiber composites. A novel image-based normal detection algorithm is devised which (i) allows for boxels in the firsts place and (ii) reduces the error in the phase-averaged stresses by around 30\% against the orientation cf. Kabel et al. (Comput Methods Appl Mech Eng 294: 168--188, 2015) even for equiaxed voxels. Further, the use of ComBo for finite strain simulations is studied in detail. An efficient and robust implementation is proposed, featuring an essential selective back-projection algorithm preventing physically inadmissible states. Various examples show the efficiency of ComBo against the original proposal by Kabel et al. (Comput Methods Appl Mech Eng 294: 168--188, 2015) and the proposed algorithmic enhancements for nonlinear mechanical problems. The general usability is emphasized by examining various Fast Fourier Transform (FFT) based solvers, including a detailed description of the Doubly-Fine Material Grid (DFMG) for finite strains. All of the studied schemes benefit from the ComBo discretization.

Abstract

Neck movement and injury during a car crash is influenced by reflexes, strength, and flexibility. This data set contains in-vivo measurements of neck kinematics and reflexes as participants are rapidly accelerated while driving a mechanically simulated car. This data set includes measurements from a mechanical driving simulator that has a limited range of movement on the order of 10’s of centimeters (ITM). Torso, head and neck kinematics are measured using infrared reflective markers and cameras while electrical activity from the neck muscles are recorded using surface electromyography. MVC tests for flexion and extension are used to normalize the EMG signals.

Abstract

Turbulent inflow methods offer new possibilities for an efficient simulation by reducing the computational domain to the interesting parts. Typical examples are turbulent flow over cavities, around obstacles or in the context of zonal large eddy simulations. Within this work, we present the current state of two turbulent inflow methods implemented in our high order discontinuous Galerkin code FLEXI with special focus laid on HPC applications. We present the recycling-rescaling anisotropic linear forcing (RRALF), a combination of a modified recycling-rescaling approach and an anisotropic linear forcing, and a synthetic eddy method (SEM). For both methods, the simulation of a turbulent boundary layer along a flat plate is used as validation case. For the RRALF method, a zonal large eddy simulation of the rear part of a tripped subsonic turbulent boundary layer over a flat plate is presented. The SEM is validated in the case of a supersonic turbulent boundary layer using data from literature at the inflow. In the course of the cluster upgrade to the HPE Apollo system at HLRS, our framework was examined for performance on the new hardware architecture. Optimizations and adaptations were carried out, for which we will present current performance data.

Abstract

Screening of electrostatic interactions in room-temperature ionic liquids and concentrated electrolytes has recently attracted much attention as surface force balance experiments have suggested the emergence of unanticipated anomalously large screening lengths at high ion concentrations. Termed underscreening, this effect was ascribed to the bulk properties of concentrated ionic systems. However, underscreening under experimentally relevant conditions is not predicted by classical theories and challenges our understanding of electrostatic correlations. Despite the enormous effort in performing large-scale simulations and new theoretical investigations, the origin of the anomalously long-range screening length remains elusive. This contribution briefly summarises the experimental, analytical and simulation results on ionic screening and the scaling behaviour of screening lengths. We then present an atomistic simulation approach that accounts for the solvent and ion exchange with a reservoir. We find that classical density functional theory (DFT) for concentrated electrolytes under confinement reproduces ion adsorption at charged interfaces surprisingly well. With DFT, we study confined electrolytes using implicit and explicit solvent models and the dependence on the solvent's dielectric properties. Our results demonstrate how the absence vs. presence of solvent particles and their discrete nature affect the short and long-range screening in concentrated ionic systems.

Abstract

The tendency of initially separated droplets to convene, named droplet grouping, is relevant in many technical applications, such as 3D ink-jet printing, medical inhalators and spray coating. Recent comparison of numerical studies to experimental data has shown that droplet shape deformation and subsequent oscillation can have a significant influence on the grouping time in droplet streams 1, 2. Therefore, this study aims at numerically investigating the effect of non-spherical droplet shapes and oscillation on droplet dynamics and grouping behavior in monodisperse droplet streams for two relevant stages: 1) at the initial stage after breakup of the liquid jet into a stream of distinct droplets and 2) after the approach and coalescence of droplets within the stream. The simulations are performed with the in-house multiphase flow solver Free Surface 3D (FS3D) based on the Volume-of-Fluid (VOF) method, which solves the incompressible Navier-Stokes equations via direct numerical simulation (DNS). The evolution of the dominant oscillation mode and the damping rate for single, translating droplets as well as the effect of nearby droplets in the stream is compared with an analytical solution. Time until grouping in the first and subsequent grouping behavior in the second considered stage are examined for varying initial deviation from a spherical droplet and different initial Reynolds numbers. The study shows that shape oscillation resulting from sufficient deformation tends to expedite grouping processes in the stage after breakup of the liquid jet, decreasing time until grouping up to 33 % for a Reynolds number of 500. The coalescence in the grouping process within the second stage also induces stretching of the droplets. The influence on follow-up grouping after droplet coalescence in streams is found to be minor, since in this case time scales of the damping of the oscillation are small compared to characteristic times for the grouping process itself.

Abstract

Liquids with a non-Newtonian rheological behavior are frequently encountered in engineering applications, where multiple phases can be involved. A rather scarcely investigated non-Newtonian behavior is thixotropy, in which the fluid’s viscosity depends on the shear rate as well as on the deformation history. In this study, a fundamental case of an oscillating droplet consisting of an ideal thixotropic fluid is numerically investigated with the in-house multiphase flow solver Free Surface 3D. The ideal thixotropic behavior is modeled by an additional kinetic equation which involves a structural breakdown and recovery term. The influence of both terms on the evolution of the spatially averaged structural parameter λ is studied. The numerical results show that λ oscillates around a steady-state value, which is reached quicker for faster recovery of the structure. The occurring maxima of λ are shifted in relation to the oscillation modes. This effect is more pronounced for a lower breakdown strength and for a slower recovery rate.

Abstract

User avatars are a critical component in collaborative Virtual Environments. A multitude of tools exist, employing various avatar types for self-representation and the representation of others. However, the optimal appearance for these avatars remains unclear. Consumer applications predominantly utilize stylized avatars, which are less prevalent in enterprise sectors. To investigate users’ avatar preferences, we conducted three user studies in both non-immersive and immersive environments, exploring whether social contexts influence virtual self-avatar selection. We generated individual avatars based on photographs of 91 participants, comprising 71 employees in the automotive industry and 20 individuals external to the company. Our findings indicate that work situations, irrespective of the occupational domain, significantly impact selfavatar choices. Conversely, we observed no correlation between the display medium or personality dimensions and avatar selection. Drawing upon our results, we provide recommendations for future avatar representation in work environments.

Abstract

We propose novel quadratic performance tests for linear discrete-time impulsive systems based on viewing these systems as feedback interconnections of some non-impulsive linear system with an impulsive operator. In order to systematically analyze such interconnections, we employ the framework of integral quadratic constraints and propose novel constraints of this kind for capturing the behavior of the involved impulsive operator. As a major benefit, the modularity of this framework permits seamless extensions to interconnections affected by heterogeneous uncertainties in a straightforward fashion. This contrasts with alternative approaches which are based on capturing the system's impulsive behavior by means of a clock. Building upon the developed analysis criteria, we characterize the existence of non-impulsive estimators for such impulsive interconnections in a lossless fashion and in terms of linear matrix inequalities. Finally, our approach is illustrated by means of several numerical examples.

Abstract

Starting from a linear fractional representation of a linear system affected by constant parametric uncertainties, we demonstrate how to enhance standard robust analysis tests by taking available (noisy) input-output data of the uncertain system into account. Our approach relies on a lifting of the system and on the construction of data-dependent multipliers. It leads to a test in terms of linear matrix inequalities which guarantees stability and performance for all systems compatible with the observed data if it is in the affirmative. In contrast to many other data-based approaches, prior physical knowledge is included at the outset due to the underlying linear fractional representation.

Abstract

It is well known that any port-Hamiltonian (pH) system is passive, and conversely, any minimal and stable passive system has a pH representation. Nevertheless, this equivalence is only concerned with the input-output mapping but not with the Hamiltonian itself. Thus, we propose to view a pH system either as an enlarged dynamical system with the Hamiltonian as additional output or as two dynamical systems with the input-output and the Hamiltonian dynamic. Our first main result is a structure-preserving Kalman-like decomposition of the enlarged pH system that separates the controllable and zero-state observable parts. Moreover, for further approximations in the context of structure-preserving model-order reduction (MOR), we propose to search for a Hamiltonian in the reduced pH system that minimizes the 2-distance to the full-order Hamiltonian without altering the input-output dynamic, thus discussing a particular aspect of the corresponding multi-objective minimization problem corresponding to $H_2$-optimal MOR for pH systems. We show that this optimization problem is uniquely solvable, can be recast as a standard semidefinite program, and present two numerical approaches for solving it. The results are illustrated with three academic examples.

Abstract

Augmented Reality (AR) is a key technology for digitization in enterprises. However, often there is a lack of stable tracking solutions when used inside manufacturing environments. Many different tracking technologies are available, yet, it can be difficult to choose the most appropriate tracking solution for different use cases with their varying conditions. In order to shed light on common tracking requirements and conditions for automotive AR use cases we conducted a use case study spanning 61 use cases within the complete product life-cycle of a large automotive manufacturer. By analyzing the gathered data we were able to note the frequency of different tracking requirements and conditions within automotive AR use cases. Based on these use cases we could also derive common factors of influence for AR tracking in the automotive industry, which show the various challenges automotive AR tracking is currently facing.

Abstract

Building on the well-known total variation, this paper develops a general regularization technique based on nonlinear isotropic diffusion (NID) for inverse problems with piecewise smooth solutions. The novelty of our approach is to be adaptive (we speak of A-NID), i.e., the regularization varies during the iterates in order to incorporate prior information on the edges, deal with the evolution of the reconstruction and circumvent the limitations due to the nonconvexity of the proposed functionals. After a detailed analysis of the convergence and well-posedness of the method, the latter is validated by simulations performed on synthetic and real data on computerized tomography.

Abstract

The evaporation of water from bare soil is often accompanied by the formation of a layer of crystallized salt, a process that must be understood in order to address the issue of soil salinization. Here, we use nuclear magnetic relaxation dispersion measurements to better understand the dynamic properties of water within two types of salt crusts: sodium chloride (NaCl) and sodium sulfate (Na2SO4). Our experimental results display a stronger dispersion of the relaxation time T1 with frequency for the case of sodium sulfate as compared to sodium chloride salt crusts. To gain insight into these results, we perform molecular dynamics simulations of salt solutions confined within slit nanopores made of either NaCl or Na2SO4. We find a strong dependence of the value of the relaxation time T1 on pore size and salt concentration. Our simulations reveal the complex interplay between the adsorption of ions at the solid surface, the structure of water near the interface, and the dispersion of T1 at low frequency, which we attribute to adsorption–desorption events.

Abstract

This paper introduces a novel representation of convolutional Neural Networks (CNNs) in terms of 2-D dynamical systems. To this end, the usual description of convolutional layers with convolution kernels, i.e., the impulse responses of linear filters, is realized in state space as a linear time-invariant 2-D system. The overall convolutional Neural Network composed of convolutional layers and nonlinear activation functions is then viewed as a 2-D version of a Lur'e system, i.e., a linear dynamical system interconnected with static nonlinear components. One benefit of this 2-D Lur'e system perspective on CNNs is that we can use robust control theory much more efficiently for Lipschitz constant estimation than previously possible.

Abstract

VR-based motion guidance systems can provide 3D movement instructions and real-time feedback for practicing movement without a live instructor. However, the precise visualization of movement paths or postures may be insufficient to learn a new motor skill, as they might make users too dependent and lead to poor performance when there is no guidance. In this paper, we propose to use enhanced error visualization, asymptotic path, increasing transparency, and haptic constraint to improve the memorability of motion guidance. Our study results indicated that adding an enhanced error feedback visualization helped the users with short-term retention.

Abstract

A large body of work has identified characteristic neural signatures of value-based decision-making. A core circuit has been shown to track the values of options under consideration, and the dynamics of neural activity associated with this circuit have been found to closely resemble the ramping evidence accumulation process believed to underpin goal-directed choice, most notably in the centroparietal positivity (CPP). However, recent neuroimaging studies suggest that value-based choices trigger multiple value-related neural signatures, some of which are unrelated to decision-making per se but instead reflect reflexive affective reactions to one’s options. Here, we use the temporal resolution of EEG to test whether choice-independent value signals could be dissociated from well-known temporal signatures of the choice process. We show that EEG activity during value-based choice can be decomposed into distinct spatiotemporal clusters, one stimulus-locked (associated with the affective salience of a choice set) and one response-locked (associated with the difficulty of selecting the best option). We show that neither of these clusters meet the criteria for an evidence accumulation signal. Instead, and to our surprise, we found that stimulus-locked activity can mimic an evidence accumulation process when aligned to the response (as with the CPP). In this dataset and a second one that uses a more traditional perceptual decision-making task, we show that the CPP’s apparent pattern of evidence accumulation disappears when stimulus-locked and response-locked signals are accounted for jointly. Collectively, our findings show that neural signatures of value can reflect choiceindependent processes that when analyzed using standard approaches, can look deceptively like evidence accumulation.Significance Statement To choose, people must evaluate their options and select between them. Selection is well described by a process of accumulating evidence up to some threshold, with an electrophysiological signature in the centroparietal positivity (CPP). However, decision-making also gives rise to value signals reflecting affective reactions and other selection-unrelated processes. Measuring EEG while participants made value-based choices, we identified two spatiotemporally distinct value signals, neither reflecting evidence accumulation. Instead, we show that evidence accumulation signals found in the CPP can arise artifactually from overlapping stimulus- and response-related activity. These findings call for a significant reexamination of established links between neural and computational mechanisms of choice, while inviting deeper consideration of the array of cognitive and affective processes that occur in parallel.Competing Interest StatementThe authors have declared no competing interest.

Abstract

This letter presents a novel approach for identifying uncorrelated atomic configurations from extensive data sets with a non-standard neural network workflow known as random network distillation (RND) for training machine-learned inter-atomic potentials (MLPs). This method is coupled with a DFT workflow wherein initial data is generated with cheaper classical methods before only the minimal subset is passed to a more computationally expensive ab initio calculation. This benefits training not only by reducing the number of expensive DFT calculations required but also by providing a pathway to the use of more accurate quantum mechanical calculations for training. The method's efficacy is demonstrated by constructing machine-learned inter-atomic potentials for the molten salts KCl and NaCl. Our RND method allows accurate models to be fit on minimal data sets, as small as 32 configurations, reducing the required structures by at least one order of magnitude compared to alternative methods.

Abstract

Eukaryotic cells can undergo different forms of programmed cell death, many of which culminate in plasma membrane rupture as the defining terminal event1--7. Plasma membrane rupture was long thought to be driven by osmotic pressure, but it has recently been shown to be in many cases an active process, mediated by the protein ninjurin-18 (NINJ1). Here we resolve the structure of NINJ1 and the mechanism by which it ruptures membranes. Super-resolution microscopy reveals that NINJ1 clusters into structurally diverse assemblies in the membranes of dying cells, in particular large, filamentous assemblies with branched morphology. A cryo-electron microscopy structure of NINJ1 filaments shows a tightly packed fence-like array of transmembrane $\alpha$-helices. Filament directionality and stability is defined by two amphipathic $\alpha$-helices that interlink adjacent filament subunits. The NINJ1 filament features a hydrophilic side and a hydrophobic side, and molecular dynamics simulations show that it can stably cap membrane edges. The function of the resulting supramolecular arrangement was validated by site-directed mutagenesis. Our data thus suggest that, during lytic cell death, the extracellular $\alpha$-helices of NINJ1 insert into the plasma membrane to polymerize NINJ1 monomers into amphipathic filaments that rupture the plasma membrane. The membrane protein NINJ1 is therefore an interactive component of the eukaryotic cell membrane that functions as an in-built breaking point in response to activation of cell death.

Abstract

Section snippets The Rode model's explanation of the FLR's ascending limb allegedly involves pushing forces in thin filaments The main Rockenfeller et al. argument to dismiss the Rode model is the wrong assertion that the Rode model describes the FLR’s ascending limb based on pushing or compressive forces in thin filaments. The thin filaments would be too soft to transmit relevant pushing forces. However, in the Rode model, thin filaments are always pulled away from the Z-disk. Furthermore, Rockenfeller et al. wrongly assert that swXBs generate pushing forces on thin filaments in the Rode model and that the The Rockenfeller model ignores geometrical changes developing on the FLR’s ascending limb The Rockenfeller model assumes a hexagonal filament lattice as in optimal overlap down to 0.4-μm half-sarcomere length (with about 1 μm being the optimal length) and that the lattice distances progressively increase with shortening due to volume constancy. These assumptions ignore several geometrical changes in the half-sarcomere that occur during shortening and hamper the predictive power of the model in the range of the FLR’s ascending limb. For example, the Rockenfeller model predicts noThe Rockenfeller model assumes effective force in the region of double thin-filament overlap Double thin-filament overlap starts to develop when thin filaments slide through the M-line near optimal half-sarcomere length (left-hand side of the FLR plateau). The Rockenfeller model assumes that XBs are formed in the region of double thin-filament overlap. This assumption leads to effective force in this region and cannot explain Trombitás and Tigyi-Sebes’ experiments (8). Trombitás and Tigyi-Sebes stretched rigor muscle fibers leading to the detachment of thin filaments from the Z-disk. The Rockenfeller model's effective overlap function is mechanically inconsistent on the FLR's ascending limb One basis of the Rockenfeller model’s force calculation is the effective filament overlap, the theoretical capacity to produce isometric force depending on half-sarcomere length without effects of interfilament spacing. Surprisingly, the effective filament overlap (their Figure 1 in (3)) is straight at about 0.8-μm half-sarcomere length. First, when thick-filament tips slide through the Z-disk at this length, the myosin heads meet thin filaments of opposite polarity and form swXBs in their

Abstract

Probabilistic (Bayesian) modeling has experienced a surge of applications in almost all quantitative sciences and industrial areas. This development is driven by a combination of several factors, including better probabilistic estimation algorithms, flexible software, increased computing power, and a growing awareness of the benefits of probabilistic learning. However, a principled Bayesian model building workflow is far from complete and many challenges remain. To aid future research and applications of a principled Bayesian workflow, we ask and provide answers for what we perceive as two fundamental questions of Bayesian modeling, namely (a) “What actually is a Bayesian model?” and (b) “What makes a good Bayesian model?”. As an answer to the first question, we propose the PAD model taxonomy that defines four basic kinds of Bayesian models, each representing some combination of the assumed joint distribution of all observable and unobservable variables (P), a posterior approximator (A), and training data (D). As an answer to the second question, we propose and discuss ten utility dimensions according to which we can evaluate Bayesian models holistically, namely, (1) causal consistency, (2) parameter recoverability, (3) predictive performance, (4) fairness, (5) structural faithfulness, (6) parsimony, (7) interpretability, (8) convergence, (9) estimation speed, and (10) robustness. Finally, we propose two example utility decision trees that describe hierarchies and trade-offs between utilities depending on the inferential goals that drive model building and testing.

Abstract

Phosphate is vital for plant and algae growth, yield, and survival, but in most environments, it is poorly available. To cope with phosphate starvation, photosynthetic organisms used their phospholipids as a phosphate reserve. In microalgae, betaine lipids replace phospholipids whereas, in higher plants, betaine lipid synthesis is lost, driving plants to other strategies. The aim of this work was to evaluate to what extent betaine lipids and PC lipids share physicochemical properties and could thus substitute each other. Using neutron diffraction and molecular dynamics simulations of two synthetic lipids, dipalmitoylphosphatidylcholine (DPPC) and dipalmitoyl-diacylglyceryl-N,N,N-trimethylhomoserine (DP-DGTS), we show that DP-DGTS bilayers are thicker, more rigid, and mutually more repulsive than DPPC bilayers. The different properties and hydration response of PC and DGTS provide an explanation for the diversity of betaine lipids observed in marine organisms and for their disappearance in seed plants.Competing Interest StatementThe authors have declared no competing interest.

