Predicting Unplanned Maintenance Needs Related to Rail Track Geometry

The present work puts forward a simple method to predict unplanned maintenance needs related to rail track geometric condition for future implementation in a Decision Support System for maintenance and renewal decisions. An exploratory analysis through logistic regression was conducted using the track geometry inspection records from the Portuguese Infrastructure Manager (REFER) databases, in order to predict spot maintenance needs depending on planned maintenance criteria and other explaining variables such as the presence of bridges and switches. Main findings showed that the standard deviation of horizontal alignment defects (filtered in the wavelength range 3-25m) is a statistically significant predictor of unplanned maintenance needs due to track geometry condition.