Spatial flywheels have failed repeatedly due to the friction failures of their bearing cages，which affects the safe operation and life of spacecrafts in orbit.Therefore，it is very important to evaluate the frictional state of the bearing cage of a flywheel in its ground performance test.Since cages also have friction in normal operation，how to identify their health state under friction has become an urgent problem to be solved.In order to solve such a problem，this paper presents a spatial flywheel bearing state recognition method based on the time-frequency characteristics of acoustic signals.Firstly，the acoustic signals of a flywheel in its acceleration process are obtained by the ground test bench.Secondly，the three-dimensional （3D） waterfall diagram is used to identify the frictional types of the cage，and the high-energy frictional acoustic data radiated by the flywheel during the acceleration process are obtained.Thirdly，a time-frequency analysis is carried out on the high-energy frictional acoustic data to construct the characteristic parameters that can reflect the frictional stability of the flywheel.Finally，the entropy-weighting technique for order preference by similarity to an ideal solution （Topsis） is used to construct the comprehensive evaluation index.The application results of multiple flywheels show that the bearing health state recognition method proposed in this paper has high accuracy and can provide effective guidance for flywheel ground test and screening.