Модель прогнозирования класса готовности датчиков термохимического газосигнализатора для предотвращения пожаров и взрывов
Year: 2024
Type: article
Abstract: Introduction. The relevance of the problem considered in the paper is in the need to improve the automation and intellectualization of organizational management of fire and explosion safety at the facilities of the fuel and energy complex. Zero drift due to the influence of a number of parameters in thermochemical sensors reduces their sensitivity, which requires the development of new approaches in the implementation of their maintenance. The sensor readiness class prediction model, based on a convolutional neural network (CNN), allows to adjust the established decision-making process, preventing dangerous situations at early stages of development. Objective. To increase the efficiency of decision makers (DM) in planning the work of teams performing diagnostics and maintenance of auxiliary equipment used , among other things, to ensure fire safety (FS). Methods. To calculate the classification values when predicting the readiness class of gas alarm sensors , a strategic planning method based on the importance of features (dynamic strategic planning using mathematical programming) was used . According to it , the readiness class was defined as the sum of the binary values of the features multiplied by the normalized value of their importance. Results and discussion. To carry out calculations and evaluate the results of the CNN application , the authors developed a programme in the Python programming language . It was used to generate a common data set from which training and test sets were selected in a ratio of 9 : 1. After their formation , the CNN was trained. Testing showed that the DM can predict the readiness class of gas alarm sensors with a probability of 89 %. Conclusions. The CNN presented in the paper allows to increase the efficiency of the DM work when planning the work of teams performing diagnostics and technical maintenance of auxiliary equipment for FS. The principle of operation of this CNN can be used to solve other similar planning and management tasks.
Authors: И. В. Самарин, А. Ю. Строгонов, А. В. Крючков
Institution: Gubkin Russian State University of Oil and Gas
Language: Russian
Subfield: Control and Systems Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W4404961882
