Electric current profile feature extraction algorithm with an implementation for point machines

Failures, operational downtime and especially accidents are highly unwanted in the transportation sector. Every such incident may cause both enormous human and financial distress. This raises the demand for smart sensor solutions which are not only able to detect faults, but also to monitor the condition of critical components in the aforementioned sector. In order to monitor the condition of a device or machine, access to relevant parameters has to be granted. A commonly accessible value in the context of machines is the electrical supply current. In this paper, a feature extraction algorithm based on the electric current profile is presented. The algorithm is able to detect and extract relevant features out of the current profile and determine whether fault indicators are present, in order to inform the relevant authorities to act before any loss occurred. The presented solution is optimized by using data reduction methods and is able to run completely locally on a low-cost embedded microcontroller. This provides multiple benefits, such as cost minimization, machine data protection and minimum data traffic.