Golikova V. Identification of stochastic objects and processes in the automated system of industrial tests for internal combustion engines.

The thesis is devoted to the solution of the actual industrial problem connected with the control improvement of automated industrial testing of internal combustion engines. For the first time the method to improve control quality and to increase quantitative indicators of automated industrial testing efficiency for the internal combustion engines is offered. It is based on stochastic models which are used to estimate technical state of the tested objects in real time regime. The method to identify scalar and multivariate stochastic processes which characterize dynamic of the engine diagnostic parameters is developed. The stationary linear combinations describing joint stochastic behavior of the elements of multivariate non-stationary processes are determined. The software to control technology of the automated industrial tests and for intellectual system of decision-making based on the testing result is designed. It combines methods to analyze multivariate non-stationary time series and adaptive filtering of linear combinations describing relations between the diagnostic parameters. The new computing algorithm to estimate the reliability of engine automated testing is proved. It allows to determine the identification errors as functions of errors of the coefficient values for the multivariate models and to forecast the quantitative indicators of the testing efficiency at the preliminary stage. The probability to make incorrect decision, when accuracy and reliability of testing results, their terms and costs are given, is estimated.