Time series data mining for railway wheel and track monitoring: a survey

Abstract

AbstractThe railway sector has witnessed a significant surge in condition-based maintenance, thanks to the proliferation of sensing technologies and data-driven methodologies, such as machine learning. However, despite the plethora of algorithms designed to detect and classify track irregularities and wheel out-of-roundness, they often fall short when put to the test in real-world scenarios. These shortcomings typically stem from their inability to meet all four critical requirements for constructing an effective maintenance plan: (R1) suitability of the condition-based maintenance strategy, (R2) availability of relevant data, (R3) proper problem formulation, and (R4) accurate evaluation of data mining methods. In response to the absence of a unified framework and standardized guidelines, this survey delves into the realm of time series sensor data and wheel-track interface components for railway structural health monitoring. This survey aims to bridge this gap by offering an extensive categorization, pinpointing existing challenges, and outlining potential directions for future research. Through these efforts, this survey provides a more thorough and targeted exploration of the subject matter, contributing to the advancement of this field.