Inter-turn short-circuit fault (ISF) degrades its reliability and may cause serious catastrophes for direct-drive permanent magnet synchronous motor (DDPMSM). Fault location technology can reduce maintenance time, increase the mean time between failure (MTBF), and then improve the reliability of DDPMSM. Hence, an intelligent fault locating system for DDPMSM is proposed in this paper. This system proposes a knowledge graph (KG) based diagnostic tool for detection and location of the fault coil. First, the fault model of the DDPMSM with multiple branches parallel winding is established, which is used to analyze the fault characteristics of motor. Second, the BDC and BRC are proposed as the fault indicator. The effectiveness and robustness of fault indicator are analyzed. Then, the KG system are designed and established according to the relationship between fault indicator and location of fault coil. Finally, the system is tested by data under different fault and operation conditions. The test results showed that the proposed fault locating system can detect and locate the fault coil in early stage. The minimum ratio of shorted turns to branch turns that can be detected is 0.52%. The minimum ratio of shorted turns to branch turns that can be located is 6.25%.