Hydrodynamic Mechanical Seal End Faces Condition Monitoring And Health Assessment

Mechanical seal is the key component of a rotary fluid machinery to prevent leakage,and it is also the powerful guarantee of safety production and environmental protection in modern process industry. Monitoring the operating state of mechanical seal end faces and assessing their health condition would help the early warning of the mechanical seal failure.Based on the analysis of working mechanism and failure mechanism, and by taking hydrodynamic mechanical seal as the research object, the acoustic emission (AE) monitoring method for seal opening speed and fluid film stability has been studied in this paper through advanced signal processing and pattern recognition. Then by extracting the features of seal wearing, the assessment method for health condition of hydrodynamic mechanical seal end faces has also been studied. The main contents and achievements are as follows:(1) The working mechanism of seal has been expounded, and the feasibility about assessing the health condition of seal end faces by using seal opening speed and film stability index has been studied. On the hydroynamic-static mechanical seals experiment table, a multi-sensor monitoring system based on eddy current, acoustic emission and vibration detection has been constructed, and finally, a series of experiments about seal opening speed, film stability and seal wearing have been operated for paper's data support.(2) After preliminary analysis, the michanism of the signal production and variation has been sxpounded, then the state of seal opening and the criterion of film stability have been denoted. In addition, the noise characteristics and the vibration characteristics at different seal speed have been studied, which would provid a basis for the research about the detections of seal opening speed and fluid film stability.(3) The detection method for opening speed of hydrodynamic mechanical seal based on AE monitoring has been studied. For the problem of the difficulties in signal denoising,which is caused by the random overlap of the useful signal and the background noise in frequency domain during the seal startup, a new modeling approach based on genetic particle filtering (GPF) and least square support vector machine (LS-SVM) is presented. First, the GPF is used for AE denoising, then some features are extracted, and LS-SVM is used for friction state modeling. On the basis, the seal opening speed can be obtained by the friction state changes at different seal speed, and the experimental data shows that the difference of the above result and reference value is less than 3 r·min-1, which confirms the effectiveness of the paper's method.(4) The detection method for film stability of hydrodynamic mechanical seal based on AE monitoring has been studied. For the problem of the terrible signal-to-noise ratio (SNR)of the AE signal and the non-uniform sample information while the seal is working, a new modeling approach based on particle swarm fast independent component analysis(PS-FastICA) and support vector data description with negative examples (SVDD-Neg) is presented. First, the PS-FastICA is used for weak signal extraction, then some features are extracted, and SVDD-Neg is used for film stability modeling. On the basis, the seal film stability can be judged by identifying a series of samples in an observation cycle, and the experimental data shows that the accuracy rate of the above result is higher than 90%, which confirms the effectiveness of the paper's method.(5) The assessment method for health condition of hydrodynamic mechanical seal end faces based on seal opening speed and film stability has been studied. Based on the analysis of several typical types of seal wear, two Elman neural networks are constructed for assessing the health condition of the seal end faces. Then for the problem of fusion failure which is caused by dealing with high conflict evidences by using D-S evidence theory, a new fusion rule is presented to fuse the two neural networks' decisions. Experimental data shows that the paper's rule not only can raise the accuracy rate of the health assessment for hydrodynamic mechanical seal end faces, and it also has stronger robustness and reliability than some other classical fusion rules in practice.