MODEL-ORIENTED METHOD FOR CONSTRUCTION OF INTELLIGENT INFORMATION SYSTEMS OF DIAGNOSING BASED ON VOLTERRA KERNELS

Abstract

The method of building an intelligent computing system for diagnostics of nonlinear dynamic objects is
offered in this paper. This method is based on integral power Volterra series using as model of objects. The diagnostic
features space is built using such models. There are discrete values of first order Volterra kernels and diagonal section
of second order Volterra kernels as well as moments of Volterra kernels.
Estimations of correct recognition probability of objects states base on chosen diagnostic features sets are
received using maximum likelihood estimation method.
Second order Volterra kernels sections give more information about diagnostic object than first order Volterra
kernels. The possibility and advantages of diagnostic model using for object as a union of first and second order
Volterra kernels is shown. These models provide the highest information level about object being diagnosed. The
highest informativeness and noise immunity is reached using union of Volterra kernels moments of the first order and
Volterra kernels diagonal sections of the second order.
All features set in noiseless conditions usually have several best solutions (features combinations) or several
solutions that are in the neighborhood of best solution. The selection of the best features sets should be carried out
considering the changes of the diagnostic quality with impact of noise. The numerical experimental results are applied
for switched reluctance motor diagnostic model building.