Publications

2019
E. Papadopoulou, C. M. Megaridis, J. H. Walther, and P. Koumoutsakos, “Ultrafast Propulsion of Water Nanodroplets on Patterned Graphene,” ACS Nano, 2019. Publisher's Version
S. M. H. Hashemi, et al., “A versatile and membrane-less electrochemical reactor for the electrolysis of water and brine,” Energ. Environ. Sci. 2019. Publisher's VersionAbstract
Renewables challenge the management of energy supply and demand due to their intermittency. A promising solution is the direct conversion of the excess electrical energy into valuable chemicals in electrochemical reactors that are inexpensive, scalable, and compatible with irregular availability of electrical power. Membrane-less electrolyzers, deployed on a microfluidic platform, were recently shown to hold great promise for efficient electrolysis and cost-effective operation. The elimination of the membrane increases the reactor lifetime, reduces fabrication costs, and enables the deployment of liquid electrolytes with ionic conductivities that surpass those allowed by solid membranes. Here, we demonstrate a membrane-less architecture that enables unprecedented throughput by 3D printing a device that combines components such as the flow plates and the fluidic ports in a monolithic part while at the same time providing tight tolerances and smooth surfaces for precise flow conditioning. We show that inertial fluidic forces are effective even in milifluidic regimes and, therefore, are utilized to control the two-phase flows inside the device and prevent cross-contamination of the products. Simulations provide insight on governing fluid dynamics of coalescing bubbles and their rapid jumps away from the electrodes and help identify three key mechanisms for their fast and intriguing return towards the electrodes. Experiments and simulations are used to demonstrate the efficiency of the inertial separation mechanism in milichannels and at higher flow rates than in microchannels. We analyze the performance of the present device for two reactions: water splitting and the Chlor-Alkali process and find product purities of more than 99% and Faradaic efficiencies of more than 90%. The present membrane-less reactor - containing more efficient catalysts - provides close to 40 times higher throughput than its microfluidic counterpart and paves the way for realization of cost-effective and scalable electrochemical stacks that meet the performance and price targets of the renewable energy sector.
B. Rogie, W. B. Markussen, J. H. Walther, and M. R. Kærn, “Numerical investigation of air-side heat transfer and pressure drop characteristics of a new triangular finned microchannel evaporator with water drainage slits,” Fluids, vol. 4, no. 4, pp. 205, 2019. Publisher's VersionAbstract

The present study investigated a new microchannel profile design encompassing condensate drainage slits for improved moisture removal with use of triangular shaped plain fins. Heat transfer and pressure drop correlations were developed using computational fluid dynamics (CFD) and defined in terms of Colburn j-factor and Fanning f-factor. The microchannels were square 2.00 × 2.00 mm and placed with 4.50 mm longitudinal tube pitch. The transverse tube pitch and the triangular fin pitch were varied from 9.00 to 21.00 mm and 2.50 to 10.00 mm, respectively. Frontal velocity ranged from 1.47 to 4.40 m·s−1. The chosen evaporator geometry corresponds to evaporators for industrial refrigeration systems with long frosting periods. Furthermore, the CFD simulations covered the complete thermal entrance and developed regions, and made it possible to extract virtually infinite longitudinal heat transfer and pressure drop characteristics. The developed Colburn j-factor and Fanning f-factor correlations are able to predict the numerical results with 3.41% and 3.95% deviation, respectively.

H. Taghavifar, A. Nemati, and J. H. Walther, “Combustion and exergy analysis of multi-component diesel-DME-methanol blends in HCCI engine,” Energy, vol. 187, pp. 115951, 2019. Publisher's VersionAbstract

A homogeneous compression ignition (HCCI) engine is taken for numerical investigation on the application of renewable fuels contained blends of methanol and DME with the base diesel fuel, which will be replaced with diesel in different percentages. First, the combustion and engine performance of the engine for two and three-component fuels will be discussed and secondly, the simultaneous effect of EGR in 20% by mass and engine speed in two blends of having maximum and minimum diesel proportion are compared and examined. The results indicate that the replacement of diesel with 20% of DME and 30% by methanol (D50M30DME20) at 1400 rpm generates a greater pressure and accumulated heat (AHRpeak = 330.569 J), whereas D80M20/2000 rpm/EGR20 gives a defective combustive performance with poor engine efficiency (IMEP = 7.21 bar). The interesting point is that the proposed optimum blend of D50 can achieve the best performance with  35% mechanical efficiency of 35%. The case of D60M10DME30 though dominates in terms of RPR = 3.177 bar/deg and ignition delay (ID = 4.54 CA) that gives the highest exergy performance coefficient (EPC = 2.063) due to its high work and lowest irreversibility.

