Abstract: The oil debris size and density information can reflect the lubrication condition of the rotating equipments, which can help to take appropriate measures to prevent premature failure. The disturbance generated when the ferromagnetic debris passed through is converted to corresponding voltage signals based on the electromagnetic oil debris sensor and then is transformed into true RMS signal by a hardware circuit. An oil debris online identification algorithm is proposed based on a locally weighted regression scatter plot smoother(LOWESS) to noise reduction and feature extraction in order to obtain the real-time information of the size and density of the ferromagnetic particles in the temperature drift, zero drift, electromagnetic interference and other noise environments. Finally, the feasibility and effectiveness of the proposed method is proved by the contrast between the detected results of the MetalSCAN3110 sensor and our method.