The aim of the work is to investigate the relationship between bearing radial clearance and vibration frequencies during service. The bearing clearance can be determined based on the energy spectrum of the rotation vibration by the application of neural networks. Experimental investigations are performed using two double-row self-aligning ball bearings and vibration sensors in order to examine this assumption. The bearing vibrations are recorded at different clearances and rotation frequencies and analyzed. The developed system can be used for monitoring or prediction of clearance during service.