Software & Systems

Journal articles №4 2014

21. Methodology of training recurrent artificial neural network with dynamic stack memory [№4 за 2014 год]
Authors: Lila V.B. (lila@i-intellect.ru) - Rostov State University of Civil Engineering, Ph.D; Puchkov E.V. (puchkoff@i-intellect.ru) - Rostov State University of Civil Engineering, Ph.D;
Abstract: The work proposes to generalize recurrent artificial neural networks by adding the delay feedback signals of the hidden layer to a few bars in the form of a dynamic stack memory. It allows providing an adaptive memorization of the past time events and creating a flexible tool to create nonlinear models. The proposed universal architecture of a recurrent artificial neural network with dynamic stack memory generalizes such networks as the multilayer perceptron, Jordan and Elman neural networks, as well as neural networks with feedback neurons. The paper presents a training methodology of standard neural network architecture for solving the problem of time series forecasting. It is based on the learning sample transformation. The feedbacks from a hidden layer or outputs of the neural network are eliminated by adding feedback signals to the training set. The opportunities of NeuroNADS neural network emulator have been expanded to implement the proposed methodology of recurrent neural network with dynamic stack memory training. The article presents a new object-oriented model and main software classes of neural network emulator. Based on the data for 2010–2011 there has been predicted the average monthly density of solar activity at a 10,7 cm wavelength for the first six months of 2012. After analyzing it was concluded that the recurrent artificial neural network with dynamic stack memory can be trained using the proposed methodology and the constructed models of artificial neural networks can be used for time series forecasting.
Keywords: recurrent neural networks, multilayer perceptron, elman neural network, jordan neural network, feedback, dynamic stack memory, time series forecasting, neuronads
Visitors: 14180

22. Artificial immune systems: review and current state [№4 за 2014 год]
Authors: Chernyshev Yu.O. (sergeev00765@mail.ru) - Don State Technical University, Ph.D; Grigoryev G.V. (grigoryev.georgii@gmail.com) - Don State Technical University; Ventsov N.N. (vencov@list.ru) - Don State Technical University, Ph.D;
Abstract: Solving optimization problems is one of the most relevant spheres of heuristic algorithms application, inspired by live or inanimate nature. This article summarizes the current state of relatively new class of bioin spired algorithms that are artificial immune systems, with their issues, prons and cons. The authors are review developments of artificial immune systems, as well as researches of immune networks, immune response, somatic hypermutation, danger theory and selection processes. Artificial immune systems attract more and more interest due to the best of available properties among bioinspired methods such as dynamic location of working elements (from evolutionary algorithms) and learning principles (from artificial neural networks). If immune networking is one of the earliest model-based processes among immune inspired researches, then danger theory and modeling of somatic hypermutation and selection can be considered relatively young research areas. However, immune networks are still pressing problem today: they are used as a key parts of bioinspired hybrid systems. Different implemetations of these systems can be applied in many science researches and industries. Application is often the same as that of heuristi c algorithms researched earlier, but now new application spheres are emerging as well that have not been researched before in this perspective – computer science (for example, data-mining). The review of current state in artificial immune systems sphere is based on work of both foreign and Russian authors.
Keywords: selection processes, danger theory, artificial immune network, immune systems, bioinspired systems
Visitors: 14095

