An Intelligent Two-Stage Fault Classification Model for Railway Turnout Systems Based on FastDTW

Abstract

The identification and classification of railway turnout faults are essential for guaranteeing train safety. Traditional diagnostic methods for these faults face challenges due to limited accuracy, stemming from the scarcity of fault samples, and often fail to provide detailed fault classification. In response to these issues, we introduce an advanced two-stage model for the classification of railway turnout faults, utilizing the FastDTW algorithm, known for its efficient approximation of DTW (dynamic time warping) with linear time and space complexity. In the first stage, we employ a Shapelets feature extraction algorithm, based on a greedy strategy, to efficiently identify the most representative segments from long sequence action curves. Progressing to the second stage, the model tackles the inherent singularities in the FastDTW algorithm by incorporating a novel curve segmentation technique, also rooted in a greedy strategy. This technique fine-tunes the fault classification process, leading to more accurate outcomes. The effectiveness and precision of our proposed model were validated empirically using a dataset of 540 faulty curves from a specific high-speed railway station, achieving an impressive classification accuracy of 97%. This substantial accuracy in fault curve classification underscores the potential of our model to significantly enhance the safety and efficiency of railway operations, marking a notable advancement in the field of railway turnout fault classification.