PREPRINTSpages: [ 1 ] [ 2 ] [ 3 ] [ 4 ] [ 5 ] [ 6 ] [ 7 ] [ 8 ] [ 9 ] [ 10 ] [ 11 ] [ 12 ] [ 13 ] [ 14 ] [ 15 ] [ 16 ] [ 17 ] [ 18 ]  show all  Balance and characteristics chemes with staggered conservative and fluxes variables on picewiseconstant initial dates (preprint IBRAE200216)
Preprint IBRAE200216
Goloviznin V.M., Karabasov S.A.
It is shown, that the “balance and characteristics” approach to construction of numerical algorithms for hyperbolic conservation laws gives on set piecewise – constant initial functions the exact decision of the elementary equation of convective transport. Two different realizations of the corresponding algorithm named as algorithm « jumping transport » are described. The new algorithm possesses the following properties: is explicit – does not demand procedures of the decision of systems of the linear equations, including procedure of “the running account”; – it is determined on constant minimally possible computing stencil – operates with grid function from two times layers within of one computational cell; – is conservative and monotonous at CFL numbers , smaller units; – on smooth function does not approximate the original equation in classical sense. Generalization of algorithm jumping transport on a case of nonuniform computational grids and on the convection – diffusion equation is given. Examples of test calculations are considered.
Bibliographical reference
Goloviznin V.M., Karabasov S.A. BALANCE AND CHARACTERISTICS CHEMES WITH STAGGERED CONSERVATIVE AND FLUXES VARIABLES ON PICEWISECONSTANT INITIAL DATES. (In Russian). Preprint IBRAE200216. Moscow: Nuclear Safety Institute, July 2002. 19 p. — Refs.: 22 items. 
  Balancecharacteristic schemes with separated conservative and flux variables (preprint IBRAE200215)
Preprint IBRAE200215
Goloviznin V.M., Karabasov S.A., Kobrinsky I.M.
The new approach is suggested for the development of difference methods with high resolution for the equation of convection with regard to diffusion. It is based on the introduction of two different types of variables – “conservative” and “flux”, corresponding to the realization of the conservation law and correct calculation of characteristic region of influence respectively. The process of creation of new algorithms consists of two stages. At the first stage, linear uniform conservative difference schemes with improved dissipative and dispertion characteristics are constructed on the minimal computing stencil. At the second stage, conservative algorithm of minimal correction of the calculated values is used for the realization of the sufficient conditions of the principle of maximum. Explicit monotone algorithms were developed, that are stable in the case of the Courant number is less then unity and have the second order of accuracy on the smooth solutions. It is shown that new algorithms have noticeable advantages in comparison with the wellknown TVDschemes, based on the limitation of fluxes.
Bibliographical reference
Goloviznin V.M., Karabasov S.A., Kobrinsky I.M. BALANCECHARACTERISTIC SCHEMES WITH SEPARATED CONSERVATIVE AND FLUX VARIABLES. (In Russian). Preprint IBRAE200215. Moscow: Nuclear Safety Institute RAS, July 2002. 25 p. — Refs.: 15 items. 
  Modelling of grain face diffusion transport and swelling in UO2 fuel (preprint IBRAE200214)
Preprint IBRAE200214
Berdyshev A.V., Veshchunov M.S.
An advanced model for the grain face transport based on the selfconsistent consideration of gas atoms diffusion, sinking to and resolution from bubbles on grain faces, is developed. An important role of grain boundary diffusion of gas atoms to edges before interlinking of intergranular bubbles, is outlined. The coalescence of face bubbles due to their random migration is considered as the main mechanism of grain face bubbles relaxation. Implementation in the MFPR code of the new model and numerical treatment of various available data on gas release from irradiated fuel, fuel swelling and grain face microstructure, show a satisfactory agreement of the code predictions with measurements.
Bibliographical reference
Berdyshev A.V., Veshchunov M.S. Modelling of grain face diffusion transport and swelling in UO_{2} fuel. Preprint IBRAE200214. Moscow: Nuclear Safety Institute RAS, June 2002. 19 p. — Refs.: 29 items. 
  Environmental data mining and modelling based on machine learning algorithms and geostatistics (preprint IBRAE200213)
Preprint IBRAE200213
Parkin R., Kanevski M., Pozdnukhov A., Timonin V., Maignan M., Yatsalo B., Canu S.
