| Title |
The Design of Big Data Railway Safety Platform Architecture and Application Model |
| Authors |
이승신(Seung-Shin Lee) ; 오염덕(Ryum-duck Oh) |
| DOI |
https://doi.org/10.5370/KIEE.2024.73.3.567 |
| Keywords |
Railway Safety Platform; Big Data Architecture; MQTT (Message Queue Transport Telemetry); Kafka; MongoDB; YOLOv5 |
| Abstract |
In this study, we propose a big data railway safety platform architecture by applying communication and database technologies and platform architectures used in many industries for real-time failure and anomaly detection of railway operations. There have been studies on big data architecture in data collection, communication, storage, and analysis areas. However, previous studies have not addressed the design of big data architecture for the safe operation of railways specifically. Therefore, in this study, in order to collect, store, and analyze data that may occur in railway operations, we designed an architecture that can be implemented by using currently available technologies from the perspective of the entire data life cycle. In particular, a combination of MQTT and Kafka was proposed as a message and event broker for the railway safety platform architecture, and MongoDB was ultimately proposed as a NoSQL database. In addition, the application model of the big data railway safety platform was presented using the designed architecture, and YOLOv5, an object detection algorithm, was used to conduct an experiment on how image data from railroad tracks can be used in anomaly detection of railway operations. The neural network trained with YOLOv5 can accurately classify eight rail components of the railway and also classify the abnormal states of the eight components relatively accurately. In subsequent research, we plan to implement this architecture as a real big data platform to expand anomaly detection experiments on railroad tracks. |