A MODERN APPROACH TO BUILDING PREDICTIVE MODELS BASED ON HISTORICAL OPERATING EXPERIENCE OF INDUSTRIAL EQUIPMENT

Chesalov Alexander Uryevich

The digital transformation of industry, the transition to a data economy, and the imperative of technological sovereignty necessitate the development and implementation of intelligent systems for predictive and prescriptive maintenance. Predictive models based on the analysis of accumulated historical operating experience of industrial equipment are becoming a key element of such systems. This article explores a modern approach for creating, training, and operating such models. It discusses a convergent architecture that combines the Industrial Internet of Things, edge artificial intelligence, cloud computing, and cognitive systems. Particular attention is paid to methodological approaches to working with uncertain and contradictory historical data based on the Dempster-Shafer evidence theory, as well as to the prospects for integrating large-scale language models. It is concluded that the synergy of these technologies forms the basis for creating self-learning industrial ecosystems capable of continuously accumulating and leveraging operational experience.

Keywords:predictive maintenance, predictive models, Dempster-Shafer theory.