Abstract

Recent experiments on weak polyelectrolyte brushes found marked shifts in the effective $pK_a$ that are linear in the logarithm of the salt concentration. Comparing explicit-particle simulations with mean-field calculations we show that for high grafting densities the salt concentration effect can be explained using the ideal Donnan theory, but for low grafting densities the full shift is due to a combination of the Donnan effect and the polyelectrolyte effect. The latter originates from electrostatic correlations that are neglected in the Donnan picture and that are only approximately included in the mean-field theory. Moreover, we demonstrate that the magnitude of the polyelectrolyte effect is almost invariant with respect to salt concentration but depends on the grafting density of the brush. This invariance is due to a complex cancellation of multiple effects. Based on our results, we show how the experimentally determined $pK_a$ shifts may be used to infer the grafting density of brushes, a parameter that is difficult to measure directly.

Abstract

Adversarial robustness and generalization are both crucial properties of reliable machine learning models. In this paper, we study these properties in the context of quantum machine learning based on Lipschitz bounds. We derive tailored, parameter-dependent Lipschitz bounds for quantum models with trainable encoding, showing that the norm of the data encoding has a crucial impact on the robustness against perturbations in the input data. Further, we derive a bound on the generalization error which explicitly depends on the parameters of the data encoding. Our theoretical findings give rise to a practical strategy for training robust and generalizable quantum models by regularizing the Lipschitz bound in the cost. Further, we show that, for fixed and non-trainable encodings as frequently employed in quantum machine learning, the Lipschitz bound cannot be influenced by tuning the parameters. Thus, trainable encodings are crucial for systematically adapting robustness and generalization during training. With numerical results, we demonstrate that, indeed, Lipschitz bound regularization leads to substantially more robust and generalizable quantum models.

Abstract

We study the question of how visual analysis can support the comparison of spatio-temporal ensemble data of liquid and gas flow in porous media. To this end, we focus on a case study, in which nine different research groups concurrently simulated the process of injecting CO2 into the subsurface. We explore different data aggregation and interactive visualization approaches to compare and analyze these nine simulations. In terms of data aggregation, one key component is the choice of similarity metrics that define the relationship between different simulations. We test different metrics and find that using the machine-learning model “S4” (tailored to the present study) as metric provides the best visualization results. Based on that, we propose different visualization methods. For overviewing the data, we use dimensionality reduction methods that allow us to plot and compare the different simulations in a scatterplot. To show details about the spatio-temporal data of each individual simulation, we employ a space-time cube volume rendering. All views support linking and brushing interaction to allow users to select and highlight subsets of the data simultaneously across multiple views. We use the resulting interactive, multi-view visual analysis tool to explore the nine simulations and also to compare them to data from experimental setups. Our main findings include new insights into ranking of simulation results with respect to experimental data, and the development of gravity fingers in simulations.

Abstract

deepsysid is a system identification toolkit for multistep prediction using deep learning and hybrid methods. The toolkit is easy to use. After you follow the instructions in the README, you will be able to download a dataset, run hyperparameter optimization and identify your best-performing multistep prediction models with just three commands. The most current version of this software is available on GitHub.

Abstract

This repository contains the necessary scripts to reproduce the results from our paper "ReLiNet: Stable and Explainable Multistep Prediction with Recurrent Linear Parameter Varying Networks". See the README file for more information. The most current version of this software is available on Github.

Abstract

Multistep prediction models are essential for the simulation and model-predictive control of dynamical systems. Verifying the safety of such models is a multi-faceted problem requiring both system-theoretic guarantees as well as establishing trust with human users. In this work, we propose a novel approach, ReLiNet (Recurrent Linear Parameter Varying Network), to ensure safety for multistep prediction of dynamical systems. Our approach simplifies a recurrent neural network to a switched linear system that is constrained to guarantee exponential stability, which acts as a surrogate for safety from a system-theoretic perspective. Furthermore, ReLiNet’s computation can be reduced to a single linear model for each time step, resulting in predictions that are explainable by definition, thereby establishing trust from a human-centric perspective. Our quantitative experiments show that ReLiNet achieves prediction accuracy comparable to that of state-of-the-art recurrent neural networks, while achieving more faithful and robust explanations compared to the model-agnostic explanation method of LIME.

Abstract

In this paper, the thermal behavior of living tissue has been modeled using a non-local three-phase lag approach. The simulation results of this model- ing have been compared with experimental results of alternating radiation with different periods on human skin, yielding satisfactory alignments. Ad- ditionally, it has been investigated that the TPL model, which is an equation with an integral term, can simulate energy accumulation within the dermal tissue. Moreover, the non-local nature of the modeling has been explored to alter the influence of phase lag terms. Furthermore, apart from numer- ically solving the equation, an analytical solution has been derived for the frequency equation, demonstrating the effects of simulation parameters on the frequency equation and simulation results. The obtained results indi cate that these parameters not only independently affect the outcomes but also interact with other parameters, leading to variations beyond their direct impacts.

Abstract

The multibody system transfer matrix method (MSTMM) is an efficient formalism for systems involving rigid bodies and simple elastic elements like beams. The paper tackles the challenge of setting up transfer matrices as building blocks for generally shaped, three-dimensional elastic bodies, and establishes a novel connection between the transfer matrix method and a large family of model order reduction (MOR) methods. An articulated manipulator arm serves as a demonstration example throughout the paper to first explain the basic concepts of MSTMM. Then, a classical floating frame of reference formulation (FFRF) of flexible bodies is adapted to MSTMM to express their equations of motion in a global inertial frame. The favorable ideas from FFRF are transferred into the typical MSTMM formulation yielding the transfer equation of a general elastic body. These ideas include: (i) splitting the overall motion into a large rigid body motion and a small elastic deformation; (ii) discretization of the deformation field by using the linear finite element (FE) method; (iii) calculating the standard input data from commercial FE data for generally shaped bodies; and (iv) MOR, to find the most suitable shape functions and speed up the simulation. Next, the procedure for solving elastic multibody system dynamics using MSTMM is illustrated. In a final numerical simulation, the proposed elastic MSTMM concept for a coupled rigid-flexible multibody system shows good agreement with a classical Newton-Euler formulation. The algorithm is based on the classical FFRF without additional assumptions and thus can be used with any MOR method available in MatMorembs.

Abstract

The accuracy of the training data limits the accuracy of bulk properties from machine-learned potentials. For example, hybrid functionals or wave-function-based quantum chemical methods are readily available for cluster data but effectively out of scope for periodic structures. We show that local, atom-centered descriptors for machine-learned potentials enable the prediction of bulk properties from cluster model training data, agreeing reasonably well with predictions from bulk training data. We demonstrate such transferability by studying structural and dynamical properties of bulk liquid water with density functional theory and have found an excellent agreement with experimental and theoretical counterparts.

Abstract

Standard kernel methods for machine learning usually struggle when dealing with large datasets. We review a recently introduced Structured Deep Kernel Network (SDKN) approach that is capable of dealing with high-dimensional and huge datasets - and enjoys typical standard machine learning approximation properties.

Abstract

Ecccentric muscle contractions are fundamental to everyday life. They occur markedly in jumping, running, and accidents. Following an initial force rise, stretching of a fully activated muscle can result in a phase of decreasing force (“Give”) followed by force redevelopment. However, how the stretch velocity affects “Give” and force redevelopment remains largely unknown. We investigated the force produced by fully activated single-skinned fibers of rat extensor digitorum longus muscles during long stretches. Fibers were pulled from length 0.85 to 1.3 optimal fiber length at a rate of 1\%, 10\%, and 100\% of the estimated maximum shortening velocity. “Give” was absent in slow stretches. Medium and fast stretches yielded a clear “Give.” After the initial force peak, forces decreased by 11.2\% and 27.8\% relative to the initial peak force before rising again. During the last half of the stretch (from 1.07 to 1.3 optimal fiber length, which is within the range of the expected descending limb of the force-length relationship), the linear force slope tripled from slow to medium stretch and increased further by 60\% from medium to fast stretch. These results are compatible with forcible cross-bridge detachment and redevelopment of a cross-bridge distribution, and a viscoelastic titin contribution to fiber force. Accounting for these results can improve muscle models and predictions of multibody simulations.NEW & NOTEWORTHY Eccentric muscle contractions are part of our daily lives. We found that force increased monotonically during slow stretches of fully activated muscle fibers, whereas higher stretch velocities resulted in an increasing drop in force after an initial increase and a final steeper rise in force. Cross-bridges cannot explain the observed force traces. This requires a viscoelastic non-cross-bridge contribution. Considering these results can improve muscle models and predictions of multibody simulations.

Abstract

Population snapshot data can be used to study heterogeneity in cell populations. Various approaches to integrating such data into computational models have been published, which enable new treatment strategies for cancer therapy, by exploiting the intra-tumor heterogeneity. A precision medicine approach for the cure of cancer could benefit from the combination of single-cell data and respective analytical methods. Here, we introduce the inverse transformation algorithm, which transforms population snapshot data to parameter distributions that are consistent with the underlying data given a dynamic model with distributed parameters. Therefore, it enables the assessment of the heterogeneity in and behavior of the whole underlying cell population. In contrast to the frequently used Approximate Bayesian Computation methods for population matching, our algorithm is a non-parametric likelihood-free approach. It directly computes a density function value for a single parameter based on density transformation methods. If the model can be described as a one-to-one map that invertibly maps parameters to measurable outputs, the inverse transformation algorithm asymptotically returns the true underlying parameter distribution. The inverse transformation algorithm is applied to snapshot data simulated via standard differential equation models for biochemical reaction networks. In particular, we evaluate our algorithm on two small test-bed models and discuss advantages and limitations in comparison to other existing approaches.

Abstract

The condition for the formation of droplet groups in liquid sprays is poorly understood. This study looks at a simplified model system consisting of two iso-propanol droplets of equal diameter, Dd0, in tandem, separated initially by a center-to-center distance, a20, and moving in the direction of gravity with an initial velocity, Vd0\>Vt, where Vt is the terminal velocity of an isolated droplet from Stokes flow analysis. A theoretical analysis based on Stokes flow around this double-droplet system is presented, including an inertial correction factor in terms of drag coefficient to account for large Reynolds numbers (&\#8811;1). From this analysis, it is observed that the drag force experienced by the leading droplet is higher than that experienced by the trailing droplet. The temporal evolutions of the velocity, Vd(t), of the droplets, as well as their separation distance, a2(t), are presented, and the time to at which the droplets come in contact with each other and their approach velocity at this time, \ΔVd0, are calculated. The effects of the droplet diameter, Dd0, the initial droplet velocity, Vd0, and the initial separation, a20 on to and \ΔVd0 are reported. The agreement between the theoretical predictions and experimental data in the literature is good.

Abstract

How myofilaments operate at short mammalian skeletal muscle lengths is unknown. A common assumption is that thick (myosin-containing) filaments get compressed at the Z-disc. We provide ultrastructural evidence of sarcomeres contracting down to 0.44 µm—approximately a quarter of thick filament resting length—in long-lasting contractions while apparently keeping a regular, parallel thick filament arrangement. Sarcomeres produced force at such extremely short lengths. Furthermore, sarcomeres adopted a bimodal length distribution with both modes below lengths where sarcomeres are expected to generate force in classic force–length measurements. Mammalian fibres did not restore resting length but remained short after deactivation, as previously reported for amphibian fibres, and showed increased forces during passive re-elongation. These findings are incompatible with viscoelastic thick filament compression but agree with predictions of a model incorporating thick filament sliding through the Z-disc. This more coherent picture of mechanical mammalian skeletal fibre functioning opens new perspectives on muscle physiology

Abstract

Line integration of stream-, streak-, and pathlines is widely used and popular for visualizing single-phase flow. In multiphase flow, i.e., where the fluid consists, e.g., of a liquid and a gaseous phase, these techniques could also provide valuable insights into the internal flow of droplets and ligaments and thus into their dynamics. However, since such structures tend to act as entities, high translational and rotational velocities often obfuscate their detail. As a remedy, we present a method for deriving a droplet-local velocity field, using a decomposition of the original velocity field removing translational and rotational velocity parts, and adapt path- and streaklines. Generally, the resulting integral lines are thus shorter and less tangled, which simplifies their analysis. We demonstrate and discuss the utility of our approach on droplets in two-phase flow data and visualize the removed velocity parts employing glyphs for context.

Abstract

We introduce a system capable of generating interactive Aug-mented Reality guitar tutorials by parsing common digital guitar tablature and by capturing the performance of an expert using a multi-camera array. Instructions are presented to the user in an Augmented Reality application using either an abstract visualization, a 3D virtual hand, or a 3D video. To support individual users at different skill levels the system provides full control of the play-back of a tutorial, including its speed and looping behavior, while delivering live feedback on the user's performance.

Abstract

We present efficient reduced basis (RB) methods for the simulation of a coupled problem consisting of a rigid robot hand interacting with soft tissue material. The soft tissue is modeled by the linear elasticity equation and discretized with the Finite Element Method. We look at two different scenarios: (i) the forward simulation and (ii) a feedback control formulation of the model. In both cases, large-scale systems of equations appear, which need to be solved in real-time. This is essential in practice for the implementation in a real robot. For the feedback-scenario, we encounter a high-dimensional Algebraic Riccati Equation (ARE) in the context of the linear quadratic regulator. To overcome the real-time constraint by significantly reducing the computational complexity, we use several structure-preserving and non-structure-preserving reduction methods. These include reduced basis techniques based on the Proper Orthogonal Decomposition. For the ARE, we compute a low-rank-factor and hence solve a low-dimensional ARE instead of solving a full dimensional problem. Numerical examples for both cases (i) and (ii) are provided. These illustrate the approximation quality of the reduced solution and speedup factors of the different reduction approaches.

Abstract

Models for the co-transport of two different colloids commonly assume a one-way coupling. This is because often a large colloid and small colloid are involved. Therefore, they assume that the spread of smaller colloid is affected by the transport of larger colloids, but not the other way around. However, a number of studies have shown that this assumption is not valid, even for large and small colloids. Therefore, in this study, a two-way coupled model is developed to simulate the co-transport of two different colloids in porous media and their effect on each other. We have considered the interactions of the two colloids with the grain surface, kinetics of heteroaggregation (of the two colloids), and heteroaggregate deposition onto the grain surface. We assumed a first-order kinetic model to represent heteroaggregate formation and its deposition on the grain surface. The model is evaluated by fitting the experimental data reported in four different papers from the literature on the co-transport of clay colloids and viruses, bacteria and graphene oxide nanoparticles, and clay colloids and graphene oxide nanoparticles. The model performance is compared with the commonly-used one-way coupled model. The two-way coupled model is found to satisfactorily simulate most of the experimental conditions reported in the above papers, except for the co-transport of montmorillonite–adenovirus, and Staphylococcus aureus- graphene oxide nanoparticles.

Abstract

This work presents a framework to synthesize structured gain-scheduled controllers for structured plants that are affected by time-varying parametric scheduling blocks. Using a so-called lifting approach, we are able to handle several structured gain-scheduling problems arising from a nested inner and outer loop configuration with partial or full dependence on the scheduling block. Our resulting design conditions are formulated in terms of convex linear matrix inequalities and permit to handle multiple performance objectives.

Abstract

EnzymeML is an XML-based data exchange format that supports the comprehensive documentation of enzymatic data by describing reaction conditions, time courses of substrate and product concentrations, the kinetic model, and the estimated kinetic constants. EnzymeML is based on the Systems Biology Markup Language, which was extended by implementing the STRENDA Guidelines. An EnzymeML document serves as a container to transfer data between experimental platforms, modeling tools, and databases. EnzymeML supports the scientific community by introducing a standardized data exchange format to make enzymatic data findable, accessible, interoperable, and reusable according to the FAIR data principles. An application programming interface in Python supports the integration of software tools for data acquisition, data analysis, and publication. The feasibility of a seamless data flow using EnzymeML is demonstrated by creating an EnzymeML document from a structured spreadsheet or from a STRENDA DB database entry, by kinetic modeling using the modeling platform COPASI, and by uploading to the enzymatic reaction kinetics database SABIO-RK.

Abstract

Abstract The water permeability of aquaporins (AQPs) varies by more than an order of magnitude even though the pore structure, geometry, as well as the channel lining residues are highly conserved. However, channel gating by pH, divalent ions or phosphorylation was only shown for a minority of AQPs. Structural and in silico indications of water flux modulation by flexible side chains of channel lining residues have not been experimentally confirmed yet. Hence, the aquaporin “open state” is still considered to be a continuously open pore with water molecules permeating in a single-file fashion. Using protein mutations outside the selectivity filter in the aqua(glycerol)facilitator GlpF of Escherichia coli we, to the best of our knowledge, for the first time, modulate the position of the highly conserved Arg in the selectivity filter. This in turn enhances or reduces the unitary water permeability of GlpF as shown in silico by molecular dynamics (MD) simulations and in vitro with purified and reconstituted GlpF. This finding suggests that AQP water permeability can indeed be regulated by lipid bilayer asymmetry and the transmembrane potential. Strikingly, our long-term MD simulations reveal that not only the conserved Arg in the selectivity filter, but the position and dynamics of multiple other pore lining residues modulate water passage through GlpF. This finding is expected to trigger a wealth of future investigations on permeability and regulation of AQPs among others with the aim to tune water permeability for biotechnological applications.

Abstract

In protein simulation or structure refinement based on values of observable quantities measured in (aqueous) solution, solvent (water) molecules may be explicitly treated, omitted, or represented by a potential of mean-solvation-force term, depending on protein coordinates only, in the force field used. These three approaches are compared for hen egg white lysozyme (HEWL). This 129-residue non-spherical protein contains a variety of secondary-structure elements, and ample experimental data are available: 1630 atom--atom Nuclear Overhauser Enhancement (NOE) upper distance bounds, 213 3 J-couplings and 200 S2 order parameters. These data are used to compare the performance of the three approaches. It is found that a molecular dynamics (MD) simulation in explicit water approximates the experimental data much better than stochastic dynamics (SD) simulation in vacuo without or with a solvent-accessible-surface-area (SASA) implicit-solvation term added to the force field. This is due to the missing energetic and entropic contributions and hydrogen-bonding capacities of the water molecules and the missing dielectric screening effect of this high-permittivity solvent. Omission of explicit water molecules leads to compaction of the protein, an increased internal strain, distortion of exposed loop and turn regions and excessive intra-protein hydrogen bonding. As a consequence, the conformation and dynamics of groups on the surface of the protein, which may play a key role in protein--protein interactions or ligand or substrate binding, may be incorrectly modelled. It is thus recommended to include water molecules explicitly in structure refinement of proteins in aqueous solution based on nuclear magnetic resonance (NMR) or other experimentally measured data.

Abstract

We present the magnetic Moment Tensor Potentials (mMTPs), a class of machine-learning interatomic potentials, accurately reproducing both vibrational and magnetic degrees of freedom as provided, e.g., from first-principles calculations. The accuracy is achieved by a two-step minimization scheme that coarse-grains the atomic and the spin space. The performance of the mMTPs is demonstrated for the prototype magnetic system bcc iron, with applications to phonon calculations for different magnetic states, and molecular-dynamics simulations with fluctuating magnetic moments.

Abstract

NMTVis is a web-based visual analytics system to analyze, understand, and correct translations generated with neural machine translation. First, a document can be translated using a neural machine translation model (we support an LSTM-based and the Transformer architecture). Afterward, users can find mistranslated sentences, explore and correct these sentences and retrain the model to generate a better translation for the whole document. Our approach targets the correction of domain-specific documents. This extended version of our visual analytics system provides additional visualization and interaction techniques as well as scripts for computer-based evaluation of our approach. You can find important information about our system here and an introduction to our system here.

