Publications

2022
P. R. Vlachas, G. Arampatzis, C. Uhler, and P. Koumoutsakos, “Multiscale simulations of complex systems by learning their effective dynamics,” Nat Mach Intell, 2022. Publisher's VersionAbstract
Predictive simulations of complex systems are essential for applications ranging from weather forecasting to drug design. The veracity of these predictions hinges on their capacity to capture effective system dynamics. Massively parallel simulations predict the system dynamics by resolving all spatiotemporal scales, often at a cost that prevents experimentation, while their findings may not allow for generalization. On the other hand, reduced-order models are fast but limited by the frequently adopted linearization of the system dynamics and the utilization of heuristic closures. Here we present a novel systematic framework that bridges large-scale simulations and reduced-order models to learn the effective dynamics of diverse, complex systems. The framework forms algorithmic alloys between nonlinear machine learning algorithms and the equation-free approach for modelling complex systems. Learning the effective dynamics deploys autoencoders to formulate a mapping between fine- and coarse-grained representations and evolves the latent space dynamics using recurrent neural networks. The algorithm is validated on benchmark problems, and we find that it outperforms state-of-the-art reduced-order models in terms of predictability, and large-scale simulations in terms of cost. Learning the effective dynamics is applicable to systems ranging from chemistry to fluid mechanics and reduces the computational effort by up to two orders of magnitude while maintaining the prediction accuracy of the full system dynamics. We argue that learning the effective dynamics provides a potent novel modality for accurately predicting complex systems.
Multiscale simulations of complex systems by learning their effective dynamics
J. Lipkova, B. Menze, B. Wiestler, P. Koumoutsakos, and J. Lowengrub, “Modelling glioma progression, mass effect and intracranial pressure in patient anatomy,” J. R. Soc. Interface, vol. 19, no. 188, 2022. Publisher's VersionAbstract
Increased intracranial pressure is the source of most critical symptoms in patients with glioma, and often the main cause of death. Clinical interventions could benefit from non-invasive estimates of the pressure distribution in the patient's parenchyma provided by computational models. However, existing glioma models do not simulate the pressure distribution and they rely on a large number of model parameters, which complicates their calibration from available patient data. Here we present a novel model for glioma growth, pressure distribution and corresponding brain deformation. The distinct feature of our approach is that the pressure is directly derived from tumour dynamics and patient-specific anatomy, providing non-invasive insights into the patient's state. The model predictions allow estimation of critical conditions such as intracranial hypertension, brain midline shift or neurological and cognitive impairments. A diffuse-domain formalism is employed to allow for efficient numerical implementation of the model in the patient-specific brain anatomy. The model is tested on synthetic and clinical cases. To facilitate clinical deployment, a high-performance computing implementation of the model has been publicly released.
Modelling glioma progression, mass effect and intracranial pressure in patient anatomy
J. Bae and P. KOUMOUTSAKOS, “Scientific multi-agent reinforcement learning for wall-models of turbulent flows,” Nature Communications, vol. 13, 2022. Publisher's VersionAbstract

The predictive capabilities of turbulent flow simulations, critical for aerodynamic design and weather prediction, hinge on the choice of turbulence models. The abundance of data from experiments and simulations and the advent of machine learning have provided a boost to turbulence modeling efforts. However, simulations of turbulent flows remain hindered by the inability of heuristics and supervised learning to model the near-wall dynamics. We address this challenge by introducing scientific multi-agent reinforcement learning (SciMARL) for the discovery of wall models for large-eddy simulations (LES). In SciMARL, discretization points act also as cooperating agents that learn to supply the LES closure model. The agents self- learn using limited data and generalize to extreme Reynolds numbers and previously unseen geometries. The present simulations reduce by several orders of magnitude the computa- tional cost over fully-resolved simulations while reproducing key flow quantities. We believe that SciMARL creates unprecedented capabilities for the simulation of turbulent flows.

