Basarab

pp. 1-22

pp. 152-174

Application of adhesives and sealants in assembly technology of modern vacuum electro-mechanical devices allows connecting parts of different materials and significantly reduces the cost of products. However, outgassing of the adhesive joints deteriorates the vacuum inside the devices and limits their service life. Prolonged vacuum outgassing adhesive joints at elevated temperatures solves this problem but labor intensity of this operation prevents its use in mass production. For devices which are switched on relatively rare or once the desired vacuum level can be achieved by switching the getter directly before use. A short-time heating of the getter removes the main residual gas component - the hydrocarbons from the inner atmosphere of a vacuum device and provide the its functionality for a long time interval. An additional internal outgassing adhesive joints in the process of manufacturing the device is not required.

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 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.

Results of calculations of incoming and outgoing traffic’s dynamic characteristics were presented in this paper along with the distribution of hardware capacities by the example of the university’s corporate network server. Lyapunov and Hurst exponents, which characterize chaotic and fractal properties of the processes, were evaluated for network traffic. Traffic’s phase diagrams were investigated and, for the first time, appearance of attractors was revealed; these attractors allow one to conduct a more detailed analysis of the impact of load on the network’s traffic capacity. A self-similarity was found for all the processes; this fact confirms the possibility of using fractal models for working with data, including solving urgent problems of predicting behavior of the time series.

A new technology of balancing geared metal resonators of cylindrical Coriolis vibratory gyroscopes (CVG) for strap down inertial navigation systems of low and medium accuracy was proposed. Removal of an unbalanced mass from the teeth surfaces was performed by a high-precision electrochemical method. The proposed balancing algorithm allows to eliminate lower forms of mass defect, one by one or simultaneously, affecting the instrument operation. The algorithm does not require complex calculations when determining the mass removed from the teeth, and it could be used with any number of teeth on the edge of a resonator. The developed technology of balancing can be easily implemented and automated, it does not require expensive equipment.

A neural network algorithm of balancing the non-ideal ring resonator of solid-state wave gyro is considered in the article. The algorithm is based on principles of combinatorial optimization and minimization of the Hopfield network energy; the algorithm allows harmonics compensation of angular mass distribution. The proposed approach can be generalized on various models of discrete balancing and combined with other tools of artificial intelligence: genetic algorithms, simulation annealing et al.