Abstract

In this work, we present a novel hybrid Discontinuous Galerkin scheme with hp-adaptivity capabilities for the compressible Euler equations. In smooth regions, an efficient and accurate discretization is achieved via local p-adaptation. At strong discontinuities and shocks, a finite volume scheme on an h-refined element-local subgrid gives robustness. Thus, we obtain a hp-adaptive scheme that exploits both the high convergence rate and efficiency of a p-adaptive high order scheme as well as the stable and accurate shock capturing abilities of a low order finite volume scheme, but avoids the inherent resolution loss through h-refinement. A single a priori indicator, based on the modal decay of the local polynomial solution representation, is used to distinguish between discontinuous and smooth regions and control the p-refinement. Our method is implemented as an extension to the open source software FLEXI. Hence, the efficient implementation of the method for high performance computers was an important criterion during the development. The efficiency of our adaptive scheme is demonstrated for a variety of test cases, where results are compared against non adaptive simulations. Our findings suggest that the proposed adaptive method produces comparable or even better results with significantly less computational costs.

Abstract

To model subsurface flow in uncertain heterogeneous or fractured media an elliptic equation with a discontinuous stochastic diffusion coefficient—also called random field—may be used. In case of a one-dimensional parameter space, Lévy processes allow for jumps and display great flexibility in the distributions used. However, in various situations (e.g. microstructure modeling), a one-dimensional parameter space is not sufficient. Classical extensions of Lévy processes on two parameter dimensions suffer from the fact that they do not allow for spatial discontinuities see for example Barth and Stein (Stoch Part Differ Equ Anal Comput 6(2):286–334, 2018). In this paper a new subordination approach is employed see also Barth and Merkle (Subordinated gaussian random fields. ArXiv e-prints, arXiv:2012.06353math.PR, 2020) to generate Lévy-type discontinuous random fields on a two-dimensional spatial parameter domain. Existence and uniqueness of a (pathwise) solution to a general elliptic partial differential equation is proved and an approximation theory for the diffusion coefficient and the corresponding solution provided. Further, numerical examples using a Monte Carlo approach on a Finite Element discretization validate our theoretical results.

Abstract

General elliptic equations with spatially discontinuous diffusion coefficients may be used as a simplified model for subsurface flow in heterogeneous or fractured porous media. In such a model, data sparsity and measurement errors are often taken into account by a randomization of the diffusion coefficient of the elliptic equation which reveals the necessity of the construction of flexible, spatially discontinuous random fields. Subordinated Gaussian random fields are random functions on higher dimensional parameter domains with discontinuous sample paths and great distributional flexibility. In the present work, we consider a random elliptic partial differential equation (PDE) where the discontinuous subordinated Gaussian random fields occur in the diffusion coefficient. Problem specific multilevel Monte Carlo (MLMC) Finite Element methods are constructed to approximate the mean of the solution to the random elliptic PDE. We prove a-priori convergence of a standard MLMC estimator and a modified MLMC—control variate estimator and validate our results in various numerical examples.

Abstract

Motivated by the subordinated Brownian motion, we define a new class of (in general discontinuous) random fields on higher-dimensional parameter domains: the subordinated Gaussian random field. We investigate the pointwise marginal distribution of the constructed random fields, derive a Lévy-Khinchin-type formula and semi-explicit formulas for the covariance function. Further, we study the pointwise stochastic regularity and present various numerical examples.

Abstract

A data-driven modeling framework for non-linear eddy viscosity models is presented. In contrast to the majority of similar approaches, it splits the multivariate regression problem into a set of univariate problems and is, furthermore, independent of a specific machine learning algorithm. Instead of inferring the closure equation from high-fidelity data as a whole, the coefficients of the tensor polynomial are learned individually. The target variables are obtained from an efficient field inversion procedure by virtue of successive tensor projections. The hypotheses for each closure coefficient are then fitted separately by employing an arbitrary regression technique. Ordinary curve fitting, neural networks and gene expression programming are considered as examples. The robustness of the model against an extrapolation and the stability of the solver are promoted by coefficient limiters and a full barycentric realizability correction. Additionally, turbulent scale consistency is guaranteed by a k-corrective frozen-RANS method, which is the subject of a companion paper (Mandler and Weigand, 2022). The proposed strategy leads to models which are well suited for the application to the same class of flows they were inferred from, namely separated channel flows. As proven by an extensive extrapolation study, the resulting neuralSST model is robust against geometry and Reynolds number modifications provided the type of flow does not drastically change. It consistently outperforms both the shear stress transport (SST) and a more complex elliptic-blending model and agrees well with the reference data.

Abstract

This paper addresses a consistency problem in data-driven turbulence modeling, which arises as the hypotheses are inferred from high-fidelity data but evaluated within a low-fidelity RANS solver. After elaborating on its origin, which is the systematic difference of the turbulent scales predicted by a low- and a high-fidelity solver, the frozen-RANS concept is thoroughly discussed as one possible mitigation strategy. Different variations of this concept are proposed varying in the way to incorporate the turbulent kinetic energy correction in the RANS solver and also in their degree to which they fulfill scale consistency. Applying these concepts to the neuralSST model, which is introduced in the companion paper (Mandler and Weigand, 2022), confirms the importance of scale consistency for an improved mean flow field prediction and that the latter can be achieved by a k-correction. This becomes particularly evident as the complexity of the test cases, which are different types of separated channel flows, increases. However, which particular strategy is employed to augment the low-fidelity prediction of the turbulent kinetic energy is of minor importance and the decision is mainly driven by the trade-off between the flexibility of the solver to modify the scale equations and the numerical stability of the resulting model.

Abstract

The multi-scale character of skeletal muscle models requires simulations with high spatial resolution to capture all relevant effects. This naturally involves high computational load that can only be tackled by parallel computations. We simulate electrophysiology and muscle contraction using a state-of-the-art, biophysical chemo-electro-mechanical model that requires meshes of the 3D domain with embedded, aligned 1D meshes for muscle fibers. We present novel algorithms to construct highly-resolved meshes with robust properties for real muscle geometries from surface triangulations. We demonstrate their use and suitability in a simulation of the biceps brachii muscle and tendons. In addition, the respective simulations showcase several functional enhancements of our simulation framework OpenDiHu.

Abstract

The dielectric constant of water/oligomer mixtures, spanning the range from pure water to pure oligomeric melts, is investigated using molecular dynamics (MD) simulations. As prototypical water-soluble organic substances we consider neutral poly-glycine, poly-ethylene glycol and charged monomeric propanic acid. As the water content is reduced, the dielectric constant decreases but does not follow an ideal mixing behavior. The deviations from ideal mixing originate primarily in the non-linear relation between the oligomer mass fraction and collective polarization effects. We find that the dielectric constant is dominated by water polarization, even if the oligomer mass fraction exceeds 50\%. By a double extrapolation of the MD simulation results to the limit of vanishing water fraction and to the limit of infinite oligomeric chain length, we estimate the orientational contribution to the dielectric constant of the pure polymeric melts. By this procedure, we obtain ε = 17 plus minus 2 for polyglycine and ε = 1 plus minus 0.3 for polyethylene glycol. The large difference is rationalized by polarization correlations of glycine units. Interestingly, we find constant temperature simulations to outperform replica exchange simulations in terms of equilibration speed.

Abstract

With more augmented reality (AR)/ devices around, real-time simulation on resource- restricted devices has attracted considerable interest in recent years, and there is greater interest in using them for things like medical visualization. There already exist simula- tion methods such as the Finte Element (FE) method for continuum-biomechanics, which is good for 3D simulation of soft tissue, but is very computationally expensive. In recent years, sparse grid (SG) surrogates have been shown to allow real-time forward sim. using continuum-biomechanical data. Furthermore, a neural network (NN) surrogate has been trained using data interpolated with the sparse grid for real-time vis. using such data. However, that proof-of-concept model is purely densely connected. This work aims to achieve the same or better performance with fewer parameters by a novel NN architecture. This work does so using a four-part model, consisting of densely connected, convolutional and locally connected layers, utilizing a point-cloud-to-voxel grid embedding for the convolutional neural network (CNN) to work. The main focus of this work has been on the process of finding and evaluating such a surrogate model. The use of CNN’s locally connected layers helped to obtain a prescise and lightweight model. In addition, a sophisticated mapping procedure has been developed, to take advantage of a CNN’s architecture into a point-cloud dataformat. The results demonstrated the effectiveness of this method. Depending on the hyperparameter configurations, the surrogate model was able to achieve frame rates of up to 2280 frames per second (fps) and a mean absolute error (mae) of 0.0082mm in the muscle surface coordinates. Furthermore, a notable reduction of trainable parameters, namely, from 43.9mio. to 104, 507 was possible, which is only 0.2% of the original value. The results also provided insight into the hyperparameter configurations that have the greatest impact on the model. In general, activation functions and channel sizes exerted the most influence on model quality. High frame-rates also help reduce the energy consumption of mobile devices powered by batteries. Due to machine learning (ML) being a huge field, this work can’t calim to have found the best solution. However, the current architecture improves upon the prove-of-concept in all desired areas. In the future, the model could be tested and executed on a mobile device or become part of a AR application with dynamic 3D objects.

Abstract

Unitary cluster expansions of the electronic wavefunction have recently gained much interest because of their use in conjunction with quantum algorithms. In this contribution, we investigate some aspects of an ansatz, using generalized two-body excitation operators, which have been considered in some recent studies on quantum algorithms for quantum chemistry. Our numerical results show that, in particular, two-body operators with effective particle–hole excitation level of one in connection with the usual particle–hole double excitation operators lead to a very accurate, yet compact representation of the wavefunction. Generalized two-body operators with effective excitation rank zero have a considerably less pronounced effect. We compare with standard and unitary coupled-cluster expansions and show that the above mentioned approach matches or even surpasses the accuracy of expansions with three-body particle–hole excitations, in particular at the onset of strong correlation. A downside of the approach is that it is rather difficult to rigorously converge it to its variational minimum.

Abstract

Gasdermins execute programmatory cell death, known as pyroptosis, by forming medium-sized membrane pores. Recently, the molecular structure of those pores as well as the diversity in their shape and size have been revealed by cryoTEM and atomic force microscopy, respectively. Even though a growth of smaller to larger oligomers and reshaping from slits to rings could be documented, the initiation of the gasdermin pore formation remains a mystery. In one hypothesis, gasdermin monomers insert into membranes before associating into oligomeric pores. In the other hypothesis, gasdermin oligomers preassemble on the membrane surface prior to membrane insertion. Here, by studying the behavior of monomeric membrane-inserted gasdermin-A3 (GSDMA3), we unveil that a monomeric gasdermin prefers the membrane-adsorbed over the membrane-inserted state. Our results thus support the hypothesis of oligomers preassembling on the membrane surface before membrane penetration. At the same time, our simulations of small membrane-inserted arcs of GSDMA3 suggest that the inserting oligomer can be small and does not have to comprise a full ring of approximately 26--30 subunits. Moreover, our simulations have revealed an astonishingly large impact of salt-bridge formation and protein surroundings on the transmembrane passage of charged residues, reducing the energetic cost by up to 53% as compared to their free forms. The here observed free energy barrier of mere 5.6 kcal/mol for the membrane insertion of monomeric GSDMA3 explains the surprising ability of gasdermins to spontaneously self-insert into cellular membranes.

Abstract

This dataset provides simulation results from a high-fidelity human body model in a pre-crash scenario as well as a surrogate model to approximate those simulation results.The high-fidelity results contain the motion of all points describing the occupant at certain points of time for several scenarios. In each scenario, the accelerations of the vehicle center of gravity act on the model. The translational accelerations in x- and y-direction as well as the angular acceleration around the z-axis are considered.In addition to the occupants motion, the corresponding time vector and used vehicle acceleration are also part of the dataset.The surrogate model was created by combining dimensionality reduction using proper orthogonal decomposition along with a long short-term memory network for regression. A standalone script to evaluate the model is provided as well.

Abstract

In this note it is shown that the famous multiplier absolute stability test of R. O'Shea, G. Zames and P. Falb is necessary and sufficient if the set of Lur'e interconnections is lifted to a Kronecker structure and an explicit method to construct the destabilizing static nonlinearity is presented.

Abstract

This dataset contains results from Grand Canonical Monte Carlo (GCMC) Simulation (data/isotherms_sim/) and experiment (data/isotherms/exp). All Data is presented in a jupyter notebook and for a fast overview without executing the notebook also as pdf-file.Furthermore the dataset contains the modified cif files of COF LZU-1, including partial charges obtained with DDEC method (data/cif/). Force field (data/forcefield/) and input files for raspa-code (data/input_files/) are also available. Results from simulations of material properties e.g. specific surface area can be found in data/material_properties/ and pore size distributions from raspa simulations are listed in data/output_psd/. The folder 'progs' simply contains functions to make the jupyter-notebook clearer.We recommend viewing the data by choosing the option "Tree".

Abstract

The exact knowledge of droplet clustering or grouping is important in many technical applications , such as fuel injection, ink-jet printing or in medical nebulizers. Despite their relevance, the fundamentals of grouping processes are yet to be fully understood. Recent studies comprise the experimental and numerical investigation of single monodisperse streams in well-defined droplet arrangements. Moving towards more complex configurations, such as multiple streams in close proximity, is a challenging task experimentally. This study aims at bridging this gap numerically with the aid of the in-house multiphase code Free Surface 3D to assess the effect of multiple droplet streams on grouping behavior. Three streams of droplets in a coplanar arrangement are simulated for which the center stream is compared to a single reference stream without the influence of neighboring droplets. To determine the impact of the outer streams on the center stream regarding grouping time, relative droplet velocity, drag forces as well as pressure and shear stress distribution, the following input parameters are changed systematically: the initial lateral distance to the center stream and the initial Reynolds number. Results show that the presence of neighboring streams with a lateral distance of less than 5 droplet diameters has a significant influence on the grouping behavior of the central stream. The process can be divided into a first part (I) where the outer streams confine the interaction with the surroundings, slowing down grouping tendencies and reducing drag forces in the middle stream; and a second part (II) where the coalescence of the outer streams imposes a substantial disturbance on the middle stream leading to an abrupt increase in the forces acting upon the droplets. Additionally, the lateral drift of the outside streams is compared qualitatively to literature studies and shows good agreement. For lateral distances of more than 10 droplet diameters, the interaction of the streams is negligible and the behavior of a single, isolated stream is retained.

Abstract

Droplet grouping is important in technical applications and in nature where more than one droplet is seen. Despite its relevance for such problems, the fundamentals of the grouping processes are not yet fully understood. Initial conditions that expedite or impede the formation of droplet groups have been studied, but a thorough investigation of the temporal and spatial evolution of the forces at play has not been conducted. In this work, the grouping process in monodisperse droplet streams is examined in detail by direct numerical simulation (DNS), for the first time, using the multiphase code Free Surface 3D. The code framework is based on the volume-of-fluid method and uses the piecewise linear interface calculation method to reconstruct the interface. A method is established to quantify the development and evolving differences of pressure and shear drag forces on each droplet in the stream using the available DNS data. The results show a linear increase in the difference between the forces, where the drag force on the leading droplet is always larger than that on the trailing droplet. A comprehensive parametric study reveals that, on the one hand, large initial inter-droplet separation and small group distances increase grouping time due to reduced difference in the drag coefficients. On the other hand, higher initial Reynolds numbers and larger irregularities in the geometrical arrangement promote droplet grouping. The flow field shows stable wake structures at initial Reynolds numbers of 300 and the onset of vortex shedding at Reynolds numbers of 500, affecting the next pair of droplets, even for larger separation distances.

Abstract

Virtual avatars are widely used for collaborating in virtual environments. Yet, often these avatars lack expressiveness to determine a state of mind. Prior work has demonstrated effective usage of determining emotions and animated lip movement through analyzing mere audio tracks of spoken words. To provide this information on a virtual avatar, we created a natural audio data set consisting of 17 audio files from which we then extracted the underlying emotion and lip movement. To conduct a pilot study, we developed a prototypical system that displays the extracted visual parameters and then maps them on a virtual avatar while playing the corresponding audio file. We tested the system with 5 participants in two conditions: (i) while seeing the virtual avatar only an audio file was played. (ii) In addition to the audio file, the extracted facial visual parameters were displayed on the virtual avatar. Our results suggest the validity of using additional visual parameters in the avatars’ face as it helps to determine emotions. We conclude with a brief discussion on the outcomes and their implications on future work.

Abstract

Visual representations are used in various digitization processes to display objects in immersive environments. In industrial environments, the self-representation can be very helpful when reviewing 3D models to conform realistic proportions, particularly in combi-nation with haptic feedback that goes beyond vibrating controllers. For instance, haptic feedback in combination with a virtual representation supports working on use cases where collision is an important part to maintain data quality. To understand the dependency of both, we conducted a pilot study with 15 users from the automotive sector to examine the influence of upper limb representations on haptic feedback in Virtual Reality. Each participant was assigned with one of three upper limb representations which was used in different scenarios using haptic feedback devices. Overall, we found that the realistic arm representation was rated highest in terms of perceived realism and achieved the best task performance.

Abstract

We show how to compose robust stability tests for uncertain systems modeled as linear fractional representations and affected by various types of dynamic uncertainties. Our results are formulated in terms of linear matrix inequalities and rest on the recently established notion of finite-horizon integral quadratic constraints with a terminal cost. The construction of such constraints is motivated by an unconventional application of the S-procedure in the frequency domain with dynamic Lagrange multipliers. Our technical contribution reveals how this construction can be performed by dissipativity arguments in the time-domain and in a lossless fashion. This opens the way for generalizations to time-varying or hybrid systems.

Abstract

We establish a framework for systematically analyzing and designing output-feedback controllers for linear impulsive and related hybrid systems that might even be affected by various types of uncertainties. In particular, the framework encompasses uncertain switched and sampled-data systems as well as networked systems with switching communication topologies. The framework is based on recently developed convex criteria involving a so-called clock for analyzing impulsive systems under dwell-time constraints. We elaborate on the extension of those criteria for dynamic output-feedback controller synthesis by means of convex optimization and generalize the so-called dual iteration to impulsive systems. The latter originally and still constitutes a promising heuristic procedure for the challenging and non-convex design of static output-feedback controllers for standard linear time-invariant systems. Moreover, for uncertain impulsive systems as modeled in terms of linear fractional representations, we generalize the nominal analysis criteria by providing novel robust analysis conditions based on a novel time-domain and clock-dependent formulation of integral quadratic constraints. Finally, by combining the insights on nominal synthesis and robust analysis, we are able to tackle challenging output-feedback designs of practical relevance, such as the design of gain-scheduled, robust or robust gain-scheduled controllers for impulsive systems. Most of the obtained analysis and synthesis conditions involve infinite-dimensional (differential) linear matrix inequalities which can be numerically solved by using relaxation methods based on, e.g., linear splines, B-splines or matrix sum-of-squares that we discuss as well.

Abstract

Prediction error quantification in machine learning has been left out of most methodological investigations of neural networks, for both purely data-driven and physics-informed approaches. Beyond statistical investigations and generic results on the approximation capabilities of neural networks, we present a rigorous upper bound on the prediction error of physics-informed neural networks. This bound can be calculated without the knowledge of the true solution and only with a priori available information about the characteristics of the underlying dynamical system governed by a partial differential equation. We apply this a posteriori error bound exemplarily to four problems: the transport equation, the heat equation, the Navier-Stokes equation and the Klein-Gordon equation.

Abstract

Abstract Evolution turned aquaporins (AQPs) into the most efficient facilitators of passive water flow through cell membranes at no expense of solute discrimination. In spite of a plethora of solved AQP structures, many structural details remain hidden. Here, by combining extensive sequence- and structural-based analysis of a unique set of 20 non-redundant high-resolution structures and molecular dynamics simulations of four representatives, key aspects of AQP stability, gating, selectivity, pore geometry, and oligomerization, with a potential impact on channel functionality, are identified. The general view of AQPs possessing a continuous open water pore is challenged and it is depicted that AQPs’ selectivity is not exclusively shaped by pore-lining residues but also by the relative arrangement of transmembrane helices. Moreover, this analysis reveals that hydrophobic interactions constitute the main determinant of protein thermal stability. Finally, a numbering scheme of the conserved AQP scaffold is established, facilitating direct comparison of, for example, disease-causing mutations and prediction of potential structural consequences. Additionally, the results pave the way for the design of optimized AQP water channels to be utilized in biotechnological applications.