23. Spectral measure playback accuracy using permutative method of modeling stochastic processes [№4 за 2014 год]
Authors: Kuznetsov B.F. (kuznetsovbf@gmail.com) - Irkutsk State Academy of Agriculture, Molodezny Settlement, Ph.D; Shishkina S.V. (svetlanashishkina@mail.ru) - National Research Irkutsk State Technical University; Borodkin D.K. (borodkin_dk@mail.ru) - Angarsk State Technical Academy, Ph.D;
Abstract: The paper considers one of the standard permutation methods of modeling stochastic processes with given spectral and proba-bilistic measures. An advantage of this method is the lack of restrictions on the combination of the probability and the spectral measure. There is a hypothesis of the reasons of appearing and structure of the general playback accuracy of autocorrelated function of modelled pro-cess using this method. The authors suggest that the modelling error consists of two parts: the variation and methodical. The error analysis method is developed. It allows making a constant value of the variation component of an error. In other words, making it independent on the autocorrelation function parameters. In turn, it allows carrying out the dependence analysis of the methodological component of the error modeling from the autocorrelation function parameters. To test this hypothesis four series of numerical experiments on modeling of stochas-tic processes were carried out using specially selected combinations of distribution laws. Distribution laws were chose n for reasons of ab-sence or presence of systematic error conditions. Based on the characteristics of the analyzed method (on the presence of lea ding and driven processes) the authors have developed a procedure for the statistical analysis of modeling results. Implementing the statistical analysis proce-dures developed by results of modeling gave no reason to reject the hypothesis of the existence of two error components: variation and me-thodical. The paper shows that using of this analysis method can significantly reduce the effect of variation in the results of systematic error analysis. It proves the method of stochastic processes modeling for solving various problems.
Keywords: slave process, master process, spectral measure, autocorrelation, error modeling, permutative method, stochastic process
Visitors: 7355

24. Software for acoustooptic processors [№4 за 2014 год]
Authors: Shibaev S.S. (sheab@list.ru) - Academy of Engineering and Technology Southern Federal University, Ph.D; Pomazanov A.V. (pav_tsure@mail.ru) - Academy of Engineering and Technology Southern Federal University, Ph.D; Volik D.P. (dvlbest@mail.ru) - Academy of Engineering and Technology Southern Federal University;
Abstract: The main purpose of acoustooptic processors which is measuring radiosignals parameters. All necessary in-formation about the signal and its parameters is enclosed in analog light Fourier distribution in analysis plane, where there is either CCD photo receiver or a photo diodes line, signal from which is then digitized with onboard means in the form of light intensity amplitudes distribution. The processing such digital signal is fully or partially entrusted to outer calculating sy stem – a personal computer. Under such conditions special software is really important. It extracts useful information out of re-ceived data stream, as well as has servicing functions. The paper considers the application structure and capabilities for the most popular class of acoustooptic processors – radiosignals frequency measurer. The program has the function of real-time reading and displaying amplitude distribution, the function of manual and automatic measuring signal frequency. It helps to carry out a few types of calibration from back-ground noise compensation to finishing with frequency scale nonlinearity account. There is also a possibility to save received distributions into files for their further analysis in such well known programs as MathCad, MatLab etc. The application al-lows controlling exterior laboratory equipment using Ethernet interface, particularly the Agilent generator or other firms. That allows testing its characteristics and taking them into account at device calibration before measurer working automati-cally.
Keywords: acoustooptic, adjustment curve, nonlinear, frequency, calibration, amplitude distribution, spectrum analyzer, application, processor, algorithm
Visitors: 8797

25. Hardware-software package for diagnostics of ionospheric condition according to characteristics of vlf transmitters [№4 за 2014 год]
Author: Skripachev V.O. (skripatchevv@inbox.ru) - Joint Stock Company "Russian Federationn Space Systems"; Polushkovsky Yu.A. (skripatchevv@inbox.ru) - Joint Stock Company "Russian Federationn Space Systems"; Nazarenko A.S. (nonimmemormei@hotmail.com) - Joint Stock Company "Russian Federationn Space Systems";
Abstract: Monitoring of ionospheric condition is very important and urgent task. The ionosphere is a sensitive en-vironment of radio wave propagation, where the responses to geophysical and geological processes appear. One met h-od of monitoring ionospheric conditions is using signal transmitters range of ver y low frequency (VLF). According to the characteristics of VLF radio signals we can judge on the presence of disturbances in the ionosphere. Low -frequency waves propagate in the near-Earth plasma. In recent decades they are of considerable interest in connection with a variety of scientific and engineering applications. So using VLF radio signals it is possible to observe the ion o-sphere D-layer response on solar flares, on the processes of upcoming strong earthquakes. To process VLF radio sig-nals the authors have developed a hardware-software package which includes an antenna, amplifier, receiver, analog-to-digital converter and a PC. The software part is built using the Qt framework and third-party software libraries QwtPlot3D and QCustomPlot. The paper includes a brief description and characteristics of each of them. The software part of the package is built using MVC design pattern. There is a list of features implemented in the software. The paper also describes the prospects of the hardware-software package development.
Keywords: software product, radiowave, ionosphere, software library, data processing, render, interface, pattern
Visitors: 11290

