DEVELOPMENT OF A REFERENCE ARCHITECTURE FOR BUILDING DIGITAL TWINS IN AUTOMATED PRODUCTION FOR WASTEWATER TREATMENT PLANT CONTROL

Abstract

Analysis of recent research and publications. Analysis of contemporary research on the integration of digital twins into production processes reveals the existence of numerous approaches aimed at improving modeling accuracy, enhancing equipment operational efficiency, and reducing development time and resource consumption. However, existing developments are predominantly oriented toward specific application tasks and do not provide a unified conceptual foundation for the phased implementation of digital twins in industrial environments. Issues related to interface standardization, interoperability between heterogeneous components, efficient data exchange, and a coherent architectural model that enables gradual scaling from pilot projects to the enterprise level require additional investigation. This justifies the relevance of developing a reference architecture for systems utilizing digital twins, particularly for such complex technological facilities as wastewater treatment plants.

Research objective. The objective of the article is to develop a reference multi-layer architecture for building digital twins in automated production for wastewater treatment plant control, which ensures modularity, flexibility, and the capacity for gradual expansion without radical changes to the system structure.

Presentation of the main research material. The article analyzes the key factors that must be considered prior to creating a digital twin: the scope and granularity of modeling (from a single asset to an entire production line), data sources (IoT devices, PLCs, SCADA, ERP/MES systems), modeling techniques (physics-based models, machine learning, or hybrid approaches), technology stack (off-the-shelf solutions, custom cloud-based developments, or combined variants), and intended purpose (prediction, optimization, quality control). The fundamental principles of architecture design are formulated: modularity, openness, scalability, simplicity of initial deployment, and incremental security implementation. Based on these principles, an architecture consisting of five layers is proposed: the Physical Layer, Data Ingestion Layer, Data Management Layer, Modeling and Simulation Layer, and Analytics and Application Layer. The first two layers are located directly at the production site, which minimizes network traffic through local data filtering and transformation, while the remaining layers are deployed in the cloud environment, providing flexibility, scalability, and cost efficiency through the pay-as-you-go model. For each layer, recommendations are provided regarding the selection of protocols (MQTT, OPC UA, HTTP), storage technologies (InfluxDB, PostgreSQL, AWS S3), and modeling tools (MATLAB, Simulink, Unity).

Conclusions. The proposed reference multi-layer architecture allows the implementation of a digital twin to commence with the basic functionality of an individual asset and to be gradually scaled to cover the entire wastewater treatment facility or enterprise. A strategy for the phased development of a digital twin for wastewater treatment plants is defined: at the initial stage, the existing PLC- and SCADA-based infrastructure is used as the data source, and modeling is built upon deterministic physics-based models of biochemical and hydraulic processes. Subsequent stages involve transitioning to hybrid modeling techniques incorporating machine learning algorithms and introducing predictive maintenance and energy consumption optimization functions. This approach minimizes initial capital expenditures and ensures the systematic accumulation of historical data for the further development of the system.