Abstract

Under ambient atmospheric conditions, a thin film of water wets many solid surfaces, including insulators, ice, and salt. The film thickness as well as its transport behavior sensitively depend on the surrounding humidity. Understanding this intricate interplay is of highest relevance for water transport through porous media, particularly in the context of soil salinization induced by evaporation. Here, we use molecular simulations to evaluate the transport properties of thin water films on prototypical salt and soil interfaces, namely NaCl and silica solid surfaces. Our results showtwo distinct regimes for water transport: at low water coverage, the film permeance scales linearly with the adsorbed amount, in agreement with the activated random walk model.For thicker water films, the permeance scales as the adsorbed amount to the power of 3, in line with the Stokes equation. By comparing results obtained for silica and NaCl surfaces, we find that, at low water coverage, water permeance at the silica surface is considerably lower than at the NaCl surface, which we attribute to difference in hydrogen bonding. We also investigate the effect of atomic surface defects on the transport properties. Finally, in the context of water transport through porous material, we determine the humidity-dependent crossover between a vapor dominated and a thin film dominated transport regimes depending on the pore size.

Abstract

This paper considers the problem of designing accelerated gradient-based algorithms for optimization and saddle-point problems. The class of objective functions is defined by a generalized sector condition. This class of functions contains strongly convex functions with Lipschitz gradients but also non-convex functions, which allows not only to address optimization problems but also saddle-point problems. The proposed design procedure relies on a suitable class of Lyapunov functions and on convex semi-definite programming. The proposed synthesis allows the design of algorithms that reach the performance of state-of-the-art accelerated gradient methods and beyond.

Abstract

Several environmental phenomena require monitoring time-dependent densities in porous media, e.g., clogging of river sediments, mineral dissolution/precipitation, or variably-saturated multiphase flow. Gamma-ray attenuation (GRA) can monitor time-dependent densities without being destructive or invasive under laboratory conditions. GRA sends gamma rays through a material, where they are attenuated by photoelectric absorption and then recorded by a photon detector. The attenuated intensity of the emerging beam relates to the density of the traversed material via Beer--Lambert's law. An important parameter for designing time-variable GRA is the exposure time, the time the detector takes to gather and count photons before converting the recorded intensity to a density. Large exposure times capture the time evolution poorly (temporal raster error, inaccurate temporal discretization), while small exposure times yield imprecise intensity values (noise-related error, i.e. small signal-to-noise ratio). Together, these two make up the total error of observing time-dependent densities by GRA. Our goal is to provide an optimization framework for time-dependent GRA experiments with respect to exposure time and other key parameters, thus facilitating neater experimental data for improved process understanding. Experimentalists set, or iterate over, several experimental input parameters (e.g., Beer--Lambert parameters) and expectations on the yet unknown dynamics (e.g., mean and amplitude of density and characteristic time of density changes). We model the yet unknown dynamics as a random Gaussian Process to derive expressions for expected errors prior to the experiment as a function of key experimental parameters. Based on this, we provide an optimization framework that allows finding the optimal (minimal-total-error) setup and demonstrate its application on synthetic experiments.

Abstract

We contribute MolecuSense, a virtual version of a physical molecule construction kit, based on visualization in Virtual Reality (VR) and interaction with force-feedback gloves. Targeting at chemistry education, our goal is to make virtual molecule structures more tangible. Results of an initial user study indicate that the VR molecular construction kit was positively received. Compared to a physical construction kit, the VR molecular construction kit is on the same level in terms of natural interaction. Besides, it fosters the typical digital advantages though, such as saving, exporting, and sharing of molecules. Feedback from the study participants has also revealed potential future avenues for tangible molecule visualizations.

Abstract

We present an integrated approach for the use of simulated data from full order discretization as well as projection-based Reduced Basis reduced order models for the training of machine learning approaches, in particular Kernel Methods, in order to achieve fast, reliable predictive models for the chemical conversion rate in reactive flows with varying transport regimes.

Abstract

Interphase mass transfer is an important solute transport process in two-phase flow in porous media. During two-phase flow, hydrodynamically stagnant and flowing zones are formed, with the stagnant ones being adjacent to the interfaces through which the interphase mass transfer happens. Due to the existence of these stagnant zones in the vicinity of the interface, the mass transfer coefficient decreases to a certain extent. There seems to be a phenomenological correlation between the mass transfer coefficient and the extent of the stagnant zone which, however, is not yet fully understood. In this study, the phase-field method-based continuous species transfer model is applied to simulate the interphase mass transfer of a dissolved species from the immobile, residual, non-aqueous phase liquid (NAPL) to the flowing aqueous phase. Both scenarios, this of a simple cavity and this of a porous medium, are investigated. The effects of flow rates on the mass transfer coefficient are significantly reduced when the stagnant zone and the diffusion length are larger. It is found that the stagnant zone saturation can be a proxy of the overall diffusion length of the terminal menisci in the porous medium system. The early-stage mass transfer coefficient continuously decreases due to the depletion of the solute in the small NAPL clusters that are in direct contact with the flowing water. The long-term mass transfer mainly happens on the interfaces associated with large NAPL clusters with larger diffusion lengths, and the mass transfer coefficient is mainly determined by the stagnant zone saturation.

Abstract

Recurrent neural networks are capable of learning the dynamics of an unknown nonlinear system purely from input-output measurements. However, the resulting models do not provide any stability guarantees on the input-output mapping. In this work, we represent a recurrent neural network as a linear time-invariant system with nonlinear disturbances. By introducing constraints on the parameters, we can guarantee finite gain stability and incremental finite gain stability. We apply this identification method to learn the motion of a four-degrees-of-freedom ship that is moving in open water and compare it against other purely learning-based approaches with unconstrained parameters. Our analysis shows that the constrained recurrent neural network has a lower prediction accuracy on the test set, but it achieves comparable results on an out-of-distribution set and respects stability conditions.

Abstract

Mechanical metamaterials such as open- and closed-cell lattice structures, foams, composites, and so forth can often be parametrized in terms of their microstructural properties, for example, relative densities, aspect ratios, material, shape, or topological parameters. To model the effective constitutive behavior and facilitate efficient multiscale simulation, design, and optimization of such parametric metamaterials in the finite deformation regime, a machine learning-based constitutive model is presented in this work. The approach is demonstrated in application to elastic beam lattices with cubic anisotropy, which exhibit highly nonlinear effective behaviors due to microstructural instabilities and topology variations. Based on microstructure simulations, the relevant material and topology parameters of selected cubic lattice cells are determined and training data with homogenized stress-deformation responses is generated for varying parameters. Then, a parametric, hyperelastic, anisotropic constitutive model is formulated as an artificial neural network, extending a recent work of the author extending a recent work of the author, Comput Mech., 2021;67(2):653-677. The machine learning model is calibrated with the simulation data of the parametric unit cell. The authors offer public access to the simulation data through the GitHub repository https://github.com/CPShub/sim-data. For the calibration of the model, a dedicated sample weighting strategy is developed to equally consider compliant and stiff cells and deformation scenarios in the objective function. It is demonstrated that this machine learning model is able to represent and predict the effective constitutive behavior of parametric lattices well across several orders of magnitude. Furthermore, the usability of the approach is showcased by two examples for material and topology optimization of the parametric lattice cell.

Abstract

The motion of two dissimilar spherical non-evaporating liquid droplets moving vertically along their line of centers is analysed mathematically. The stimulus for the study is evidence from the literature that, under appropriate operating conditions, droplet grouping can occur. Carefully controlled experiments, aimed at providing a paradigm for understanding this phenomenon, considered the configuration of the current work. The governing non-linear ordinary differential equation describing the behaviour of the distance between the droplets is solved analytically for the first time, using the homotopy analysis method. In order to successfully implement the latter, the fact that the two consecutive droplets considered are mildly different in size was exploited. This enabled a considerable reduction of the complicated expressions for the drag forces acting on the two droplets. Whilst retaining generality, further simplification was achieved using curve fitting for some of the key expressions appearing in the reduced form of the drag forces. The resulting nonlinear equation was then tractable for straightforward application of the homotopy analysis method. For almost identical droplets, validation of the solution is afforded by experimental data from the literature. Comparison between the predictions of the new analytical solution and a numerical solution of the relevant ODE yielded excellent agreement thereby ratifying the proposed combined approach. The results indicate that the larger droplet of the pair approaches the leading droplet more slowly than if the two droplets are equal in size.

Abstract

Symplectic model order reduction is a structure-preserving reduction technique for Hamiltonian systems. Apart from theoretical results like the preservation of stability, it has been demonstrated to give improved numerical results compared to classical MOR techniques. A key element in this procedure is the choice of a good symplectic reduced order basis (ROB). In our work, we introduce so-called canonizable Hamiltonian systems in energy coordinates. For such systems with the assumption of a periodic solution, we derive a globally optimal symplectic ROB in the sense of the proper symplectic decomposition (PSD). To this end, we show that the proper orthogonal decomposition (POD) of a canonizable Hamiltonian system is automatically symplectic from which we deduce optimality of the PSD. To verify our findings numerically, we consider a reproduction experiment for the linear wave equation.

Abstract

The framework of port-Hamiltonian (pH) systems is a powerful and broadly applicable modeling paradigm. In this paper, we extend the scope of pH systems to time-delay systems. Our definition of a delay pH system is motivated by investigating the Kalman-Yakubovich-Popov inequality on the corresponding infinite-dimensional operator equation. Moreover, we show that delay pH systems are passive and closed under interconnection. We describe an explicit way to construct a Lyapunov-Krasovskii functional and discuss implications for delayed feedback.

Abstract

The present study proposes a methodology for predicting the vertical light nonaqueous-phase liquids (LNAPLs) distribution within an aquifer by considering the influence of water table fluctuations. The LNAPL distribution is predicted by combining (1) information on air/LNAPL and LNAPL/water interface elevations with (2) the initial elevation of the water table without LNAPL effect. Data used in the present study were collected during groundwater monitoring undertaken over a period of 4 months at a LNAPL-impacted observation well. In this study, the water table fluctuations raised the free LNAPL in the subsurface to an elevation of 206.63 m, while the lowest elevation was 205.70 m, forming a thickness of 0.93 m of LNAPL-impacted soil. Results show that the apparent LNAPL thickness in the observation well is found to be three times greater than the actual free LNAPL thickness in soil; a finding that agrees with previous studies reporting that apparent LNAPL thickness in observation wells typically exceeds the free LNAPL thickness within soil by a factor estimated to range between 2 and 10. The present study provides insights concerning the transient variation of LNAPL distribution within the subsurface and highlights the capability of the proposed methodology to mathematically predict the actual LNAPL thickness in the subsurface, without the need to conduct laborious field tests. Practitioners can use the proposed methodology to determine by how much the water table should be lowered, through pumping, to isolate the LNAPL-impacted soil within the unsaturated zone, which can then be subjected to in situ vadose zone remedial treatment.

Abstract

As an extension to the well-established stationary elliptic partial differential equation (PDE) with random continuous coefficients we study a time-dependent advection-diffusion problem, where the coefficients may have random spatial discontinuities. In a subsurface flow model, the randomness in a parabolic equation may account for insufficient measurements or uncertain material procurement, while the discontinuities could represent transitions in heterogeneous media. Specifically, a scenario with coupled advection and diffusion coefficients that are modeled as sums of continuous random fields and discontinuous jump components are considered. The respective coefficient functions allow a very flexible modeling, however, they also complicate the analysis and numerical approximation of the corresponding random parabolic PDE. We show that the model problem is indeed well-posed under mild assumptions and show measurability of the pathwise solution. For the numerical approximation we employ a sample-adapted, pathwise discretization scheme based on a finite element approach. This semi-discrete method accounts for the discontinuities in each sample, but leads to stochastic, finite-dimensional approximation spaces. We ensure measurability of the semi-discrete solution, which in turn enables us to derive moments bounds on the mean-squared approximation error. By coupling this semi-discrete approach with suitable coefficient approximation and a stable time stepping, we obtain a fully discrete algorithm to solve the random parabolic PDE. We provide an overall error bound for this scheme and illustrate our results with several numerical experiments.

Abstract

This dataset contains data of 125 1-hour simulations of ship motion during various sea states performing random maneuvers in 4 degrees of freedom (surge-sway-yaw-roll). The original ship is a patrol ship developed by Perez et al. 1. We have extended it with a set of two symmetrically placed rudder propellers. Additionally, we simulate wind forces according to Isherwood's wind model 2. Wind-induced waves are generated with the JONSWAP spectrum 3 and the corresponding wave forces are then computed using wave force response amplitude operators (ROA).Implementations of the ship model, Isherwood's wave model, wave force ROAs and the JONSWAP spectrum can be found in the Marine Systems Simulator toolbox by Fossen and Perez 4.The dataset is split into a routine operation set (96 hours) and into an Out-Of-Distribution (OOD) set (29 hours). The routine operation set is split into train-validation-test with a 60-10-30 split, while the OOD set is used solely for testing.The dataset is used for the evaluation of nonlinear system identification methods for multi-step predictions. The following inputs and outputs are considered for the identification problem. Inputs are the shaft speeds of both propellers, their azimut angles, wind angle of attack, and wind speed. Measured states or outputs are surge velocity, sway velocity and roll rate, as well as yaw angle and roll angle.Please see the README.txt file for details regarding the file structure of this dataset and a description of the variables in the .tab files.This research is funded by Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) under Germany's Excellence Strategy - EXC 2075 - 390740016. We acknowledge the support by the Stuttgart Center for Simulation Science (SimTech).1 T. Perez, A. Ross, and T. I. Fossen, “A 4-DOF SIMULINK model of acoastal patrol vessel for manoeuvring in waves,” in IFAC MCMC, 2006.2 R. M. Isherwood, “Wind resistance of merchant ships,” The RoyalInstitution of Naval Architects, 1972.3 K. Hasselmann, T. Barnett, E. Bouws, H. Carlson, D. Cartwright, K. Enke,J. Ewing, H. Gienapp, D. Hasselmann, P. Kruseman, A. Meerburg,P. Muller, D. Olbers, K. Richter, W. Sell, and H. Walden, “Measurementsof wind-wave growth and swell decay during the joint north sea waveproject (JONSWAP),” Deut. Hydrogr. Z., vol. 8, pp. 1-95, 01 1973.4 T. I. Fossen and T. Perez, “Marine Systems Simulator (MSS),” https://github.com/cybergalactic/MSS, 2004, last accessed: 2022-06-14.

Abstract

In a previous study, acoustic waves were shown to have the potential to induce grouping of particles in exhaust systems 1. In the present study, three-dimensional CFD simulations of single isopropanol droplet streams moving in air subjected to an acoustic field were performed with ANSYS Fluent. The Eulerian-Lagrangian approach was utilized based on the classical work of Gor’kov 2, 3 with the generalization of Settnes and Bruus 4. The simulations reveal a grouping dynamics behavior of the droplets in acoustic standing waves (ASWs). The distances between the droplets in the stream are mainly affected by the droplet residence time which is the time a droplet spends in the ASW. The grouping dynamics also depend on the number of pressure nodes in the field.

Abstract

TET dioxygenases convert 5-methylcytosine (5mC) preferentially in a CpG context into 5-hydroxymethylcytosine (5hmC) and higher oxidized forms, thereby initiating DNA demethylation, but details regarding the effects of the DNA sequences flanking the target 5mC site on TET activity are unknown. We investigated oxidation of libraries of DNA substrates containing one 5mC or 5hmC residue in randomized sequence context using single molecule readout of oxidation activity and sequence and show pronounced 20 and 70-fold flanking sequence effects on the catalytic activities of TET1 and TET2, respectively. Flanking sequence preferences were similar for TET1 and TET2 and also for 5mC and 5hmC substrates. Enhanced flanking sequence preferences were observed at non-CpG sites together with profound effects of flanking sequences on the specificity of TET2. TET flanking sequence preferences are reflected in genome-wide and local patterns of 5hmC and DNA demethylation in human and mouse cells indicating that they influence genomic DNA modification patterns in combination with locus specific targeting of TET enzymes.

Abstract

The typical characteristic of a thin porous layer is that its thickness is much smaller than its in-plane dimensions. This often leads to physical behaviors that are different from three-dimensional porous media. The classical Richards equation is insufficient to simulate many flow conditions in thin porous media. Here, we have provided an alternative approach by accounting for the dynamic capillarity effect. In this study, we have presented a set of one-dimensional in-plane imbibition and subsequent drainage experiments in a thin fibrous layer. The X-ray transmission method was used to measure saturation distributions along the fibrous sample. We simulated the experimental results using Richards equation either with classical capillary equation or with a so-called dynamic capillarity term. We have found that the standard Richards equation was not able to simulate the experimental results, and the dynamic capillarity effect should be taken into account in order to model the spontaneous imbibition. The experimental data presented here may also be used by other researchers to validate their models.

Abstract

In this work, we present a novel higher-order smooth artificial viscosity method for the discontinuous Galerkin spectral element method and related high order methods. A neural network is used to detect the need for stabilization. Inspired by techniques from image edge detection, the neural network locates discontinuities inside mesh elements on a sub-cell level. Once the sub-cell positions of the shock fronts have been identified, the use of radial basis functions enables the construction of a high order smooth artificial viscosity field on quadrilateral meshes. We show the superiority of using higher order smooth artificial viscosity over piecewise linear approaches in particular on coarse meshes. The capabilities of the novel method are illustrated with typical applications.

Abstract

Artificial neural networks (NNs) are one of the most frequently used machine learning approaches to construct interatomic potentials and enable efficient large-scale atomistic simulations with almost ab initio accuracy. However, the simultaneous training of NNs on energies and forces, which are a prerequisite for, e.g., molecular dynamics simulations, can be demanding. In this work, we present an improved NN architecture based on the previous GM-NN model Zaverkin V.; Kästner, J. J. Chem. Theory Comput. 2020, 16, 5410−5421, which shows an improved prediction accuracy and considerably reduced training times. Moreover, we extend the applicability of Gaussian moment-based interatomic potentials to periodic systems and demonstrate the overall excellent transferability and robustness of the respective models. The fast training by the improved methodology is a prerequisite for training-heavy workflows such as active learning or learning-on-the-fly.

Abstract

We perform a numerical and experimental study of immiscible two-phase flows within predominantly 2D transparent PDMS microfluidic domains with disordered pillar-like obstacles, that effectively serve as artificial porous structures. Using a high sensitivity pressure sensor at the flow inlet, we capture experimentally the pressure dynamics under fixed flow rate conditions as the fluid–fluid interface advances within the porous domain, while also monitoring the corresponding phase distribution patterns using optical microscopy. Our experimental study covers 4 orders of magnitude with respect to the injection flow rate and highlights the characteristics of immiscible displacement processes during the transition from the capillarity-controlled interface displacement regime at lower flow rates, where the pores are invaded sequentially in the form of Haines jumps, to the viscosity-dominated regime, where multiple pores are invaded simultaneously. In the capillary regime, we recover a clear correlation between the recorded inlet pressure and the pore-throat diameter invaded by the interface that follows the Young–Laplace equation, while during the transition to the viscous regime such a correlation is no longer evident due to multiple pore-throats being invaded simultaneously (but also due to significant viscous pressure drop along the inlet and outlet channels, that effectively mask capillary effects). The performed experimental study serves for the validation of a robust Level-Set model capable of explicitly tracking interfacial dynamics at sub-pore scale resolutions under identical flow conditions. The numerical model is validated against both well-established theoretical flow models, that account for the effects of viscous and capillary forces on interfacial dynamics, and the experimental results obtained using the developed microfluidic setup over a wide range of capillary numbers. Our results show that the proposed numerical model recovers very well the experimentally observed flow dynamics in terms of phase distribution patterns and inlet pressures, but also the effects of viscous flow on the apparent (i.e. dynamic) contact angles in the vicinity of the pore walls. For the first time in the literature, this work clearly shows that the proposed numerical approach has an undoubtable strong potential to simulate multiphase flow in porous domains over a wide range of Capillary numbers.

