SYNTHESIS OF THERMAL DIAGNOSTIC EXPERT COMPONENTS WITH AN ARTIFICIAL NEURON

Abstract

The article notes the growing popularity of digital programmable technology in diagnostic monitoring systems of electromechanical equipment (EME) for various purposes due to the ability to monitor the technical condition of operating devices in real time. The main reasons that restrain the use of DMS with artificial neural networks in the municipal sphere are considered. It has been noted the directions of improvement of popular means of thermal parameters monitoring and hardware solutions to increase the initial data validity used in the possible EMO fault identification. The purpose of this work was to study and develop components for the formation of initial information, including artificial neurons, which make it possible to increase the reliability of possible fault identification accompanied by heating of individual parts of the operated electromechanical equipment. Based on the adopted algorithm for approximating the initial data arrays, the priority of using the logistic function for modeling the rate of temperature change in the EME was justified. It have been proposed the electronic model structure of an artificial neuron (AN) and an algorithm for generating information output signal, depending on the rate of change of a controlled parameter at a technological object. It have been presented the electronic modeling results in the Simulink environment and the physical implementation of the AN electronic model, which confirmed the suitability of the proposed device in the diagnostic thermal expert of the EME technical condition during its operation in real time. Electronic experiments with AN made it possible to obtain a calibration characteristic for a practical assessment of the tendency for the development of non-standardized thermal events that may cause possible faults in certain parts of the equipment. It have been considered possible options for using AN in local thermal diagnostic tools for the analysis and assessment of events indicating the feasibility of performing unscheduled maintenance or preceding possible and unknown electromechanical equipment faults. It has been presented the results of experiments and simulation of thermal processes, confirming the expandability of the functional diagnostic devices properties with neural network systems, which popularity is constantly growing.