Traction motor plays an important role in the reliable operation of high-speed train, so it is necessary to detect the fault of traction motor. The traction system of high-speed train is complex and the operating conditions are variable, resulting in different fault severity and fault type, which is difficult to detect. And, most of the fault detection methods are based on the time series of traction motor, and few are based on the symbol sequence. This paper presents a probabilistic finite state automata (PFSA)-based fault detection method for traction motor of high-speed trains. Three-phase current of traction motor is first converted into a symbol sequence via symbolic aggregate approximation (SAX). PFSA model is constructed to represent the symbol sequence, and state transitions in the model are described through D-Markov machine. The state transition counting matrix of PFSA model is used to detect the occurrence of faults. The proposed method is tested on a benchmark of the traction motor of a high-speed train. Through the simulation results, it is verified the effectiveness of the proposed method.