E. Wagemann, J. H. Walther, E. R. Cruz-Chú, and H. A. Zambrano, “Water flow in silica nanopores coated by carbon nanotubes from a wetting translucency perspective,” Journal of Physical Chemistry, vol. 123, no. 42, pp. 25635–25642, 2019. Publisher's VersionAbstract

Nearly frictionless water transport makes carbon nanotubes promising materials for use as conduits in nanofluidic applications. Here, we conduct molecular dynamics simulations of water flow within amorphous silica nanopores coated by a (39,39) single-walled carbon nanotube (SWCNT). Our atomistic models describe the interaction between water and pore walls based on two possible scenarios, translucency and opacity to wetting of a SWCNT. Simulation results indicate that the SWCNT coating enhances water flow through silica pores ca. 10 times compared to predictions from the classical Hagen–Poiseuille relation. By varying the strength of the water–pore interaction, we study the relationship between surface wettability and hydrodynamic slippage. We observe an increase in the slip length for higher values of water contact angle. Moreover, cases with SWCNT opacity and translucency to wetting display a substantial difference in the computed slippage, showing that the water contact angle is not the only factor that determines the slip boundary condition under nanoconfinement. We attribute this disparity to the corrugation of the potential energy landscape at the inner pore wall. The present study provides a theoretical framework for the use of carbon nanotube-based coatings in designing more efficient nanofluidic conduits.

C. S. Hemmingsen, et al., “Multiphase coupling of a reservoir simulator and computational fluid dynamics for accurate near-well flow,” Journal of Petroleum Science and Engineering, vol. 178, pp. 517–527, 2019. Publisher's VersionAbstract

Near-well flow analysis is an important tool for gaining detailed insight of the flow behaviour and for improving well design and production optimization of real reservoirs. One challenge of accurate numerical modelling of the flow field in the vicinity of the well is related to the scale disparity factor in space and time. The numerical scale gap between the reservoir and the wellbore justifies the representation of a well as a point or line sink/source term in traditional reservoir models. However, standard numerical techniques for reservoir simulation are incapable of resolving the near-singular character of the pressure field in the vicinity of the well. Under the assumption that all length scales have impact on flow patterns, we present a proof-of-concept study aimed at improving the quality of the numerical simulation by considering the geometry and fluid flow near the wellbore in a fully connected system, thus accounting for the fine scale phenomena by means of a hybrid Navier-Stokes/Darcy wellbore model coupled with a full scale reservoir model. A weak coupling method based on fixed-point iterations, that preserves the mass flux transport across the coupled interface, while adjusting productivity indices, is demonstrated via numerical experiments. Several different numerical experiments are performed to demonstrate the versatility and the improved well performance insight that the coupled method offers, including horizontal well inflow profile, influence of formation damage and optimal well configuration.

