Abstract: In response to the fault vibration signal of upper hinge point of soft yoke mooring system susceptible to interference of other signals, the theory of minimum entropy deconvolution was proposed. With the thrust roller bearing in mooring leg of Mingzhu FPSO as the study object, the paper designed a large scale model experiment. The model experiment includes three parts, i.e. single point platform, soft yoke mooring system and mooring support. In the experiment, the platform with six degrees of freedom replaces the floating body to do the yaw movement. By comparing the acceleration signals of the intact thrust roller bearing and the faulty thrust roller bearing, the bearing fault was analyzed, and the feasibility of the minimum entropy deconvolution (MED) analysis method and the correctness of the model experiment were verified. The results show that the MED method can highlight the pulse impact components of vibration signals, and judge and recognize the initial working state of bearing. Meanwhile, the model device considering bearing diagnosis also provides an experimental platform for professional teaching and scientific research for colleges and universities.