Abstract:The degradation of stator winding turn insulation is one of the primary causes of inverter-fed machine failures. Online monitoring of the stator winding turn insulation degradation can detect potential safety hazards in time，but faces the challenge of weak turn insulation degradation characteristics. In order to improve the sensitivity of turn insulation degradation，a turn insulation condition monitoring method for inverter-fed machine using fractional-order Fourier transform combined with convolutional neural network（FrFT-CNN）is proposed. Firstly， the sensitivity of the high-frequency switching oscillating current to the turn insulation change is analyzed in the fractional domain. Then，the one-dimensional convolutional neural network model suitable for turn insulation state monitoring of inverter motors is designed. The experimental results show that the FrFT-CNN method significantly improves the accuracy of turn insulation state monitoring compared with the traditional convolutional neural networks method，and also exhibits the advantage of higher stability.