Flow-State Identification of Oil-Based Magnetic Fluid Seal Based on Acoustic Emission Technology

Abstract

At present, most research studies on the changing process of the magnetic fluid seal are analyzed with the pressure signal of each chamber or the magnetic fluid flow photos taken by a camera, which need to change the seal structure. Based on nondestructive acoustic emission technology, a flow-state identification model of the oil-based magnetic fluid seal using the grey wolf optimizer and random forest is proposed in this study. The acoustic emission signal and pressure signal are collected at the same time under static conditions in the two-stage pole shoes oil-based magnetic fluid seal experiment. Through power spectrum analysis of the acoustic emission signal with the aid of pressure signal, the changing process before seal failure is divided into three states: no magnetic fluid flow, the first pole shoe magnetic fluid flow, and two pole shoes magnetic fluid flow together. Then, the time- and frequency-domain features of acoustic emission signal samples are extracted to form feature vectors as inputs, and the flow-state identification model is established based on the grey wolf optimizer and random forest. The experimental results show that the testing accuracy and F1 scores (the index representing the precision and recall at the same weight) of three states are close to or higher than 90%. The effectiveness of oil-based magnetic fluid seal flow-state identification model based on non-destructive acoustic emission technology is proved.