Abstract

Many engineering technologies such as electronic cooling and thermal desalination exemplify the enhancement of evaporation and boiling heat transfer by surface modification. Nevertheless, the core parameters of heat transfer such as critical heat flux and heat transfer coefficient are associated with surface wettability and morphology. Herein, for the first time, a metal–organic framework (MOF) film, viz. HKUST-1, was integrated into a metallic woven mesh (macroporous support) for enhancing liquid rewetting and capillary-driven evaporation and boiling heat transfer. Compared to bare copper mesh, this architecture was found to significantly increase the critical heat flux by 205% and the heat transfer coefficient by 90%. The complex coupled two-phase (liquid and gas) transport process involving capillary wicking, evaporation, adsorption and desorption were critically examined by analysing the dynamics of multiple interfaces during horizontal wicking. Relying upon visible colorimetric changes, HKUST-1 sustained on the copper woven mesh could expedite quantitative analysis of the coupled capillary evaporation process. In principle, this is primed to offer fundamental insights into the mechanisms of transport phenomena. Introduction of such previously unreported hierarchical porous structures could also potentially advance the state-of-the-art of passive thermal management technologies. In essence, a new route to elicit superhydrophilic surfaces emerges, paving new ways for understanding the intrinsic mechanisms of phase-change heat transfer.

Abstract

Enzymatically induced calcite precipitation (EICP) in porous media can be used as an engineering option to achieve precipitation in the pore space, for example, aiming at a targeted sealing of existing flow paths. This is accomplished through a porosity and consequent permeability alteration. A major source of uncertainty in modeling EICP is in the quantitative description of permeability alteration due to precipitation. This report presents methods for investigating experimentally the time-resolved effects of growing precipitates on porosity and permeability on the pore scale, in a poly-di-methyl-siloxane microfluidic flow cell. These methods include the design and production of the microfluidic cells, the preparation and usage of the chemical solutions, the injection strategy, and the monitoring of pressure drops for given fluxes for the determination of permeability. EICP imaging methods are explained, including optical microscopy and X-ray microcomputed tomography (XRCT), and the corresponding image processing and analysis. We present and discuss a new experimental procedure using a microfluidic cell, as well as the general perspectives for further experimental and numerical simulation studies on induced calcite precipitation. The results of this study show the enormous benefits and insights achieved by combining both light microscopy and XRCT with hydraulic measurements in microfluidic chips. This allows for a quantitative analysis of the evolution of precipitates with respect to their size and shape, while monitoring their influence on permeability. We consider this to be an improvement of the existing methods in the literature regarding the interpretation of recorded data (pressure, flux, and visualization) during pore morphology alteration.

Abstract

Turbulent channel flow with a porous wall is investigated using direct numerical simulation, where the porous media domain consists of regular or random circular cylinder arrays. We compare the statistics and structure of the mean flow and turbulence in the channel flow with a bulk Reynolds number of 2500 and two porosities (u ¼ 0:6 and 0.8) for the porous media. It is shown that the random interface significantly affects the dynamics of turbulence and the time-averaged flow. More intense mixing is observed near the random interface due to augmented form-induced shear stresses. Due to the strong dependence of induced flow direction on the interface geometry, we segmented the flow field into two types of areas based on the slope angle formed by the top-layer cylinders: the windward area and leeward area. The conditional average of turbulence kinematic energy budget over each type of area reveals their respective role in turbulence transportation more explicitly. In addition, we use finite-time Lyapunov exponents to inspect the Lagrangian coherent structures in the flow fields, which reveal the preferential fluid trajectories in the random porous medium geometry.

Abstract

The interaction between the flow above and below a permeable wall is a central topic in the study of porous media. While previous investigations have provided compelling evidence of the strong coupling between the two regions, few studies have quantitatively measured the directionality, i.e. cause-and-effect relations, of this interaction. To shed light on the problem, interface-resolved direct numerical simulations of channel flow over a cylinder array for porosity $=0.4$–$0.9$ are performed, and the friction Reynolds number of the top smooth wall is controlled to be $Re_180$. We use transfer entropy as a marker to evaluate the causal interaction between the free turbulent flow and the porous medium. Correlation analysis and linear coherence spectra are also leveraged to complete the study. Our results show that the permeability of the porous medium has a profound impact on the intensity, time scale and spatial extent of surface–subsurface interactions. The interaction of the free flow and porous medium is further decomposed into two coupling directions, namely, top-down and bottom-up. For low-porosity cases, top-down and bottom-up interactions are strongly asymmetric, the former being mostly influenced by small near-wall eddies, and the latter reflecting the onset of Kelvin–Helmholtz type instabilities due to the perturbation from the porous medium. As the porosity increases, both top-down and bottom-up interactions are dominated by shear-flow instabilities.

Abstract

The intrinsic permeability is a crucial parameter to characterise and quantify fluid flow through porous media. However, this parameter is typically uncertain, even if the geometry of the pore structure is available. In this paper, we perform a comparative study of experimental, semi-analytical and numerical methods to calculate the permeability of a regular porous structure. In particular, we use the Kozeny–Carman relation, different homogenisation approaches (3D, 2D, very thin porous media and pseudo 2D/3D), pore-scale simulations (lattice Boltzmann method, Smoothed Particle Hydrodynamics and finite-element method) and pore-scale experiments (microfluidics). A conceptual design of a periodic porous structure with regularly positioned solid cylinders is set up as a benchmark problem and treated with all considered methods. The results are discussed with regard to the individual strengths and limitations of the used methods. The applicable homogenisation approaches as well as all considered pore-scale models prove their ability to predict the permeability of the benchmark problem. The underestimation obtained by the microfluidic experiments is analysed in detail using the lattice Boltzmann method, which makes it possible to quantify the influence of experimental setup restrictions.

Abstract

We introduce an ML-driven approach that enables interactive example-based queries for similar behavior in ensembles of spatiotemporal scientific data. This addresses an important use case in the visual exploration of simulation and experimental data, where data is often large, unlabeled and has no meaningful similarity measures available. We exploit the fact that nearby locations often exhibit similar behavior and train a Siamese Neural Network in a self-supervised fashion, learning an expressive latent space for spatiotemporal behavior. This space can be used to find similar behavior with just a few user-provided examples. We evaluate this approach on several ensemble datasets and compare with multiple existing methods, showing both qualitative and quantitative results.

Abstract

We present implicit visualization of 2D vector field topology, and show its utility for validating and guiding approaches for periodic orbit extraction. Instead of following the traditional approach by explicit extraction of the topological skeleton, we investigate its implicit visualization by approaches that label the regions that are separated by the skeleton. While such approaches perform well for gradient fields, they fail, in particular, to visualize periodic orbits. This motivates us to complement the label-based approach with a closely related distance-based metric. We show that our approach is able to reveal periodic orbits, also in configurations in which the state-of-the-art techniques for periodic orbit extraction fail, and demonstrate their utility for interactive extraction of all periodic orbits of a 2D vector field.

Abstract

The scenario of an impacting drop onto a film is highly relevant in many natural and technical systems. A fundamental and often required parameter of these interactions is the so called splashing threshold above which secondary droplets are generated. For interactions with differing liquid properties for the film and the impacting drop a general splashing threshold is, however, still unknown because an experimental determination is difficult to achieve. For this reason, we investigate the suitability of a numerical determination of this threshold by means of direct numerical simulation using the multiphase flow solver Free Surface 3D (FS3D). Simulations across an already existing splashing threshold are performed stemming from an empirical correlation. In order to determine the necessary grid resolution for accurately reproducing the corresponding impact regime, all interactions are simulated for several grids. A detailed grid study shows that only by using very high grid resolutions the threshold can be reproduced with a sufficient accuracy, whereas the use of coarser resolutions leads to a significant underestimation of the threshold. Additionally, simulations of highly resolved impact phenomena on thin films depend heavily on the efficient solution of the problem with most of the computational costs affiliated to solving the Pressure Poisson Equation within the FS3D framework. Therefore, the implemented multigrid solver was optimized employing advanced tree structured communication during coarsening and refinement on the levels during the solution cycle. A performance analysis of FS3D using the original and the improved multigrid solver shows that the implemented tree structured communication leads to a remarkable speed-up.

Abstract

Values of 3J-couplings as obtained from NMR experiments on proteins cannot easily be used to determine protein structure due to the difficulty of accounting for the high sensitivity of intermediate 3J-coupling values (4--8 Hz) to the averaging period that must cover the conformational variability of the torsional angle related to the 3J-coupling, and due to the difficulty of handling the multiple-valued character of the inverse Karplus relation between torsional angle and 3J-coupling. Both problems can be solved by using 3J-coupling time-averaging local-elevation restraining MD simulation. Application to the protein hen egg white lysozyme using 213 backbone and side-chain 3J-coupling restraints shows that a conformational ensemble compatible with the experimental data can be obtained using this technique, and that accounting for averaging and the ability of the algorithm to escape from local minima for the torsional angle induced by the Karplus relation, are essential for a comprehensive use of 3J-coupling data in protein structure determination.

Abstract

Summary This paper proposes the novel past-faults fault prediction algorithm Linespots, based on the Bugspots algorithm. We analyse the predictive performance and runtime of Linespots compared with Bugspots with an empirical study using the most significant self-built dataset as of now, including high-quality samples for validation. As a novelty in fault prediction, we use Bayesian data analysis and Directed Acyclic Graphs to model the effects. We found consistent improvements in the predictive performance of Linespots over Bugspots for all seven evaluation metrics. We conclude that Linespots should be used over Bugspots in all cases where no real-time performance is necessary.

Abstract

Of relevance to energy storage, electrochemistry and catalysis, ionic and dipolar liquids display unexpected behaviours---especially in confinement. Beyond adsorption, over-screening and crowding effects, experiments have highlighted novel phenomena, such as unconventional screening and the impact of the electronic nature---metallic versus insulating---of the confining surface. Such behaviours, which challenge existing frameworks, highlight the need for tools to fully embrace the properties of confined liquids. Here we introduce a novel approach that involves electronic screening while capturing molecular aspects of interfacial fluids. Although available strategies consider perfect metal or insulator surfaces, we build on the Thomas--Fermi formalism to develop an effective approach that deals with any imperfect metal between these asymptotes. Our approach describes electrostatic interactions within the metal through a `virtual' Thomas--Fermi fluid of charged particles, whose Debye length sets the screening length $łambda$. We show that this method captures the electrostatic interaction decay and electrochemical behaviour on varying $łambda$. By applying this strategy to an ionic liquid, we unveil a wetting transition on switching from insulating to metallic conditions.

Abstract

We present a convex solution for the design of generalized accelerated gradient algorithms for strongly convex objective functions with Lipschitz continuous gradients. We utilize integral quadratic constraints and the Youla parameterization from robust control theory to formulate a solution of the algorithm design problem as a convex semidefinite program. We establish explicit formulas for the optimal convergence rates and extend the proposed synthesis solution to extremum control problems.

Abstract

We consider a class of stochastic evolution equations that include in particular the stochastic Camassa–Holm equation. For the initial value problem on a torus, we first establish the local existence and uniqueness of pathwise solutions in the Sobolev spaces Hs with s>3/2. Then we show that strong enough nonlinear noise can prevent blow-up almost surely. To analyze the effects of weaker noise, we consider a linearly multiplicative noise with non-autonomous pre-factor. Then, we formulate precise conditions on the initial data that lead to global existence of strong solutions or to blow-up. The blow-up occurs as wave breaking. For blow-up with positive probability, we derive lower bounds for these probabilities. Finally, the blow-up rate of these solutions is precisely analyzed.

Abstract

In this letter, we investigate the verification of dissipativity properties for polynomial systems without an explicitly identified model but directly from noise-corrupted measurements. Contrary to most data-driven approaches for nonlinear systems, we determine dissipativity properties over all finite time horizons using noisy input-state data. To this end, we propose two noise characterizations to deduce two data-based set-membership representations of the ground-truth system. Each representation then serves as a framework to derive computationally tractable conditions to verify dissipativity properties with rigorous guarantees from noise-corrupted data using sum of squares (SOS) optimization.

Abstract

Medical inhalers have been used for the treatment of a wide range of respiratory diseases including COVID-19. In this study, we investigate the annular liquid jet breakup with a coaxial supersonic gas jet using large eddy simulations. This contributes to a further understanding and improvement of medical inhaler designs. The liquid is sucked in by the low pressure as a result of the high velocity gas jet and breaks up due to the interaction with the gas jet. This type of spray nozzle configuration is commonly used in medical inhalers. Two different gas nozzle diameters are studied. The simulated liquid structure is compared with preliminary, qualitative experimental results. The gas jet pressure and radial velocity of the liquid are found to be coupled and the interaction between them plays an important role in formation of the liquid structure. The effect of gas nozzle diameter on flow rate, mean radius of the liquid, and mean radial velocity, as well as its oscillation behavior has been investigated. The power development of dominate frequencies of averaged radial liquid velocity along the flow direction is shown. The growth of the instabilities can be observed from these results.

Abstract

The interface between data-driven learning methods and classical simulation poses an interesting field offering a multitude of new applications. In this work, we build on the notion of physics-informed neural networks (PINNs) and employ them in the area of shallow-water equation (SWE) models. These models play an important role in modeling and simulating free-surface flow scenarios such as in flood-wave propagation or tsunami waves. Different formulations of the PINN residual are compared to each other and multiple optimizations are being evaluated to speed up the convergence rate. We test these with different 1-D and 2-D experiments and finally demonstrate that regarding a SWE scenario with varying bathymetry, the method is able to produce competitive results in comparison to the direct numerical simulation with a total relative $L_2$ error of $8.9e-3$.

Abstract

Conventional macroscale two-phase flow equations for porous media (such as Darcy's law and Richards Equation) require a constitutive relation for capillary pressure (Pc). The capillary pressure relation significantly impacts the behavior and prediction of fluid flow in porous media, and needs to accurately characterize the capillary forces. In a typical laboratory experiment, a functional macroscopic capillary pressure-saturation (Pc-Sw) relationship is measured as the difference between the pressures of the non-wetting-phase reservoir at the inlet (Pnw) and wetting-phase reservoir at the outlet (Pw) of a porous medium. It is well-known that this traditional macroscopic capillary pressure definition is valid only at equilibrium conditions and if the phases are connected. Under non-equilibrium (dynamic) conditions, when the fluids are moving, the macroscopic capillary pressure measured in experiments implicitly includes the pressure head caused by viscous effects. The goal of the present effort is to understand how well the traditional macroscopic capillary pressure definition represents the pore-scale capillary forces under different flow conditions. Using direct numerical simulations (DNS) of two-phase flow in a porous medium, we evaluate the capillary pressure at the pore-scale, and compare it to the macroscopic capillary pressure, Pc(Sw), that is typically measured in experiments using pressure transducers. The pore-scale capillary pressure is the pressure difference across the interface between two fluids as the fluids move through a porous medium; the interface pressure differences at fluid-fluid invasion front are averaged across all the pores of the porous medium to yield a representative pore-scale capillary pressure curve, referred to as the interface capillary pressure. The pore-scale interface capillary pressure represents the “true” capillary forces in the system, since depends only on the pore morphology (shape) and interfacial energy of the two fluids, and does not account for the viscous dissipation. In experiments it is difficult to measure the interface capillary pressure jump without accounting for the viscous pressure head, which is at least an order of magnitude larger. Upscaling the pore-scale capillary pressure is an essential step for complete characterization of capillary-dominant two-phase flow in a porous medium at the laboratory scale. Drainage is simulated under equilibrium (quasi-static) and non-equilibrium (dynamic) conditions for various capillary numbers. The Navier–Stokes (NS) equations are solved in the pore space using the open-source finite-volume computational fluid dynamics (CFD) code, OpenFOAM. The Volume-of-Fluid (VOF) method is employed to track the evolution of the fluid–fluid interfaces, and a contact angle is used to account for the effect of wall adhesion. The simulations are first validated with published experimental data for dynamic and quasi-static drainage in a micromodel. From the microscale simulations, the interface capillary pressure is determined, and compared to the macroscopic capillary pressure under equilibrium and non-equilibrium conditions. Our results show the traditionally-measured macroscopic capillary pressure curves exhibit a strong dependence on the capillary number under dynamic flow conditions. In contrast, the interface capillary pressure-saturation relation, which relies on pore-scale pressure differences at the invasion front, is almost invariant of flow conditions (dynamic and quasi-static).

Abstract

We present version 3 of the open-source simulator for flow and transport processes in porous media DuMux. DuMux is based on the modular C++ framework Dune (Distributed and Unified Numerics Environment) and is developed as a research code with a focus on modularity and reusability. We describe recent efforts in improving the transparency and efficiency of the development process and community-building, as well as efforts towards quality assurance and reproducible research. In addition to a major redesign of many simulation components in order to facilitate setting up complex simulations in DuMux, version 3 introduces a more consistent abstraction of finite volume schemes. Finally, the new framework for multi-domain simulations is described, and three numerical examples demonstrate its flexibility.

Abstract

Sparse grids enable accurate function approximation in higher-dimensional settings with a moderate number of grid points. In this paper, we introduce generalized versions of various efficient sparse grid algorithms in a unified notation: For local basis functions, transformations between function values at grid points and interpolation coefficients can be efficiently realized using the well-known unidirectional principle (Balder and Zenger, SIAM Journal on Scientific Computing, 17(3):631--646, May 1996; Bungartz, Finite Elements of Higher Order on Sparse Grids, Universität München, Aachen, November 1998). For general basis functions without spatial adaptivity, Avila and Carrington show in (The Journal of Chemical Physics, 143(21):214108, 2015) and follow-up work that the unidirectional principle can instead be applied to an equivalent incremental basis, and we show that a second application of the unidirectional principle can be used to compute interpolation coefficients for the original, non-incremental basis. Our Open Source implementation for the latter case is able to compute interpolation coefficients for up to 20 million sparse grid points in one second on a single CPU core. We also discuss the efficient application of linear operators occurring in derivative evaluations of interpolants and in the solution of partial differential equations on sparse grids with finite elements, e.g., using the UpDown scheme (Bungartz, Dünne Gitter und deren Anwendung bei der adaptiven Lösung der dreidimensionalen Poisson-Gleichung, PhD thesis, Technische Universität München, 1992; Pflüger, Spatially Adaptive Sparse Grids for High-Dimensional Problems, Verlag Dr. Hut, München, February 2010; Zeiser, Journal of Scientific Computing, 47(3):328--346, 2011; Bungartz et al., Journal of Computational and Applied Mathematics, 236(15):3741--3750, 2012).

Abstract

The dual iteration was introduced in a conference paper in 1997 by Iwasaki as an iterative and heuristic procedure for the challenging and non-convex design of static output-feedback controllers. We recall in detail its essential ingredients and go beyond the work of Iwasaki by demonstrating that the framework of linear fractional representations allows for a seamless extension the dual iteration to output-feedback designs of tremendous practical relevance such as the design of robust or robust gain-scheduled controllers. In the paper of Iwasaki the dual iteration is solely based on, and motivated by algebraic manipulations resulting from the elimination lemma. We provide a novel control theoretic interpretation of the individual steps, which paves the way for further generalizations of the powerful scheme to situations where the elimination lemma is not applicable. Exemplary, we extend the dual iteration to the multi-objective design of static output-feedback H∞-controllers, which guarantee that the closed-loop poles are contained in an a priori specified generalized stability region. We demonstrate the approach with numerous numerical examples inspired from the literature.

Abstract

For a partially unknown linear systems, we present a systematic control design approach based on generated data from measurements of closed-loop experiments with suitable test controllers. These experiments are used to improve the achieved performance and to reduce the uncertainty about the unknown parts of the system. This is achieved through a parametrization of auspicious controllers with convex relaxation techniques from robust control, which guarantees that their implementation on the unknown plant is safe. This approach permits to systematically incorporate available prior knowledge about the system by employing the framework of linear fractional representations.

