Русский
!

Conference publications

Abstracts

XXI conference

Information support systems for diagnostics of the objects of different nature on the basis of Volterra models

Grigorenko S.N., Pavlenko V.D.

Ukraine, 65044, Odessa, av. Shevchenko, 1, ONPU. Department of KSU.

1 pp. (accepted)

A new method is proposed model diagnostics, based on non-parametric identification object of the control (OC) and building decision rules classify in space diagnostic signs with the help of methods for pattern recognition learning, different from using known as source of diagnostic information nonparametric nonlinear dynamical models OC in the form of multidimensional Volterra kernels [1], which are identified by the results of experimental studies OC “input - output”. It allows to realize a universal approach when building space diagnostic signs in the system of the recognition of the states of different nature OC, improve the reliability of diagnosis.

Developed software tools that implement the proposed methods and algorithms information support of the automated systems of indirect control and diagnostics of conditions of objects of different nature, which is integrated into the the package of applied programs “IMIDA” (Instrumental Milieu for Intellectual Data Analysis).

Established that the identification of a valve-jet drives in the form of model Volterra, kernels of the 1nd order, practically independent of changes in the value of the controlled parameter, however diagonal cross-section kernels 2nd order varies considerably in size and can therefore be used as a source of primary data at construction of diagnostic models.



© 2004 Designed by Lyceum of Informational Technologies №1533