SCIENCE & EDUCATION

A modeling problem of information risks in computer-aided systems was considered in this paper. It was shown that a usage of fuzzy models is a promising approach to solving this problem because their development requires far less information about the system. Analysis of various types of fuzzy models was carried out; their main features were emphasized. The Mamdani model was recognized as the most suitable for estimating information risks. Combination of fuzzy and neural network modeling was proposed; that means the creation of neuro-fuzzy networks by transforming the Mamdani fuzzy model to a self-learning neural network, which operates on the basis of fuzzy logical apparatus and fuzzy sets. An example of a neuro-fuzzy network for an assessment of information risks was presented.

This work is a part of investigation devoted to the development of design methods for ontological knowledge bases, oriented at the decision support in the enterprise information systems. An approach to finding solutions in knowledge bases using document’s metadata was considered. Document’s metadata as well as search queries were represented as the frames of design and search patterns respectively. Slots of those patterns corresponded to the concepts’ roles in the utilized ontology. Specified roles divided concepts of ontology, a document and a query into clusters. Semantic networks for those clusters were defined in such a way that search queries of a document were represented as a set of semantic networks, corresponding to the slots of design and search patterns. The relevancy of a document was estimated by a set of metrics which formalized the nearness of semantic networks. A problem formulation of multi-criteria estimation of the relevancy of documents in the enterprise ontological knowledge base and an adaptive solution method were presented in this paper.

This paper describes a method for solving the inverse dynamic problem with the use of the systems for simulation of complex technical objects. Mathematical basis for solving that problem were presented. The problem-solving procedure contains two stages. Numerical simulation of a mechanical system under given external influence is performed at the first stage; this influence is represented as dependence of position coordinates of the system’s several elements on time. During this simulation reciprocal deflections of the operating mechanism’s application points are recorded. At the second stage previously applied forces are removed. Simulation of the system under the influence of displacements in the operating mechanisms determined at the first stage is carried out. In the process created forces are registered. An example of solving the inverse dynamic problem for a tripod was presented.

This paper deals with the problem of multi-criterion optimization. It was assumed that the Pareto set is a solution to this problem. A promising approach to solving this problem is the adaptive weighted sum (AWS) method. Just like the classic weighted sum method based on the additive convolution of partial optimality criteria, the adaptive method also is based on the  convolution. But the AWS method also proposes adaptation of weighting coefficients during iterations on the basis of the current location of the search subarea. This method works with meta-models of criterion functions in order to reduce computational costs. Results of our investigations showed that the AWS method provides a high quality Pareto-approximation in case of a convex, although, maybe, unconnected Pareto frontier. For problems with a concave Pareto frontier this method doesn’t always provide sufficient quality of solution or it does but this takes too much time. In some cases the AWS method could provide an unacceptable solution caused by a special method of constraint satisfaction. The purpose of this work is to overcome specified disadvantages of the AWS method. A problem formulation for multi-criterion optimization was presented along with basic approaches to its solving. Several modifications of the ASW method proposed by the authors were considered. A brief description of the developed software implementing the method and its modifications were presented. Performance study of this software was carried out.

In this paper a mathematical model based on directed graphs was presented. This model allows one to apply a quantitative approach to risk assessment and vulnerability detection in information networks, which is an important step during the security system design. A utility function and a heuristic for informed memory-bounded search SMA* algorithm were also proposed. These results could be used for security audit and countermeasures planning.

The problem of hierarchical representation of a computer network for carrying out its multiscale analysis was considered. An original problem was reduced to the known set cover problem (SCP). Since this problem is NP-difficult, approximate heuristic techniques were useful for its solving. We proposed an algorithm based on the recurrent Hopfield neural network and its combination with the simulated annealing algorithm (so-called Boltzmann machine). A greedy algorithm was used for a construction of hierarchy levels. This approach was evaluated with a use of the test problems from OR-Library. Proposed approach could be used in the systems of traffic analysis and anomaly detection in computer networks.