P. Karnakov, F. Wermelinger, M. Chatzimanolakis, S. Litvinov, and P. Koumoutsakos, “A High Performance Computing Framework for Multiphase, Turbulent Flows on Structured Grids,” in PASC '19: Proceedings of the Platform for Advanced Scientific Computing Conference, PASC '19, 2019, pp. 1-9. Publisher's VersionAbstract
We present a high performance computing framework for mul- tiphase, turbulent flows on structured grids. The computational methods are validated on a number of benchmark problems such as the Taylor-Green vortex that are extended by the inclusion of bubbles in the flow field. We examine the effect of bubbles on the turbulent kinetic energy dissipation rate and provide extensive data for bubble trajectories and velocities that may assist the develop- ment of engineering models. The implementation of the present solver on massively parallel, GPU enhanced architectures allows for large scale and high throughput simulations of multiphase flows.
A High Performance Computing Framework for Multiphase, Turbulent Flows on Structured Grids
G. Arampatzis, D. Wälchli, P. Weber, H. Rästas, and P. Koumoutsakos, “(Μ,Łambda)-CCMA-ES for Constrained Optimization with an Application in Pharmacodynamics,” in PASC '19: Proceedings of the Platform for Advanced Scientific Computing Conference, PASC '19, 2019, pp. 1-9. Publisher's VersionAbstract
We present the algorithm CCMA-ES, an extension to CMA-ES, an evolution strategy that has shown to perform well in a broad range of black-box optimization problems. The (µ, λ)-CMA-ES effectively handles nonlinear nonconvex functions but faces difficulties in constrained optimization problems. We introduce viability boundaries to improve the search for an initial point in the valid domain and adapt the covariance matrix using normal approximations to maintain the inequality constraints. Using benchmark problems from 2006 CEC we compare the performance of CCMA-ES with a state of the art optimization algorithm (mViE) showing favorable results. Finally, CCMA-ES is applied to a pharmacodynamics problem describing tumor growth, and we demonstrate that CCMA-ES outperforms mViE in terms of the objective function value and total function evaluations.
(Μ,Łambda)-CCMA-ES for Constrained Optimization with an Application in Pharmacodynamics
G. Novati and P. Koumoutsakos, “Remember and Forget for Experience Replay,” Proceedings of the 36th International Conference on Machine Learning, vol. 97. pp. 4851-4860, 2019. Publisher's VersionAbstract
Proceedings of the 36th International Conference on Machine Learning Experience replay (ER) is a fundamental component of off-policy deep reinforcement learning (RL). ER recalls experiences from past iterations to compute gradient estimates for the current policy, increasing data-efficiency. However, the accuracy of such updates may deteriorate when the policy diverges from past behaviors and can undermine the performance of ER. Many algorithms mitigate this issue by tuning hyper-parameters to slow down policy changes. An alternative is to actively enforce the similarity between policy and the experiences in the replay memory. We introduce Remember and Forget Experience Replay (ReF-ER), a novel method that can enhance RL algorithms with parameterized policies. ReF-ER (1) skips gradients computed from experiences that are too unlikely with the current policy and (2) regulates policy changes within a trust region of the replayed behaviors. We couple ReF-ER with Q-learning, deterministic policy gradient and off-policy gradient methods. We find that ReF-ER consistently improves the performance of continuous-action, off-policy RL on fully observable benchmarks and partially observable flow control problems.
C. Wen, N. Karvounis, J. H. Walther, Y. Yan, Y. Feng, and Y. Yang, “An efficient approach to separate CO2 using supersonic flows for carbon capture and storage,” Applied Energy, vol. 238, pp. 311–319, 2019. Publisher's VersionAbstract

The mitigation of CO2 emissions is an effective measure to solve the climate change issue. In the present study, we propose an alternative approach for CO2 capture by employing supersonic flows. For this purpose, we first develop a computational fluid dynamics (CFD) model to predict the CO2 condensing flow in a supersonic nozzle. Adding two transport equations to describe the liquid fraction and droplet number, the detailed numerical model can describe the heat and mass transfer characteristics during the CO2 phase change process under the supersonic expansion conditions. A comparative study is performed to evaluate the effect of CO2 condensation using the condensation model and dry gas assumption. The results show that the developed CFD model predicts accurately the distribution of the static temperature contrary to the dry gas assumption. Furthermore, the condensing flow model predicts a CO2 liquid fraction up to 18.6% of the total mass, which leads to the release of the latent heat to the vapour phase. The investigation performed in this study suggests that the CO2 condensation in supersonic flows provides an efficient and eco-friendly way to mitigate the CO2 emissions to the environment.

M. M. Hejlesen, G. Winckelmans, and J. H. Walther, “Non-singular green’s functions for the unbounded poisson equation in one, two and three dimensions,” Applied Mathematics Letters, vol. 89, pp. 28–34, 2019. Publisher's VersionAbstract

In this paper, we derive the non-singular Green’s functions for the unbounded Poisson equation in one, two and three dimensions using a spectral cut-off function approach to impose a minimum length scale in the homogeneous solution. The resulting non-singular Green’s functions are relevant to applications which are restricted to a minimum resolved length scale (e.g. a mesh size h) and thus cannot handle the singular Green’s function of the continuous Poisson equation. We furthermore derive the gradient vector of the non-singular Green’s function, as this is useful in applications where the Poisson equation represents potential functions of a vector field.

