School of Mechanical Engineering Seminar
Wednesday, June 20, 2018 at 14:00
Tokhna Building of Mechanical Engineering, Room 106

Stability of Stratified Two-Phase Flows in Inclined Channels

Ilya Barmak

 PhD student of Prof. Neima Brauner and Prof. Amos Ullmann

Stratified two-phase channel flow can be achieved only for certain operating conditions, for which the flow is stable and no undesired effects are encountered. The knowledge of these conditions is essential for the design and operation of pipelines and other process equipment. The main goal of this research is to acquire this knowledge by means of linear modal and non-modal stability analyses in a form that can provide insights that are useful from the practical point of view.

Modal analysis is a traditional approach that examines the linear stability with respect to an arbitrary wavenumber perturbation by solving numerically the system of Orr-Sommerfeld equations defined in each sublayer. The results are summarized in the form of stability maps showing the operational conditions at which a stratified-smooth flow pattern is stable. In inclined flows, up to three distinct base states with different holdups exist, so that the stability analysis has to be carried out for each branch separately. Although the conducted modal stability analysis considered only 2D perturbations, we proved the sufficiency of 2D analysis for exploring the stability of stratified horizontal and inclined two-phase channel flows also with respect to 3D perturbations (i.e., Squire theorem) .

The modal analysis, however,  predicts instability only in the limit of infinite time. The non-modal (temporal) perturbations approach has been proven to be more relevant for relatively short times. Using this approach, we are looking for the so-called “optimal perturbations” that exhibit the maximum gain for transfer of energy from the mean flow to the perturbations. We found that the optimal perturbations may lead to the onset of shear instability and further non-linear transition on the route to turbulence in one of the phases, and/or to the growth of the interface displacement and thereby to flow pattern transition within the region that is predicted to be linearly stable by the modal analysis.

Particle Trapping and Conveying Using an Optical Archimedes’ Screw

:By

Sahar Froim

MSc student under the supervision of Dr. Alon Bahabad

Abstract

Trapping and manipulation of particles using laser beams has become an important tool in diverse fields of research.

In recent years, particular interest has been devoted to the problem of conveying optically trapped particles over extended distances either downstream or upstream of the direction of photon momentum flow. In this work, we proposed and experimentally demonstrated an optical analog of the famous Archimedes’ screw where the rotation of a helical intensity beam is transferred to the axial motion of optically trapped micrometer-scale, airborne, carbon-based particles. With this optical screw, particles were easily conveyed with controlled velocity and direction, upstream or downstream of the optical flow, over a distance of half a centimeter. Our results offer a very simple optical conveyor that could be adapted to a wide range of optical trapping scenarios.

On Thursday, June 14, 2018, 15:00

Room 011, Kitot building

Speaker: Adi Hayat

M.Sc. student under the supervision of Prof. Daniel Cohen-Or

Sunday, June 17^th, 2018 at 15:30

Room 206, Wolfson Mechanical Bldg., Faculty of Engineering

Generative Low-Shot Network Expansion​

Conventional deep learning classifiers are static in the sense that they are trained on a predefined set of classes and learning to classify a novel class typically requires re-training. In our work, we address the problem of Low-Shot network-expansion learning. We introduce a learning framework which enables expanding a pre-trained (base) deep network to classify novel classes when the number of examples for the novel classes is particularly small. We present a simple yet powerful distillation method where the base network is augmented with additional weights to classify the novel classes, while keeping the weights of the base network unchanged. We term this learning hard distillation, since we preserve the response of the network  on the old classes to be equal in both the base and the expanded network. We show that since only a small number of weights needs to be trained, the hard distillation excels for low-shot training scenarios. Furthermore, hard distillation avoids detriment to classification performance on the base classes. Finally, we show that low-shot network expansion can be done with a very small memory footprint by using a compact generative model of the base classes training data with only a negligible degradation relative to learning with the full training set.