26. On the approach to the development of the radar system constructing methodology [№4 за 2014 год]
Authors: Semenov S.A. (sergeysemenov53@gmail.com) - Military Academy of the Aerospace Defence, Ph.D; Coban A.Y. (alenachugrei@mail.ru) - Military Academy of the Aerospace Defence;
Abstract: The problem of constructing a long-range radar system is a complex semistructured task due to the high complexity of its structures. Therefore, the existing methodological apparatus of its construction is a set of techniques that are not full y intercon-nected. Each of those techniques is a tool to structure the system construction at some point using a variety of indicators and criteria. For example, on the stages of the functional image of information processing that is created on the basis of standard policies and procedures when designing and constructing elements of the group. For example, on the stages of the functional image of infor-mation processing that is created on the basis of standard policies and procedures in the designing and constructing group elements. Moreover, the said steps don’t formalize a common goal, an integrated measure of the achievement and its single method of con-structing. As a result, there is no possibility for comparative evaluation of the potential and real system efficiency. Automated infor-mation processing procedures that implement functionality are not enough. For targeted construction, the article suggests an ap-proach to the development of a common methodology for all phases of long-range radar system construction in the formalization of the productions and abstract problem solving with emphasis on the construction phase of ref inement options in shaping the image of the data collection and processing functional subsystem.
Keywords: technique, constructing, system, objective function, effectively, an indicator, information loss, business computing
Visitors: 8675

27. Developing a software module for an automatic choice of sparse solver in strength analysis [№4 за 2014 год]
Author: Stepin N.E. (N_i_k_i1989@mail.ru) - Lomonosov Moscow State University;
Abstract: The paper presents a software module which includes a variety of different algorithms and methods: direct and iterative sparse solvers for symmetric and non-symmetric matrices, various preconditioners, different formats for storing sparse matrices in RAM, HPC technologies based on OpenMP and CUDA. A method for solving incompressible materials problems based on Uzawa algorithm was implemented in this software module. An algorithm for an optimal sparse solver choice depending on mechanical problem, its dimensions and available hardware resources was designed and implemented. A user can limit some options and set parameters that influence a choice of a sparse solver or directly specify which solver to use. In fact, the software module is some kind of a wrapper over individual sparse solvers which takes a matrix and a right hand side as an input as well as some solver’s settings, and then decides which solver to run based on the algorithm, confi g-ures it and converts a matrix to a corresponding format (different solvers may have different optimal storage formats for a sparse matrix). A series of numerical experiments that confirms the validity of the criteria used in the algorithm was per-formed by the author.
Keywords: theory of elasticity, incompressible materials, finite element method, uzawa algorithm, iterative methods for sparse systems of linear algebraic equations
Visitors: 9018

28. A performance research of a number of iterative linear solvers within elastoplastic problem [№4 за 2014 год]
Authors: Tolmachev A.V. (tolmachev.arseny@gmail.com) - Institute of Engineering Science of the Ural Branch of the RAS; Konovalov A.V. (avk@imach.uran.ru) - Institute of Engineering Science of the Ural Branch of the Russian Academy of Sciences, Ph.D; Partin A.S. (dmitriy-v-k@yandex.ru) - Institute of Engineering Science of the Ural Branch of the RAS, Ph.D;
Abstract: An elastoplastic problem with large plastic deformations is substantially non-linear both geometrically and physically. Most of the finite element solution time is being spent solving linear system in order to find generalized speed vector in the FE grid nodes. To decrease solution time it is crucial to use parallel computations, especially cluster computers. Arising matrix is non-symmetric and has large dimension banded and sparse within band. Direct solvers create fill-in within a band and therefore induce heavy costs both in memory and computation time. We performed a computational experiment based on the test problem of parallelepiped compression with plates on “Uran” cluster located in Institute of Mathematics and Mechanics UB RAS in order to analyze the performance of SOR, BiCGStab and GMRES solvers. We estimated linear solver computation time variance that is induced by random latencies in data transfer within the cluster system. Experiment results showed that SOR requires more time than BiCGStab and GMRES thus is inefficient. GMRES is the fastest on a small nu m-ber of processors. However BiCGStab is more scalable and requires less time than GMRES when using large number of pro-cessors. Solve time deviation does not exceed 5 % from the average. An ILUC preconditioner is the most efficient one execu-tion time-wise, however a SAINV preconditioner requires the least number of BiCGStab iterations.
Keywords: elastic-plastic problem, finite element method, parallel iterative linear solvers
Visitors: 9702

