The abnormal state of supporting bearings significantly affects the directional accuracy of spacecraft flywheel systems. A self-sensing triboelectric nanogenerator (TENG) offers a desirable route for on-orbit health monitoring and can potentially improve the intelligence level of spacecraft. Here, an extremely compact and lightweight TENG (CL-TENG) for the nonguided clearance of bearings is proposed and fabricated for the condition monitoring of a flywheel assembly. Using the bearing radial space allows the CL-TENG to perceive the revolution and whirling behavior of the cage; therefore, the output of the CL-TENG is simultaneously affected by the bearing speed and load. A test platform for capturing the kinematic of the cage is established to validate the effectiveness of the CL-TENG based on bearing skidding, cage rotation, and whirling instability. A negative exponential correlation between the output voltage and dynamic whirling clearance index is shown to be the basis of cage whirling sensing. The internal clearance and materials of the CL-TENG are optimized to enhance the output performance while preventing wear on the flexible electrodes. The application of the CL-TENG to an actual flywheel system in a simulated space environment and nonstationary operating conditions demonstrate its advantages in detecting the abnormal operating state of the cage.