Диагностирование электромеханических систем на основе нейросетевых технологий

Abstract

The authors suggest a block diagram and hardware implementation of equipment for electromechanical systems (EMS) technical condition diagnostics. EMS include an acquisition unit, encoding unit, a unit that forms a group control signal of EMS technical condition, its transmission to a communication channel and to a receiving unit for the EMS technical condition group control signal, primary processing unit, extraction and neural network analysis units. The main element of EMS diagnosing neural network system hardware facilities is the neural network device, which shows the output response of a decision neural network regarding EMS technical condition. The article proves the feasibility of using neural chips in embedded diagnostics systems of EMS technical condition. At the same time neurocomputer and transputer computing systems are useful when the unit of collection, coding, generation, transmission of EMS technical condition group control signal is in one zone, and the receiving unit of the EMS technical condition group control signal, its primary processing, extraction and neural network analysis is in another zone. Suggested special software for diagnosing EMS states includes software and mathematical modules for preliminary statistical processing of measured EMS parameters arrays, modules for models synthesis, as well as modules for using synthesized models depending on the class of technical condition control tasks. The paper shows a research on the possibility of applying software of neural networks options for diagnosing complex EMS with a given reliability. There are results of this study.