To increase the quality of railway service it is important to detect and repair faults in rail tracks before they hinder safe railway service. Nowadays, the tracks are periodically inspected by specialized trains. However, this inspection can not be performed often enough to guarantee the integrity of the tracks. This often leads to halted services at inconvenient times such as morning rush hour.
Televic has developed a mechatronics sensor system that can be equipped on any train and which allows high frequency monitoring of the tracks. However, to extract upcoming track faults this data needs to be continuously monitored by highly experienced workers.
This thesis will try to use anomaly detection algorithms to automate this process. The thesis student will first select the most promising anomaly detection algorithm and then use this algorithm to develop and verify a tool that performs this automatic anomaly detection. The thesis student will verify his system will real-life data.