Scientific multi-agent reinforcement learning for wall-models of turbulent flows
P. R. Vlachas, J. Zavadlav, M. Praprotnik, and P. Koumoutsakos, “Accelerated Simulations of Molecular Systems through Learning of Effective Dynamics,” J. Chem. Theory Comput. vol. 18, no. 1, pp. 538-549, 2022. Publisher's VersionAbstract
Simulations are vital for understanding and predicting the evolution of complex molecular systems. However, despite advances in algorithms and special purpose hardware, accessing the time scales necessary to capture the structural evolution of biomolecules remains a daunting task. In this work, we present a novel framework to advance simulation time scales by up to 3 orders of magnitude by learning the effective dynamics (LED) of molecular systems. LED augments the equation-free methodology by employing a probabilistic mapping between coarse and fine scales using mixture density network (MDN) autoencoders and evolves the non-Markovian latent dynamics using long short-term memory MDNs. We demonstrate the effectiveness of LED in the Müller–Brown potential, the Trp cage protein, and the alanine dipeptide. LED identifies explainable reduced-order representations, i.e., collective variables, and can generate, at any instant, all-atom molecular trajectories consistent with the collective variables. We believe that the proposed framework provides a dramatic increase to simulation capabilities and opens new horizons for the effective modeling of complex molecular systems.
Accelerated Simulations of Molecular Systems through Learning of Effective Dynamics
2021
K. Daniilidis, et al., “Robotics in the AI era: a vision for a hellenic robotics initiative,” Found. Trends Mach. Learn. vol. 9, no. 3, pp. 201-265, 2021. Publisher's VersionAbstract

In January 2021, the Hellenic Institute of Advanced Study (HIAS) assembled a panel including world leading roboticists from the Hellenic diaspora, who volunteered their scientific expertise to provide a vision for Robotics in Greece. This monograph, entitled “Robotics in the Artificial Intelligence (AI) era,” will hopefully trigger a dialogue towards the development of a national robotics strategy. Our vision is that Robotics in the AI era will be an essential technology of the future for the safety and security of the Hellenic nation, its environment and its citizens, for modernizing its economy towards Industry 4.0, and for inspiring and educating the next generation workforce for the challenges of the 21st century. To contribute towards making this vision a reality, after reviewing global trends in robotics and assessing the Greek robotics ecosystem, we arrived at the following key findings and recommendations: Firstly, we think that Greece should develop a national Hellenic Robotics Initiative that serves as the nation’s long-term vision and strategy across the entire Greek robotics ecosystem. Also, certain societal drivers should be key in the areas of focus. Safety and security is an area of national importance necessitating a national initiative, while agrifood, maritime and logistics provide opportunities for internationally leading innovation.

We recommend the establishment of a mission-driven, government-funded, organization advancing unmanned vehicles in societal drivers of national importance, and Greece should leverage its unique geography and become a living testbed of robotics innovation turning the country into a development site for exportable technologies. In our opinion, universities should create Centers of Excellence in robotics and AI as well as consider innovation-leading research institutes such as the Italian Institute of Technology. We recommend investing in robotics education using Maker Spaces in order to prepare the workforce with 21st century skills to become Industry 4.0 innovators. Furthermore, we believe that the government should collect, measure, and analyze data on the robotics industry, robotics uses, labor shifts, and brain gain and promote awareness via a Hellenic Robotics Day. And finally, the government should regulate robot safety without stifling innovation, provide safe experimentation areas and mechanisms for certifying safety of locally developed robots. This monograph has many additional suggestions that enhance the above main recommendations. As authors, we advocate bringing the robotics ecosystem together in order to sharpen and expand these findings to an ambitious, long-term, and detailed national strategy and roadmap for robotics in the AI era.

