CMOS inverter electromagnetic reliability prediction method based on neural network

The invention discloses a neural network-based CMOS (complementary metal oxide semiconductor) inverter electromagnetic reliability prediction method. The method comprises the following steps: 1, modeling through a TCAD device, and completing physical model building and basic characteristic verification of the CMOS inverter; 2, screening parameters influencing the electromagnetic reliability characteristic of the CMOS inverter; 3, changing a parameter input combination through a control variable method, and carrying out a simulation experiment to obtain a sample set; 4, establishing a neural network to predict damage types of the CMOS inverter under different conditions, and respectively predicting the two electromagnetic damage quantization parameters in the step 2 under the condition of different damage types; 5, training a neural network by using the sample set data; and step 6, finally performing function verification and effect evaluation on the trained network by using test set data. According to the met