Improved SVD-SSDA on Spark for real-time subway train tracking based on ultra-weak FBG arrays

Tracking and ensuring the safety of trains is an important issue in subway operation management. Under the long-distance monitoring requirements, extracting features in real-time from large-scale stream data to track trains is a large and time-consuming task. With the support of the dynamic and distributed monitoring capabilities of ultra-weak fiber Bragg grating (UWFBG) arrays, this paper proposes a method combining the singular value decomposition (SVD) and the sequential similarity detection algorithm (SSDA) to handle the stream data to track trains in real-time. First, the vibration signal is denoised and is converted into a grayscale image using sliding window. Then, to improve the efficiency of recognition, the singular value features and the texture features are combined to build a template library for gray-scale image matching on the basis of SVD and SSDA. The details of SVD-SSDA deployment on Spark are illustrated to ensure real-time performance. Finally, the experimental results on the actual train data indicate that SVD-SSDA on Spark using ultra-weak FBG arrays can effectively identify the data stream and satisfy the requirements for real-time train tracking.

Automated summary

This paper proposes a method combining singular value decomposition (SVD) and sequential similarity detection algorithm (SSDA), with the support of ultra-weak fiber Bragg grating (UWFBG) arrays, to track trains in real-time from large-scale stream data. Experimental results indicate that the proposed method can effectively satisfy the requirements for real-time train tracking.