Robotics in the AI era: a vision for a hellenic robotics initiative
L. Amoudruz and P. Koumoutsakos, “Independent Control and Path Planning of Microswimmers with a Uniform Magnetic Field,” Advanced Intelligent Systems, vol. 2100183, 2021. Publisher's VersionAbstract
Artificial bacteria flagella (ABFs) are magnetic helical microswimmers that can be remotely controlled via a uniform, rotating magnetic field. Previous studies have used the heterogeneous response of microswimmers to external magnetic fields for achieving independent control. Herein, analytical and reinforcement learning control strategies for path planning to a target by multiple swimmers using a uniform magnetic field are introduced. The comparison of the two algorithms shows the superiority of reinforcement learning in achieving minimal travel time to a target. The results demonstrate, for the first time, the effective independent navigation of realistic microswimmers with a uniform magnetic field in a viscous flow field.
Independent Control and Path Planning of Microswimmers with a Uniform Magnetic Field
Y. Yang, N. Karvounis, J. H. Walther, H. Ding, and C. Wen, “Effect of area ratio of the primary nozzle on steam ejector performance considering nonequilibrium condensations,” Energy, vol. 237, pp. 121483, 2021. Publisher's VersionAbstract
The formation and evaporation of nanodroplets in steam ejectors is neglected in many numerical simulations. We analyse the influence of a primary nozzle on steam ejector performances considering phase change processes. The numerical model is validated in detail against experimental data of supersonic nozzles and steam ejectors available in the literature. The results show that the first nonequilibrium condensation is observed within the primary nozzle, while under-expanded supersonic flow causes a second nucleation-condensation process to achieve a large liquid fraction of 0.26 in the steam ejector. The compression process of the supersonic flow results in a steep decrease of the degree of subcooling leading to droplet evaporations. The condensation and evaporation processes repeat alternatively depending on the flow behaviour in the mixing section. The increasing area ratio leads to the transition of the flow structure from under-expanded flows to over-expanded flows in the mixing section. The droplet diameter is about 7 nm in the constant section and the entrainment ratio can reach approximately 0.75 for an area ratio of 8, which achieves a good performance of the steam ejector.
P. Gunnarson, I. Mandralis, G. Novati, P. Koumoutsakos, and J. Dabiri, “Learning efficient navigation in vortical flow fields,” Nature Communications, vol. 12, no. 1, 2021. Publisher's VersionAbstract

Efficient point-to-point navigation in the presence of a background flow field is important for robotic applications such as ocean surveying. In such applications, robots may only have knowledge of their immediate surroundings or be faced with time-varying currents, which limits the use of optimal control techniques. Here, we apply a recently introduced Reinforcement Learning algorithm to discover time-efficient navigation policies to steer a fixed- speed swimmer through unsteady two-dimensional flow fields. The algorithm entails input- ting environmental cues into a deep neural network that determines the swimmer’s actions, and deploying Remember and Forget Experience Replay. We find that the resulting swimmers successfully exploit the background flow to reach the target, but that this success depends on the sensed environmental cue. Surprisingly, a velocity sensing approach significantly out- performed a bio-mimetic vorticity sensing approach, and achieved a near 100% success rate in reaching the target locations while approaching the time-efficiency of optimal navigation trajectories. 

 

Learning efficient navigation in vortical flow fields
S. M. Martin, D. Wälchli, G. Arampatzis, A. E. Economides, P. Karnakov, and P. Koumoutsakos, “Efficient and scalable software framework for Bayesian uncertainty quantification and stochastic optimization,” Comput. Method. Appl. M. 2021. Publisher's VersionAbstract

We present Korali, an open-source framework for large-scale Bayesian uncertainty quantification and stochastic optimization. The framework relies on non-intrusive sampling of complex multiphysics models and enables their exploitation for optimization and decision-making. In addition, its distributed sampling engine makes efficient use of massively-parallel architectures while introducing novel fault tolerance and load balancing mechanisms. We demonstrate these features by interfacing Korali with existing high-performance software such as APHROS, LAMMPS (CPU-based), and MIRHEO (GPU-based) and show efficient scaling for up to 512 nodes of the CSCS Piz Daint supercomputer. Finally, we present benchmarks demonstrating that Korali outperforms related state-of-the-art software frameworks.

Korali: Efficient and scalable software framework for Bayesian uncertainty quantification and stochastic optimization
E. Papadopoulou, J. Zavadlav, R. Podgornik, M. Praprotnik, and P. Koumoutsakos, “Tuning the Dielectric Response of Water in Nanoconfinement through Surface Wettability,” ACS Nano, 2021. Publisher's VersionAbstract

The tunable polarity of water can be exploited in emerging technologies including catalysis, gas storage, and green chemistry. Recent experimental and theoretical studies have shown that water can be rendered into an effectively apolar solvent under nanoconfinement. We furthermore demonstrate, through molecular simulations, that the static dielectric constant of water can be modified by changing the wettability of the confining material. We find the out-of-plane dielectric response to be highly sensitive to the level of confinement and can be reduced up to 40× , in accordance with experimental data. By altering the surface wettability from superhydrophilic to super- hydrophobic, we observe a 36% increase for the out-of-plane and a 31% decrease for the in-plane dielectric constants. Our findings demonstrate the feasibility of tunable water polarity, a phenomenon with great potential for scientific and technological impact.

