Statistical Process Control for Modern Switch Failure Detection

Kurzfassung

A highly available infrastructure is a premise for capable railway operation of high quality. Therefore maintenance is necessary to keep railway infrastructure elements available. Especially switches are critical because they connect different tracks and allow a train to change its moving direction without stopping. Their inspection, maintenance and repair have been identified as a cost driver. Especially switch failures are responsible for a comparable high number of failures and delay minutes. The reduction of failures would not only safe maintenance costs, but let more trains arrive on time and hence increase the attractiveness of the railway transport. Therefore, the Institute of Transportation Systems (TS) in cooperation with the German Railways (DB AG) is exploring ways to apply statistical process control to monitor the condition of switches and their degradation process to reduce failures and thus maintenance costs. Currently, infrastructure managers use commercially available switch diagnostic systems. They are based on measuring the electrical power consumption of the switch engine and comparing it to a manually predefined threshold indicating the failure. But most of these systems do not reach a satisfying accuracy, because they miss too many failures or produce too many false alerts. TS has identified three main aspects to overcome these issues. 1) Instead of interpreting a single feature of the measured signals TS derives multiple statistical features from the signals each defining a different characteristic, so that more useful information is gained for more precise failure detection. 2) The features are used to develop a novel multidimensional correlation model which is robust to non-failure related changes in the original measurements. Therefore false alerts are reduced. 3) The manual adaption of any threshold is eliminated from the process via a new self-adapting algorithm based on clustering techniques. This safes maintenance staffs the effort for recalibrating thresholds when external parameters change, e.g. major temperature shifts with the change of seasons.