A Novel Approach to Detect the Bearing Fault in Rail Networks Using Big Data Classification and Delivery through Edge Processing

Abstract

The proposed model aims to introduce a novel method to gather sensor data from wheel-bearings and process the data to detect the fault and this can be implemented as a part of the Railway Condition Monitoring system (RCM). In the proposed architecture, a method is implemented to check the sensor data against external factors to increase the accuracy of the system. This is done through the use of Edge detection techniques and using it to recognize the track shapes by checking images of the tracks. Furthermore, it would use edge processing to reduce the load on the network infrastructure. The edge processing also allows for onboard fault detection all the while simplifying the data for transfer across the network. Along with these, the model also uses SVM with PCA for higher accuracy in the classification of the data by reducing the need to eliminate variables. The extracted feature of the data would also be sent to the cloud or the data centre for big data analysis. The proposed model’s efficiency is further increased using the Laplacian Score which was implemented to further organize the data. Through the proposed model, the load on the network infrastructure was reduced, at the same time, it was able to make an accurate prediction for predictive maintenance and provides increased safety for the passengers of the train.