APPLICATION OF EXPERT METHODS AND NEURAL NETWORKS FOR FORECASTING THE REMAINING RESOURCE OF TECHNICAL OBJECTS
DOI:
https://doi.org/10.34185/1991-7848.itmm.2024.01.057Keywords:
expert methods, neural network, prognostication, technical object, residual resourceAbstract
In the work, a comparative analysis of the use of an expert system based on a knowledge base with an expert system based on a neural network for determining the residual resource of technical objects is carried out. It is suggested to use expert methods and neural networks to build the dependence between the input parameters and the residual resource. At the same time, the data used to estimate the remaining resource can be based on both expert assessments and be obtained as a result of technical diagnostics of a technical object. The task of determining the residual resource based on measurements and expert assessments can be formulated as the task of approximating the function of many variables. These variables are the input parameters for building the neural network. Variables include data obtained during technical diagnostics during operation (exceeding parameters, change of environment, dynamic loads) and external factors. Next, some mapping is built in such a way that for each possible input image, an output is formed that characterizes the residual resource of the technical object.
References
Poshyvalov V.P. On the methodology for predicting the service life of structures at the design stage using probabilistic approaches / V. P. Poshevalov, Yu. F. Daniev, L. V. Reznichenko, I. I. Telegina / Mathematical problems of technical mechanics - 2018: Int. Sci. Conf., Kviten, 2018, Dnipro, Ukraine: mat. conf. – Kiev, Cherkasy, Kamyansk, 2018. - P. 31. 2.Kriesel D. A brief introduction to Neural Networks [Electronic resource], 2007. 286 p. [http://www.dkriesel.com/en/science/neural_networks].
Galushkin A.I. Neural networks: basic theory. - M.: Hotline - Telecom, 2010. - 496 p.
Lyubimova T.V., Gorelova A.V. Solving the forecasting problem using neural networks//Innovative Science. 2015, No. 4. P.39-42.
