In this paper, the dynamic current (IDDT) of CMOS circuit is studied, and a preprocessing method for IDDT is proposed. the fault characteristics are obtained from the standard deviation and the skewness of the IDDT. This paper uses CMOS NAND gate circuit simulation, the method of wavelet transformation pretreatment methods were compared, and combined with the probabilistic neural network classification method of fault diagnosis. The simulation results demonstrate the effectiveness of the proposed method based on the IDDT standard deviation and the skewness of the method, and the average correct rate of the fault diagnosis of the transistor bridge fault, the open circuit fault and the transistor parameters is 90.7%.