Publications

A computational fluid dynamics study of the scavenging process in a large two-stroke marine engine is presented in this work. Scavenging which is one of the key processes in the two-stroke marine engines, has a direct effect on fuel economy and emissions. This process is responsible for fresh air delivery, removing the combustion products from the cylinder, cooling the combustion chamber surfaces and providing a swirling flow for better air-fuel mixing. Therefore, having a better understanding of this process and the associated flow pattern is crucial. This is not achievable solely by experimental tests for large engines during engine operation due to the difficulties of measuring the flow field inside the cylinder. In this study, the axial and tangential velocities are compared and validated with the experimental results obtained from Particle Image Velocimetry (PIV) tests [1]. The simulations are conducted using both Unsteady Reynolds Averaged Navier Stokes (URANS) and Large Eddy Simulation (LES) turbulence models. We observe in general, there is a good agreement between the numerical and experimental results. The flow inside the cylinder is studied in different locations related to the bottom of the scavenging ports during the period with open exhaust valve. Moreover, the replacement of combustion products with fresh scavenge air is analysed. The effective flow angle is calculated for the air flow through the scavenging ports. It is found that the effective flow angle is different from the geometrical angle of the ports (20°). Results illustrate better performance of LES, especially in the prediction of the tangential velocity which is crucial for the simulation of an accurate swirl and air-fuel mixing inside the marine engines. LES predicts a uniform profile for the tangential velocity at the top of cylinder which is consistent with the experimental results while URANS predicts a solid body rotation.

Inspired by the recently reported translucency of monolayer graphene (GE) to wetting, atomistic simulations are employed to evaluate water flow enhancement induced by GE deposited on the inner surfaces of hydrophilic nanochannels. The flow in the coated channels exhibits a slip length of approximately 3.0 nm. Moreover, by contrasting the flow rates in channels with coated walls against flow rates in the corresponding uncoated channels, an “effective” flow enhancement from 3.2 to 3.7 is computed. The probability density function of the water dipole orientation indicates that the flow enhancement is related to a thinner structured water layer at the solid–liquid interface. This study provides quantitative evidence that GE employed as coating reduces substantially hydraulic losses in hydrophilic nanoconfinement.

We present a novel strategy to improve load balancing for large scale Bayesian inference problems. Load imbalance can be particularly destructive in generation based uncertainty quantification (UQ) methods since all compute nodes in a large-scale allocation have to synchronize after every generation and therefore remain in an idle state until the longest model evaluation finishes. Our strategy relies on the concurrent scheduling of independent Bayesian inference experiments while sharing a group of worker nodes, reducing the destructive effects of workload imbalance in population-based sampling methods.

To demonstrate the efficiency of our method, we infer parameters of a red blood cell (RBC) model. We perform a data-driven calibration of the RBC's membrane viscosity by applying hierarchical Bayesian inference methods. To this end, we employ a computational model to simulate the relaxation of an initially stretched RBC towards its equilibrium state. The results of this work advance upon the current state of the art towards realistic blood flow simulations by providing inferred parameters for the RBC membrane viscosity.

We show that our strategy achieves a notable reduction in imbalance and significantly improves effective node usage on 512 nodes of the CSCS Piz Daint supercomputer. Our results show that, by enabling multiple independent sampling experiments to run concurrently on a given allocation of supercomputer nodes, our method sustains a high computational efficiency on a large-scale supercomputing setting.

Liquid penetration analysis in porous media is of great importance in a wide range of applications such as ink jet printing technology, painting and textile design. This article presents an investigation of droplet impingement onto metallic meshes, aiming to provide insights by identifying and quantifying impact characteristics that are difficult to measure experimentally. For this purpose, an enhanced Volume-Of-Fluid (VOF) numerical simulation framework is utilised, previously developed in the general context of the OpenFOAM CFD Toolbox. Droplet impacts on metallic meshes are performed both experimentally and numerically with satisfactory degree of agreement. From the experimental investigation three main outcomes are observed—deposition, partial imbibition, and penetration. The penetration into suspended meshes leads to spectacular multiple jetting below the mesh. A higher amount of liquid penetration is linked to higher impact velocity, lower viscosity and larger pore size dimension. An estimation of the liquid penetration is given in order to evaluate the impregnation properties of the meshes. From the parametric analysis it is shown that liquid viscosity affects the adhesion characteristics of the drops significantly, whereas droplet break-up after the impact is mostly controlled by surface tension. Additionally, wettability characteristics are found to play an important role in both liquid penetration and droplet break-up below the mesh.

