A fractional calculus technique for on-line detection of oil debris

The quantity and size of the metallic debris falling off machine components due to fatigue, corrosion and pitting provide valuable information about the machine structural integrity. For this reason, in-line oil debris sensors have been widely adopted for on-line fault detection. However, the effectiveness of such sensors relies on how the acquired data are processed. The existing method for identifying debris or particle signatures is threshold based. Specifying a proper threshold is application dependent and knowledge demanding. Even if a 'proper' threshold can be determined, it is useful only when the signal-to-noise ratio (SNR) is sufficiently high. However, the signal is often contaminated by noise, making the threshold irrelevant. As such, we propose a fractional calculus technique consisting of two detectors to enhance the detection process. The first detector is used to extract the 'particle-like' waveform. Nevertheless, if the signal is weak, the signature obtained by this detector cannot be definitely classified as a true particle signature. The second detector is hence used to confirm or dismiss the signature authenticity depending on whether it can locate the three inflection points—a key feature of a particle signature. This technique does not require any threshold and therefore can avoid the difficulties present in the existing method. Both simulated and experimental results have shown that the proposed technique is robust in a noisy environment and can effectively detect small particles. Furthermore, this technique does not require additional filtering or de-noising steps, and signal processing can be simplified accordingly.