Evolving comprehensible neural network trees using genetic algorithms

This paper presents a comprehensible neural network tree (CNNTREE). CNNTREE is a proposed general modular neural network structure, where each node in this tree is a comprehensible expert neural network (CENN). One advantage of using CNNTREE is that it is a “gray box”; because it can be interpreted easily for symbolic systems; where each node in the CNNTREE is equivalent for symbolic operator in the symbolic system. Another advantage of CNNTREE is that it can be trained as any normal multi layer feed forward neural network. An evolutionary algorithm is given for designing the CNNTREE. Back propagation is also checked as local learning algorithm that fits for real time learning constraints. The tree generalization and training performance are examined using experiments with a digit recognition problem. The article is published in the original. Elsayed Mostafa. Received the B.Sc. degree in electrical (Communication) Eng., Cairo University at 1967. Dipl.-Ing, and Dr-Ing. from Stuttgart University at 1977, 1981 respectively. He is a member of ECS and EEES. He is currently a professor of electronic circuits, Faculty of Engineering, University of Helwan. Amr Kamel. Graduated from Computer Department, Faculty of Engineering of Ain Shams University, Egypt in 1999, and studying M.Sc. degree in computer engineering from the Faculty of Engineering of Helwan University. His special fields of interest include neural networks and genetic algorithms. Alaa Hamdy. Was born in Giza in Egypt, on August 17, 1966. He graduated from the Telecommunications and Electronics Department, Faculty of Engineering and Technology of Helwan University, Cairo, Egypt in 1989. He received the M.Sc. degree in computer engineering from the same university in 1996 and the Ph.D. degree from the Faculty of Electrical Engineering, Poznan University of Technology, Poland in 2004. Currently he is working as a lecturer in the Faculty of Engineering of Helwan University. His special fields of interest, include image processing, pattern analysis, and machine vision.     

Etude de l'influence du mode d'anodisation sur la cinétique de croissance des films d'alumine en milieu acide sulfurique

The kinetics of growth of Al₂O₃ oxide films made by anode oxidation of ALuminium 1050A with two anodization techniques has been investigatedan compared. The first technique is the classical method based on continuous tension during the complete electrlysis duration, and the second one is the modern method ofpulsed tensions, whcih consists of periodically imposing to the material two anodc tensions E₁ and E₂ (E₁ > E₂) during time T₁ and T₂ until the end on anodization. We have shown that the second technique allows the fabrication of films more rapidly for the same electrolysis duraton. Energy and electrolysis time are saved, making it attractive from an economical viewpoint. This advantage arises from the positive effect of excess heat disiption gathered during the imposition of tension, E₁, when E₂ is imposed thereafter (recovery effect). The logging of in situ transient responses in the case of pulsed anodization is a significant asset to explain the growth film kinetics obtained with this method.     

Optimal Sizing of a DBD Ozone Generator Using Response Surface Modeling

Dielectric barrier discharge (DBD) reactors used as ozone generators are well known today and widely used for water treatment and air disinfection. The purpose of this article is to propose an experimental procedure based on the response surface modeling in order to optimize the geometrical dimensions of the cylindrical shape ozone generator, i.e., the discharge gap and the electrodes length. Because an effective ozone generator is expected to give high ozone concentration with a minimum of power requirements, the applied high voltage was associated with the geometrical parameters to carry out a composite centered faces design. Obtained results indicate that for an efficient ozone generator, length of the electrodes needs to be optimized while the discharge gap should be minimized.     