This paper presents a comparison of full-scale computational fluid dynamics (CFD) simulations with speed trial measurements for a ro-ro vessel and a general cargo vessel. Significant work has been done on validating CFD simulation in model scale. However, in full-scale very few publicly available studies have been conducted due to limited access of validation data. The present study includes extensive validation and verification of both resistance, propeller open-water and self-propulsion simulations in both model and full-scale. The self-propulsion simulations include modelling of the free surface and rotation of the 3D propeller. Full-scale resistance and propeller open-water as well as model scale self-propulsion simulations show good agreement with towing tank measurements and predictions. However, the full-scale self-propulsion simulations using the traditional approach of including the roughness as a point force estimated by an empirical formula significantly underestimate the power from the speed trial measurements. By including the effect of hull and propeller roughness directly into the CFD model, by modifying the wall functions, the discrepancy between CFD and speed trial measurements decreases significantly. This indicates that inclusion of a roughness model directly into the CFD simulation could be a more accurate method than the traditional approach of using empirical formulas originally designed for towing tank extrapolation.

The fluid flow and heat and mass transfer in a supersonic separator are not understood well due to the complicated interaction of the supersonic flow, swirling flow, phase transition and shock waves. In the present study, we develop a wet steam model to investigate the flow structure inside a supersonic separator with the co-existence of non-equilibrium condensation and shock waves. A study of the effect of the inlet subcooling and inlet saturation on the condensation behaviour is conducted to evaluate the performance of the supersonic separation with a focus on the shock wave. The numerical result shows that the degree of supersaturation of the water vapour can reach a maximum value of 4.28 within the designed supersonic separator and generate a peak nucleation rate of approximately 1021 kg m−3 s−1. The occurrence of the shock wave changes the equilibrium thermodynamic state, which leads to the re-evaporation of the condensed droplet. Higher inlet subcooling and inlet saturation not only shift downstream the position of the shock wave, but also induce an earlier condensation and higher liquid fraction. For the present nozzle, when the inlet subcooling and inlet saturation are about 34 K and 0.28 respectively, the shock wave intersects the region of the intense nucleation process, the non-equilibrium condensation process is terminated due to the increase of the pressure and temperature downstream the shock wave. Stronger swirling flow results in non-uniform distribution of the static pressure and decreases the nucleation rate of water vapour. The high swirling flow with a maximum swirl velocity of 150 m/s weakens the liquid fraction by 25% compared to the no swirling flow. This indicates that it is important to balance the swirling flow and condensation process to achieve an efficient performance of the supersonic separator.

In this Letter we show that a bifurcation cascade and fully sustained turbulence can share the phase space of a fluid flow system, resulting in the presence of competing stable attractors. We analyze the toroidal pipe flow, which undergoes subcritical transition to turbulence at low pipe curvatures (pipe-totorus diameter ratio) and supercritical transition at high curvatures, as was previously documented. We unveil an additional step in the bifurcation cascade and provide evidence that, in a narrow range of intermediate curvatures, its dynamics competes with that of sustained turbulence emerging through subcritical transition mechanisms.

Medical imaging plays a central role in cancer therapy, however scans cannot detect the full extent of infiltrative brain tumors. Post-mortem and histological studies show that tumor cells can be found even 2 cm beyond the tumor outlines visible on the medical scans. Current radiotherapy planning is handling these uncertainties in a rather rudimentary fashion.