Abstract

We provide a convex solution to the robust gain-scheduled estimation problem based on integral quadratic constraints with dynamic D-scalings for both the uncertain and the scheduled component. This closes an important gap since so far merely static scalings or multipliers could be used for the scheduled component in estimation problems. To this end, we provide novel synthesis criteria in terms of linear matrix inequalities for the design of nominal gain-scheduled estimators which allow for a direct combination with available results on robust estimation. We illustrate the benefit of our design approach by means of a numerical example.

Abstract

The Riemann problem is one of the basic building blocks for numerical methods in computational fluid mechanics. Nonetheless, there are still open questions and gaps in theory and modeling for situations with complex thermodynamic behavior. In this series, we compare numerical solutions of the macroscopic flow equations with molecular dynamics simulation data. To enable molecular dynamics for sufficiently large scales in time and space, we selected the truncated and shifted Lennard-Jones potential, for which also highly accurate equations of state are available. A comparison of a two-phase Riemann problem is shown, which involves a liquid and a vapor phase, with an undergoing phase transition. The loss of hyperbolicity allows for the occurrence of anomalous wave structures. We successfully compare the molecular dynamics data with two macroscopic numerical solutions obtained by either assuming local phase equilibrium or by imposing a kinetic relation and allowing for metastable states.

Abstract

We present a data-driven visual analysis approach for the in-depth exploration of large numbers of droplets. Understanding droplet dynamics in sprays is of interest across many scientific fields for both simulation scientists and engineers. In this paper, we analyze large-scale direct numerical simulation datasets of the two-phase flow of non-Newtonian jets. Our interactive visual analysis approach comprises various dedicated exploration modalities that are supplemented by directly linking to ParaView. This hybrid setup supports a detailed investigation of droplets, both in the spatial domain and in terms of physical quantities . Considering a large variety of extracted physical quantities for each droplet enables investigating different aspects of interest in our data. To get an overview of different types of characteristic behaviors, we cluster massive numbers of droplets to analyze different types of occurring behaviors via domain-specific pre-aggregation, as well as different methods and parameters. Extraordinary temporal patterns are of high interest, especially to investigate edge cases and detect potential simulation issues. For this, we use a neural network-based approach to predict the development of these physical quantities and identify irregularly advected droplets.

Abstract

This paper considers the problem of designing accelerated gradient-based algorithms for optimization and saddle-point problems. The class of objective functions is defined by a generalized sector condition. This class of functions contains strongly convex functions with Lipschitz gradients but also non-convex functions, which allows not only to address optimization problems but also saddle-point problems. The proposed design procedure relies on a suitable class of Lyapunov functions and on convex semi-definite programming. The proposed synthesis allows the design of algorithms that reach the performance of state-of-the-art accelerated gradient methods and beyond.

Abstract

We present a novel projection-based model reduction framework for parametric linear time-invariant systems that allows interpolating the transfer function at a given frequency point along parameter-dependent curves as opposed to the standard approach where transfer function interpolation is achieved for a discrete set of parameter and frequency samples. We accomplish this goal by using parameter-dependent projection spaces. Our main result shows that for holomorphic system matrices, the corresponding interpolatory projection spaces are also holomorphic. The coefficients of the power series representation of the projection spaces can be computed iteratively using standard methods. We illustrate the analysis on three numerical examples.

Abstract

Previous laboratory experiments with the kinetic interface sensitive (KIS) tracers have shown promising results with respect to the quantification of the fluid-fluid interfacial area (IFA) under dynamic, two-phase flow conditions. However, pore-scale effects relevant to two-phase flow (e.g., the formation of hydrodynamically stagnant/immobile zones) are not yet fully understood, and quantitative information about how far these effects influence the transport of the tracer reaction products is not yet available. Therefore, a pore-scale numerical model that includes two-phase, reactive flow, and transport of the KIS tracer at the fluid-fluid interface is developed. We propose a new method to quantitatively analyze how the mass of the KIS-tracer reaction product in the flowing water is affected by the presence of the immobile zones. The model employs the phase field method (PFM) and a new continuous mass transfer formulation, consistent with the PFM. We verify the model with the analytical solutions of transport involving advection, reaction and diffusion processes. The model is tested for two-phase flow conditions in a conceptual 2D slit. The applicability of the model is demonstrated in NAPL/water drainage scenarios in a conceptual porous domain, comparing the results in terms of the spatial distribution of the phases and solute concentration. Furthermore, we distinguish the mobile and immobile zones based on the local Péclet number, and the corresponding IFA, and solute mass in these two zones is quantified. Finally, we show that the solute mass in flowing water can be employed to selectively determine the mobile part of the IFA.

Abstract

Porous media surface roughness strongly influences the transport of solutes during drainage due to the formation of thick water films (capillary condensation) on the surface of the porous medium. For interfacially-reacted, water-based solutes, these water films increase both the solute production at the fluid-fluid interface, due to the increased number of fluid-fluid interfaces, and the loss of the solute due to retention in the water films. This study applied a pore-scale, direct numerical simulation with the phase-field method-based continuous solute transport model to simulate the reactive transport of the kinetic interfacial sensitive tracer. The study is implemented during primary drainage in a 2D slit with rough solid walls, where the fractal geometries of the solid surfaces were generated numerically. The moving interfacial area is found to be changing non-monotonically with the root mean square of surface roughness. With increasing root mean square roughness, the average film thickness increases linearly, whereas the film-associated interfacial area per smooth surface area converges to a value slightly larger than one. The retention of the solute mass produced by the moving meniscus in the water film is observed, and this is described by a film-associated mobile mass retention term. An implicit relation between the mobile interfacial area and the solute mass in flowing zones is found. Finally, it is found that the film-associated mobile mass retention term is linearly related to the root mean square roughness.

Abstract

Abstract Accurately predicting bare-soil evaporation requires the proper characterization of the near-surface atmospheric conditions. These conditions, dependent on factors such as surface microtopography and wind velocity, vary greatly and therefore require high-resolution datasets to be fully incorporated into evaporation models. These factors are oftentimes parameterized in models through the aerodynamic resistance (ra), in which the vapor roughness length (z0v) and the momentum roughness length (z0m) are two crucial parameters that describe the transport near the soil-atmosphere interface. Typically, when evaluating bare-soil evaporation, these two characteristic lengths are assumed equal, although differences are likely to occur especially in turbulent flows over undulating surfaces. Thus, this study aims to investigate the relationship between z0v and z0m above undulating surfaces to ultimately improve accuracy in estimating evaporation rate. To achieve this goal, four uniquely designed wind tunnel—soil tank experiments were conducted considering different wind speeds and undulation spacings. Particle image velocimetry (PIV) was used to measure the velocity field above the undulating surface in high resolution. Using the high-fidelity data set, the logarithmic ratio of z0v to z0m is determined and used to estimate ra. Results confirm that these lengths differ significantly, with the logarithmic ratio roughly ranging from −15 to −5 under the conditions tested. PIV-measured results demonstrate this ratio is closely tied to the mass and momentum transport behaviors influenced by surface undulations. Using the data-integrated formulation of ra, predictions of evaporation rate were prepared for both the laboratory and lysimeter experiments, demonstrating the efficacy of the proposed approach in this study.

Abstract

The combination of machine learning with control offers many opportunities, in particular for robust control. However, due to strong safety and reliability requirements in many real-world applications, providing rigorous statistical and control-theoretic guarantees is of utmost importance, yet difficult to achieve for learning-based control schemes. We present a general framework for learning-enhanced robust control that allows for systematic integration of prior engineering knowledge, is fully compatible with modern robust control and still comes with rigorous and practically meaningful guarantees. Building on the established Linear Fractional Representation and Integral Quadratic Constraints framework, we integrate Gaussian Process Regression as a learning component and state-of-the-art robust controller synthesis. In a concrete robust control example, our approach is demonstrated to yield improved performance with more data, while guarantees are maintained throughout.

Abstract

Abstract Model Predictive Control (MPC), also referred to as Moving Horizon Control (MHC), experienced a massive growth of interest over the last decades due to increasing computing power and the ability to naturally incorporate state and input constraints. As a model-based controller, the performance of Model Predictive Control heavily relies upon the quality of the model used. One can leverage measurement data to improve upon a nominal model, e.g., based on first principles, using Gaussian Processes, or other regression techniques. We instead opt for the introduction of (nonlinear) rate constraints to force the system to stay within regions of the state space where a simple, possibly linear, model describes the underlying dynamics well, while still robustly accounting for small model mismatches introduced with the simple model. These rate constraints can again be learned using any parametric or non-parametric regression method, making the approach very flexible. As an exemplary system, we choose an omnidirectional mobile robot moving in an environment with changing friction coefficients, putting state-dependent limits on the maximal rate of change of its speed.

Abstract

In article number 2007982, Holger Steeb, Sabine Ludwigs, and co-workers present humidity-triggered bending actuators based on simple polymer bilayers and their in-depth mechanical characterization. A loop consisting of the mechanical characterization and a simple analytical model allows the prediction of deformations of in principle any complex geometry and material combination, for example, for moving parts and curvatures in soft robotics.

Abstract

In this study, the complexity of a steady-state flow through porous media is revealed using confocal laser scanning microscopy (CLSM). Micro-particle image velocimetry (micro-PIV) is applied to construct movies of colloidal particles. The calculated velocity vector fields from images are further utilized to obtain laminar flow streamlines. Fluid flow through a single straight channel is used to confirm that quantitative CLSM measurements can be conducted. Next, the coupling between the flow in a channel and the movement within an intersecting dead-end region is studied. Quantitative CLSM measurements confirm the numerically determined coupling parameter from earlier work of the authors. The fluid flow complexity is demonstrated using a porous medium consisting of a regular grid of pores in contact with a flowing fluid channel. The porous media structure was further used as the simulation domain for numerical modeling. Both the simulation, based on solving Stokes equations, and the experimental data show presence of non-trivial streamline trajectories across the pore structures. In view of the results, we argue that the hydrodynamic mixing is a combination of non-trivial streamline routing and Brownian motion by pore-scale diffusion. The results provide insight into challenges in upscaling hydrodynamic dispersion from pore scale to representative elementary volume (REV) scale. Furthermore, the successful quantitative validation of CLSM-based data from a microfluidic model fed by an electrical syringe pump provided a valuable benchmark for qualitative validation of computer simulation results.

Abstract

Turbulence transportation across permeable interfaces is investigated using direct numerical simulation, and the connection between the turbulent surface flow and the pore flow is explored. The porous media domain is constructed with an in-line arranged circular cylinder array. The effects of Reynolds number and porosity are also investigated by comparing cases with two Reynolds numbers (Re≈3000,6000) and two porosities (φ=0.5,0.8). It was found that the change of porosity leads to the variation of flow motions near the interface region, which further affect turbulence transportation below the interface. The turbulent kinetic energy (TKE) budget shows that turbulent diffusion and pressure transportation work as energy sink and source alternatively, which suggests a possible route for turbulence transferring into porous region. Further analysis on the spectral TKE budget reveals the role of modes of different wavelengths. A major finding is that mean convection not only affects the distribution of TKE in spatial space, but also in scale space. The permeability of the wall also have an major impact on the occurrence ratio between blow and suction events as well as their corresponding flow structures, which can be related to the change of the Kármán constant of the mean velocity profile.

Abstract

The microscopic structure of porous walls modulates the turbulent flow above. The standard approach, the volume-averaged modelling of the porous wall, does not resolve the pore structure. To systematically link geometric characteristics with flow properites, direct numerical simulations are conducted which are fully-resolving the microscopic structure. A high-order spectral/hp element solver is adopted to solve the incompressible Navier-Stokes equations. Resolving the full energy-spectra relies on a zonal polynomial refinement based on a conforming mesh. A low and a high porosity case with in-line arrays of cylinders are analysed for two Reynolds numbers. The peak in the streamwise energy spectra is shifted towards the pore unit length for both cases. Proper Orthogonal Decomposition (POD) shows that the fluctuations in the porous wall are linked to the structures above. Q2 structures are linked with blowing events and Q4 structures with suction events in the first pore row. The numerical solver Nektar exhibits an excellent scalability up to 96k cores on “Hazel Hen” where a slightly improved performance is observed on the brand new HPE “Hawk” system. Strong scaling tests indicate an efficiency of 70% with around 5, 000 mesh-nodes per core, which indicates a high potential for an adequate use of a HPC platform to investigate turbulent flows above porous walls while resolving the pore structure.

Abstract

Experimental and field studies reported a significant discrepancy between the cleanup and contamination time scales, while its cause is not yet addressed. Using high-resolution fast synchrotron x-ray computed tomography, we characterized the solute transport in a fully saturated sand packing for both contamination and cleanup processes at similar hydrodynamic conditions. The discrepancy in the time scales has been demonstrated by the nonuniqueness of hydrodynamic dispersion coefficient versus injection rate (Péclet number). Observations show that in the mixed advection-diffusion regime, the hydrodynamic dispersion coefficient of cleanup is significantly larger than that of the contamination process. This nonuniqueness has been attributed to the concentration-dependent diffusion coefficient during the cocurrent and countercurrent advection and diffusion, present in contamination and cleanup processes. The new findings enhance our fundamental understanding of transport processes and improve our capability to estimate the transport time scales of chemicals or pollution in geological and engineering systems.

Abstract

Physical motion models offer interpretable predictions for the motion of vehicles. However, some model parameters, such as those related to aero- and hydrodynamics, are expensive to measure and are often only roughly approximated reducing prediction accuracy. Recurrent neural networks achieve high prediction accuracy at low cost, as they can use cheap measurements collected during routine operation of the vehicle, but their results are hard to interpret. To precisely predict vehicle states without expensive measurements of physical parameters, we propose a hybrid approach combining deep learning and physical motion models including a novel two-phase training procedure. We achieve interpretability by restricting the output range of the deep neural network as part of the hybrid model, which limits the uncertainty introduced by the neural network to a known quantity. We have evaluated our approach for the use case of ship and quadcopter motion. The results show that our hybrid model can improve model interpretability with no decrease in accuracy compared to existing deep learning approaches.

Abstract

We establish a local theory, i.e., existence, uniqueness and blow-up criterion, for a general family of singular SDEs in Hilbert spaces. The key requirement relies on an approximation property that allows us to embed the singular drift and diffusion mappings into a hierarchy of regular mappings that are invariant with respect to the Hilbert space and enjoy a cancellation property. Various nonlinear models in fluid dynamics with transport noise belong to this type of singular SDEs. By establishing a cancellation estimate for certain differential operators of order one with suitable coefficients, we give the detailed constructions of such regular approximations for certain examples. In particular, we show novel local-in-time results for the stochastic two-component Camassa–Holm system and for the stochastic Córdoba–Córdoba–Fontelos model.

Abstract

With the ongoing advances in the area of cloud computing, Internet of Things, Industry 4.0, and the increasing prevalence of cyber-physical systems and devices equipped with sensors, the amount of data generated every second is rising steadily. Thereby, the gathering of data and the creation of added value from this data is getting easier and easier. However, the increasing volume of data stored in the cloud leads to new challenges. Analytics software and scalable platforms are required to evaluate the data distributed all over the internet. But with distributed applications and large data sets to be handled, the network becomes a bottleneck. Therefore, in this work, we present an approach to automatically improve the deployment of such applications regarding the placement of data processing components dependent on the data flow of the application. To show the practical feasibility of our approach, we implemented a prototype based on the open-source ecosystem OpenTOSCA. Moreover, we evaluated our prototype using various scenarios.

Abstract

Recent experiments have reported anomalously large screening lengths of interactions between charged surfaces confining concentrated electrolytes and ionic liquids. Termed underscreening, this effect was ascribed to bulk properties of dense ionic systems. Herein, we study bulk ionic screening with extremely large-scale molecular dynamics simulations, allowing us to assess the range of distances relevant to the experiments. Our results yield two screening lengths satisfying distinct scaling relations. However, with an accuracy of 10−5kBT in interionic potentials of mean force, we find no signs of underscreening, suggesting that other than bulk effects might be at play in the experiments.

Abstract

We investigate how Mixed Reality (MR) can be used to guide human body motions, such as in physiotherapy, dancing, or workout applications. While first MR prototypes have shown promising results, many dimensions of the design space behind such applications remain largely unexplored. To better understand this design space, we approach the topic from different angles by contributing three user studies. In particular, we take a closer look at the influence of the perspective, the characteristics of motions, and visual guidance on different user performance measures. Our results indicate that a first-person perspective performs best for all visible motions, whereas the type of visual instruction plays a minor role. From our results we compile a set of considerations that can guide future work on the design of instructions, evaluations, and the technical setup of MR motion guidance systems.

Abstract

The mechanism of dynamic wetting and the fluid dynamics during the onset of droplet mobilization driven by a microchannel flow are not clearly understood. In this work, we use microparticle tracking velocimetry to visualize the velocity distribution inside the droplet both prior to and during mobilization. Time-averaged and instantaneous velocity vectors are determined using fluorescent microscopy for various capillary numbers. A circulating flow exists inside the droplet at a subcritical capillary number, in which case the droplet is pinned to the channel walls. When the capillary number exceeds a critical value, droplet mobilization occurs, and this process can be divided into two stages. In the first stage, the location of the internal circulation vortex center moves to the rear of the droplet and the droplet deforms, but the contact lines at the top walls remain fixed. In the second stage, the droplet rolls along the solid wall, with fixed contact angles keeping the vortex center in the rear part of the droplet. The critical capillary number for the droplet mobilization is larger for the droplet fluid with a larger viscosity. A force-balance model of the droplet, considering the effect of fluid properties, is formulated to explain the experimental trends of advancing and receding contact angles with the capillary number. Numerical simulations on internal circulations for the pinned droplet indicate that the reversed flow rate, when normalized by the inlet flow rate and the kinematic viscosity ratio of the wetting and nonwetting phases, is independent of the capillary number and the droplet composition.

Abstract

Quantum computing has the potential to significantly impact many application domains, as several quantum algorithms are promising to solve problems more efficiently than possible on classical computers. However, various complex pre- and post-processing tasks have to be performed when executing a quantum circuit, which require immense mathematical and technical knowledge. For example, calculations on today’s quantum computers are noisy and require an error mitigation task after the execution. Hence, integrating classical applications with quantum circuits is a difficult challenge. In this paper, we introduce a modeling extension for imperative workflow languages to enable the integration of quantum computations and ease the orchestration of classical applications and quantum circuits. Further, we show how the extension can be mapped to native modeling constructs of extended workflow languages to retain the portability of the workflows. We validate the practical feasibility of our approach by applying our proposed extension to BPMN and introduce Quantum4BPMN.

Abstract

While neural networks with attention mechanisms have achieved superior performance on many natural language processing tasks, it remains unclear to which extent learned attention resembles human visual attention. In this paper, we propose a new method that leverages eye-tracking data to investigate the relationship between human visual attention and neural attention in machine reading comprehension. To this end, we introduce a novel 23 participant eye tracking dataset - MQA-RC, in which participants read movie plots and answered pre-defined questions. We compare state of the art networks based on long short-term memory (LSTM), convolutional neural models (CNN) and XLNet Transformer architectures. We find that higher similarity to human attention and performance significantly correlates to the LSTM and CNN models. However, we show this relationship does not hold true for the XLNet models -- despite the fact that the XLNet performs best on this challenging task. Our results suggest that different architectures seem to learn rather different neural attention strategies and similarity of neural to human attention does not guarantee best performance.

Abstract

A discretization is proposed for models coupling free flow with anisotropic porous-medium flow. Our approach employs a staggered-grid finite volume method for the Navier-Stokes equations in the free-flow subdomain and a MPFA finite volume method to solve Darcy flow in the porous medium. After appropriate spatial refinement in the free-flow domain, the degrees of freedom are conveniently located to allow for a natural coupling of the two discretization schemes. In turn, we automatically obtain a more accurate description of the flow field surrounding the porous medium. Numerical experiments highlight the stability and applicability of the scheme in the presence of anisotropy and second-order grid convergence was found in both domains, verifying our approach.

