Mechanical seal is a key component in industrial production, and its safe and reliable operation are very important for the continuity of industrial production. There are many problems with the maintenance time of mechanical seals. Premature replacement of seals will increase production costs, and vice versa, it will cause security risks due to failure. Given the lack of an effective state monitoring method for the performance degradation caused by face wear of contact mechanical seals in the process of use, a method for measuring the tribological behavior of mechanical seals by face vibration acceleration was proposed. The test rig of the mechanical seal was built, the original waveform data of vibration acceleration of the stationary ring face was collected, and the tribological performance test of the mechanical seal was carried out. By replacing four rotating rings with different face surface roughness to simulate the operation to failure test of mechanical seal, the relationship between sensitive characteristic parameters of seal face acceleration and the increase of wear degree of the face was explored. By changing the rotational speed, the variation law of the sensitive characteristic parameters of the seal face acceleration with the rotational speed was explored, and then the evolution law of the sensitive characteristic parameters of the mechanical seal face vibration with the tribological regime was discussed. Based on the test data, the performance degradation evaluation method based on k-medoid clustering was attempted to establish, and the evaluation standard of mechanical seal performance degradation was established. The results showed that the face vibration acceleration monitoring method could be used to monitor the tribological regime of mechanical seals. Fuzzy entropy, mean entropy and permutation entropy were three face vibration-sensitive characteristic parameters that were sensitive to tribological regimes. With the increase of face wear degree, the face tribological regime gradually transiteds from the mixed friction state to the boundary tribological regime. The fuzzy entropy, mean value, and entropy of face vibration acceleration showed good sensitivity to the increase of face wear degree, among which the fuzzy entropy was more sensitive to the change of face wear degree than the mean value and entropy of arrangement. The fuzzy entropy, mean value, and permutation entropy of the face vibration of the stationary ring increased linearly with the increase of rotating speed under the mixed friction state of the mechanical seal. The fundamental frequency amplitude of the face vibration waveform increaseed with the increase of rotating speed, and the change of the fundamental frequency amplitude of the radial vibration waveform was more sensitive. The failure evaluation method of the mechanical seal was established. The three-dimensional vector coordinates of the original mean value of axial, radial, and tangential face vibration data were calculated in real-time. The Euclidean distance between the coordinates and the cluster center corresponding to the degenerate state was calculated. When the mechanical seal performance degradation level reached IV, it indicated that the mechanical seal might fail. The above conclusions could be used in engineering applications of seals. It was proved that the face vibration acceleration signal contained abundant information about the change of microstructure in the dynamic process and was sensitive to the change of microscopic motion in the elastic deformation. The face vibration acceleration measurement method monitoring of mechanical seals could help workers assess the status of seals and determine whether and what interventions were needed. This had great engineering value for the safe operation and life management of seals.