Neuroinformatics methodology in a decision support system for industrial process management

Modern organizational and technical systems—as a set of interconnected technical means and the personnel responsible for their operation and intended use—reflect all the trends in digitalization and automation of human activity occurring in the era of the fourth industrial revolution. The complexity of the interrelations between system components and influencing factors determines the complexity of the functions implemented by such systems, while simultaneously increasing the cost of erroneous design decisions. The purpose of this work is to illustrate current trends and an example of a solution for overcoming exponential explosion problems while taking into account multiple factors using neuroinformatics tools to improve efficiency and minimize errors in making optimal decisions on managing multidimensional production processes in organizational and technical systems. An analysis of the subject area revealed the feasibility of solving optimization problems using dynamic neural networks with feedback. In particular, dynamic-static networks have been identified as the most appropriate architectures for solving linear programming problems, due to the clear interpretation of neural network solutions and the ease of implementing inequality constraints. A software implementation for the solution to this problem is described. The experimental dependencies of the performance indicators for classifying the states of production processes, which are subsequently used in the control loop of the technological process of petrochemical production, are presented.