Abstract

We present the first method to anticipate averted gaze in natural dyadic interactions. The task of anticipating averted gaze, i.e. that a person will not make eye contact in the near future, remains unsolved despite its importance for human social encounters as well as a number of applications, including human-robot interaction or conversational agents. Our multimodal method is based on a long short-term memory (LSTM) network that analyses non-verbal facial cues and speaking behaviour. We empirically evaluate our method for different future time horizons on a novel dataset of 121 YouTube videos of dyadic video conferences (74 hours in total). We investigate person-specific and person-independent performance and demonstrate that our method clearly outperforms baselines in both settings. As such, our work sheds light on the tight interplay between eye contact and other non-verbal signals and underlines the potential of computational modelling and anticipation of averted gaze for interactive applications.

Abstract

We introduce an approach for the visual analysis of eye movement data from two people playing competitive virtual board games. Our approach provides methods to temporally synchronize and spatially register gaze and mouse recordings from two eye tracking devices. Analysts can examine such fused data visually with a combination of techniques: attention maps and gaze plots as well as a temporal summary of the distance between gaze positions and mouse events of the two players. We show different game scenarios from the competitive game Go, which is especially complex for analyzing strategies of individual players, to demonstrate our methods. In general, our visual analysis approach can provide analysts with insights into strategies, learning processes, and means of communication between people.

Abstract

In the present work, we explore the capability of artificial neural networks (ANN) to predict the closure terms for large eddy simulations (LES) solely from coarse-scale data. To this end, we derive a consistent framework for LES closure models, with special emphasis laid upon the incorporation of implicit discretization-based filters and numerical approximation errors. We investigate implicit filter types that are inspired by the solution representation of discontinuous Galerkin and finite volume schemes and mimic the behavior of the discretization operator, and a global Fourier cutoff filter as a representative of a typical explicit LES filter. Within the perfect LES framework, we compute the exact closure terms for the different LES filter functions from direct numerical simulation results of decaying homogeneous isotropic turbulence. Multiple ANN with a multilayer perceptron (MLP) or a gated recurrent unit (GRU) architecture are trained to predict the computed closure terms solely from coarse-scale input data. For the given application, the GRU architecture clearly outperforms the MLP networks in terms of accuracy, whilst reaching up to 99.9% correlation between the networks' predictions and the exact closure terms for all considered filter functions. The GRU networks are also shown to generalize well across different LES filters and resolutions. The present study can thus be seen as a starting point for the investigation of data-based modeling approaches for LES, which not only include the physical closure terms, but account for the discretization effects in implicitly filtered LES as well.

Abstract

Modern cytometry methods allow collecting complex, multi-dimensional data sets from heterogeneous cell populations at single-cell resolution. While methods exist to describe the progression and order of cellular processes from snapshots of such populations, these descriptions are limited to arbitrary pseudotime scales. Here we describe MAPiT, an universal transformation method that recovers real-time dynamics of cellular processes from pseudotime scales by utilising knowledge of the distributions on the real scales. As use cases, we applied MAPiT to two prominent problems in the flow-cytometric analysis of heterogeneous cell populations: (1) recovering the kinetics of cell cycle progression in unsynchronised and thus unperturbed cell populations, and (2) recovering the spatial arrangement of cells within multi-cellular spheroids prior to spheroid dissociation for cytometric analysis. Since MAPiT provides a theoretic basis for the relation of pseudotime values to real temporal and spatial scales, it can be used broadly in the analysis of cellular processes with snapshot data from heterogeneous cell populations.

Abstract

A method for efficient computational homogenization of hyperelastic materials under finite strains is proposed. Multiple spatial scales are homogenized in a recursive procedure: starting on the smallest scale, few high fidelity FE computations are performed. The resulting fields of deformation gradient fluctuations are processed by a snapshot POD resulting in a reduced basis (RB) model. By means of the computationally efficient RB model, a large set of samples of the homogenized material response is created. This data set serves as the support for the Concentric Interpolation (CI) scheme, interpolating the effective stress and stiffness. Then, the same procedure is invoked on the next larger scale with this CI surrogating the homogenized material law. A three-scale homogenization process is completed within few hours on a standard workstation. The resulting model is evaluated within minutes on a laptop computer in order to generate fourth-scale results. Open source code is provided.

Abstract

It is hypothesized that the nonlinear muscle characteristic of biomechanical systems simplify control in the sense that the information the nervous system has to process is reduced through off-loading computation to the morphological structure. It has been proposed to quantify the required information with an information-entropy based approach, which evaluates the minimally required information to control a desired movement, i.e., control effort. The key idea is to compare the same movement but generated by different actuators, e.g., muscles and torque actuators, and determine which of the two morphologies requires less information to generate the same movement. In this work, for the first time, we apply this measure to numerical simulations of more complex human movements: point-to-point arm movements and walking. These models consider up to 24 control signals rendering the brute force approach of the previous implementation to search for the minimally required information futile. We therefore propose a novel algorithm based on the pattern search approach specifically designed to solve this constraint optimization problem. We apply this algorithm to numerical models, which include Hill-type muscle-tendon actuation as well as ideal torque sources acting directly on the joints. The controller for the point-to-point movements was obtained by deep reinforcement learning for muscle and torque actuators. Walking was controlled by proprioceptive neural feedback in the muscular system and a PD controller in the torque model. Results show that the neuromuscular models consistently require less information to successfully generate the movement than the torque-driven counterparts. These findings were consistent for all investigated controllers in our experiments, implying that this is a system property, not a controller property. The proposed algorithm to determine the control effort is more efficient than other standard optimization techniques and provided as open source.

Abstract

We derive novel criteria for designing stabilizing dynamic output-feedback controllers for a class of aperiodic impulsive systems subject to a range dwell-time condition. Our synthesis conditions are formulated as clock-dependent linear matrix inequalities (LMIs) which can be solved numerically, e.g., by using matrix sum-of-squares relaxation methods. We show that our results allow us to design dynamic output-feedback controllers for aperiodic sample-data systems and illustrate the proposed approach by means of a numerical example.

Abstract

Solute transport in unsaturated porous materials is a complex process, which exhibits some distinct features differentiating it from transport under saturated conditions. These features emerge mostly due to the different transport time scales at different regions of the flow network, which can be classified into flowing and stagnant regions, predominantly controlled by advection and diffusion, respectively. Under unsaturated conditions, the solute breakthrough curves show early arrivals and very long tails, and this type of transport is usually referred to as non-Fickian. This study directly characterizes transport through an unsaturated porous medium in three spatial dimensions at the resolution of 3.25 μm and the time resolution of 6 s. Using advanced high-speed, high-spatial resolution, synchrotron-based X-ray computed microtomography (sCT) we obtained detailed information on solute transport through a glass bead packing at different saturations. A large experimental dataset (>50 TB) was produced, while imaging the evolution of the solute concentration with time at any given point within the field of view. We show that the fluids’ topology has a critical signature on the non-Fickian transport, which yet needs to be included in the Darcy-scale solute transport models. The three-dimensional (3D) results show that the fully mixing assumption at the pore scale is not valid, and even after injection of several pore volumes the concentration field at the pore scale is not uniform. Additionally, results demonstrate that dispersivity is changing with saturation, being twofold larger at the saturation of 0.52 compared to that at the fully saturated domain.

Abstract

The present work aims at the identification of the effective constitutive behavior of �5 aluminum grain boundaries (GB) for proportional loading by using machine learning (ML) techniques. The input for the ML approach is high accuracy data gathered in challenging molecular dynamics (MD) simulations at the atomic scale for varying temperatures and loading conditions. The effective traction-separation relation is recorded during the MD simulations. The raw MD data then serves for the training of an artificial neural network (ANN) as a surrogate model of the constitutive behavior at the grain boundary. Despite the extremely fluctuating nature of the MD data and its inhomogeneous distribution in the traction-separation space, the ANN surrogate trained on the raw MD data shows a very good agreement in the average behavior without any data-smoothing or pre-processing. Further, it is shown that the trained traction-separation ANN captures important physical properties and is able to predict traction values for given separations not contained in the training data. For example, MD simulations show a transition in traction-separation behaviour from pure sliding mode under shear load to combined GB sliding and decohesion with intermediate hardening regime at mixed load directions. These changes in GB behaviour are fully captured in the ANN predictions. Furthermore, by construction, the ANN surrogate is differentiable for arbitrary separation and also temperature, such that a thermo-mechanical tangent stiffness operator can always be evaluated. The trained ANN can then serve for large-scale FE simulation as an alternative to direct MD-FE coupling which is often infeasible in practical applications.

Abstract

Strategies for the generation of periodic discrete structures with identical two-point correlation—called 2PC-equivalent—are developed. It is shown that starting from a set of 2PC-equivalent root structures, 2PC-equivalent child structures of arbitrary resolution and number of phases (e.g. material phases) can be generated based on phase extension through trivial embeddings, kernel-based extension and phase coalescence. Proofs are provided by means of discrete Fourier transform theory. A Python 3 implementation is offered for reproduction of examples and future applications.

Abstract

The feasibility of synthesizing unnatural DNA/RNA has recently been demonstrated, giving rise to new perspectives and challenges in the emerging field of synthetic biology, DNA data storage, and even the search for extraterrestrial life in the universe. In line with this outstanding potential, solid-state nanopores have been extensively explored as promising candidates to pave the way for the next generation of label-free, fast, and low-cost DNA sequencing. In this work, we explore the sensitivity and selectivity of a graphene/h-BN based nanopore architecture towards detection and distinction of synthetic Hachimoji nucleobases. The study is based on a combination of density functional theory and the non-equilibrium Green's function formalism. Our findings show that the artificial nucleobases are weakly binding to the device, indicating a short residence time in the nanopore during translocation. Significant changes in the electron transmission properties of the device are noted depending on which artificial nucleobase resides in the nanopore, leading to a sensitivity in distinction of up to 80%. Our results thus indicate that the proposed nanopore device setup can qualitatively discriminate synthetic nucleobases, thereby opening up the feasibility of sequencing even unnatural DNA/RNA.

Abstract

Exchange processes between a turbulent free flow and a porous media flow are sensitive to the flow dynamics in both flow regimes, as well as to the interface that separates them. Resolving these complex exchange processes across irregular interfaces is key in understanding many natural and engineered systems. With soil–water evaporation as the natural application of interest, the coupled behavior and exchange between flow regimes are investigated numerically, considering a turbulent free flow as well as interfacial forms and obstacles. Interfacial forms and obstacles will alter the flow conditions at the interface, creating flow structures that either enhance or reduce exchange rates based on their velocity conditions and their mixing with the main flow. To evaluate how these interfacial forms change the exchange rates, interfacial conditions are isolated and investigated numerically. First, different flow speeds are compared for a flat surface. Second, a porous obstacle of varied height is introduced at the interface, and the effects the flow structures that develop have on the interface are analyzed. The flow parameters of this obstacle are then varied and the interfacial exchange rates investigated. Next, to evaluate the interaction of flow structures between obstacles, a second obstacle is introduced, separated by a varied distance. Finally, the shape of these obstacles is modified to create different wave forms. Each of these interfacial forms and obstacles is shown to create different flow structures adjacent to the surface which alter the mass, momentum, and energy conditions at the interface. These changes will enhance the exchange rate in locations where higher velocity gradients and more mixing with the main flow develop, but will reduce the exchange rate in locations where low velocity gradients and limited mixing with the main flow occur.

Abstract

Bubbly turbulent flow in a channel is investigated using interface-resolved direct numerical simulation. An efficient coupled level-set volume-of-fluid solver based on a fast Fourier transform algorithm is implemented to enable a high resolution and fast computation at the same time. Up to 384 bubbles are seeded in the turbulent channel flow corresponding to 5.4\% gas volume fraction. Bubbles are clustered in the channel center due to the downward flow direction. The bubbles induce additional pseudo-turbulence in the channel center and are also able to attenuate the energy in the boundary layer by reducing the shear production. Turbulent kinetic energy budget indicates a significant buoyancy production in the channel center. A local equilibrium between buoyancy production and dissipation is observed here besides the shear production peak in the boundary layer. Comparing the local production and dissipation indicates a coexistence of boundary layer turbulence near the wall and bubble-induced pseudo-turbulence in the channel center. The liquid phase and gas phase are coupled through the complex liquid–gas interface. Local flow topology analysis is depicted in the liquid phase around the bubbles as well as in the gas phase. The flow topology of the liquid phase and the gas phase differs from each other significantly. Local dissipation is more dominant in the liquid phase near the bubble interface, whereas local enstrophy is preferred in the gas phase. In the liquid phase, a high dissipation event is preferred close to the interface, whereas a high enstrophy event is dominant away from the interface.

Abstract

Hamiltonian systems are central in the formulation of non-dissipative physical systems. They are characterized by a phase-space, a symplectic form and a Hamiltonian function. In numerical simulations of Hamiltonian systems, algorithms show improved accuracy when the symplectic structure is preserved 10. For structure-preserving model order reduction (MOR) of Hamiltonian systems, symplectic MOR 17, 14, 11, 3, 16 can be used. It is based on a reduced-order basis that is symplectic, which requires symplectic basis generation techniques. In our work, we discuss greedy algorithms for symplectic basis generation. We complement the procedure presented in 14 with ideas of the POD-greedy 9, which results in a new greedy symplectic basis generation technique, the PSD-greedy. Inspired by POD-greedy, we use compression techniques in the greedy iterations to enrich the basis iteratively. We prove that this algorithm computes a symplectic basis when symplectic techniques are used for compression. In the numerical experiments, we compare the discussed methods for a linear elasticity problem. The results show that improvements of up to one order of magnitude in the relative reduction error are achievable with the new basis generation technique compared to the existing greedy approach from 14.

Abstract

Hyperbolic conservation laws are utilized to describe a variety of real-world applications, which require the consideration of the influence of uncertain parameters on the solution to the problem. To extend these models, one is often interested in including discontinuities in the state space to the flux function of the conservation law. This paper studies the solution of a stochastic nonlinear hyperbolic partial differential equation (PDE), whose flux function contains random spatial discontinuities. The first part of the paper defines the corresponding stochastic adapted entropy solution and required properties for existence and uniqueness are addressed. The second part contains the numerical simulation of the nonlinear hyperbolic problem as well as the estimation of the expectation of the problem via the multilevel Monte Carlo method.

Abstract

We present a framework for systematically combining data of an unknown linear time-invariant system with prior knowledge on the system matrices or on the uncertainty for robust controller design. Our approach leads to linear matrix inequality (LMI) based feasibility criteria which guarantee stability and performance robustly for all closed-loop systems consistent with the prior knowledge and the available data. The design procedures rely on a combination of multipliers inferred via prior knowledge and learnt from measured data, where for the latter a novel and unifying disturbance description is employed. While large parts of the paper focus on linear systems and input-state measurements, we also provide extensions to robust output-feedback design based on noisy input-output data and against nonlinear uncertainties. We illustrate through numerical examples that our approach provides a flexible framework for simultaneously leveraging prior knowledge and data, thereby reducing conservatism and improving performance significantly if compared to black-box approaches to data-driven control.

Abstract

The stable and accurate approximation of discontinuities such as shocks on a finite computational mesh is a challenging task. Detection of shocks or strong discontinuities in the flow solution is typically achieved through a priori troubled cell indicators, which guide the subsequent action of an appropriate shock capturing mechanism. Arriving at a stable and accurate solution often requires empirically based parameter tuning and adjustments of the indicator settings to the discretization and solution at hand. In this work, we propose to separate the task of shock detection and shock capturing more strongly and aim to develop a shock indicator that is robust, accurate, requires minimal user input and is suitable for high order element-based methods like discontinuous Galerkin and flux reconstruction methods. The novel indicator is learned from analytical data through a supervised learning strategy; its input is given by the high order solution field, its output is an element-local map of the shock position. We use state of the art methods from edge detection in image analysis based on deep convolutional multiscale networks and deep supervision to train the indicators. The resulting networks are then used as black box indicators, showing their robustness and accuracy on well established canonical testcases. All simulations are run ab initio using the developed indicators, showing that they provide also stability during the strongly transient phases. In particular for high order schemes with large cells and considerable inner-cell resolution capabilities, we demonstrate how the additional accurate prediction of the position of the shock front can be exploited to guide inner-element shock capturing strategies.

Abstract

Abstract We investigate the influence of near-surface wind conditions on subsurface gas transport and on soil-atmosphere gas exchange for gases of different density. Results of a sand tank experiment are supported by a numerical investigation with a fully coupled porous medium-free flow model, which accounts for wind turbulence. The experiment consists of a two-dimensional bench-scale soil tank containing homogeneous sand and an overlying wind tunnel. A point source was installed at the bottom of the tank. Gas concentrations were measured at multiple horizontal and vertical locations. Tested conditions include four wind velocities (0.2/1.0/2.0/2.7 m/s), three different gases (helium: light, nitrogen: neutral, and carbon dioxide: heavy), and two transport cases (1: steady-state gas supply from the point source; 2: transport under decreasing concentration gradient, subsequent to termination of gas supply). The model was used to assess flow patterns and gas fluxes across the soil surface. Results demonstrate that flow and transport in the vicinity of the surface are strongly coupled to the overlying wind field. An increase in wind velocity accelerates soil-atmosphere gas exchange. This is due to the effect of the wind profile on soil surface concentrations and due to wind-induced advection, which causes subsurface horizontal transport. The presence of gases with pronounced density difference to air adds additional complexity to the transport through the wind-affected soil layers. Wind impact differs between tested gases. Observed transport is multidimensional and shows that heavy as well as light gases cannot be treated as inert tracers, which applies to many gases in environmental studies.

Abstract

We present a machine learning-based approach for detecting and visualizing complex behavior in spatiotemporal volumes. For this, we train models to predict future data values at a given position based on the past values in its neighborhood, capturing common temporal behavior in the data. We then evaluate the model's prediction on the same data. High prediction error means that the local behavior was too complex, unique or uncertain to be accurately captured during training, indicating spatiotemporal regions with interesting behavior. By training several models of varying capacity, we are able to detect spatiotemporal regions of various complexities. We aggregate the obtained prediction errors into a time series or spatial volumes and visualize them together to highlight regions of unpredictable behavior and how they differ between the models. We demonstrate two further volumetric applications: adaptive timestep selection and analysis of ensemble dissimilarity. We apply our technique to datasets from multiple application domains and demonstrate that we are able to produce meaningful results while making minimal assumptions about the underlying data.

Abstract

Generalized Pythagoras trees were developed for visualizing hierarchical data, producing organic, fractal-like representations. However, the drawback of the original layout algorithm is visual overlap of tree branches. To avoid such overlap, we introduce an adapted drawing algorithm using ellipses instead of circles to recursively place tree nodes representing the subhierarchies. Our technique is demonstrated by resolving overlap in diverse real-world and generated datasets, while comparing the results to the original approach.

Abstract

We investigate the influence of uncertain input parameters on the aeroacoustic feedback of cavity flows. The so-called Rossiter feedback requires a direct numerical computation of the acoustic noise, which solves hydrodynamics and acoustics simultaneously, in order to capture the interaction of acoustic waves and the hydrodynamics of the flow. Due to the large bandwidth of spatial and temporal scales, a high order numerical scheme with low dissipation and dispersion error is necessary to preserve important small scale information. Therefore, the open-source CFD solver FLEXI, which is based on a high-order discontinuous Galerkin spectral element method, is used to perform the aforementioned direct simulations of an open cavity configuration with a laminar upstream boundary layer. To analyse the precision of the deterministic cavity simulation with respect to random input parameters we establish a framework for uncertainty quantification. In particular, a non-intrusive spectral projection method with Legendre and Hermite polynomial basis functions is employed in order to treat uniform and normal probability distributions of the uncertain input. The results indicate a strong, non-linear dependency of the acoustic feedback mechanism on the investigated uncertain input parameters. An analysis of the stochastic results offers new insights into the noise generation process of open cavity flows and reveals the strength of the implemented uncertainty quantification framework.

