Research on friction state monitoring method of liquid film seal during start-up based on acoustic emission

Abstract

This paper proposes a method based on acoustic emission to achieve nondestructive testing of the friction state of a liquid film seal. However, the acoustic emission signal is susceptible to noise, and extracting the feature signal is challenging. To address these issues, we propose a signal processing method based on singular value decomposition and adaptive variational mode decomposition (SVD-AVMD). Additionally, we investigate the acoustic emission mechanism of the liquid film seal using the strength theory of shear strain energy and CEB contact model. We design a liquid film seal friction experiment based on acoustic emission technology to extract the characterization signal of the friction state of the liquid film seal using SVD-AVMD. Finally, we apply a convolutional neural network to identify the friction state of the liquid film seal. The results demonstrate that SVD-AVMD has a significantly better ability to capture the center frequency of each mode component and to recover each mode component compared to simple variational mode decomposition. SVD-AVMD can filter out background noise while retaining the most useful information to the maximum extent possible. Furthermore, the root mean square of the extracted AE signal is consistent with the description of the AE mechanism, achieving the goal of studying and judging the friction state of the liquid film seal end face. Finally, we show that combining the convolutional neural network and acoustic emission time-frequency characteristics can improve the recognition accuracy of the friction state of the liquid film seal.