Design optimization of supersonic jet pumps using high fidelity flow analysis

Supersonic jet pumps are simple devices with no moving parts, where a high velocity (primary) flow is used to pump a second fluid. In this paper, Computational Fluid Dynamics (CFD) is combined with an optimization framework in order to develop a tool for the rapid generation of jet pump designs. A key feature of the problem formulation is the transformation of the jet pump design parameters in terms of geometric ratios. This approach dramatically reduces the number of unrealistic designs covered by the Design of Experiments. Optimal Latin Hypercubes for surrogate model building and model validation points are constructed using a permutation genetic algorithm and design points are evaluated using CFD. Surrogate models of primary and entrained flow rates are built using a Moving Least Squares approach. A series of optimizations for various pump sizes are performed using a genetic algorithm and Sequential Quadratic Programming, with responses calculated from the surrogates. This approach results in a set of optimized designs, from which pumps for a wide range of flow rates can be interpolated.     

Dropout prediction in e-learning courses through the combination of machine learning techniques

In this paper, a dropout prediction method for e-learning courses, based on three popular machine learning techniques and detailed student data, is proposed. The machine learning techniques used are feed-forward neural networks, support vector machines and probabilistic ensemble simplified fuzzy ARTMAP. Since a single technique may fail to accurately classify some e-learning students, whereas another may succeed, three decision schemes, which combine in different ways the results of the three machine learning techniques, were also tested. The method was examined in terms of overall accuracy, sensitivity and precision and its results were found to be significantly better than those reported in relevant literature.     

Gradient based hyper-parameter optimisation for well conditioned kriging metamodels

In this work a two step approach to efficiently carrying out hyper parameter optimisation, required for building kriging and gradient enhanced kriging metamodels, is presented. The suggested approach makes use of an initial line search along the hyper-diagonal of the design space in order to find a suitable starting point for a subsequent gradient based optimisation algorithm. During the optimisation an upper bound constraint is imposed on the condition number of the correlation matrix in order to keep it from being ill conditioned. Partial derivatives of both the condensed log likelihood function and the condition number are obtained using the adjoint method, the latter has been derived in this work. The approach is tested on a number of analytical examples and comparisons are made to other optimisation approaches. Finally the approach is used to construct metamodels for a finite element model of an aircraft wing box comprising of 126 thickness design variables and is then compared with a sub-set of the other optimisation approaches.     

Heavy-ion-induced luminescence of amorphous SiO2 during nanoparticle formation

Silica glass was implanted with negative 60 keV Cu ions at an ion flux from 5 to 75 μA/cm² up to a fluence of 1 × 10¹⁷ ions/cm² at initial sample temperatures of 300, 573 and 773 K. Spectra of ion-induced photon emission (IIPE) were collected in situ in the range from 250 to 850 nm. Optical absorption spectra of implanted specimens were ex situ measured in the range from 190 to 2500 nm.

IIPE spectra showed a broad band centered around 560 nm (2.2 eV) that was assigned to Cu⁺ solutes. The band appeared at the onset of irradiation, increased in intensity up to a fluence of about 5 × 10¹⁵ ions/cm² and then gradually decreased indicating three stage of the ion beam synthesis of nanoclusters: accumulation of implants, nucleation and growth nanoclusters. The IIPE intensity normalized on the ion flux is independent on the ion flux below 20 μA/cm²at higher fluences. The intensity of the band increased with increasing samples temperature, when optical absorption spectra reveal the increase of Cu nanoparticles size.     

Dynamic force microscopy imaging of native membranes

We employed magnetic ACmode atomic force microscopy (MACmode AFM) as a novel dynamic force microscopy method to image surfaces of biological membranes in their native environments. The lateral resolution achieved under optimized imaging conditions was in the nanometer range, even when the sample was only weakly attached to the support. Purple membranes (PM) from Halobacterium salinarum were used as a test standard for topographical imaging. The hexagonal arrangement of the bacteriorhodopsin trimers on the cytoplasmic side of PM was resolved with 1.5nm lateral accuracy, a resolution similar to images obtained in contact and tapping-mode AFM. Human rhinovirus 2 (HRV2) particles were attached to mica surfaces via nonspecific interactions. The capsid structure and 2nm sized protein loops of HRV2 were routinely obtained without any displacement of the virus. Globular and filamentous structures on living and fixed endothelial cells were observed with a resolution of 5-20nm. These examples show that MACmode AFM is a favorable method in studying the topography of soft and weakly attached biological samples with high resolution under physiological conditions.