The paper presents some contemporary approaches to the spatial environmental data analysis, processing and presentation. The main topics are concentrated on the decision–oriented problems of environmental and pollution spatial data mining and modelling. The set of tools used consists of machine learning algorithms (MLA) – Multilayer Perceptron and Support Vector Regression, and recently developed geostatistical predictive and simulation models. The innovative part of the report deals with integrated/hybrid models, including ML Residuals Kriging predictions and ML Residuals Sequential Gaussian simulations. ML algorithms efficiently solve problems of spatial nonstationarity, which are difficult for geostatistical approach, but geostatistical tools are widely and successfully applied to characterise the performance of the ML algorithms, analysing the quality and quantity of the spatially structured information extracted from data by ML. Moreover, mixture of ML data driven and geostatistical model based approaches are attractive for decisionmaking process..
Bibliographical reference
Parkin R., Kanevski M., Pozdnukhov A., Timonin V., Maignan M., Yatsalo B., Canu S. ENVIRONMENTAL DATA MINING AND MODELLING BASED ON MACHINE LEARNING ALGORITHMS AND GEOSTATISTICS. Preprint IBRAE200213. Moscow: Nuclear Safety Institute RAS, 2002. 13p. — Refs.: 9 items. 
  Reseach of opportunities of practical use neural networks for identification of the basic parameters of the fractional diffusion (preprint IBRAE200212)
Preprint IBRAE200212
Goloviznin V., Kiselev V., Korotkin I., Semenov V., Hromov A., Yurkov Y.
The inverse problem of identification of fractional diffusion parameters is considered according to practical measurements. For the decision of this problem it is offered to use neural networks. The brief review neural networks is given The various ways of representation of the information on an input in a neural network are considered. Results of research of noise influence and quantity of training examples on accuracy of definition of parameters fractional diffusion also are given. As results of practical measurements it is used numerical decisions of a direct problem.
Bibliographical reference
Goloviznin V., Kiselev V., Korotkin I., Semenov V., Hromov A., Yurkov Y. RESEACH OF OPPORTUNITIES OF PRACTICAL USE NEURAL NETWORKS FOR IDENTIFICATION OF THE BASIC PARAMETERS OF THE FRACTIONAL DIFFUSION. (In Russian). Preprint ¹ IBRAE200212. Moscow: Nuclear Safety Institute RAS, 2002. 37 p. — Refs.: 6 items. 
  Stochastic approximation ratio’s property for oscillating processes (preprint IBRAE200211)
Preprint IBRAE200211
Visochansckiy V.B., Islamov R.T.
Results of stochastic approximation ratio’s property analysis for oscillating processes are presented in this paper. Determination of stochastic approximation ratio’s property in the variation of oscillating function parameters is a purpose of the research. Dependence of stochastic approximation ratio’s on such parameters as, function’s mean value, magnitude and phase is carried out for standard function. Additional investigation of dependence of stochastic approximation ratio’s on range of definition of is developed.
Bibliographical reference
Visochansckiy V.B., Islamov R.T. Stochastic approximation ratio’s property for oscillating processes. (In Russian). Preprint IBRAE200211. Moscow: Nuclear Safety Institute RAS, 2002. 21 p. — Refs.: 3 items. 
  Computational methods for onedimensional fractional diffusion equations (preprint IBRAE200210)
Preprint IBRAE200209
Goloviznin V.M., Kiselev V.P., Korotkin I.A.
In the work the computing algorithms for the numerical decision of a fractional diffusion primal problem in a onedimensional case have been developed and analyzed. Fractional diffusion essentially differs from classical diffusion by behavior of substance concentration on large distances from the initial data source. In the publication the review of basic definitions of fractional derivatives is given. On the basis of this definitions difference methods of the first and second orders of approximation have been constructed. Explicit, partially implicit unconditionally stable schemes and a method based on Fourior transform are also given. The numerous examples of calculations represent computing properties of new algorithms, their detailed comparison is carried out, the second order of convergence of the decision of stationary boundary value problem by a method of an establishment is shown. The given algorithms are supposed to be used first of all for the adjustment of methodical questions of an inverse problem decision — estimation of unknown values of fractional diffusion parameters by the results of experiments on location.