K. M. Pang, M. Jangi, X. Bai, J. Schramm, J. H. Walther, and P. Glarborg, “Effects of ambient pressure on ignition and flame characteristics in diesel spray combustion,” Fuel, vol. 237, pp. 676–685, 2019. Publisher's VersionAbstract

This work reports on numerical investigation of effects of ambient pressure (Pam) on spray combustion under engine-like conditions. Three cases with different Pam of 42, 85 and 170 bar at a fixed ambient temperature of 1000 K are considered. Zero-dimensional calculations are first performed for autoignition of stagnant adiabatic homogenous mixtures to evaluate performance of the selected diesel surrogate fuel models and to identify the Pam effects on the most reactive mixture. An Eulerian-based transported probability density function model is then chosen for the three-dimensional computational fluid dynamics study. The results show the predicted ignition delay times and flame lift-off lengths are in reasonably good agreement with experiment, with the relative difference below 28%. The current work reveals that low-temperature reactions occur across a wide range of mixture fraction but a noticeable rise of temperature (>100 K above ambient temperature) is detected first on the fuel-lean side of the stoichiometric line in all three cases. The high-temperature ignition occurs first on the fuel-rich side in the 42 and 85 bar cases, where the igniting mixture appears to be more fuel-rich in the latter case. As Pam is further increased to 170 bar, the igniting mixture becomes more fuel-lean and the high-temperature ignition occurs on the fuel-lean side. The ignition behavior is found to depend on both physical and chemical processes. At 170 bar, the reaction rate increases and the associated transition from low- to high-temperature ignition is relatively fast, as compared to the transport of warmer products from the lean zone into the fuel-rich mixture. Also, within the fuel-rich region, the local temperature is low due to liquid fuel vaporization and the condition is not appropriate for ignition. These collectively cause the high-temperature ignition to occur on the fuel-lean side. Analyses on the quasi-steady spray flame structures reveal that, apart from poorer air entrainment due to reduced lift-off length, the higher rich-zone temperature and lower scalar dissipation rate also lead to a higher peak soot volume fraction at higher Pam.

