Weld Battering of Thermite Welds and Flash Butt Welds based on Statistical Evaluations

Research output: Contribution to journalArticlepeer-review

Abstract

Continuous welded rails are an improvement over bolted joints, but still some issues occur associated with welds. One of these issues, rail surface irregularities, is addressed in this paper. When a train passes the uneven rail surface, dynamic impact loads emerge. These loads increase the risk of weld breakage and cause ballast damage, resulting in significant long-term costs. There are two mechanisms driving weld irregularities: Initial irregularities caused by imperfect weld fabrication, and irregularities that grow over time due to the material hardness being adjusted by heating and cooling effects. The latter mechanism is called weld battering and differs for thermite and flash butt welds. A statistical evaluation based on 20 years of data and over 2000 welds yields that battering rates are 25 times higher for thermite welds. Further evaluations reveal that various track design features either amplify or attenuate weld battering. Track radii, steel grade, the passing vehicle collective, sleeper type, rail profile, the width of the extraneous material, the wear rate of the surrounding rail and the manufacturing quality are found to be influential features. The age of a weld and the daily load in gross tonnes have no significant influence. These results indicate that weld battering must be described as local wear and that dynamic loads are not the driving force.

Original languageEnglish
Article number109457
JournalEngineering Failure Analysis
Volume173
DOIs
Publication statusPublished - 15 May 2025

Keywords

  • Asset management
  • Flash butt welding
  • Irregularities
  • Railway track
  • Thermite welding

ASJC Scopus subject areas

  • General Materials Science
  • General Engineering

Fingerprint

Dive into the research topics of 'Weld Battering of Thermite Welds and Flash Butt Welds based on Statistical Evaluations'. Together they form a unique fingerprint.

Cite this