Industrial machinery diagnosis by means of normalized time-frequency maps

The development of intelligent and autonomous monitoring systems applied to rotating machinery represents the evolution towards the automatic industrial plants supervision. In this paper, an original method to detect camshaft defaults from the monitoring of the motor phase current is presented. This method is based on the short-time Fourier transform in order to analyze the spectral variations over each cycle of the system. The time-frequency maps are then normalized using statistical techniques in order to create a reference of the healthy functioning of the system. Normalized time-frequency maps allow the detection of changes from the reference that are statistically significant. The method is evaluated on data from an industrial packing machine at three different speeds and for two noise levels. It obtains excellent results with 100% correct detections and 0% false alarms in each case. Results are compared to those obtains with classical spectral approaches.