In this study, we applied 1D convolutional neural networks (1D-CNNs) to permanent magnet synchronous motor (PMSM) fault diagnosis on 12 common PMSM fault types, namely, normal motor (Class A), poor dynamic balance of rotor (Class B), bent shaft (Class C), magnet demagnetization (Class D), uneven air gap (Class E), rotor misalignment (Class F), stator coil three-phase imbalance (Class G), stator coil layer short circuit (Class H), poor lubrication of bearing (Class I), damaged inner ring of bearing (Class J), damaged bearing ball (Class K), and poor assembly (Class L). First, a vibration spectrum analyzer was used to measure and capture the vibration signals of a faulty motor. Then, the 1D-CNN was utilized to analyze and diagnose the captured data. The results showed that the proposed 1D-CNN method can identify 11 motor fault types with an accuracy of up to 99.7%, higher than the 96.1% accuracy of 2D convolutional neural networks (2D-CNNs). In addition, the fault diagnosis system developed in this study can perform a rapid motor fault diagnosis with a small amount of training data, significantly reducing the detection cost for PMSM fault diagnosis.