Session: 02-05-01: IoT and Digital Twins

Paper Number: 166442

Real-Time Monitoring Platform for Commercial Refrigerators Using Digital Twin Aggregates 

The reliable operation of industrial refrigeration systems is essential in the food and beverage industry, where even minor failures can lead to significant economic losses, increased energy consumption, and food spoilage. Traditional maintenance strategies, such as reactive maintenance, which involves addressing failures after they occur, and preventive maintenance, which follows scheduled maintenance intervals, often result in excessive downtime and inefficient resource allocation. The emergence of Industry 4.0 technologies, particularly Digital Twin architectures and AI-driven predictive maintenance, has introduced new opportunities for enhancing real-time monitoring and failure prevention in refrigeration systems. However, despite their potential, existing solutions rarely integrate Digital Twin technology with predictive analytics to provide a comprehensive maintenance framework. This research presents a centralized real-time monitoring platform that leverages Digital Twin technology to enhance predictive maintenance for industrial refrigerators. The proposed system continuously synchronizes with real-world sensor data, creating a dynamic virtual representation of refrigeration units. This integration allows for real-time monitoring of operational conditions and facilitates early-stage fault detection, failure classification, and time-to-failure estimation. By incorporating machine learning algorithms and time-series forecasting models, the system predicts potential breakdowns before they occur, enabling proactive interventions to minimize unplanned downtime. The methodological approach is based on a multi-stage data-driven framework that processes sensor data to extract statistical, frequency, and time-domain features, which are then analyzed to detect anomalies using models such as Isolation Forests. Machine learning techniques, such as Long Short-Term Memory (LSTM) networks, classify failure types, while time-series forecasting models predict the remaining useful life of refrigeration components. The predictive analytics are integrated into a Digital Twin simulation to validate the model’s predictions and optimize maintenance scheduling. This hybrid approach enhances the accuracy of failure forecasts and ensures a more effective decision-making process for maintenance planning. The research is currently in its implementation phase, using a publicly available dataset that contains real-world data collected on industrial refrigerators. Although results are not yet available, the system is expected to improve failure prediction accuracy, reduce unplanned downtime, and optimize maintenance scheduling by shifting from a reactive to a data-driven predictive approach. The anticipated benefits include increased operational efficiency, lower maintenance costs, and enhanced system reliability. By integrating Digital Twin technology with predictive analytics, this research contributes to the advancement of intelligent maintenance systems, demonstrating how real-time monitoring and AI-driven forecasting can revolutionize maintenance practices in industrial refrigeration. Future work will focus on improving the integration with the digital twin, refining predictive models, and validating the system’s effectiveness in real-world industrial settings.

Presenting Author: Alessandro Stefanone Politecnico di Milano

Presenting Author Biography: He is a Ph.D. candidate at Politecnico di Milano, specializing in the integration of intelligent services and tools in the product development process. His research focuses on modeling industrial information through knowledge graphs and incorporating AI algorithms and Digital Twins within PLM systems.

Authors:

Alessandro Stefanone Politecnico di Milano
Matteo Sala Politecnico di Milano
Roberto Zanetti Politecnico di Milano
Matteo Pirotta Politecnico di Milano
Marco Rossoni Politecnico di Milano
Giorgio Colombo Politecnico di Milano