S. S. Asadzadeh, P. S. Larsen, H. U. Riisgård, and J. H. Walther, “Hydrodynamics of the leucon sponge pump,” Journal of The Royal Society Interface, vol. 16, no. 150, pp. 20180630, 2019. Publisher's VersionAbstract
Leuconoid sponges are filter-feeders with a complex system of branching inhalant and exhalant canals leading to and from the close-packed choanocyte chambers. Each of these choanocyte chambers holds many choanocytes that act as pumping units delivering the relatively high pressure rise needed to overcome the system pressure losses in canals and constrictions. Here, we test the hypothesis that, in order to deliver the high pressures observed, each choanocyte operates as a leaky, positive displacement-type pump owing to the interaction between its beating flagellar vane and the collar, open at the base for inflow but sealed above. The leaking backflow is caused by small gaps between the vaned flagellum and the collar. The choanocyte pumps act in parallel, each delivering the same high pressure, because low-pressure and high-pressure zones in the choanocyte chamber are separated by a seal (secondary reticulum). A simple analytical model is derived for the pump characteristic, and by imposing an estimated system characteristic we obtain the back-pressure characteristic that shows good agreement with available experimental data. Computational fluid dynamics is used to verify a simple model for the dependence of leak flow through gaps in a conceptual collar–vane–flagellum system and then applied to models of a choanocyte tailored to the parameters of the freshwater demosponge Spongilla lacustris to study its flows in detail. It is found that both the impermeable glycocalyx mesh covering the upper part of the collar and the secondary reticulum are indispensable features for the choanocyte pump to deliver the observed high pressures. Finally, the mechanical pump power expended by the beating flagellum is compared with the useful (reversible) pumping power received by the water flow to arrive at a typical mechanical pump efficiency of about 70%.
S. S. Asadzadeh, et al., “Hydrodynamic functionality of the lorica in choanoflagellates,” Journal of The Royal Society Interface, vol. 16, no. 150, pp. 20180478, 2019. Publisher's VersionAbstract
Choanoflagellates are unicellular eukaryotes that are ubiquitous in aquatic habitats. They have a single flagellum that creates a flow toward a collar filter composed of filter strands that extend from the cell. In one common group, the loricate choanoflagellates, the cell is suspended in an elaborate basket-like structure, the lorica, the function of which remains unknown. Here, we use Computational Fluid Dynamics to explore the possible hydrodynamic function of the lorica. We use the choanoflagellate Diaphaoneca grandis as a model organism. It has been hypothesized that the function of the lorica is to prevent refiltration (flow recirculation) and to increase the drag and, hence, increase the feeding rate and reduce the swimming speed. We find no support for these hypotheses. On the contrary, motile prey are encountered at a much lower rate by the loricate organism. The presence of the lorica does not affect the average swimming speed, but it suppresses the lateral motion and rotation of the cell. Without the lorica, the cell jiggles from side to side while swimming. The unsteady flow generated by the beating flagellum causes reversed flow through the collar filter that may wash away captured prey while it is being transported to the cell body for engulfment. The lorica substantially decreases such flow, hence it potentially increases the capture efficiency. This may be the main adaptive value of the lorica.
2018
W. Byeon, Q. Wang, R. Kumar Srivastava, and P. Koumoutsakos, “ContextVP: Fully Context-Aware Video Prediction,” in Computer Vision – ECCV 2018, Springer, 2018, pp. 781–797. Publisher's Version
Z. Y. Wan, P. R. Vlachas, P. Koumoutsakos, and T. P. Sapsis, “Data-assisted reduced-order modeling of extreme events in complex dynamical systems,” PLoS ONE, vol. 13, no. 5, pp. 1-22, 2018. Publisher's VersionAbstract
The prediction of extreme events, from avalanches and droughts to tsunamis and epidemics, depends on the formulation and analysis of relevant, complex dynamical systems. Such dynamical systems are characterized by high intrinsic dimensionality with extreme events having the form of rare transitions that are several standard deviations away from the mean. Such systems are not amenable to classical order-reduction methods through projection of the governing equations due to the large intrinsic dimensionality of the underlying attractor as well as the complexity of the transient events. Alternatively, data-driven techniques aim to quantify the dynamics of specific, critical modes by utilizing data-streams and by expanding the dimensionality of the reduced-order model using delayed coordinates. In turn, these methods have major limitations in regions of the phase space with sparse data, which is the case for extreme events. In this work, we develop a novel hybrid framework that complements an imperfect reduced order model, with data-streams that are integrated though a recurrent neural network (RNN) architecture. The reduced order model has the form of projected equations into a low-dimensional subspace that still contains important dynamical information about the system and it is expanded by a long short-term memory (LSTM) regularization. The LSTM-RNN is trained by analyzing the mismatch between the imperfect model and the data-streams, projected to the reduced-order space. The data-driven model assists the imperfect model in regions where data is available, while for locations where data is sparse the imperfect model still provides a baseline for the prediction of the system state. We assess the developed framework on two challenging prototype systems exhibiting extreme events. We show that the blended approach has improved performance compared with methods that use either data streams or the imperfect model alone. Notably the improvement is more significant in regions associated with extreme events, where data is sparse.
P. R. Vlachas, W. Byeon, Z. Y. Wan, T. P. Sapsis, and P. Koumoutsakos, “Data-driven forecasting of high-dimensional chaotic systems with long short-term memory networks,” P. Roy. Soc. A-Math. Phy. vol. 474, no. 2213, pp. 20170844, 2018. Publisher's VersionAbstract
We introduce a data-driven forecasting method for high-dimensional chaotic systems using long shortterm memory (LSTM) recurrent neural networks. The proposed LSTM neural networks perform inference of high-dimensional dynamical systems in their reduced order space and are shown to be an effective set of nonlinear approximators of their attractor. We demonstrate the forecasting performance of the LSTM and compare it with Gaussian processes (GPs) in time series obtained from the Lorenz 96 system, the Kuramoto–Sivashinsky equation and a prototype climate model. The LSTM networks outperform the GPs in short-term forecasting accuracy in all applications considered. A hybrid architecture, extending the LSTM with a mean stochastic model (MSM–LSTM), is proposed to ensure convergence to the invariant measure. This novel hybrid method is fully data-driven and extends the forecasting capabilities of LSTM networks.
S. Verma, G. Novati, and P. Koumoutsakos, “Efficient collective swimming by harnessing vortices through deep reinforcement learning,” P. Natl. Acad. Sci. pp. 201800923, 2018. Publisher's Version
S. Wu, P. Angelikopoulos, J. L. Beck, and P. Koumoutsakos, “Hierarchical Stochastic Model in Bayesian Inference for Engineering Applications: Theoretical Implications and Efficient Approximation,” ASCE-ASME J. Risk Uncertain. Eng. Sys. B, vol. 5, no. 1, pp. 011006, 2018. Publisher's Version
G. Arampatzis, D. Wälchli, P. Angelikopoulos, S. Wu, P. Hadjidoukas, and P. Koumoutsakos, “Langevin Diffusion for Population Based Sampling with an Application in Bayesian Inference for Pharmacodynamics,” SIAM J. Sci. Comput. vol. 40, no. 3, pp. B788–B811, 2018. Publisher's Version

Pages