Aiming at the real-time diagnosis and accurate fault tracing of motor over-current in traction drive system, a real-time diagnosis method based on fault time-series feature pattern recognition is proposed. Firstly, the common fault sources leading to over-current of traction motor are analyzed. Then, by summarizing the change law of event time-series strongly related to working conditions, the diagnosis rule base of motor over-current fault time-series feature pattern is designed off-line. Then, the relevant sensor signals and system working condition information are collected online in real time, the relevant feature quantities and event flags are calculated, and the patterns and diagnostic rules in the off-line designed diagnosis rule base are matched in real time, so as to realize the real-time diagnosis and accurate tracing of traction motor over-current fault. Finally, the effectiveness of the proposed method is verified by field fault cases testing.