МОДЕЛИРОВАНИЕ ОТКАЗОВ ЭЛЕКТРОННЫХ СХЕМ С ПОМОЩЬЮ НЕЙРОННЫХ СЕТЕЙ

Abstract

The object of researches is electronic circuits.
 For elements of the circuit there are defined characteristics
 of input and output signals in a working condition
 and at a state of non-operability. The subject of
 researches is a reliability of electronic circuits (EC).
 The purpose of the work consists in the automation of
 reliability tests at the expense of the failure simulation
 of electronic circuit elements with the aid of artificial
 neural networks (ANN). There is developed a method
 for carrying out EC reliability tests with the use of automation
 means. During tests one simulates different
 failures of circuit elements. For element failure simulation
 there are used ANN trained fragments. The ANN
 fragments are trained with the use of the selection of
 input and output signals of the element in a working
 condition and at a state non-operability. For the signal
 formation of a working condition a signal generator is
 used. For the signal formation of a state of nonoperability
 the signals from outputs of a noise generator
 are added. To reduce time and costs for training
 there is offered for use the ANN of a special switch
 type which allows copying, replicating, modifying
 ANN, training and forming ANN from its fragments.