Bibliographical reference
Goloviznin V.M., Kiselev V.P., Korotkin I.A. COMPUTATIONAL METHODS FOR ONEDIMENSIONAL FRACTIONAL DIFFUSION EQUATIONS. Preprint IBRAE200210. Moscow: Nuclear Safety Institute, May 2002. 35 p. — Refs.: 14 items. 
  Classification of environmental data with kernel based algorithms (preprint IBRAE200209)
Preprint IBRAE200209
Pozdnukhov A., Timonin V., Kanevski M., Savelieva E., Chernov S.
Soil type classification is an important problem from different points of view. Vertical migration of radionuclides in soils can be mentioned as an example. The process of migration depends on a number of different properties corresponding both to radionuclides and soils. All soil properties are strongly connected with a soil type. Official soil type maps are not good enough to be used for migration problems. Real soil type is more variable value, than it is usually presented in official maps. Soil type mapping can be improved by using additional information obtained during radionuclide concentration measurement.
In this work the classification problem is solved by machine learning methods such as probabilistic neural networks PNN (supervised learning algorithm) and Support Vector Machines SVM. The advantages of both methods are general nonlinear modelling that avoids the direct modelling of spatial correlation structure. The methods are compared with the nearest neighbour method, the simplest approach to spatial classification.
Bibliographical reference
Pozdnukhov A., Timonin V., Kanevski M., Savelieva E., Chernov S. CLASSIFICATION OF ENVIRONMENTAL DATA WITH KERNEL BASED ALGORITHMS. Preprint IBRAE200209. Moscow: Nuclear Safety Institute RAS, 2002. 23 p. — Refs.: 10 items. 
  Sciencebased modelling of Chernobyl fallout trend (preprint IBRAE200208)
Preprint IBRAE200208
Kanevski M., Savelieva E., Demyanov V., Chernov S., Sorokovikova O., Belikov V.
This work is devoted to the science based modeling of ^{137}Cs contamination due toChernobyl fallout, this fallout is an example of data with nonlinear trend and high local variability and that’s why they are very efficient for testing nonlinear estimators. Science based atmospheric dispersion modeling is performed on the base of the Lagrange model atmospheric transport model and with the help of software NOSTRADAMUS, developed at IBRAE. Nevertheless of detailed preliminary preparation of model parameters the final result shows only the main orientation of the trace. It does not reproduce the spotted structure presented in measured data. Such situation is caused by not exact knowledge of parameters important for atmospheric modeling – Chernobyl example seems to be too difficult for pure sciencebased modeling. The residuals of sciencebased model were studied using Artificial Neural Networks (ANN). Such hybridization allows to improve the results by introducing some “hot spots” distributed not in the main trace.
Bibliographical reference
Kanevski M., Savelieva E., Demyanov V., Chernov S., Sorokovikova O., Belikov V. SCIENCEBASED MODELLING OF CHERNOBYL FALLOUT TREND. Preprint IBRAE200208. Moscow: Nuclear Safety Institute RAS, 2002. 23 p. — Refs.: 15 items. 
  GEOSOM application for data analysis using the selforganizing maps (preprint IBRAE200207)
Preprint IBRAE200207
Kanevski M., Chernov S., Demianov V., Savelieva E., Timonin V., Trouttse A.
In this work the program application GeoSOM described which was developed and realized to data analyze using SelfOrganizing Kohonen’s Maps. The paper contains the theory of SOM method which had been realized. There is given the example of data analysis using GeoSOM application.
Bibliographical reference
Kanevski M., Chernov S., Demianov V., Savelieva E., Timonin V., Trouttse A. GEOSOM APPLICATION FOR DATA ANALYSIS USING THE SELFORGANIZING MAPS. Preprint IBRAE200207. Moscow: Nuclear Safety Institute RAS, 2002. 16 p. — Refs.: 2 items. 

pages: [ 1 ] [ 2 ] [ 3 ] [ 4 ] [ 5 ] [ 6 ] [ 7 ] [ 8 ] [ 9 ] [ 10 ] [ 11 ] [ 12 ] [ 13 ] [ 14 ] [ 15 ] [ 16 ] [ 17 ] [ 18 ]  show all
