Increasing the Stability of Organisational and Technical Systems Through the Use of Machine Learning

The stability of technical and organisational-technical systems is an integral factor in the functioning of a wide range of complex technical and organisational-technical systems, such as energy systems, industrial facilities, transport systems, and others. While it depends on various factors, there are currently no means to support decision-making for automated dispatch control systems users in case of failures. The article analyses the effects of failures and breakdowns in complex technical and organisational-technical systems. It provides examples of major technogenic disasters. The study examines the human factor influence on the emergence of technogenic disasters. Currently, automated dispatch control systems do not have any means to facilitate decision-making by providing the dispatcher with a clear plan of action to troubleshoot and eliminate a possible disaster. The authors propose a concept of support means for decision-making in response to crises using automated dispatch systems for managing complex organisational-technical systems based on machine learning and artificial intelligence. The means is based on machine learning methods and has a two-component structure: an artificial neural network that defines the type of failure and a programme component that selects the algorithm. The neural network is implemented in the Python programming language, and the user interface is in the C# language. The neural network fulfils the classification task defining the failure based on an array of logical variables that demonstrate the change of state of system elements. The programme component uses the database created with SQL language to define the algorithm of actions for the dispatcher. The authors describe the dataset for the model in case of its use in energy systems. There is also a description of the model components. The model’s distinguishing features are adaptivity due to machine learning methods and the possibility of further model training to increase the classification accuracy based on its use cases. Keywords: machine learning, automated dispatch control systems, decision-making support, system stability.