Vol 26, No 3 (2018)

The practical experience of synthesis of the model for the water-oil emulsion separation process is considered. In modern literature, the process does not have an accurate analytical description, and its parameters are nonlinearly depend on each other. The source of the data for the model is a series of laboratory experiments conducted using certified oil test methods. The resulting tables have gaps due to an irregular step in the values of the parameters. To simulate a nonlinear process, the Levenberg-Marquardt method was used to train the neural network. An analysis was made that determined the quality criteria for the model. To automate the modeling process, the MATLAB software complex was used, as well as the Neural Network Toolbox. The training of the neural network on the original data set showed that, due to missing parameter values and a small training sample size, it is not possible to obtain an adequate neural network model without preliminary data preparation. To fill in the missing data values, the method of interpolation with Hermite splines was used in the conditions of constraints by the minimum and maximum values of the parameters from the data set. With the help of this method, it was possible to expand the training data set for the neural network by 40 times. Further training of the neural network showed that the reproduction of the training sample is performed with high accuracy, and the quality criteria of the model show acceptable values. Based on the obtained four-dimensional model, it is possible to conduct various types of analysis, both expert and automated. The results of the analysis can be the basis for issuing recommendations on the organization and conducting of the technological process of oil treatment, as well as the synthesis of an adaptive control system with a reference model.

The procedure for selecting the quality indicators of the technological operation of applying polymer insulation in the manufacture of wired communication cables has been considered. It is shown that the quality indicators depend on the type of products, the type of polymer insulation applied, the working frequency band of the signal transmitted through the cable. Ways to ensure the guaranteed maintenance of the performance characteristics of the cable being manufactured during its service life have been proposed. Problems are formulated to solve when automating technological processes of superimposing various types of polymer insulation: continuous, chemically foamed and physically foamed.

The is devoted to ontological data analysis. This is a complex of models and methods designed to extract the formal ontology of the studied knowledge domain from the initial empirical material. Such ontologies are describing the knowledge domain of the aggregate of concepts and relationships. According to observational data and experiments, the initial empirical materials are characterized by incompleteness and inconsistency, which are conditioned by the realities of the accumulation of empirical information. This leads to the fact that the required for ontology’s building formal context - the correspondence “objects-properties” - can be previously presented only within the framework of some multi-valued logic. It needs to be approximated in ordinary binary logic, since effective methods for the derivation of formal concepts are developed only for unambiguous (binary) formal contexts. In order to solve this problem correctly, in the general case it is necessary to know the properties existence constraints of the objects in the studied knowledge domain. But from where can be a priori known such subtle information about the knowledge domain that is only being studied? The main idea of the authors is that this knowledge is a product of a priori hypothesizing of the study's subject. Choosing certain properties to be measured from objects of multidimensional observation and experiments, the subject actually formulates hypothetical concepts about the studied knowledge domain, and using a fundamental cognitive procedure - conceptual scaling of measurable properties - he develops these hypothetical ideas. With such a formation of a system of measurable properties, well-defined properties existence constraints that determine the structure of this system are fixed. Only part of the subject's hypotheses is confirmed a posteriori, and, therefore, certain properties existence constraints are true. The paper identifies and analyzes the fundamental patterns of the structural organization of the system of measurable properties, and thus proposes an exhaustive model of properties existence constraints in the ontological data analysis.

Based on the analysis of known approaches for the synthesis of optimal signal reconstruction algorithms, involving the use of regularizing procedures in solving incorrect inverse problems, and related computational problems, a method for constructing digital filters for solving inverse problems of signal recovery, time series and images using the approximation approach. A method for developing a model of the weight function of the inverse filter, based on the criterion of the minimum quadratic error of the weight function mismatch of the path "direct - inverse filter". The problem of restoration of blurred and defocused images in space systems of remote sensing of the Earth is formulat ed, the possibility of reducing the amount of calculations in the processing of two-dimensional data sets by moving to the solution of the one-dimensional problem is shown. Along with a General approach for the synthesis of the inverse filter, a variant of the algorithm with a number of restrictions imposed on the weight function of the inverse filter is presented. On the basis of the considered method, an approach to the problem of restoring blurred images with a known point scattering function is proposed. Approbation of algorithms on model examples and at processing of the real images received at remote sensing of the Earth is carried out. To quantify the quality of recovery, a relative mean square measure of the difference between the reference and recovered signals (images) was used. The results of testing show that the use of this approach allows to reduce the error of recovery, which gives an advantage in solving problems of approximation and data recovery.

The brief description of the method for measuring of radial clearances between the blade tips and stator of the compressor, using the single-coil eddy-current sensor with self-compensation of temperature effect on the sensor is given. The quality of neighboring blades effect on the results of radial clearances measuring is presented as well as the effect of temperature influence on controlled and neighboring blades with a fixed radial clearance. The results of computing experiments obtained on the base of existing models of electro-magnetic interaction between a sensitive element and the blade are considered. These results contain quantitative estimates of the neighboring blades effect and temperature effect on the controlled and neighboring blades as a function of radial clearances.

The paper proposes a method for increasing the reliability of operation of electric submersible installations (ESI) for oil production, based on the use of quality management mechanisms, such as the diagrams of K. Ishikawa and V. Pareto, as well as the method for analyzing the types and consequences of potential failures. FMEA analysis (Failure Mode and Effects Analysis). Each of these mechanisms is strategically aimed at solving specific problems - K. Ishikawa's diagram (or the fish bone diagram) allows us to graphically determine the most significant cause-effect relationships between the factors and consequences in the problem under study; the FMEA method allows to analyze the technological irregularities, to forecast potentially possible failures (defects), to establish the causes and consequences of their occurrence, to assess the emerging risks and to take measures to eliminate, prevent and reduce damage from their occurrence; on the basis of the FMEA analysis, the construction of the Pareto diagram is carried out, which makes it possible to prioritize the actions necessary to solve the problem. As a result, "narrow" locations in the operation of the ESI are identified and recommendations for their elimination are developed.

Despite the advantages of renewable distributed energy, an increase in its volume leads to serious changes in the characteristics of the electrical network. This is especially noticeable in low and medium voltage networks. To date, the main method of voltage regulation in Russia is to maintain specified voltage levels at control points of the distribution network. The existing centralized system for regulating the operating parameters does not ensure the effectiveness of voltage control in networks. An alternative to the traditional method of voltage regulation is an intelligent power system with an actively-adaptive electrical network. The effectiveness of multi-agent regulation is higher than that of traditional methods of voltage regulation in distribution networks, this is due to system adaptability and real-time operation. The introduction of an intelligent adaptive voltage regulation system reduces power losses, improves the quality of supply to consumers, increases the reliability of the system and its level of throughput.

The hybrid model of back-to-back HVDC is based on the concept of hybrid simulation and includes mathematical models of transformers, harmonic filters, phase reactors, direct current circuit and physical models of a static voltage converter, as well as model of an automatic control and protection system whose operation algorithms implemented at the digital level. The principle of realization of the mathematical models is presented, by example of the harmonic filter model, as well as the physical model of the static voltage converter. The developed hybrid model is integrated into the simulation complex, based on the concept of hybrid simulation, and in which a test scheme of a two-machine electric power system is prepared. In this scheme, the case of a short circuit and power control through the back-to-back HVDC are reproduced. The obtained simulation results coincide with the results obtained in the MatLAB Simulink, which indicates the compliance of the developed model to the digital model. In this case, the properties and capabilities of the concept of hybrid modeling, and as a result the developed model and simulation complex, allow using them in modeling the large scale electric power systems, including in real time and for an unlimited time interval, unlike many digital hard- and software complexes.