Application of Artificial Intelligence for the Implementation of Mismatch Negativity Potential Algorithms in Industrial Automated Predictive Maintenance Systems

Problem statement. One of the urgent problems of industrial automation is that the operation of the few predictive maintenance systems available on the Russian market is usually based on the collection and analysis of equipment data without considering the joint impact of internal and external factors. In the current economic conditions, it is necessary to make a reasonable choice and apply new technologies of artificial intelligence for research and realization of basic principles of mismatch negativity potential, which will open new horizons for increasing efficiency and reliability of industrial automated systems of predictive or prescriptive maintenance of multistage technological processes. Modeling of automatic reactions to environmental changes and prediction of failures will allow to develop adaptive systems that will significantly reduce the risks of failures and accidents, as well as contribute to optimization of production resources and reduction of operating costs. Purpose. To study the possibility of using artificial intelligence technologies to implement algorithms based on the potential of mismatch negativity (MMN) and the possibility of their application in industrial automated systems of predictive or prescriptive maintenance, as well as to develop a basic MMN algorithm and implement it in the Python programming language. Results. An algorithm implementing the basic principles of mismatch negativity potential has been developed. The practical necessity of using such an algorithm, which is based on neurophysiological mechanisms of sensory information processing in the human brain, for detecting anomalies in the operation of industrial equipment caused by external factors such as temperature, humidity, vibrations, and electromagnetic interference was determined, which allows solving the following tasks of industrial automation: anomaly detection, modeling of environmental impact, optimization of operational processes, prediction of failures, adaptation to changes in the environment. The basic architecture of the automated system is proposed, which takes into account the need to use software algorithms of mismatch negativity potential. It consists of modules of data verification, model training, anomaly detection, predictive model, visualization and module of integration with other industrial information and automated systems. The paper also presents the program code for the implementation of the basic MMN algorithm in Python language. Practical significance. The results of the study can be used to design industrial automated systems of predictive or prescriptive maintenance, in which accuracy and decision time play an important role.