Mono‐, Di‐ and Tetra‐Cationic Surfactants as Carbon Steel Corrosion Inhibitors

Corrosion inhibition of three new synthesized cationic surfactants, N‐(2‐(((Z)‐4‐(pyridin‐4‐yl)but‐3‐en‐1‐yl)amino)ethyl)‐N‐(2‐((E)‐(pyridin‐4‐ylmethylene)amino)ethyl)dodecan‐1‐aminium bromide I(4N), N¹,N²‐didodecyl‐N¹‐((Z)‐4‐(pyridin‐4‐yl)but‐3‐en‐1‐yl)‐N²‐(2‐((E)‐(pyridin‐4‐ylmethylene)amino)ethyl)ethane‐1,2‐diaminium bromide II(4N) and 1‐dodecyl‐4‐((E)‐((2‐(dodecyl(2‐(dodecyl((Z)‐4‐(1‐dodecylpyridin‐1‐ium‐4‐yl)but‐3‐en‐1‐yl)ammonio)ethyl)ammonio)ethyl)imino)methyl)pyridin‐1‐ium bromide IV(4N) on carbon steel was investigated by weight loss, electrochemical impedance spectroscopy and polarization measurements. Results show that the synthesized cationic surfactants inhibit corrosion of carbon steel in 1 M HCl. The inhibitive action occurs by virtue of adsorption on the metal surface following a Langmuir adsorption isotherm model. Polarization curves reveal that the investigated cationic surfactants can be classified as mixed inhibitor types. The variations in the corrosion inhibition efficiency between three cationic surfactants are correlated with their chemical structures, with more hydrophobic surfactants yielding higher inhibition efficiency.     

Effect of mutation schemes on convergence to optimum design of steel frames

In this paper an attempt is made to modify a distributed genetic algorithm (DGA) to minimize the weight of steel frames. The main aspects of modifications include the twin analogy and a number of mutation schemes. Although the mutation is a secondary operator in the GA, it plays an influential role in diversifying the population and exploring more feasible design space. In this research, four different mutation schemes are examined and their effects on convergence to optimum solutions are evaluated. Using three benchmark examples, it is demonstrated that mutation operator can efficiently influence on the speed of convergence and consequently saving computation time. As coding a program is required, the DO-DGA software has been developed to handle the algorithm. DO-DGA applies modified DGA, including different mutation schemes, to obtain the global optimum solution.     

Prediction-based Deadbeat Control for Grid-connected Inverter with L-filter and LCL-filter

Abstract—In this article, an improved deadbeat control algorithm suitable for digital signal processor-based circuit implementation is proposed. The control algorithm allows the derivation of a nearly sine wave output current with a fixed switching frequency of a current-controlled voltage source inverter. Two low-pass output filters configurations are considered in this study: a simple inductance filter and an LCL-filter. By taking advantage of prior knowledge of the state variables’ shape, the improved deadbeat control algorithm is based on a simple prediction model to derive the expected duty cycle needed to switch on and off the power switches. The control study of the grid-connected inverter with L and LCL output filters has been considered using a co-simulation approach with (Powersim Inc., Rockville, Maryland, USA) and MATLAB software (The MathWorks, Natick, Massachusetts, USA). The obtained results show the improvement of both shape quality and tracking accuracy of the output current quantified by low ripple content and a nearly unity power factor.     

Dyeing of cationised cotton with acid dyes

The production of cationised cotton fabrics, via Sandene 8425, 1, 1-dimethyl-3-hydroxy azetidinium chloride and 1, 1-diethyl-3-hydroxy azetidinium chloride, greatly enhances the exhaustion and fixation of acid dyes. This paper shows that these dyes can be used in the absence of salt in a neutral medium. Wash fastness, light fastness and other properties were studied.     

Characterisation of 4043 aluminium alloy deposits obtained by wire and arc additive manufacturing using a Cold Metal Transfer process

ABSTRACT

4043 aluminium deposits were elaborated using a 3D print device equipped with a Cold Metal Transfer welding source. Two sets of process parameters leading to different average powers were compared in order to establish the relations between the powers and energies produced and the geometrical characteristics of the deposits. The effects of the travel speed and layer superposition on the transfer mechanisms as well as on the geometrical characteristics of the deposits were discussed for both sets of parameters. Finally, the formed microstructures were analysed and the porosity defects were quantified and discussed with regard to the heat input characteristics and the solidification conditions.     

Convective flow and heat transfer in a tall porous cavity side-cooled with temperature profile

Natural convection in an air-filled (Prandtl number = 0.7) porous cavity with profiled side cooling and constant bottom heating is investigated over the Rayleigh number range of 1×10⁴ to 1×10⁸ at two Darcy numbers: 1×10^-4 and 1×10^-6. The aspect ratio based on cavity height was varied from 0.5 to 0.1 to investigate penetration length according to linear or sinusoidal temperature profile. The general non-Darcy model adopted in this work was validated against experimental and theoretical results in the literature and Nusselt number was predicted within less than 3% in the worst case. The effect of left wall imposed temperature profile was investigated in detail. Different convective regimes were observed depending on the imposed profile. An active region was found to take place with the linear temperature profile and with extent proportional to Rayleigh number as predicted by scale analysis.