Predicting unplanned maintenance needs related to rail track geometry

摘要

The present work puts forward a simple method to predict unplanned maintenance needsrelated to rail track geometric condition for future implementation in a Decision SupportSystem for maintenance and renewal decisions. An exploratory analysis through logisticregression was conducted using the track geometry inspection records from the PortugueseInfrastructure Manager (REFER) databases, in order to predict spot maintenance needsdepending on planned maintenance criteria and other explaining variables such as thepresence of bridges and switches. Main findings showed that the standard deviation ofhorizontal alignment defects (filtered in the wavelength range 3-25m) is a statisticallysignificant predictor of unplanned maintenance needs due to track geometry condition.