<![CDATA[This study proposed an anomaly detection model for wayside units in Communication-Based Train Control (CBTC) systems using the Random Forest algorithm. The primary goal was to identify deviations in technical parameters such as voltage, temperature, humidity, and signal strength (RSL) that may indicate potential failures in the system. Data were collected from IoT-based sensors deployed on MRT Jakarta’s CBTC wayside units and transmitted via HTTP to a cloud database for further processing. The Random Forest model was trained using labeled data and evaluated using unseen test data. The evaluation metrics, accuracy, precision, recall, and F1-score, reached 100%, indicating that the model correctly identified both normal and anomalous conditions without misclassification. Further analysis showed that high humidity, excessive panel temperature, and low RSL values were the most frequent anomaly indicators. Based on this, the system also generated maintenance recommendations, making it not only reactive but also proactive in supporting condition-based maintenance (CBM). The results demonstrated that the Random Forest-based anomaly detection system is highly effective, scalable, and reliable for real-time monitoring of railway infrastructures. This approach can serve as a foundation for future development of smart maintenance systems in other safety-critical domains.]]>
Copyrights © 2025
