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

Abstract Industrial refrigeration systems are essential to preserving the cold chain in sectors such as food and beverage, where equipment failures can result in significant economic losses, increased energy consumption, and product spoilage. However, traditional Supervisory Control and Data Acquisition systems often offer only localized monitoring, limiting fleet-wide visibility and hindering predictive maintenance efforts. This study introduces a scalable and interoperable monitoring platform for commercial refrigeration units, built around Digital Twin aggregates to enable centralized real-time data acquisition and anomaly detection. The system features a multi-tier architecture that integrates OPC UA data exchange protocol for standardized data exchange and utilizes a Long Short-Term Memory (LSTM) encoder-decoder model for semi-supervised anomaly detection. Sensor data is processed to extract time-domain features, and the LSTM model, trained solely on healthy operational data, detects deviations indicative of faults in real time. A functional Digital Twin platform was implemented as a web-based application. Evaluation using a publicly available dataset of industrial refrigeration cycles demonstrates strong performance, achieving an area under the Receiver Operating Characteristic (ROC) curve of 0.95 and an area under the Precision-Recall curve of 0.72 for the fault detection task. While results are promising, current limitations include the detection of only a single generic fault type based exclusively on power consumption data.