Dynamic Mask Mel-frequency Cepstrum Method for Condition Monitoring of Dry Gas Seal

The dry gas seal is a crucial component for rotating equipment, and non-invasive condition monitoring is essential for ensuring its safe operation. This paper addresses the challenge of low signal-to-noise ratio (SNR) in externally acquired acoustic emission (AE) signals by proposing a dynamic mask Mel-frequency cepstrum (DMMFC) method. First, a dynamic mask library is constructed by analyzing the frequency-domain characteristics of AE signals recorded under normal operating conditions. This process facilitates the removal of dominant periodic components that are unrelated to face friction. Subsequently, a Mel-domain triangular filter bank is designed based on the spectral centroid of the residual signal, followed by the calculation of the Mel-frequency cepstrum. Experimental results from both axial and radial external measurements demonstrate that the proposed method effectively filters background noise and accurately characterizes the face friction behavior of the dry gas seal system.