TECHNICAL AND APPLIED ASPECTS OF CREATING DIGITAL TWINS FOR INDUSTRIAL ENTERPRISES

Abstract

The object of the study is the process of developing a digital twin of an industrial enterprise operating in a heterogeneous manufacturing environment. The research addresses the problem of integrating heterogeneous technological equipment, organizing streaming processing of production data, and implementing predictive analytics of equipment technical condition without full modernization of the existing production infrastructure. The relevance of this work is driven by the need to improve production efficiency, reduce unplanned equipment downtime, and support the digital transformation of industrial enterprises within the context of Industry 4.0. As a result of the study, a five-layer microservice architecture of an industrial digital twin was developed. A multi-protocol data acquisition and streaming processing system supporting OPC-UA, FANUC FOCAS2, and Modbus RTU was implemented, along with a predictive analytics subsystem based on Isolation Forest, LSTM, and LSTM Autoencoder algorithms. The proposed solution enabled integration of 28 out of 34 units of technological equipment without replacing the existing hardware infrastructure, real-time processing of up to 62,000 messages per second, and prediction of bearing remaining useful life with an MAE of 8.3 hours over a 48-hour forecasting horizon. These results were achieved through the combination of a containerized Kubernetes-based microservice architecture, time–frequency analysis methods for industrial signals, and deep learning algorithms for time-series analysis. Unlike existing approaches, the proposed solution enables the integration of equipment of different generations within a unified digital platform and the implementation of a digital twin without full modernization of the production infrastructure. The practical evaluation of the digital twin’s effectiveness was conducted in an experimental simulation manufacturing environment replicating the operation of a machine-building workshop. The obtained results confirmed the effectiveness of the proposed approach: the Overall Equipment Effectiveness (OEE) increased from 72.4% to 78.1%, the number of simulated unplanned downtimes decreased by 63%, and maintenance costs were reduced by 26.6%. The estimated total annual economic effect is approximately 3.5 million UAH, while the projected payback period is about 1.1 years. The obtained results are based on an effectiveness evaluation model developed within the experimental study and can be used to assess the feasibility of implementing digital twins in industrial enterprises with heterogeneous equipment fleets.