Abstract: The accuracy of pipeline analog-to-digital converters is greatly affected by non-ideal factors such as operational amplifier gain error, comparator offset voltage and capacitor mismatch. In order to address the challenges posed by non-ideal factors leading to low performance in pipeline analog-to-digital converters and the suboptimal calibration efficiency and performance of Back Propagation neural networks, a genetic algorithm is proposed for optimizing the calibration algorithm of Back Propagation neural networks. Back Propagation neural network combined with genetic algorithm is used to learn and extract the error features contained in the digital output of pipeline analog-to-digital converters, and the pipeline analog-to-digital converters is calibrated. The simulation results show that for a 12 bit 160 MS/s pipeline analog-to-digital converters, the effective number of bits and the spurious free dynamic range of the pipeline analog-to-digital converters after calibration are increased from 7.01 bit and 38.22 dB to 10.97 bit and 82.65 dB, respectively.