Application of Kohonen's Algorithm in Electrical Diagnostics of Analog Circuits of Radioelectronic Devices

Abstract

The article presents the method of radioelectronic devices diagnosing, based on mathematical modeling of electrical characteristics and circuit operation modes, as well as on application of the Kohonen network for data processing and clustering. The developed method allows to automatically diagnose the technical condition of an electronic device by analyzing the amplitude-frequency characteristics in the control points. TVhen implementing the method for circuit modeling and obtaining the family of amplitude-frequency characteristics, the computer-aided design system NI Multisim was used, and the high-level language Python was used for creating the program code of functional modules. On the basis of the experimental data the fault database of the electronic device was formed, the processing of the information arrays is carried out by the artificial neural network, based on the self-organizing Kohonen maps. Fault classification model allows to automate the process of diagnosis and to identify the technical state of an electronic device.