Нейросетевая архитектура вывода решений в опасных ситуациях на сложном технологическом объекте

Abstract

Preventing and neutralize dangerous situations tasks are relevant in the operation of complex technological objects. Complex technological objects present, in particular, in the support city systems (heat, water, energy, gas supply systems), at large industrial, mining, or processing enterprises. The development of dangerous situations at such facilities can lead to undesirable or even catastrophic consequences. The decision-making process to neutralize (prevent) an emerging dangerous situation is aimed at finding an action program that should transfer the current emergency situation into a target, standard situation. The article examines the possibility of implementing a case-based reasoning method for retrieving a solution using a neural network in order to prevent and neutralize dangerous situations at a complex technological object of city infrastructure. The authors consider the situation as a set of elements states of a complex object and the relationships between elements. To solve the tasks, the work examines two neural network architectures: a model that builds upon the multilayer perceptron and the "comparator - adder" architecture.Experiments have shown that the proposed neural network architecture "comparator - adder" showed higher accuracy than the multilayer perceptron for the considered tasks of comparing situations. The obtained result continues the well-known research in the integration of machine learning methods and methods of knowledge-based systems field. It serves as the basis for the further development of decision inference hybrid models for intelligent control of complex objects.