The rolling bearings in moment wheel assemblies (MWAs) or control moment gyros (CMGs) are not only the core components in spacecrafts but also prone to failure. Therefore, a high reliability is the critical characteristic for spacecraft bearings, and long-life testing on the ground is one of the main means for bearing reliability assessment. In practical applications, a convenient and reliable method is required for monitoring the health status of abnormal bearings in MWAs during the long-life test. In this paper, a monitoring method based on the clustering fusion of normal operation acoustic parameters is proposed for the identification of abnormal bearings. Firstly, the characteristics of MWA's acoustic signal and its feasibility as a monitoring medium are clarified based on tests and modal analysis. Then statistical parameters and sound quality parameters are introduced to characterize the changes caused by bearing faults, and the root mean square (RMS), kurtosis, and sharpness parameters are selected to construct the feature vectors. The K-medoids clustering technology is used to fuse the characteristic parameters, and the safety distance for normal bearing operation can be obtained by a suitable method. Finally, the abnormal index is presented based on the excess rate and excess distance to judge the abnormal states of several types of bearings through tests. The research results indicate that the presented monitoring method based on the clustering of the normal operation acoustic parameters can not only identify various faults of the spacecraft bearing (ball defects, outer ring defects, cage instability, etc.) effectively but also give a quantitative evaluation of the severity of the abnormality.