Tuning the Dielectric Response of Water in Nanoconfinement through Surface Wettability
E. Papadopoulou, C. M. Megaridis, J. H. Walther, and P. Koumoutsakos, “Nanopumps without Pressure Gradients: Ultrafast Transport of Water in Patterned Nanotubes,” J. Phys, Chem. B, 2021. Publisher's VersionAbstract
The extreme liquid transport properties of carbon nanotubes present new opportunities for surpassing conventional technologies in water filtration and purification. We demonstrate that carbon nanotubes with wettability surface patterns act as nanopumps for the ultrafast transport of picoliter water droplets without requiring externally imposed pressure gradients. Large-scale molecular dynamics simulations evidence unprecedented speeds and accelerations on the order of 1010 g of droplet propulsion caused by interfacial energy gradients. This phenomenon is persistent for nanotubes of varying sizes, stepwise pattern configurations, and initial conditions. We present a scaling law for water transport as a function of wettability gradients through simple models for the droplet dynamic contact angle and friction coefficient. Our results show that patterned nanotubes are energy-efficient nanopumps offering a realistic path toward ultrafast water nanofiltration and precision drug delivery.
I. Mandralis, P. Weber, G. Novati, and P. Koumoutsakos, “Learning swimming escape patterns for larval fish under energy constraints,” Phys. Rev. Fluids, vol. 6, pp. 093101, 2021. Publisher's VersionAbstract
Swimming organisms can escape their predators by creating and harnessing unsteady flow fields through their body motions. Stochastic optimization and flow simulations have identified escape patterns that are consistent with those observed in natural larval swimmers. However, these patterns have been limited by the specification of a particular cost function and depend on a prescribed functional form of the body motion. Here, we deploy reinforcement learning to discover swimmer escape patterns for larval fish under energy constraints. The identified patterns include the C-start mechanism, in addition to more energetically efficient escapes. We find that maximizing distance with limited energy requires swimming via short bursts of accelerating motion interlinked with phases of gliding. The present, data efficient, reinforcement learning algorithm results in an array of patterns that reveal practical flow optimization principles for efficient swimming and the methodology can be transferred to the control of aquatic robotic devices operating under energy constraints.
K. Larson, et al., “Data-driven prediction and origin identification of epidemics in population networks,” Royal Society Open Science, 2021. Publisher's VersionAbstract
Effective intervention strategies for epidemics rely on the identification of their origin and on the robustness of the predictions made by network disease models. We introduce a Bayesian uncertainty quantification framework to infer model parameters for a disease spreading on a network of communities from limited, noisy observations; the state-of-the-art computational framework compensates for the model complexity by exploiting massively parallel computing architectures. Using noisy, synthetic data, we show the potential of the approach to perform robust model fitting and additionally demonstrate that we can effectively identify the disease origin via Bayesian model selection. As disease-related data are increasingly available, the proposed framework has broad practical relevance for the prediction and management of epidemics.
J. C. Ong, K. M. Pang, X. - S. Bai, M. Jangid, and J. H. Walther, “Large-eddy simulation of n-dodecane spray flame: Effects of nozzle diameter on autoignition at varying ambient temperatures,” Proceedings of the Combustion Institute, vol. 38, no. 2. pp. 3427-3434, 2021. Publisher's VersionAbstract
In the present study, large-eddy simulations (LES) are used to identify the underlying mechanism that governs the ignition phenomena of spray flames from different nozzle diameters when the ambient temperature (Tam) varies. Two nozzle sizes of 90µm and 186µm are chosen. They correspond to the nozzle sizes used by Spray A and Spray D, respectively, in the Engine Combustion Network. LES studies of both nozzles are performed at three Tam of 800K, 900K, and 1000K. The numerical models are validated using the experimental liquid and vapor penetration, mixture fraction (Z) distribution, as well as ignition delay time (IDT). The ignition characteristics of both Spray A and Spray D are well predicted, with a maximum relative difference of 14% as compared to the experiments. The simulations also predict the annular ignition sites for Spray D at Tam ⩾ 900K, which is consistent with the experimental observation. It