Condition Assessment on Mechanical Seal Face Wear Based on Incomplete Prior Knowledge

The prior knowledge of the mechanical seal wear condition during the normal running process is very difficult to obtain, however, the opening and closing condition prior knowledge of mechanical seal is easy to obtain. According to these characteristics, taking the acoustic emission signal as the monitoring signal to detect and assess the mechanical seal face wear condition based on the method of Factor Hidden Markov Model (FHMM). Empirical mode decomposition (EMD) method was applied to extract the original signal feature. Using the contact condition prior normalized feature information of mechanical seal face to establish the FHMM. The logarithmic likelihood degree between the unknown condition feature vector and the monitoring model established can be calculated, then compared it with the seal face performance index to evaluate the wear conditions of a mechanical seal face. The experiment show that this method realized the wear condition assessment of mechanical seal using small number of samples.