A comparative study of efficient initialization methods for the k-means clustering algorithm

K-means is undoubtedly the most widely used partitional clustering algorithm. Unfortunately, due to its gradient descent nature, this algorithm is highly sensitive to the initial placement of the cluster centers. Numerous initialization methods have been proposed to address this problem. In this paper, we first present an overview of these methods with an emphasis on their computational efficiency. We then compare eight commonly used linear time complexity initialization methods on a large and diverse collection of data sets using various performance criteria. Finally, we analyze the experimental results using non-parametric statistical tests and provide recommendations for practitioners. We demonstrate that popular initialization methods often perform poorly and that there are in fact strong alternatives to these methods.     

Influence of the geometric factors of the experimental device used in suspension polymerization on the properties of poly(styrene‐co‐divinylbenzene) microparticles

The prediction of the final particle size for reactive systems such as the reactions of suspension polymerization is a complex matter. Thus, the preparation of very small microparticles is specially challenging, probably because of the coalescence of the polymeric beads taking place during the later stages of the polymerization. In this work, very small gel‐type styrene‐co‐divinylbenzene beads were synthesized by using a previously determined set of experimental synthesis conditions in which the stabilization of the dispersion of the monomeric droplets was ensured, and, under these conditions, the factors related to the geometry of the experimental device were modified to determine their actual effect on the final size of the microparticles. From the experimental results, a very simple and useful model was obtained that was able to predict the final size of the microparticles as a function of the values of the geometric factors of the reactor. This model indicates that the most influential factors in the final size of the microparticles are the liquid depth inside the reactor and the stirrer diameter; thus, an increase in the liquid depth produces larger particles, and, conversely, the particle size decreases when using larger stirrer diameters. Additionally, the model permits the design of polymerization experiments aimed at obtaining microparticles with a diameter smaller than 50 μm. © 2011 Wiley Periodicals, Inc. J Appl Polym Sci, 2012     

A rigorous computational model for hydrogen production from bio-ethanol to feed a fuel cell stack

A pseudo dynamic rigorous model of a bio-ethanol processor system (BPS) to produce hydrogen for feeding a Proton Exchange Membrane Fuel Cell (PEM-FC) is presented. The main contribution of this work is to give the overall set of differential and algebraic equations (DAE), assumptions and the way to computationally implement it. This model is able for testing the dynamic behavior of this integrated process, obtaining a reduced order linear model and checking any plant-wide control structure design. It is implemented in two programs, HYSYS and MATLAB, properly communicated to coordinate the calculations performed on each one. A part of the process considered with a faster dynamic than the rest of the units of the plant are simulated in HYSYS environment working in steady state mode. Then, auxiliary equipments and the heat exchangers network are in HYSYS which is called by MATLAB every integration interval for doing the simulation of the complete system. On the other side, the PEM-FC and the dynamic models of the plug flow reactors are developed in MATLAB workspace. Hence, strictly speaking this model must be considered as “pseudo” dynamic. The linearized and reduced order model is developed by applying system identification techniques on the rigorous model. Therefore, accounting the main objectives of the process and the most critical disturbances, a preliminary control structure can be well-tested here. Several results are presented in this work analyzing the obtained performances for opened and closed loop modes.     

Numerical studies of a detonation analogue

Abstract

As an aid to deciding a computational strategy for problems involving strong detonation waves, we have applied a variety of numerical techniques to a mathematical problem devised by Fickett, which exhibits many of the essential computational difficulties and possesses analytical solution corresponding to overdriven and underdriven reacting flows. Classical shock capturing methods such as those of MacCormack and Godunov, do not produce acceptable solutions. Better solutions can be provided by Flux Difference Methods (Roe's method) and by Random Choice Methods (still in one dimension), especially by using a new variant that yields second order accuracy. Adaptive gridding techniques for these methods are currently being investigated with encouraging preliminary results.     

