Определение разладок в тяговом приводе с использованием критерия Стьюдента

Abstract

Electric traction drive control systems are characterized by a significant number of technical and economic parameters which are time-varying. And it is practically impossible to assess them rapidly by man. At the moment there is no single systematic approach to the rolling stock diagnosis. Therefore it is proposed to perform a rapid assessment of the technical and economic parameters of electric traction drives using a neural network based on Student’scriterion. In general, the Student’s criterion uses the assumption of normality of the observed random variable. It is robust, i.e. insensitive to random variable moderate deviations of a normal distribution. The synthesis method of neural network to monitor replacement of hypotheses and models to change the mathematical expectation and variance based on Student’s criterion, that is developed by authors, can be used in those cases where the Student’s criterion assumption is not satisfied for normality of the observed random variable and other principle of the random variable distribution is known.