29. Fuzzy sliding mode control system of unstable dynamic system cart-pole based on soft computing [№4 за 2014 год]
Authors: Nefedov N.Yu. (nefnukem@gmail.com) - Dubna Internacional University for Nature, Socitty and Man; Ulyanov, S.V. (ulyanovsv46_46@mail.ru) - Dubna State University – Institute of System Analysis and Control, Dubna, Joint Institute for Nuclear Research – Laboratory of Information Technology (Professor), Ph.D;
Abstract: The paper studies sliding mode control for cart-pole unstable dynamic system using soft computing optimizer. The authors consider three major approaches to eliminate oscillations in the sliding mode and describe the advantages and disadvantages of these approaches. The paper shows the knowledge base designing algorithm for a sliding mode controller in a new set of tools to create robust control systems – Optimizer Knowledge Bases on soft computing. It also compares knowledge bases quality created by an expert and intelligent control systems: soft-computing optimizer and ANFIS. The arti-cle gives necessary formal definitions and the corresponding illustrations. This control system showed better robustness, comparing to systems created with other modern tools.
Keywords: fuzzy logic, sliding mode control, knowledge base, soft computing, robustness, intelligent control systems
Visitors: 11266

30. A software package designed to solve the potential theory problems by the boundary element method [№4 за 2014 год]
Authors: (fedotov@imach.uran.ru) - ; Spevak L.F. (lfs@imach.uran.ru) - Institute of Engineering Science of the Ural Branch of the RAS, Ph.D; Nefedova O.A. (nefedova@imach.uran.ru) - Institute of Engineering Science of the Ural Branch of the RAS;
Abstract: The paper develops the approach to the implementation of the boundary element method aimed at the paralleling of computations for solving two-dimensional problems on steady-state potential flows. The key feature of the applied solution algorithms is the exact computation of all integrals over boundary elements by analytical formulae derived by the authors. This noticeably increases solution accuracy and reduces calculation time. The designed soft-ware package is intended for solving two-dimensional problems of the potential theory in arbitrary geometry with specified boundary conditions for a required function or flow. A possibility of solving homogeneous problems (wit h-out internal sources) is implied, as well as problems with specified source functions. A source function is specified by a separate program module. Special attention is given to the case when the source function is harmonic. This enables all the calculations in solving an inhomogeneous problem to be brought to the area boundary. A graphic editor is in-cluded in the software package to specify the geometry of the area under study and to introduce problem parameters and boundary conditions. An area is specified by its external boundary consisting of rectilinear and circular portions. Boundary conditions and the number of boundary elements are specified for each portion. Besides, by means of the graphic editor, inside the calculated area a zone can be specified where the values of the required mesh are to be calcu-lated. This zone is also specified by a boundary consisting of rectangular and circular portions. The mesh fineness in the zone is specified arbitrarily, and the recalculation for a new mesh does not require a new solution to the initial problem. The software package was implemented on the Uran supercomputer installed at IMM, UB RAS. Problems on heat distribution in square and elliptic plates for various source functions are considered as examples.
Keywords: parallel computing, stationary problems of the potential theory, source function, boundary element method, analytical integration
Visitors: 9794

◄ ← Preview | 1 | 2 | 3 | 4 | 5 | Next → ►