Extraction of oil debris signature using integral enhanced empirical mode decomposition and correlated reconstruction

Oil debris sensors can provide important machine health information by detecting metallic particles in lubricating oil. However, the debris signal generated by the sensor is often contaminated by background noise, thereby leading to misleading detection results. Though some methods have been proposed to detect particle existence by noise reduction, they often cause severe signal distortion and hence the volume of the debris cannot be accurately estimated. This paper presents an integral enhanced empirical mode decomposition (EMD) and correlated reconstruction (CR) approach to extract particle signature with minimal distortion. The integral transform is used to boost the detectability of the weak signature and suppress the high-frequency noise. EMD with a mode cell thresholding strategy is then adopted to remove the trend components caused by the integral transform and further purify the integral enhanced signal. The trend components of decomposed intrinsic-mode functions (IMFs) are identified and removed using a high-pass filter with variable cutoff frequency. Some of the de-trended IMFs are selected based on a synthesized correlation coefficient to reconstruct the particle signature with minimal distortion. It should be also pointed out that the only pre-specified parameter in the proposed method is the level of de-trending accuracy, and no other parameters are needed to be pre-selected for its execution. The proposed approach is validated using both simulated and experimental data. The results show that the proposed method can effectively extract weak particle signatures from noisy data mixtures.