Detection of damaged components in aircraft engines by visual analysis of spectrograms in frequency and order domains remains a necessary and efficient processing. Even if automatic algorithms have been developed, fine tuning of those algorithms to detect damaged components expected signature still requires supervision by a vibration specialist. Moreover, novelty detection algorithms cannot replace supervised analysis for vibrations sources identification. In order to facilitate visual analysis of spectrograms, two algorithms of morphing between spectrograms are described. The first algorithm presented in this paper allows a smooth transition between frequency domain and order domain. It is composed of the following steps: i) creation of a virtual constant rotating speed, ii) interpolation of a set of virtual rotating speeds between original rotating speed curve and the virtual constant rotating speed, iii) synchronous resampling of vibration signal with previously computed virtual rotating speeds. The second algorithm allows a smooth transition between two different order domains, and is composed of the following steps: i) interpolation of a set of virtual rotating speeds between the two original rotating speed curves, ii) synchronous resampling of vibration signal with previously computed virtual rotating speeds. With this process, frequency and order resolutions are preserved, while intermediate representations help to understand how a vibration signature is transformed from one representation to another. Several examples are given from a set of real data from aircraft engines in order to show the effectiveness of this new processing, besides pedagogical purpose, in the application to aircraft engine Health Monitoring.
展开▼