is found that the mixture with Z ⩽ 0.2 at the spray periphery is more favorable for ignition to occur than the overly fuel-rich mixture of Z > 0.2 formed in the core of spray. This leads to the annular ignition sites at higher Tam. Significantly longer IDT for Spray D is obtained at Tam of 800K due to higher scalar dissipation rates (χ) during high temperature (HT) ignition. The maximum χ during HT ignition for Spray D is larger than that in Spray A by approximately a factor of 5. In contrast, at K, the χ values are similar between Spray A and Spray D. This elucidates the increase in the difference of IDT between Spray D and Spray A as Tam decreases. This may explain the contradicting findings on the effects of nozzle diameters on IDT from literature.
A. Nemati, M. V. Jensen, K. M. Pang, and J. H. Walther, “Conjugate heat transfer simulation of sulfuric acid condensation in a large two-stroke marine engine - the effect of thermal initial condition,” Applied Thermal Engineering, vol. 195, pp. 117075, 2021. Publisher's VersionAbstract
In the present study, conjugate heat transfer (CHT) calculations are applied in a computational fluid dynamics (CFD) simulation to simultaneously solve the in-cylinder gas phase dynamics and the temperature field within the liner of the engine. The effects of different initial temperatures with linear profiles across the liner are investigated on the wall heat transfer as well as on the sulfuric acid formation and condensation. The temporal and spatial behavior of sulfuric acid condensation on the liner suggests the importance of CHT calculations under large two-stroke marine engine relevant conditions. Comparing the mean value of the heat transfer through the inner and outer sides of the liner, an initial temperature difference of 15 K with a linear profile is an appropriate initial condition to initiate the temperature within the liner. Moreover, the effect of the amount of water vapor in the air on the sulfuric acid formation and condensation is studied. The current results show that the sulfuric acid vapor formation is more sensitive to the variation of the water vapor amount than the sulfuric acid condensation.
M. P. Brenner and P. Koumoutsakos, “Editorial: Machine Learning and Physical Review Fluids: An Editorial Perspective,” Physical Review Fluids, vol. 6, no. 7, pp. 070001, 2021. Publisher's Version Editorial: Machine Learning and Physical Review Fluids: An Editorial Perspective
J. C. Ong, K. M. Pang, M. Jangi, X. - S. Bai, and J. H. Walther, “Numerical study of the influence of turbulence-chemistry interaction model on {URANS} Simulations of diesel spray flame structures under marine engine-like conditions,” Energy Fuels, vol. 35, no. 14, pp. 11457–11467, 2021. Publisher's VersionAbstract
The present work performs unsteady Reynolds-averaged Navier–Stokes simulations to study the effect of turbulence–chemistry interaction (TCI) on diesel spray flames. Three nozzle diameters (d0) of 100, 180, and 363 μm are considered in the present study. The Eulerian stochastic fields (ESF) method (with the TCI effect) and well-stirred reactor (WSR) model (without the TCI effect) are considered in the present work. The model evaluation is carried out for ambient gas densities (ρam) of 30.0 and 58.5 kg/m3. The ESF method is demonstrated to be able to reproduce the ignition delay time (IDT) and lift-off length (LOL) with an improved accuracy than that from the WSR method. Furthermore, TCI has relatively more influence on LOL than on IDT. A normalized LOL (LOL*) is introduced, which considers the effect of d0, and its subsequent effect on the fuel-richness in the rich premixed core region is analyzed. The RO2 distribution is less influenced by the TCI effect as ambient density increases. The ESF model generally predicts a longer and wider CH2O distribution. The difference in the spatial distribution of CH2O between the ESF and WSR model diminishes as d0 increases. At ρam = 30.0 kg/m3, the ESF method results in a broader region of OH with lower peak OH values than in the WSR case. However, at ρam = 58.5 kg/m3, the variation of the peak OH value is less susceptible to the increase in d0 and the presence of the TCI model. Furthermore, the influence of TCI on the total OH mass decreases as d0 increases. The total NOx mass qualitatively follows the same trend as the total OH mass. This present work clearly shows that the influence of TCI on the global spray and combustion characteristics becomes less prominent when d0 increases.
A. Khosronejad, S. Kang, F. Wermelinger, P. Koumoutsakos, and F. Sotiropoulos, “A computational study of expiratory particle transport and vortex dynamics during breathing with and without face masks,” Physics of Fluids, vol. 33, no. 6, pp. 066605, 2021. Publisher's VersionAbstract

