Method for Construction the Diagnostic Features Space of Switched Reluctance Motors Based on Integral Dynamic Models

Abstract

The work is devoted to the problem of construction the diagnostic models for the nonlinear dy-namic objects. The aim of the work is to improve the reliability and fast operation in diagnosis of the states of electrical motors of under conditions of an a priori uncertainty. The a priori un-certainty results from an insufficient study of the processes, which occur in the objects of diag-nosis due to the operation in a wide range of external conditions and the presence of a great amount of disturbing effects along with environmental interferences. This aim is achieved by the development of the technical diagnosis method based on the information models of the nonline-ar dynamic objects of diagnosis, which are obtained using the nonparametric identification pro-cedure. As the information models of diagnostic objects, the integral nonparametric dynamic models based on multidimensional weight functions are considered. The most significant results consist in obtaining the method with a further development in construction of a space of diag-nostic features of the nonlinear dynamic objects based on the correlation analysis as a stage of the features’ filtration. The latter ensures the maximum diagnostic reliability. Significance of the obtained results: the application of the proposed method allows both high reliability of the object diagnosis under the priory uncertainty, and improvement of the diagnostic procedure fast opera-tion owing to the feature filtration. The proposed method was tested using the data of the diag-nosis of the switched reluctance motors.

Keywords

• nonlinear dynamic objects
• a
• diagnostic models
• model reduction
• feature selection
• correlation analysis