Wear prediction for overhead contact wires

Abstract

Overhead contact wires supply power to trains and wear down as they are used. In this thesis, the wear is analysed. Several steps are taken to prepare the data. An algorithm is created to match measurements to specific wires. A window-based change detection algorithm is implemented to divide wires into parts with a similar thickness. Replaced wires and deviating measurements are filtered out using rules. Linear regression is used to average out measurements and find the wear rate. Those results are used to look into the lifetime of the wires and various possible correlations. Almost half of the wires are filtered out in the preparation steps. The wear of most wires remaining wires can be predicted with reasonable accuracy for the next couple of years. The average wear rate was found to be lower than expected, and little wires need to be replaced because of overall wear. New wires wear faster, but no other direct relations were found.