Выбор системы информативных параметров для дискриминации альтернативных состояний топливной системы дизеля

Abstract

Измерительные сигналы, отражающие локальные процессы сложных промышленных агрегатов, несут важную информацию о долговременной функциональной стабильности таких динамических объектов. Однако, обнаружить такую информацию в сигналах с априори неизвестными вероятностными моделями нестационарности – это проблема. Ее возможное решение – создание информационных технологий параметризации и нормирования случайных спектральных изменений сигналов при существенных ограничениях на время наблюдения. Такие технологии снизят риски принятия решений при контроле и диагностике функциональных состояний промышленных, транспортных, технологических объектов. Выявление закономерностей в случайных моделях нестационарности – это получение принципиально новой дополнительной информации о функциональных свойствах динамического объекта, способствующей решению многих проблемных задач идентификации объектов и оптимального синтеза информационных компьютеризированных систем в условиях априорной неопределенности.Measuring signals reflecting local processes of complex industrial aggregates contain important information about the long-term functional stability of such dynamic objects. However, it is a problem to detect such information in signals with a priori unknown transient probabilistic models. The possible solution is to create information technologies for the parametrization and normalization of random spectral changes in signals with significant limitations during observation. Such technologies will reduce the risks of decision making when monitoring and diagnosing the functional conditions of industrial, transportation and production facilities. Identifying regularities in transient random models means the acquisition of fundamentally new additional information on the functional properties of a dynamic object. That leads to the solution of many problems concerning objects identification and the optimal synthesis of computerized information systems under conditions of a prior uncertainty. The article deals with mathematical models of the informative parameters construction in the form of cross-spectral correlation coefficients, which characterize the patterns of spectral non-transiency of high-frequency vibration signals. The efficiency of primary measuring transformations is shown in the form of statistics of accumulated sums, which adequately describe localized instantaneous velocities and vibration signals acceleration. The results of the research provided a number of scientific results. The cumulative analysis of probabilistic model of the transient vibration signal allowed broadening the information capabilities of the known single-model T-statistics, which is used to detect changes in the instantaneous power of Gaussian random signals. For the first time mathematical models of improved V- and W-statistics were obtained. They are based on the T- statistics and take into account changes in cumulants of the fourth order, which makes it possible to use these models for vibrodiagnostics of non-Gaussian random measuring signals. Correlation analysis of a two-dimensional system wavelet spectrum of random V- and W-statistics allowed developing a mathematical model of the cross-spectral correlation coefficient, which carry diagnostic information on changes in the transient vibration signal wavelet spectra. The possibility of increasing the expected amount of information as a result of vibrodiagnostics is proved by taking into account the effects of spectral vibration signals transiency. The proposed probabilistic model of the cross-spectral correlation coefficient, in the form of a multiple Taylor series, allowed determining mathematical models according to the impact of changes in frequency, energy and stochastic properties of the V-statistics wavelet spectrum on cross-spectral correlation coefficient. This made it possible to control the effects of transient spectral periodic random vibration signals using the indicated effects for obtaining additional diagnostic information. The system of informative parameters developed on the basis of two-dimensional (on a scale and shift) model of the cross-spectral correlation coefficient for V- and W-statistics showed a high control and diagnostic efficiency, providing not only a decrease (up to 8 times) of the average risks, but also an increase in the statistical power of the decision making rule, especially for controlling prefault conditions dealing with random vibration signals. Using the algebraic model of dispersive analysis of cross-spectral covariates, according to frequency and time, allowed constructing an expanded system of a priori independent coefficients of cross-spectral correlation, providing a separate control over the functional (slow) and random (fast) changes in the transient spectral vibration signals. Such selectivity of vibration control makes it possible to expand the number of controlled technical conditions, without reducing the overall possibility of decision-making. The hardware, algorithmic and software tools were developed and implemented, providing structural optimization of the computerized system for functional situational modeling and information provision as a part of armored vehicle in simulator complexes, which ensured an increase in the accuracy during the estimation of control signal levels.