A quick search method for determining corresponding points from vehicle-vibration data

This paper presents a method for reducing the enormous computation time required by adaptive dynamic time warping (ADTW). A previous method has been proposed that converts a time-axis sequence, which is obtained from passenger trains, to a space-axis sequence with a high degree of accuracy. However, ADTW has the drawback of enormous computation time for converting the time-axis sequence to space-axis sequence because it applies DTW to search for an optimal corresponding point for two vibration sequences. Therefore, an efficient corresponding-point-search processing algorithm is desirable for realizing quasi-real-time processing using passenger trains, which can monitor the rail condition several times a day. In this paper, we propose space-series sequence active search (SAS) as a high-speed processing method for obtaining a corresponding-point-search algorithm. SAS is a method that matches shifting features. It is based on matching the appearance positions of the intervals of a similar signal by shifting one sample at a time. Computer simulation shows that the computation time of SAS is 1/126 times that of ADTW.