In this article data on performance and effectiveness of the GRACE-S family of stream ciphers based on generalised cellular automata and expander graphs were discussed. Altera FPGA was used as a platform for hardware implementation. The performance of GRACE-S stream ciphers was compared with the performance of stream ciphers that had won the eSTREAM competition. According to the performed tests, the performance of the GRACE–S family of stream ciphers significantly (up to 60 times) exceeds the performance of the best-known stream ciphers.

This paper presents an analysis of existent nearest neighbor algorithms, and a description of a new nearest neighbor algorithm. A parallel version of the proposed algorithm was developed; its implementation using parallel computations was also described. Experimental results on estimating efficiency of the developed program implementation were provided. The proposed implementation is a C++ library. Possible application domains are geo-systems, liquid motion simulation and other areas where it is required to determine the nearest neighbors in a large bulk of data.

The author considers a possibility of using matched filtering of UWB signals. Maximum achievable characteristics of UWB systems make them promising for many application domains such as data transmission systems, medical systems, security systems, radars, etc. Processing is possible with the use of special filtration devices created on the basis of irregular transmission lines with the T-wave, in particular, on the basis of matched filters, which is the main topic of this paper. For a UWB sounding signal duration is much shorter than duration of a signal reflected from the target; therefore, it could be taken as a sampling period. In this case, application of preliminary matched filtering of the reflected signal, followed by traditional processing, is possible.

In this paper the author proposes a new approach to description of mechanical linkages which determine a certain position of machine’s or device’s elements in a technical system. The location relation that exists between parts of any machine or device was considered to be many-placed and is presented as a hyper-graph. A mathematical description of the product assembly sequence was presented as a contraction of vertices of the hyper-graph. Necessary conditions of contractibility for this type of graphs were proposed. A new structure - a lattice of all s-sets - was introduced. It was shown that this lattice could be used for generation of various design decisions at the stage of technological preparation of assembly production.

A new method was proposed to solve terminal control problems for multidimensional affine systems. The system under consideration is supposed to be equivalent to a regular system of a quasi-canonical form. A necessary and sufficient condition for existence of a solution of terminal control problems for transformed systems was formulated. A sufficient condition for solvability of terminal control problems was proved for quasi-canonical systems with nonlinear subsystem dimension not exceeding control dimension. An algorithm was designed to construct a solution of terminal control problems for this class of systems. A numerical example was presented to illustrate the proposed algorithm.

This paper deals with solving a tracking problem for a scheduled change in the angle of attack initiated by a target direction system with regard to longitudinal dynamics of an air-to-air missile. Control system design was performed with the use of an integrator back-stepping method. The author considers a model of longitudinal dynamics which takes into account dynamics of rocket control organs along with dependence of aero-dynamical coefficients on the modulus of the angle of attack. According to the results of the numerical simulation it is possible to draw a conclusion about working capacity of the control algorithm based on the investigated simplified model of longitudinal dynamics. Control of technical systems is a possible application domain.

In this paper the authors formulated a problem of optimising the composite micro-structure, which is a problem of minimisation of the composite density under given restrictions on the components of effective elasticity and heat-conduction tensors, and effective limits of the composite strength. The authors proposed a method for solving an optimal design problem of a composite material reinforced with micro-spheres and designed for thermal shielding. The method of asymptotical averaging (homogenisation method) was used for calculating effective composite characteristics such as elastic and heat-conduction modulus tensors and strength limits; the finite element method was used for solving local 3D heat conduction and elasticity theory problems on the composite’s periodicity cells. The finite element method was implemented with the use of GCD software developed at “FN-11” department of Bauman MSTU. The Hooke Jeeves algorithm was used for solving optimisation problems. Numerical implementation of the proposed method of micro-structure optimisation was developed by the example of a composite based on glass micro-spheres.