Abstract

We develop novel criteria for robust stability of a class of hybrid systems that are affected by piecewise constant uncertain parameters with dwell-time constraints. These criteria result from clock-dependent Lyapunov arguments that are based on several new robust stability conditions for continuous-time linear time-invariant systems which are affected by parametric passive uncertainties and do not admit any discrete dynamics. In contrast to other approaches, our stability analysis results rely on separation techniques from robust control involving dynamic resetting scalings. Moreover, these techniques are expected to pave the way for covering much more general uncertainties within the framework of integral quadratic constraints. Building upon the new insights on robust stability, we are able to propose an output-feedback gain-scheduling design methodology similarly to our recently obtained synthesis result involving dynamic resetting $D$-scalings. Most of the obtained conditions are expressed as infinite-dimensional (differential) linear matrix inequalities which can be numerically solved by using relaxation methods based on, e.g., linear splines or matrix sum-of-squares. We apply our results to obtain novel stability criteria and to design distributed controllers for networked systems over undirected switching topologies. Finally, the approach is illustrated with several numerical examples.

Abstract

We propose a novel approach for output-feedback controller design for linear systems affected by arbitrary fast time-varying uncertainties. Our approach is based on consecutively solving two related robust synthesis problems which by themselves admit convex solutions. The first one is robust full information controller design, while the subsequent one is similar to the robust estimator design problem. The latter includes a homotopy parameter that serves to construct a solution of the output-feedback synthesis problem from a given solution of the full information problem. In contrast to alternative approaches, ours requires to solve just two linear matrix inequalities and generates guaranteed bounds for the achieved performance. We illustrate our results with several numerical examples taken from the literature.

Abstract

To automate the deployment of applications several deployment technologies have been developed. However, the management of deployed applications is only partially covered by existing approaches: While management functionalities such as scaling components or changing their configurations are covered directly by cloud providers or configuration management technologies such as Chef, holistic management processes that affect multiple components probably deployed in different environments cannot be automated using these approaches. For example, testing all deployed components and their communication or backing up the entire application state that is scattered across different components requires custom management logic that needs to be implemented manually, e. g., using scripts. However, a manual implementation of such management processes is errorprone, time-consuming, and requires immense technical expertise. Therefore, we propose an approach that enables automatically generating executable management workflows based on the declarative deployment model of an application. This significantly reduces the effort for automating holistic management processes as no manual implementation is required. We validate the practical feasibility of the approach by a prototypical implementation based on the TOSCA standard and the OpenTOSCA ecosystem.

Abstract

The effect of water content on the static and dynamic properties of the deep eutectic solvent glyceline is studied using molecular dynamics (MD) simulations. Static properties are additionally calculated using the PC-SAFT equation of state. Force fields calibrated on water-free glyceline show predictive power for density and water activity over the entire composition range. In contrast, the PC-SAFT approach using pseudo one-component or two-component modelling strategies performed better for the density or the water activity, respectively. The MD simulations show that at low water content, the hydrogen-bond network between glycerol molecules as well as between glycerol and the cholinium cation is hardly affected by the water molecules while at higher water content, glycerol–glycerol hydrogen bonds are replaced by glycerol–water hydrogen bonds indicating the formation of an aqueous solution accompanied by a strong decrease of the shear viscosity. At the same time, the thermodynamic activity of water increases such that the MD simulations are able to guide the optimal composition with respect to requirements in biocatalytic applications such as low viscosity and low water activity. The combined application of PC-SAFT to efficiently predict static properties and molecular dynamics simulations to predict static and dynamic properties offers a powerful framework in solvent design applications.

Abstract

The optimization of the controllability Gramian of a given linear continuous-time system via proper design of its input matrix is a rather unexplored problem in the control literature. The most well-studied version of this problem is related to the optimal placement of actuators in a large-scale system. In this case, the input matrix is constrained to be a matrix whose columns are elements of the canonical basis. Inspired by this problem, in this paper we investigate two Gramian optimization problems in which the input matrix is subject to a Frobenius norm (or "power") constraint. Specifically, we consider the minimization of the maximum eigenvalue of the controllability Gramian and its dual, namely the maximization of the minimum eigenvalue of the Gramian, subject to a fixed Frobenius norm constraint on the input matrix. Surprisingly, these two problems admit the same optimal value, which is a simple function of the trace of the system state matrix. On the other hand, the properties of the optimal input matrices turn out to be much more difficult to capture. In this paper, we also establish several characterizations of these optimal matrices.

Abstract

Neck movement and injury during a car crash is influenced by reflexes, strength, and flexibility. This data set contains in-vivo measurements of neck kinematics and reflexes as participants are rapidly accelerated while driving a mechanically simulated car. This data set includes measurements from a mechanical driving simulator that has a limited range of movement on the order of 10’s of centimeters (ITM). Torso, head and neck kinematics are measured using infrared reflective markers and cameras while electrical activity from the neck muscles are recorded using surface electromyography. MVC tests for flexion and extension are used to normalize the EMG signals.

Abstract

Nowadays, active human body models are becoming essential tools for the development of integrated occupant safety systems. However, their broad application in industry and research is limited due to the complexity of incorporated muscle controllers, the long simulation runtime, and the non-regular use of physiological motor control approaches. The purpose of this study is to address the challenges in all indicated directions by implementing a muscle controller with several physiologically inspired control strategies into an open-source extended Hill-type muscle model formulated as LS-DYNA user-defined umat41 subroutine written in the Fortran programming language. This results in increased usability, runtime performance and physiological accuracy compared to the standard muscle material existing in LS-DYNA. The proposed controller code is verified with extensive experimental data that include findings for arm muscles, the cervical spine region, and the whole body. Selected verification experiments cover three different muscle activation situations: (1) passive state, (2) open-loop and closed-loop muscle activation, and (3) reflexive behaviour. Two whole body finite element models, the 50th percentile female VIVA OpenHBM and the 50th percentile male THUMS v5, are used for simulations, complemented by the simplified arm model extracted from the 50th percentile male THUMS v3. The obtained results are evaluated additionally with the CORrelation and Analysis methodology and the mean squared error method, showing good to excellent biofidelity and sufficient agreement with the experimental data. It was shown additionally how the integrated controller allows simplified mimicking of the movements for similar musculoskeletal models using the parameters transfer method. Furthermore, the Hill-type muscle model presented in this paper shows better kinematic behaviour even in the passive case compared to the existing one in LS-DYNA due to its improved damping and elastic properties. These findings provide a solid evidence base motivating the application of the enhanced muscle material with the internal controller in future studies with Active Human Body Models under different loading conditions.

Abstract

Abstract Significant trends in the vehicle industry are autonomous driving, micromobility, electrification and the increased use of shared mobility solutions. These new vehicle automation and mobility classes lead to a larger number of occupant positions, interiors and load directions. As safety systems interact with and protect occupants, it is essential to place the human, with its variability and vulnerability, at the center of the design and operation of these systems. Digital human body models (HBMs) can help meet these requirements and are therefore increasingly being integrated into the development of new vehicle models. This contribution provides an overview of current HBMs and their applications in vehicle safety in different driving modes. The authors briefly introduce the underlying mathematical methods and present a selection of HBMs to the reader. An overview table with guideline values for simulation times, common applications and available variants of the models is provided. To provide insight into the broad application of HBMs, the authors present three case studies in the field of vehicle safety: (i) in-crash finite element simulations and injuries of riders on a motorcycle; (ii) scenario-based assessment of the active pre-crash behavior of occupants with the Madymo multibody HBM; (iii) prediction of human behavior in a take-over scenario using the EMMA model.

Abstract

The subject of this paper is a generalized Camassa–Holm equation under random perturbation. We first establish local existence and uniqueness results as well as blow-up criteria for pathwise solutions in the Sobolev spaces Hs with s>3/2. Then we analyze how noise affects the dependence of solutions on initial data. Even though the noise has some already known regularization effects, much less is known concerning the dependence on initial data. As a new concept we introduce the notion of stability of exiting times and construct an example showing that multiplicative noise (in Itô sense) cannot improve the stability of the exiting time, and simultaneously improve the continuity of the dependence on initial data. Finally, we obtain global existence theorems and estimate associated probabilities.

Abstract

The present work aims at the identification of the effective constitutive behavior of �5 aluminum grain boundaries (GB) for proportional loading by using machine learning (ML) techniques. The input for the ML approach is high accuracy data gathered in challenging molecular dynamics (MD) simulations at the atomic scale for varying temperatures and loading conditions. The effective traction-separation relation is recorded during the MD simulations. The raw MD data then serves for the training of an artificial neural network (ANN) as a surrogate model of the constitutive behavior at the grain boundary. Despite the extremely fluctuating nature of the MD data and its inhomogeneous distribution in the traction-separation space, the ANN surrogate trained on the raw MD data shows a very good agreement in the average behavior without any data-smoothing or pre-processing. Further, it is shown that the trained traction-separation ANN captures important physical properties and is able to predict traction values for given separations not contained in the training data. For example, MD simulations show a transition in traction-separation behaviour from pure sliding mode under shear load to combined GB sliding and decohesion with intermediate hardening regime at mixed load directions. These changes in GB behaviour are fully captured in the ANN predictions. Furthermore, by construction, the ANN surrogate is differentiable for arbitrary separation and also temperature, such that a thermo-mechanical tangent stiffness operator can always be evaluated. The trained ANN can then serve for large-scale FE simulation as an alternative to direct MD-FE coupling which is often infeasible in practical applications.

Abstract

A variety of methods is available to quantify uncertainties arising within the modeling of flow and transport in carbon dioxide storage, but there is a lack of thorough comparisons. Usually, raw data from such storage sites can hardly be described by theoretical statistical distributions since only very limited data is available. Hence, exact information on distribution shapes for all uncertain parameters is very rare in realistic applications. We discuss and compare four different methods tested for data-driven uncertainty quantification based on a benchmark scenario of carbon dioxide storage. In the benchmark, for which we provide data and code, carbon dioxide is injected into a saline aquifer modeled by the nonlinear capillarity-free fractional flow formulation for two incompressible fluid phases, namely carbon dioxide and brine. To cover different aspects of uncertainty quantification, we incorporate various sources of uncertainty such as uncertainty of boundary conditions, of parameters in constitutive relations, and of material properties. We consider recent versions of the following non-intrusive and intrusive uncertainty quantification methods: arbitrary polynomial chaos, spatially adaptive sparse grids, kernel-based greedy interpolation, and hybrid stochastic Galerkin. The performance of each approach is demonstrated assessing expectation value and standard deviation of the carbon dioxide saturation against a reference statistic based on Monte Carlo sampling. We compare the convergence of all methods reporting on accuracy with respect to the number of model runs and resolution. Finally, we offer suggestions about the methods’ advantages and disadvantages that can guide the modeler for uncertainty quantification in carbon dioxide storage and beyond.

Abstract

This work proposes a data-driven method for enabling the efficient, stable time-parallel numerical solution of systems of ordinary differential equations (ODEs). The method assumes that low-dimensional bases that accurately capture the time evolution of the dynamical-system state are available. The method adopts the parareal framework for time parallelism, which is defined by an initialization method, a coarse propagator that advances solutions on a coarse time grid, and a fine propagator that operates on an underlying fine time grid. Rather than employing usual approaches for initialization and coarse propagation, we propose novel data-driven techniques that leverage the available time-evolution bases. The coarse propagator is defined by a forecast (proposed in K. Carlberg, J. Ray, and B. van Bloemen Waanders, Comput. Methods Appl. Mech. Engrg., 289 (2015), pp. 79--103) applied locally within each coarse time interval, which comprises the following steps: (1) apply the fine propagator for a small number of time steps, (2) approximate the state over the entire coarse time interval using gappy proper orthogonal decomposition (POD) with the local time-evolution bases, and (3) select the approximation at the end of the time interval as the propagated state. We also propose both local-forecast and global-forecast initialization techniques. The method is particularly well suited for POD-based reduced-order models (ROMs), as the time-evolution bases can be extracted from readily available data, i.e., the right singular vectors arising during POD computation. In addition to performing analyses related to the method's accuracy, speedup, stability, and convergence, we also numerically demonstrate the method's performance. Here, numerical experiments on ROMs for a nonlinear convection--reaction problem demonstrate the method's ability to realize near-ideal speedups.

Abstract

We consider feedback interconnections composed of a linear system whose input-output behavior is constrained by a relation. It is shown how to analyze stability of such interconnections by suitably refining the general framework of integral quadratic constraints. For the special case of static multivalued monotone nonlinearities, we reveal that one can employ Zames-Falb multipliers in order to substantially reduce conservatism if compared with passivity-based approaches and as illustrated with a numerical example. Our results are expected to open up an avenue towards a systematic stability analysis of data-integrated systems.

Abstract

We derive novel criteria for robust stability and output-feedback gain-scheduling controller synthesis for a class of switched systems that are affected by piecewise constant parameters with dwell-time constraints. Our findings are based on clock-dependent Lyapunov arguments and rely, in contrast to other approaches, on separation techniques from robust control involving dynamic multipliers. The obtained conditions are expressed as infinite-dimensional linear matrix inequalities which can be solved by, e.g., using sum-of-squares relaxation methods. We illustrate our results with a numerical example.

Abstract

Robust stability of a class of linear time-invariant systems affected by piecewise constant parameters with dwell-time constraints is considered. In contrast to other approaches the proposed result relies on separation techniques within the framework of integral quadratic constraints and is based on a novel version of the swapping lemma. In particular, our result allows to take additional knowledge on the variation of the parameter into account. The obtained conditions are expressed as infinite-dimensional linear matrix inequalities which can be solved, e.g., by using sum-of-squares relaxation methods. We illustrate the proposed approach with a numerical example.

Abstract

We present the results of atomistic molecular dynamics simulations with regard to specific ion effects in water, methanol and N,N-dimethylacetamide (DMAc). As a reference system, we introduce rigid and rod-like models of polyanions and polycations in combination with alkali metal cations and halide anions as counterions. Pronounced specific ion effects can be observed in terms of the individual anion and cation condensation behavior. The outcomes of our simulations thus reveal significant deviations from standard electrostatic mean-field theories. A detailed investigation of the individual energy contributions shows that ion–dipole interactions play a pivotal role in rationalizing the findings. The corresponding deviations in terms of the cation and anion distribution can be brought into agreement with the donor and acceptor numbers of the solvents, which thus highlights the importance of solvent–ion interactions in addition to electrostatic attraction.

Abstract

We introduce a framework to synthesize block-diagonal H∞-controllers with a parametric and a block-triangular dynamic component by convex optimization. This is motivated by a recent H∞-design approach for systems with delayed interconnections that are structured according to a strongly-connected communication graph. Our framework allows the extension to synthesis for interconnections described by multiple strongly-connected graphs that are themselves coupled through a nested delay structure.

Abstract

The tumour necrosis factor (TNF) ligand family member TNF-related apoptosis-inducing ligand (TRAIL) induces apoptosis predominantly in tumour cells, but not in normal tissues, representing therefore an attractive candidate for cancer therapy. The human TRAIL/TRAIL receptor system is very complex, four different membrane receptors bind the ligand. Two of these receptors, TRAIL-R1 and TRAIL-R2, transmit apoptotic but also non-apoptotic signals, whereas the other two, TRAIL-R3 and TRAIL-R4, act as inhibitors. Most tumour cells co-express several TRAIL receptors, allowing receptor interference. Several molecular mechanisms have been proposed by which TRAIL-R3 and TRAIL-R4 may counteract pro-apoptotic TRAIL signalling at the plasma membrane level, but possibly also intracellularly. A detailed understanding of the role of the individual TRAIL receptors and their interplay will be advantageous for the development of new TRAIL receptor agonists for cancer therapy. In fact, new TRAIL formulations will be needed since first clinical studies with soluble TRAIL or receptor agonistic antibodies showed only limited success. This review summarizes the complex TRAIL/TRAIL receptor system and the mechanisms by which TRAIL-R3 and TRAIL-R4 may interfere with TRAIL-mediated apoptosis induction. In addition, we discuss the prognostic and predictive value of TRAIL receptor expression in patients' tumour material.

Abstract

The need for extended liver resection is increasing due to the growing incidence of liver tumors in ageing societies. Individualized surgical planning is the key for identifying the optimal resection strategy and to minimize the risk of postoperative liver failure and tumor recurrence. Current computational tools provide virtual planning of liver resection by taking into account the spatial relationship between the tumor and the hepatic vascular trees, as well as the size of the future liver remnant. However, size and function of the liver are not necessarily equivalent. Hence, determining the future liver volume might misestimate the future liver function, especially in cases of hepatic comorbidities such as hepatic steatosis. A systems medicine approach could be applied, including biological, medical, and surgical aspects, by integrating all available anatomical and functional information of the individual patient. Such an approach holds promise for better prediction of postoperative liver function and hence improved risk assessment. This review provides an overview of mathematical models related to the liver and its function and explores their potential relevance for computational liver surgery. We first summarize key facts of hepatic anatomy, physiology, and pathology relevant for hepatic surgery, followed by a description of the computational tools currently used in liver surgical planning. Then we present selected state-of-the-art computational liver models potentially useful to support liver surgery. Finally, we discuss the main challenges that will need to be addressed when developing advanced computational planning tools in the context of liver surgery.

Abstract

We consider systems that consist of infinitely many identical subsystems over an infinite undirected graph. Stability analysis for such systems is performed and embedded into multiplier theory from robust control. This allows to apply recently developed gain-scheduling methods for dynamic multipliers in order to design stabilizing controllers and yields LMI conditions for their existence that are less conservative than existing ones.

Abstract

In this work we investigate the interplay of polynomial de-aliasing and sub-grid scale models for large eddy simulations based on discontinuous Galerkin discretizations. It is known that stability is a major concern when simulating underresolved turbulent flows with high order nodal collocation type discretizations. By changing the interpolatory character of the nodal collocation type discretization to a projection based discretization by increasing the number of quadrature points (polynomial de-aliasing), one is able to remove the aliasing induced stability problems. We focus on this effect and on the consequence for large eddy simulations with explicit subgrid scale models. Often, subgrid scale models have to achieve two possibly conflicting tasks in a single simulation: firstly stabilizing the numerics and secondly modeling the physical effect of the missing scales. Within a discontinuous Galerkin approach, it is possible to use either a fast (but potentially aliasing afflicted) nodal collocation discretization or a projection-based (but computationally costly) variant in combination with an explicit subgrid scale model. We use this framework to investigate the effect on the appropriate model parameter of a standard Smagorinsky subgrid scale model and of a Variational Multiscale Smagorinsky formulation. For this we first consider the 3-D viscous Taylor-Green vortex example to investigate the impact on the stability of the method and second the turbulent flow past a circular cylinder to investigate and compare the accuracy of the results. We show that the aliasing instabilities of collocative discretizations severely limit the choice of the model constant, in particular for high order schemes, while for de-aliased DG schemes, the closure model parameters can be chosen independently from the numerical scheme. For the cylinder flow, we also find that for the same model settings, the projection-based results are in better agreement with the reference DNS than those of the collocative scheme.

Abstract

This study focuses on a two-scale, continuum multicomponent model for the description of blood perfusion and cell metabolism in the liver. The model accounts for a spatial and time depending hydro-diffusion--advection--reaction description. We consider a solid-phase (tissue) containing glycogen and a fluid-phase (blood) containing glucose as well as lactate. The five-component model is enhanced by a two-scale approach including a macroscale (sinusoidal level) and a microscale (cell level). The perfusion on the macroscale within the lobules is described by a homogenized multiphasic approach based on the theory of porous media (mixture theory combined with the concept of volume fraction). On macro level, we recall the basic mixture model, the governing equations as well as the constitutive framework including the solid (tissue) stress, blood pressure and solutes chemical potential. In view of the transport phenomena, we discuss the blood flow including transverse isotropic permeability, as well as the transport of solute concentrations including diffusion and advection. The continuum multicomponent model on the macroscale finally leads to a coupled system of partial differential equations (PDE). In contrast, the hepatic metabolism on the microscale (cell level) was modeled via a coupled system of ordinary differential equations (ODE). Again, we recall the constitutive relations for cell metabolism level. A finite element implementation of this framework is used to provide an illustrative example, describing the spatial and time-depending perfusion--metabolism processes in liver lobules that integrates perfusion and metabolism of the liver.