Over the last three decades, molecular simulation has become ubiquitous in scientific fields ranging from molecular biology to chemistry and physics. It serves as a tool to rationalize experimental results, providing access to the dynamics of a system at the atomistic and molecular level, and predictions of macroscopic properties of materials. As computational hardware and software capabilities increase, molecular simulations are becoming increasingly more important as a tool to complement experiments and have become an invaluable asset for insight, prediction, and decision making by scientists and engineers. This increased importance is associated with an ever-increasing need to interpret quality of the predictions of the complex molecular systems. In the context of Virtual Measurements, as proposed by Irikura et al., we remark that for the output of a molecular simulation to be considered equivalent to an experimental measurement, it must include both a value of the quantity of interest (QoI) and a quantification of its uncertainty. In turn, uncertainty can be defined as a “parameter, associated with the result of a measurement, that characterizes the dispersion of the values that could reasonably be attributed to the measurand.” Uncertainty quantification (UQ) is essential for building confidence in model predictions and helping model-based decisions. Monitoring uncertainties in computational physics/chemistry has become a key issue, notably for multiscale modeling. Reliable predictions at coarser scales imply the rational propagation of uncertainties from finer scales; however, accounting for such uncertainties requires access to a significant computational budget.

Particle methods are a natural way of modelling flow problems.

In this contribution, we review recent developments and applications of a dynamic clustering algorithm SWINGER tailored for the multiscale molecular simulations of biomolecular systems. The algorithm on-the-fly redistributes solvent molecules among supramolecular clusters. In particular, we focus on its applications in combination with the adaptive resolution scheme, which concurrently couples atomistic and coarse-grained molecular representations. We showcase the versatility of our multiscale approach on a few applications to biomolecular systems coupling atomistic and supramolecular water models such as the well-established MARTINI and dissipative particle dynamics models and provide an outlook for future work.

The identification of cut marks and other bone surface modifications (BSM) provides evidence for the emergence of meat-eating in human evolution. This most crucial part of taphonomic analysis of the archaeological human record has been controversial due to highly subjective interpretations of BSM. Here, we use a sample of 79 trampling and cut marks to compare the accuracy in mark identification on bones by human experts and computer trained algorithms. We demonstrate that deep convolutional neural networks (DCNN) and support vector machines (SVM) can recognize marks with accuracy that far exceeds that of human experts. Automated recognition and analysis of BSM using DCNN can achieve an accuracy of 91% of correct identification of cut and trampling marks versus a much lower accuracy rate (63%) obtained by trained human experts. This success underscores the capability of machine learning algorithms to help resolve controversies in taphonomic research and, more specifically, in the study of bone surface modifications. We envision that the proposed methods can help resolve on-going controversies on the earliest human meat-eating behaviors in Africa and other issues such as the earliest occupation of America.

We investigate the collapse of a cloud composed of 12500 gas bubbles in a liquid through large-scale simulations. The gas bubbles are discretized by a diffuse interface method, and a finite volume scheme is used to solve on a structured Cartesian grid the Euler equations. We investigate the propagation of the collapse wave front through the cloud and provide comparisons to existing models such as Mørch’s ordinary differential equations and a homogeneous mixture approach. We analyze the flow field to examine the evolution of individual gas bubbles and in particular their associated microjet. We find that the velocity magnitude of the microjets depends on the local strength of the collapse wave and hence on the radial position of the bubbles in the cloud. At the same time, the direction of the microjets is influenced by the distribution of the bubbles in its vicinity. We envision that the present, state-of-the-art, large-scale simulations will serve the further development of low-order models for bubble collapse.

Controlled gliding is one of the most energetically efficient modes of transportation for
natural and human powered fliers. Here we demonstrate that gliding and landing strategies
with different optimality criteria can be identified through deep-reinforcement-learning
without explicit knowledge of the underlying physics. We combine a two-dimensional
model of a controlled elliptical body with deep-reinforcement-learning (D-RL) to achieve
gliding with either minimum energy expenditure, or fastest time of arrival, at a predetermined location. In both cases the gliding trajectories are smooth, although energy/time
optimal strategies are distinguished by small/high frequency actuations. We examine the
effects of the ellipse’s shape and weight on the optimal policies for controlled gliding.
We find that the model-free reinforcement learning leads to more robust gliding than
model-based optimal control strategies with a modest additional computational cost. We
also demonstrate that the gliders with D-RL can generalize their strategies to reach
the target location from previously unseen starting positions. The model-free character
and robustness of D-RL suggests a promising framework for developing robotic devices
capable of exploiting complex flow environments.