This paper deals with a co-evolutionary algorithm of global optimisation, Co-PSO, based on particle swarm optimisation. Implementation of this algorithm developed in MATLAB was described in this work. This software is oriented to a parallel operation of the given number of PSO algorithms at the logical level. These algorithms use various neighbourhood topologies of particles and/or various values of their free parameters. Results of a comprehensive numerical experiment on performance study of Co-PSO and its implementation were also presented. The experiment was based on Rosenbrock, Himmelblau and Rastrigin benchmark functions. Experimental results show superiority of the Co-PSO algorithm over the canonical PSO. A three-criterion problem of optimal control for a spacecraft during its reentry was solved with the use of Co-PSO. Methods of additive scalar convolution and reduction of the optimal control problem to a nonlinear programming problem were used in this work.

This article deals with the identification method for a linear dynamic element with known transfer function using an experimental frequency locus with random measurement errors. A transfer function of an element was selected as a model. It was proposed to search for the solution to the identification problem in the class of hodographs, defined by the element’s model. Search for unknown coefficients of a transfer function of the element’s model was carried out by minimizing the proposed proximity measure of the experimental element’s locus and the model’s locus. As a result, the specified problem was reduced to a system of linear equations. An illustrative computing experiment for a second-order element showed that an error of the transfer function’s coefficients was comparable with the range of measurement errors of experimental samples of this element’s locus.

Telemetric information recording systems are usually designed for display, preprocessing and record of incoming data stream to any kind of storage media in real time  In case a computer is part of a system its HDD or flash drive could serve as a storage device. In that case, the data stream is stored as an ordinary file. Telemetry data recording process can take for a long time, while the amount of recorded data could grow up to tens of gigabytes or even more. It looks reasonable to compress these data while saving them to a storage media and keep them there in archived files. Later when post-processing is applied to the recorded data, it is often interesting to work with particular fragments of a record, while fast moving of the “read pointer” within a compressed file and extracting of particular data blocks is required. In other words we need to provide fast read-only random access to the compressed data. OS “built-in” and/or well-known compression software utilities (such as Zip) generally have the same disadvantage: they have to decompress archive file from the very beginning up to a specified point. It is impossible to start unpacking a file from an arbitrary position due to the fact that the deflate algorithm that is in the core of these utilities always needs to know some prehistory in the sequence of uncompressed bytes to proceed with decompression. It is of no importance for small files, because unpacking routines are fast and highly optimized, plus unpacked data can be easily cached. In the meantime, for large telemetry records the problem remains. In this paper the author proposes an effective method that allows to make read-only random access to the packed data faster. This is achieved by inserting some supplementary information into the archive file while it is being recorded. This extra information allows starting decompression from a set of "reference points" within the archived file, but not only from zero index byte. Required modifications in the famous zlib compression library are described.

The article deals with the statistical analysis of a database of component and spare part supplies of the network of repair providers in automotive industry. Tests of regression, variance, correlation, factor analysis and other methods of multivariate statistical analysis were performed. Some patterns of the order placement, allowing one to build forecasting procedures for sales volume in a more substantiated way were obtained.

This paper examines the operating reliability of motor vehicles. It was shown that the structure of maintenance operations and repair influences the cost of performance assurance. A method of preventive replacements was proposed in this work; this method transfers part of potential repair operations to the scheduled maintenance. Such decision contributes to increasing the operating reliability and the scheduled maintenance’s price. An optimization model, which allows one to determine the rational amount of preventive replacements, was also proposed in this work.

The authors consider an affine third order system which describes action of a chemical batch reactor with a ternary working mixture. The terminal control problem for this system with restrictions on state variables was investigated. Solution to this problem is based on the orbital linearization approach by which the original problem is transformed to a terminal problem for a non-stationary second-order system in a canonical form. For the transformed terminal problem conditions for solution existence were obtained; a method of obtaining this solution was also proposed. Operability of the proposed method was illustrated by mathematical simulation.

Questions of introducing a uniform technology of the selection of students for master courses into the educational process were considered by the example of creating a multi-agent system (MAS). A system being designed would solve three problems: determination of an optimal specialization of master courses on the basis of analysis of the competence model of a bachelor graduate, classification of students by groups depending on the level of their motivation and individual training results in a bachelor degree course, and also formation of master’s educational route if a student has bachelor’s non-formed competences which are necessary for further training. The architecture of MAS was described. Functions and a method of  agents’ interaction were also presented.