A METHODOLOGY FOR CONSTRUCTING DIGITAL TWINS OF TECHNOLOGICAL PROCESSES AND ASSESSING THEIR EFFICIENCY USING THE EXAMPLE OF KAMAZ PJSC

The article systematizes the theoretical and practical aspects of creating digital twins in mechanical engineering. A refined classification of maturity levels is proposed: digital model, digital shadow, and full-fledged twin with feedback. The architecture of hybrid models combining physical equations with neural network correction to account for tool wear and process non-stationarity is considered. Special attention is paid to the step-by-step transition from passive monitoring to active control. Using the example of KAMAZ PJSC, the practical results of implementing predictive analytics systems are analyzed, including a 15-25 % reduction in downtime and an increase in Overall Equipment Effectiveness (OEE). Recommendations are formulated for enterprises at different stages of digital transformation of production processes.