A Review of Research on Detection and Evaluation of Rail Surface Defects

EasyChair Preprint 7244

Abstract

Defects on the rail surface will accelerate the wear of the wheels. At the same time, when the wheel is periodically hitting and rolling surface defects, the defects will gradually develop into the interior, which greatly increases the possibility of train derailment and cause serious safety accidents. Timely inspection of the railway tracks to find defects as early as possible is an important condition for ensuring the safe operation of railways, also prolongs the service life of railways, because most of the rolling contact fatigue (RCF) can be eliminated during the rail grinding process. Such defects appear as spalling and cracks in the initial stage of the rail surface. Manual detection has been difficult to meet the large-scale railway operating mileage. A more efficient automatic detection method is indispensable. This article reviews the latest research and exploration on the defect inspection of rail surface in recent years. In the article, there is not only the application of traditional ultrasonic and acceleration detection methods, but also the contribution of computer vision and deep learning to the detection of defects on the rail surface. The new detection technology can even classify and evaluate the damage, further improving the efficiency of the detection system. The emerging research on defect state prediction to reduce inspection costs is interesting.

Keyphrases: CNN, detect, machine learning, rail surface defects

@booklet{EasyChair:7244,
  author    = {Lei Kou},
  title     = {A Review of Research on Detection and Evaluation of Rail Surface Defects},
  howpublished = {EasyChair Preprint 7244},
  year      = {EasyChair, 2021}}