A Wear Debris Detection Method Based on Sparse Decomposition

A debris detection method for oil monitoring is proposed for the triple-coils inductive sensor. Regarding the high false alarms rate in the detection results, time domain analysis is carried out firstly to enhance the difference between debris signatures and noise signatures, in which signal energy is selected as the most effective index. Hereafter, a sliding window with fixed width is added to the signal energy values, in which sparse decomposition is performed and the potential debris signatures are determined. The evaluation criteria is the ratio of the energy of potential debris signature to the total energy within the sliding window. Signatures exceeding the threshold are considered as debris signatures. The monitoring data comes from the test runs of an aero engine of a type, all the debris are detected successfully even if one of them appears during the change in working condition. In summary, this method shows satisfactory results in both detection accuracy and the inhibition of false alarms. Moreover, it is applicable for signals from various sensing principles, including but not limited to the inductance, since the detection performance is only related to the signal itself.