A drill set working condition monitoring system is introduced and investigated to improve the performance of auto-balancing machine. The experiment setup to acquire data of armature currents of spindle and servo motors is constructed. The features of the armature currents of both spindle motor and servo motor in time domain, frequency domain and time-frequency domain are extracted respectively, and they are fused separately by two distinct RBF ANNs to get the primary fusing results. The primary results are fused by the third RBF ANN to get a comprehensive result. Experiment results demonstrate that the servo motor current has a closer relation with drill set working condition than that of spindle motor, and two successive fusing operations can achieve more reliable recognition with a correctness up to 86.67%.