PLANT (SUBSYSTEM) STATE CONTROL
AT INCOMPLETE MEASUREMENT INFORMATION
ON THE PARAMETER SET DETERMINING ITS
DYNAMICS.
III. RECURSIONS FOR PLANT DYNAMICS AND PLANT
PARAMETERS MEASUREMENT USING NEURAL NETWORKS
G. F. Malykhina, A. V. Merkusheva
A sufficiently great number of measurement
problems arises from the necessity to estimate multidimensional (vector)
characteristic of the controlled plant state and, as a matter of fact, relates
to the problems of measuring dynamic system parameters. The current state
of a dynamic system (DS) is represented by a set of parameters, the value
of which depends on the system whole preceding behavior, and determines the
evolution of the future behavior. Control of such object by means of measurement
systems (MS) is essentially complicated by the fact that lack some state
parameters do not influence measurement sensors, i.e. by incomplete measurement
infor-mation. The problem is solved by transforming the equations describing
the object dynamics and MS into recur-sive form. The use of neural network
(NN), algorithms with adaptive reaction to changing input data, allows one
to simulate (without direct digital methods) the dynamics of object-state
parameters that does not influence MS sensors.