Oil debris signal analysis based on empirical mode decomposition for machinery condition monitoring

Analysis of lubricating oil is a direct and reliable approach to machinery condition monitoring. An estimate of the amount of fatigue induced metallic debris in the lubricating oil of a mechanical system can help us plan maintenance schedules. A reliably designed preventive maintenance system can reduce lost productivity and prevent catastrophic failures by timely replacement or maintenance of mission critical mechanical components. Oil-debris sensors can provide the required information on the amount of metallic debris in oil return lines. These sensors generate a signal signature similar to a single full period sine wave with the passage of a metallic particle. As such, the output of these sensors can be analyzed and an estimate of the health state of mechanical system can be obtained. However, these sensors are sensitive to vibrations of the structure where the sensor is mounted. This sensitivity leads to the distortion of the signal output. Such signals are difficult to interpret and could be misleading. As such, an imperative step towards successful machinery fault detection is signal enhancement. In this paper, we apply empirical mode decomposition (EMD) technique to extract particle signatures from the output of oil-debris sensors contaminated with vibration induced signal components. To reduce the computational burden, the acquired signal is lowpass filtered prior to the application of the EMD. The proposed algorithm has been tested using both simulated and experimental data and has shown to be effective.