Abstract: A fault diagnosis method for early-stage inter-turn short circuit faults in the stator of a three-phase induction motor is proposed. This method is based on a one-dimensional multi-scale convolutional neural network utilizing a convolutional attention mechanism. Firstly, The research focuses on the fault diagnosis of inter-turn short circuit in the stator of an induction motor, with three-phase currents as the subject of study. A two-dimensional model of inter-turn short circuit in the induction motor is established using ANSYS. The changing characteristics of the currents under different speeds and various fault severities are analyzed. These characteristics are incorporated into the CBAM-1DMSCNN model, enabling precise stator inter-turn short circuit fault diagnosis. Subsequently, an experimental platform for inter-turn short circuit faults is established. The experiments simulated various operating conditions with different speeds and fault severities, validating the effectiveness of the CBAM-1DMSCNN model in diagnosing inter-turn short circuit faults. The fault diagnosis accuracy reached 99.4%. Finally, when compared to traditional inter-turn short circuit fault diagnosis models, the feature extraction capability and fault diagnosis accuracy of this model are significantly enhanced.