The effect of high temperature heat treatment on the structure and properties of anodic aluminum oxide

Nanoporous anodic aluminum oxide (AAO) membranes can be fabricated with highly controllable thickness and porosity, making them ideal for filtration applications. Use of these membranes is currently limited largely due to their size and overall fragility. The objective of this research was to improve mechanical properties of AAO membranes through use of high temperature heat treatment to induce phase transformations in the material. A repeatable two-step anodization process was developed for consistent sample fabrication and heat treatments were performed at 900 °C and 1200 °C in air. The pore morphology and phase composition of the as-anodized and heat-treated membranes were then observed using scanning electron microscopy (SEM) and transmission electron microscopy (TEM). Microhardness testing was utilized to evaluate the mechanical behavior of the membranes before and after heat treatment. As-anodized AAO membranes were determined to be amorphous, and membranes heat-treated to 900 °C and 1200 °C were transformed to crystalline phases while retaining their original porous structure. Heat treatment to 900 °C resulted in formation of the γ-alumina transition phase in the skeleton regions of the membrane and nanocrystalline regions of α-alumina throughout the structure, while heat treatment to 1200 °C completely transformed the material to the stable α-alumina structure. The microhardness testing showed an increase in hardness from 2.5 ± 0.4 GPa to 4.7 ± 1.0 GPa in the transformation from amorphous to α-alumina.     

WAF Method and Splitting Procedure for Simulating Hydro- and Thermal-Peaking Waves in Open-Channel Flows

Hydro- and thermal-peaking waves, generated by hydroelectric power generation, have a strong impact on the ecological integrity of aquatic ecosystems. In order to reduce such effects, mitigation procedure must be studied and implemented. To this end a one-dimensional model which solves the coupling of hydrodynamics with heat transport is developed. The solution is obtained advancing simultaneously the hydrodynamic and thermal module with the same accuracy. For the numerical solution of the governing advection-reaction/diffusion problem a splitting procedure is adopted: the advection-reaction part is solved by means of the weight average flux (WAF) finite volume explicit method, while the diffusion part is solved using a nonlinear version of the implicit Crank-Nicolson method. The WAF method is extended to second-order in the presence of reaction terms. Numerical results are presented for different test examples, which demonstrate the accuracy and robustness of the scheme and its applicability in predicting temperature transport by shallow water flows. Application to the Adige River (Northern Italy) of this framework proves that the model is an effective tool for designing hydro- and thermal-peaking waves mitigation procedures.     

Prostaglandin E2 Enhances Gap Junctional Intercellular Communication in Clonal Epithelial Cells

Prostaglandins are a group of lipids that produce diverse physiological and pathological effects. Among them, prostaglandin E2 (PGE2) stands out for the wide variety of functions in which it participates. To date, there is little information about the influence of PGE2 on gap junctional intercellular communication (GJIC) in any type of tissue, including epithelia. In this work, we set out to determine whether PGE2 influences GJIC in epithelial cells (MDCK cells). To this end, we performed dye (Lucifer yellow) transfer assays to compare GJIC of MDCK cells treated with PGE2 and untreated cells. Our results indicated that (1) PGE2 induces a statistically significant increase in GJIC from 100 nM and from 15 min after its addition to the medium, (2) such effect does not require the synthesis of new mRNA or proteins subunits but rather trafficking of subunits already synthesized, and (3) such effect is mediated by the E2 receptor, which, in turn, triggers a signaling pathway that includes activation of adenylyl cyclase and protein kinase A (PKA). These results widen the knowledge regarding modulation of gap junctional intercellular communication by prostaglandins.     

Electrocoagulation in wastewater containing arsenic: Comparing different process designs

Arsenic removal from wastewater is a key problem for copper smelters. This work shows results of electrocoagulation of aqueous solutions containing arsenic with three different process designs and operating parameters.Three types of electrocoagulation reactors were tested and compared: (a) a modified flow continuous reactor, (b) a turbulent flow reactor and (c) an airlift reactor. All used iron as sacrificial anodes. The results showed that the electrocoagulation process of a 100 mg/L As(V) solution could decrease the arsenic concentration to less than 2 mg/L in the effluent with a current density of 1.2 A/dm² with both the modified flow and the airlift reactor. The removal of arsenic with the turbulent flow reactor did not reach the same level but the Fe-to-As ratio (mol/mol) achieved in the coagulation process was in this case lower (approximately 7) than with the other two reactors.In addition, it seems that increasing the current density beyond a maximum value, the electrocoagulation process would not improve any further. This could probably be explained by passivation of the anode.