Predicting interlocking malfunctions for the railway network in Germany

Transportation research procedia, vol. 72, pp. 3481–3488

Abstract

Interlockings generate a major part of all malfunctions on the German railway network infrastructure. Therefore, predictions of future malfunctions are essential to optimize the distribution of available maintenance and renewal funds. This work compares different regression methods to predict the number of malfunctions on individual interlockings. These methods are trained and tested on a data set provided by the German infrastructure manager ‘DB Netz AG’. Using the obtained models, the importance of available feature data to the prediction quality is analyzed.

Authors 6

1. Tamme Emunds corresponding Aachen Institute of Transport Science

   RWTH Aachen University

   Affiliation as printed

   Institute of Transport Science, RWTH Aachen University, Mies-van-der-Rohe-Str, 1, 52074 Aachen, Germany

2. Christopher Wink Aachen Institute of Transport Science

   RWTH Aachen University

   Affiliation as printed

   Institute of Transport Science, RWTH Aachen University, Mies-van-der-Rohe-Str, 1, 52074 Aachen, Germany

3. Martin Lempp
   Affiliation as printed

   DB Netz AG, Adam-Riese-Str.11-13, 60327 Frankfurt a.M., Germany

4. Karim Monem
   Affiliation as printed

   DB Netz AG, Adam-Riese-Str.11-13, 60327 Frankfurt a.M., Germany

5. Björn Dickenbrok
   Affiliation as printed

   DB Netz AG, Adam-Riese-Str.11-13, 60327 Frankfurt a.M., Germany

6. Nils Nießen Aachen Institute of Transport Science

   RWTH Aachen University

   Affiliation as printed

   Institute of Transport Science, RWTH Aachen University, Mies-van-der-Rohe-Str, 1, 52074 Aachen, Germany