Online detection and classification of changing harmonic signal content is a common issue in vehicle control, both regarding mechanical and electrical oscillations. The present paper develops a fast and numerically efficient algorithm that uses wavelet packets to detect, in real time operation, dominant oscillations as they occur in active pantograph control for electrical rail vehicles. The dominant oscillations in the investigated application are strongly time-variant, as they are influenced by train speed, catenary and rail system construction parameters, and other external factors such as wind. The paper introduces a transient indicating ratio testing in wavelet domain which defines a distance measure between two consecutive measured signals.
The defined distance illuminates the difference between two signals in time-frequency domain in order to recognize when and where a transient occurs. The effectiveness of the algorithm is demonstrated in an open-loop manner on actual measured pantograph data.