We present high-fidelity numerical simulations of expiratory biosol transport during normal breathing under indoor, stagnant air conditions with and without a facile mask. We investigate mask efficacy to suppress the spread of saliva particles that is underpinnings existing social distancing recommendations. The present simulations incorporate the effect of human anatomy and consider a spectrum of saliva particulate sizes that range from 0.1 to 10 μm while also accounting for their evaporation. The simulations elucidate the vorticity dynamics of human breathing and show that without a facile mask, saliva particulates could travel over 2.2 m away from the person. However, a non-medical grade face mask can drastically reduce saliva particulate propagation to 0.72 m away from the person. This study provides new quantitative evidence that facile masks can successfully suppress the spreading of saliva particulates due to normal breathing in indoor environments.

A computational study of expiratory particle transport and vortex dynamics during breathing with and without face masks
M. Rønne, J. H. L. Allan, and Walther, J.H., “The nose-up effect in twin-box bridge flutter –- experimental observations and theoretical model,” Wind and structures, vol. 32, no. 4, pp. 293–308, 2021. Publisher's VersionAbstract
For the past three decades a significant amount of research has been conducted on bridge flutter. Wind tunnel tests for a 2000 m class twin-box suspension bridge have revealed that a twin-box deck carrying 4 m tall 50% open area ratio wind screens at the deck edges achieved higher critical wind speeds for onset of flutter than a similar deck without wind screens. A result at odds with the well-known behavior for the mono-box deck. The wind tunnel tests also revealed that the critical flutter wind speed increased if the bridge deck assumed a nose-up twist relative to horizontal when exposed to high wind speeds – a phenomenon termed the “nose-up” effect. Static wind tunnel tests of this twin-box cross section revealed a positive moment coefficient at 0⁰ angle of attack as well as a positive moment slope, ensuring that the elastically supported deck would always meet the mean wind flow at ever increasing mean angles of attack for increasing wind speeds. The aerodynamic action of the wind screens on the twin-box bridge girder is believed to create the observed nose-up aerodynamic moment at 0⁰ angle of attack. The present paper reviews the findings of the wind tunnel tests with a view to gain physical insight into the “nose-up” effect and to establish a theoretical model based on numerical simulations allowing flutter predictions for the twin-box bridge girder.
B. U. Anabaraonye, J. R. Bentzon, I. Khaliqdad, K. L. Feilberg, S. I. Andersen, and J. H. Walther, “The influence of turbulent transport in reactive processes: a combined numerical and experimental investigation in a Taylor-Couette reactor,” Chemical engineering journal, vol. 421, pp. 129591, 2021. Publisher's VersionAbstract
Turbulent reactive flows are ubiquitous in industrial processes. Decoupling transport effects from intrinsic chemical reactions requires an in-depth understanding of fluid flow physics; computational fluid dynamics (CFD) methods have been widely used for this purpose. Most CFD simulations of reactive liquid-phase flows, where the Schmidt numbers are large, rely on isotropic eddy viscosity models. However, the assumption of turbulent isotropy in most stirred reactors and wall-bounded flows is fundamentally incorrect and leads to erroneous re- sults. Here, we apply a systematic CFD approach to simulate liquid-phase diffusive and convective transport phenomena that occur in a Taylor-Couette (TC) reactor. We resolve the turbulent flow by extracting statistics from large eddy simulation which is used to tune the anisotropic Reynolds stress model. In addition, we con- ducted a series of turbulent precipitation and mixing studies in a TC reactor that was designed and fabricated in- house. The numerical model is successfully validated against a published torque correlation and it is found to accurately describe the advection and diffusion of chemical species. The validated model is then used to demonstrate key flow properties in the reactor. We define new local turbulent Pecl ́et numbers to characterize the relative increase in diffusivity from turbulent advection and observe a 29% increase in the turbulent contribution as Reynolds number is doubled. Both reactive simulations and experiments show an increase in overall reaction rates with increased turbulence. The results from reactive simulations provide a deeper understanding of flow- kinetics interactions at turbulent conditions.

Pages